the effects of raingage network density on weather … · 6th international symposium on...

6TH INTERNATIONAL SYMPOSIUM ON HYDROLOGICAL APPLICATIONS OF WEATHER RADAR MELBOURNE, AUSTRALIA. 2-4 FEBRUARY 2004

THE EFFECTS OF RAINGAGE NETWORK DENSITY ON WEATHER RADAR PRECIPITATION DATA CORRECTION THROUGH GROUND DATA F. Amiryazdani, M.Khalili, M.Kalantarzadeh, F.Golkar National Cloud Seeding Research Center of IRAN [email protected] ABSTRACT: Although the advantages of data concerning estimated precipitation via weather radars, in comparison with the rain gage networks are numerous which can be counted as: coverage of extensive areas, easy reading, quick access to real time precipitation data and high time and spatial resolution; there are many error sources that may affecting the radar estimated precipitation data such as: radar type, distance from the radar, precipitation characteristics and etc. So, for the application of radar precipitation data, one has to correct them by the ground data continuously. Usage of such information in order to correct the radar estimates are confronted many limitations like: inadequacy of pluviometer network coverage, inefficient density network and systematic errors. Endeavor has been made in this article to introduce a practical example to show the density effects of a rain gage network on the correction of estimated precipitation via weather radar. This example has enjoyed the received data about accumulated precipitation from a 3.2cm wavelength Russian weather radar (MRL-2) per a cloud seeding season (Jan.-April 2001) in the central part of Iran, as well as the simultaneously daily precipitation data from 55 neighboring rain gage stations with intensity less than 20 mm/day.

1

mailto:[email protected]


ACCURACY VERIFICATION OF SPACE AND GROUND-BASED RADAR ESTIMATES OF RAIN RATE

E. Amitai and L. Liao

NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA [email protected] ABSTRACT: The distribution of rain rates (R) is of great concern for many fields. For example, hydrological applications such as flood forecasting depend on an accurate representation of the excess rainfall--driven by R--that does not infiltrate into the soil. Better estimation of the spatial probability distribution function (PDF) of R is also crucial for better evaluation of rainfall products from ground and space-based radars. For example, the evaluation of Tropical Rainfall Measuring Mission (TRMM) Ground Validation products, in particular, the instantaneous ground-based radar rainfall products, are based on comparing PDFs of R derived from the radar estimates above the gauges to those derived from the gauges themselves. Understanding the PDFs of R will be even more important for space-based radars (e.g., TRMM, Global Precipitation Measurement [GPM] mission). The evaluation of instantaneous rainfall products and rain rate estimates from space is quite a challenge. On one hand, scatter plots of direct comparisons with ground-based radar estimates (pixel by pixel) are extremely noisy for several reasons (e.g. sample volume discrepancies, timing and navigation mismatches, uncertainties in the ground-based radar estimates), and therefore of little use for validation. On the other hand, the evaluation process based on the integral properties of R, e.g. rain accumulations over various time or space intervals, suffers from large uncertainties due to the satellite revisit time, which is of the order of hours or days. Therefore, an alternative method of analyzing PDF of radar R might be very attractive for evaluating satellite-based precipitation products, such as those from TRMM Precipitation Radar (PR). Comparisons of PDFs of R and cumulative distribution functions (CDFs) from TRMM PR and co-located WSR-88D radar observations are presented. The comparisons are made before and after rain type classification using observations made during more than 30 months in central Florida. The rain type classification allows better evaluation of the algorithms at different conditions, and potentially will allow for “extrapolation” of the uncertainties to regions not covered by validation data sets, but characterized by the same rain types. A PDF analysis using gauge and ground-based radar data is used to demonstrate an optimal gauge adjustment technique for improving rain rate distribution estimates for ground-based radars. The uncertainties in the ground-based radar PDF are reduced upon adjustment to the gauge-based R distribution. This paper demonstrates the use of PDFs for validation of R estimates from space and ground based radars. It also discusses how well the gauge-based PDFs represent the actual R distribution at the scale of a radar pixel- a key factor in using ground based radar PDFs.

2



EXPERIMENTAL INVESTIGATION OF AN X-BAND POLARIMETRIC ALGORITHM FOR ATTENUATION CORRECTION AND MICROPHYSICAL

RETRIEVALS Marios Anagnostou Emmanouil N. Anagnostou

Jothiram Vivekanandan Civil and Environmental Engineering, Univeristy of Connecticut, Storrs, CT 06269, USA [email protected] ABSTRACT: This research explores the synergy of dual-frequency and polarization radar observations of convective systems with varying size, precipitation type, and intensity. These measurements were collected during the International H2O Project (May 15 – June 20 2002) by the University of Connecticut’s mobile X-band Polarimetric (XPOL) radar and NCAR’s S-band Polarimetric (SPOL) radar. The objective of our study is to use the low-attenuated S-band measurements and the more definitive S-band Polarimetric retrievals of raindrop size distribution (DSD) to evaluate an X-band attenuation correction and DSD-profiling technique. The X-band rainfall estimation technique is based on algorithms that couple along-a radar-ray profiles of attenuated horizontal polarization reflectivity (ZH), attenuated differential reflectivity (ZDR), and differential propagation phase shift (ФDP). Based on a three-parameter “normalized” gamma DSD model, relationships are derived that correct X-band reflectivity profiles for specific and differential attenuation using jointly ZH and ФDP profiles, while simultaneously retrieving variations of the “normalized” intercept DSD parameter. Consequently, using a constrained relation between the shape and slope parameters of Gamma DSD model along with the retrieved “normalized” intercept parameter value and rain water content we estimate all three parameters of Gamma DSD for discrete space intervals along a radar ray. The ZH and ZDR attenuation correction and DSD retrievals are assessed based on the coincident SPOL data. An existing algorithm that combines non-attenuated ZH and ZDR measurements based on a same constrained relation between the shape and slope parameters of Gamma DSD model is used to retrieve DSD parameters from SPOL observations, thus facilitating the XPOL validation in terms of both ZH and ZDR attenuation correction and DSD retrievals.

3



A SINGLE DOPPLER RADAR RETRIEVAL OF HORIZONTAL WIND FIELD BASED ON A KALMAN FILTER

Hervé Andrieu and Brice Boudevillain Laboratoire Central des Ponts-et-Chaussées (LCPC), route de Bouaye, BP 4129, 44341 Bouguenais, France. [email protected] ABSTRACT: This paper presents a simple method to retrieve horizontal wind fields by the combination of a classical advection procedure and measurements of radial velocity provided by a single Doppler radar. The proposed method is formulated in the framework of a Kalman filter. The state variables are the two Cartesians components of the horizontal wind. The observations come from radial velocity measurements and from the horizontal wind obtained with radar reflectivity by a cross-correlation method. The model assumes that the horizontal wind structure between two radar scans remains stationary. The model error covariance matrix is computed according to the correlation coefficient between the observed radial velocity field and the predicted field projected on the radials. The measurement error covariance matrix for horizontal wind obtained by cross-correlation method is computed according to the correlation coefficient between the observed radial velocity field and the horizontal wind field obtained by cross-correlation method and projected on the radials. In a first feasibility study, radar reflectivity factor and Doppler velocity are simulated from outputs from a three-dimensional physically based mesoscale meteorological model. The proposed method seems to give encouraging results with a low computational cost.

4



EVALUATION OF A VIL-BASED VERY-SHORT TERM RAINFALL FORECASTING MODEL WITH ACTUAL RADAR DATA

Hervé Andrieu and Brice Boudevillain Laboratoire Central des Ponts-et-Chaussées (LCPC), route de Bouaye, BP 4129, 44341 Bouguenais, France. [email protected] ABSTRACT: This paper presents an evaluation of a very short-term rainfall forecasting method making use of voluminal radar data. Voluminal radar data allow to estimate the VIL that is the total rain water content of the atmosphere. The forecasting method is based on the conceptual modelling of the VIL evolution. It combines a simple mass balancing of water within air columns and the spatial advection of the variables using information from consecutive time steps. The method was initially tested using radar data simulated from model outputs in order to highlight its interest independently from measurement errors. The results clearly shows: i) that the information of VIL may be useful for quantitative rainfall prediction, ii) the influence of taking into account or not the ice phase in the VIL estimation. The forecasting model is now tested with actual radar data taken several rain events recorded during the intensive observation period of the Mesoscale Alpine Programme.

5



SMALL MOBILE X BAND RADARS IN SMALL HILLY CATCHMENTS: FIELD

EXPERIENCES Geoff Austin, Andrew Peace and John Nicol, Alan Seed Atmospheric Physics Group University of Auckland New Zealand ABSTRACT: The University of Auckland small mobile X band weather radar has been deployed in a variety of meteorological and hydrological field experiments over the last decade. There have been a number of successful outcomes as well as some problems encountered during this period. Examples of the data collected from a sub tropical cyclone landfall, strong orographic rainfall, a winter small catchment study and an even an erupting volcano will be described. In the light of this experience a new radar has been developed and the results of preliminary tests of this system are presented.

6


USE OF RAIN FORECAST IN HYDROLOGICAL APPLICATIONS Shaukat Ali Awan Flood Forecasting Division, Pakistan Meteorological Department [email protected] ABSTRACT: The S-band QPM Doppler Radar (3GHz radar) gave useful information about the rainfall centroid particularly over the areas across boarder. The Doppler effect was used to keep on constant track of movement both of convergence and divergence pattern. The three dimensional cloud image helped in locating the heavy rainfall over a particular sub-basin, which tremendously improve the accuracy of flood forecast during the monsoon season of 2002. During monsoon season (July to October) due to heavy rainfall the Indus basin is vulnerable to flooding and particularly the people living in the otherwise dry river beds are swept away because of flash flows. In Indus river basin the two tributaries i.e. river Sutlej and Ravi receive flood flows during monsoon season and the upper catchments of both the rivers lie across the boarder in India. Under Indus River Treaty the flood flows are received daily. The people living in this region reside with their cattle in the flood plain, which is generally dry during October to June. The flood forecasting division issued significant flood forecasts 48 hours in advance in August 2001 and got the riverbed evacuated through the Relief department and District authorities. Similarly in August 2001 in river Ravi significant flood forecast was issued 24 hours in advance and the people alongwith their cattle were shifted to safer location and as such the amount to be spent on post rehabilitation and the damage to the life, property and cattle of the poor people/farmers was saved. The issuance of early flood forecast and warning and its implementation through effective liaison with district administration authorities had a salutary effect and the confidence with the public improved in flood forecasting authority, which was appreciated both in press and in the Government.

7



RADAR INTER-COMPARISONS Shaukat Ali Awan Flood Forecasting Division, Pakistan Meteorological Department [email protected] ABSTRACT: The C-band QPM Radar (5GHz radar) at Islamabad and network with other four radars gave not only area coverage by one radar but the other radar as well, while both being installed at a distance of 250 kilometers away with area over lapping by both the radars. The radar networking particularly by QPM monitoring in a cloud cell from two different stand points gives useful information and has supplementary impact and particularly it has been noted that in outer most radar zone the second radar gives much better tracking and measurement level. In certain cases it so happened that the cloud cluster in the shadow of other cloud is missed by one radar because of signal attenuation in its path, however the same is detected by the other radar, which is in network and giving a composite picture, helps in locating the missing cloud cluster. In February 2003 a tornado formation was detected which lasted for 35 minutes because of a squall line development, only its signatures in the C-Band QPM radar making it extremely useful and reliable equipment for issuance of warnings and alerts, thus making the forecasting service trustworthy public utility.

8



VALIDATION OF A NOWCASTING TECHNIQUE FROM A HYDROLOGICAL PERSPECTIVE

Marc Berenguer Carles Corral, Daniel Sempere-Torres and Alan Seed Grup de Recerca Aplicada en Hidrometeorologia (GRAHI). Universitat Politècnica de Catalunya, Barcelona (Spain). [email protected] ABSTRACT: Nowcasting precipitation is a key point to anticipate risks in flood warning systems. In this environment, weather radars are very useful because of the high resolution of their measurements both in time and space. The aim of this study is to assess the performance of a recently proposed nowcasting technique (SPROG) from a hydrological point of view. This technique is based on the advection of radar precipitation fields and its main point is that the forecasted fields get smoothed as the forecasting time increases, to filter out the smallest scales of the field when they become unpredictable. The evaluation of the forecasted precipitation fields is done in two different ways: a) comparing them against the actually measured precipitation fields and b) according to the concept of “hydrological validation”, comparing the hydrographs calculated by a distributed rainfall-runoff model simulating operational conditions (using the forecasted precipitation fields) against the hydrographs calculated by the model with the entire series of radar measurements. This part of the study has been carried out in the framework of the Besòs basin flood forecasting system.

9



A PROBABILISTIC PRECIPITATION NOWCAST SCHEME DESIGNED TO

ACCOUNT FOR UNCERTAINTIES IN ADVECTION VELOCITY AND DEVELOPMENT

Neill E. Bowler, Alan Seed and Clive E. Pierce Met Office, Joint Centre for Hydro-Meteorological Research, Maclean Building, Crowmarsh Gifford, Wallingford, Oxfordshire, OX10 8BB, UK. [email protected] ABSTRACT: Precipitation nowcasting methodologies exploit high-resolution meteorological observations and simple extrapolation techniques to generate short range (up to six hours ahead), quantitative precipitation forecasts with high spatial (< 5 km) and temporal (< 1 hour) resolutions. These nowcasts can be particularly valuable in the production of fluvial flood forecasts and warnings. A probabilistic precipitation nowcasting scheme has been developed which aims to quantify the uncertainties in the motion and evolution of radar observed precipitation fields. In this scheme uncertainties in the evolution of radar observed precipitation fields are modelled by considering uncertainties in the field motion separately from those attributable to field evolution in Lagrangian space. The uncertainties in the evolution of the precipitation pattern are modelled using S-PROG, which recognises and exploits the fact that the temporal persistence of precipitation features is scale dependent: small-scale precipitation features tend to be shorter lived than larger scale features. In the probabilistic scheme, knowledge of the loss of deterministic predictive skill as a function of scale is used to progressively introduce stochastic features from the smallest scales upwards. An ensemble of possible forecast scenarios is produced, making the system ideal for hydrological applications. A recent performance evaluation of the scheme suggests that the stochastic scheme possesses predictive skill at lead times in excess of six hours.

12



CELL DETECTION AT FAR RANGES SUPPORTING TB-Z RELATIONSHIPS R.V. Calheiros M.A. Loma and R. Machaado

Meteorological Research Institute/IPMET-UNESP and National Institute for Space Research/INPE [email protected] ABSTRACT: Detection of cells at the outermost range was effected since the early days of operation of the radar surveillance facility in Bauru, Central Sao Paulo Brazil (22010’30“S, 51022’22”W, 460 m). With the presently operating system, an S-Band Doppler (1992) which substituted for the original C-Band (1974), the range impacted top of the highly developed summer thunderstorms at 400 to 450 km from the radar is usually detected. While having a large potential for heavy rain indication in a considerably extended area around the radar, the fact that the tip region of the core of the storm is seen significantly limits its application, e.g. for flood warnings. In the search to mitigate that limitation a “cloning-like” procedure for radar cells at long ranges was idealized and is being developed at the Meteorological Research Institute-home of the Baun radar. The technique is based on the synergestic use of satellite and radar, resulting in a “rebuilt” cell at far ranges by surrounding the tip of the core (range corrected) with the corresponding satellite observation. The satellite data used are the brightness temperature (Tb) tramsformed into Z(R) values. Basic to the procedure are, then, appropriate Tb-Z relationships. A preliminary test has been performed, with the Baun radar and the HSB(Humidity Sensor for Brazil) channel 4 (183.3GHz) data for the Tb-Z conversion. While intended primarily for the summer period, with intense convective activity, the procedure has undergone this first test with the restriction of satellite data availability for the transition dry-to-rainy period only. In this paper, the development of the first Tb-Z relationships involving the summer is presented. Also new relationships for a larger number of events in the transition season are shown. On the other hand, because microwave sensors can, until now, be exclusively on board polar orbiters the frequency of sampling in time is low. One alternative is the use of geostationary measurements e.g. in the IR band, to interpolate between the microwave observations. A preliminary verification of the relationship between the IR measurements and Z is also described in this work. First results suggest that the derived probability matched Tb-Z relationships indicate that the appropriately saturation-like behaviour at the upper and lower ranges of values. For the IR data, results indicate compatibility with those obtained previously in the area of the Baun radar. Attempts to identify stratifications based on rain type, now being carried out, are described. Also considered are impact effects expected in the use of the Baun radar for determination of flooding conditions at the Paraiba valley, a sensitive region (large population, economic activity) in the radar furthest range.

13



DUAL POLARIZATION RADAR ESTIMATES OF RAINFALL: RECENT ADVANCES

V.Chandrasekar and V.N.Bringi Colorado State University ABSTRACT: Polarization diversity measurements of rainfall have advanced significantly providing more understanding of the rainfall microphysics, thereby contributing to the advancement of the rainfall estimation process. The application of polarimetric radar data to the retrieval of raindrop size distribution parameters and rain rate in samples of convective and stratiform rain types is presented. Data from the Colorado State University (CSU), CHILL, NCAR S-band polarimetric (S-Pol), and NASA Kwajalein radars are analyzed for the statistics and functional relation of these parameters with rain rate. Surface drop size distribution measurements using two different disdrometers (2D video and RD-69) from a number of climatic regimes are analyzed and compared with the radar retrievals in a statistical and functional approach. The understanding of the advances in rainfall microphysics is used to develop a polarimetrically based reflectivity rainfall ( Z-R ) relation from radar data. The ability of dual polarized radars to estimate the DSD parameters provides an opportunity to address an important question whether the knowledge of DSD matters in the rainfall estimation process or it is predominantly a statistical/ engineering regression problem. Rainfall estimation process at radar scales can be seen from two fundamental approaches namely, physical approach or statistical/engineering approach. Based on each approach algorithms have been developed in the past. Physical approaches can track small scale rainfall variability, which cannot be done by engineering approaches such as Neural network based radar rainfall estimation techniques. Z-R algorithms which have been developed historically as regression solutions have been placed on firmer physical basis recently, using the scaling and normalization concepts in raindrop size distributions. This paper will provide a synthesis of physically based approach and statistical/engineering approaches using data from dual-polarization radar observations.

14


REAL-TIME RADAR RAINFALL ESTIMATES S. Chumchean A. Seed and A. Sharma

School of Civil and Environmental Engineering, The University of New South Wales ABSTRACT: This paper presents a method to estimate radar rainfall in real-time. We propose a procedure for automated correction of mean field bias using hourly rainfall accumulation from the rain gauge network, which aims to reduce the difference between radar and rain gauge rainfall. The underlying philosophy for estimation of radar rainfall in real-time used in this study is that the estimated radar rainfall must first be corrected for the reflectivity measurement error and the Z-R conversion error based on the physical methods, and then a statistical method will be used to remove the average difference (mean field bias) between radar estimates at the rain gauge locations and the corresponding gauge rainfall amounts. Kalman filter techniques are used for the statistical approach to estimate mean field bias in real-time. The logarithm mean field radar rainfall bias is modeled as an AR1 model with a stationary variance. A radar rainfall error variance model is used to estimate an observation error variance of each hour, which depends on rainfall intensity, number and location of rain gauges that are used to estimate the mean field bias of that hour. The 6-month radar and rain gauge data record from the Kurnell radar in Sydney, Australia are used to test an efficiency of the proposed method. The result of the analysis presented here are also used to assess the number of rain gauges required in radar calibration that best compromises the accuracy of the measured radar rainfall and the rain gauge operational costs.

15



EMPIRICAL MODELING OF THE UNCERTAINTIES IN RADAR RAINFALL ESTIMATES

Grzegorz J. Ciach Mekonnen Woldemariam, IIHR – Hydroscience & Engineering The University of Iowa [email protected] ABSTRACT: Mathematical modeling of the way in which radar rainfall (RR) approximates the truth is one of the ways to quantify the RR uncertainties that has not been explored so far. In this approach one can represent RR in the form of an “observation equation,” that is, as a function of the corresponding true rainfall and a random error process. The error process describes the cumulative effect of all the sources of RR uncertainties that are variable in space and time. We present the first results of our work on the identification and estimation of this relationship. These results are based on the Level II reflectivity data from the WSR-88D radar in Tulsa, Oklahoma, and rainfall measurements from 23 surrounding Oklahoma Mesonet raingauges. The data sample consists of the corresponding pairs of radar and raingauge hourly rainfall that cover 50 rainy days selected from the period of 1994-95. About 2300 of these pairs have the raingauge rainfall values no less than 0.5 mm. The RR estimates were obtained using the standard NEXRAD Z-R relationship (A=300, b=1.4), and clutter/AP removal and VPR correction procedures were applied prior to the uncertainty analysis. Accumulation intervals from one hour to one day were analyzed using this sample. The raingauge accumulations were used as an approximation of the true rainfall in this study. The RR uncertainty model that we explored is factorized into a deterministic distortion that is a function of true rainfall and a multiplicative error factor that is a positively-defined random variable. The standard deviation of the error factor depends on the true rainfall, however, its expectation is always equal to one. With this constraint, the deterministic distortion function can be defined as the conditional mean of RR conditioned on the true rainfall. We show that the error factor can be modeled with satisfactory accuracy using gamma distribution. We use nonparametric regression to estimate the deterministic distortion and the error factor variance as functions of the true rainfall. The results show that the deterministic distortion is a nonlinear function of the true rainfall that indicates systematic overestimation of week rainfall and underestimation of strong rainfall (conditional bias). The standard deviation of the error factor is a decreasing function of the true rainfall that ranges from about 0.8 for weak rainfall to about 0.3 for strong rainfall. Both the deterministic distortion and the random error seem to be relatively invariant in respect to the accumulation interval.

16



ON THE IMPACT OF ASSIMILATING DOPPLER RADAR WIND DATA INTO

AN OPERATIONAL NUMERICAL WEATHER PREDICTION MODEL C. G. Collier F. Rihan and I. Roulstone Telford Institute of Environmental Systems, School of Environment & Life Sciences, University of Salford [email protected] ABSTRACT: Perhaps the greatest challenge in the production of extended forecasts of flash floods is the development of reliable quantitative precipitation forecasts (QPFs). Research in this area is continuing on a number of fronts. Nowcasting systems have been specified which seek to forecast the development of meteorological features sometimes through the use of statistically based descriptions of rainfall patterns, or using output from mesoscale models to move and develop radar measured precipitation fields. Such work has met with some success, although difficulties remain in coping with fine scale meteorological development which impacts forecast accuracy at space and time scales relevant to hydrologists. One must therefore question how far it may be possible to take nowcasting - based approaches for flash flood forecasting. An alternative, or perhaps complementary, approach is to use QPFs generated by high resolution Numerical Weather Prediction (NWP) models. At present the European Medium Range Weather Forecasting (ECMWF) centre uses an Ensemble Prediction System (EPS) to produce precipitation forecasts up to 10 days ahead, which are finding use as an initial warning of floods in large river catchments. However, these forecasts are too coarse for forecasting flash floods over small to medium catchments. Currently several Meteorological Centres are developing operational high resolution NWP models having grid lengths of a few kilometres. These systems promise, in the future, to provide QPFs which might rival nowcasting systems in terms of accuracy and timeliness. However, they will only be successful if the model initial conditions accurately reflect the mesoscale dynamics underpinning the occurrence of significant rainfall. Over the last few years research has been undertaken to assimilate radar reflectivity data into NWP models. Positive impacts on the forecasts have been observed, although they tend to die away after a few hours into the forecast. Whilst work continues in this area, an equally profitable approach might be to assimilate radar radial wind data into the NWP models. The assimilation of wind data may more directly impact model dynamics, and therefore generate more sustained benefits. Work along these lines is being carried out. However, there are important challenges which have to be overcome. In this paper we consider how the assimilation of radar radial winds may be approached within the framework provided by a four dimensional variational (4DVAR) analysis system. We outline a project, currently in progress, being undertaken by the University of Salford in collaboration with the UK Met Office, which seeks to implement such a procedure in the Met Office high resolution mesoscale model. It is planned to use data from the Chilbolton S-band Doppler radar system located in central southern England. The approach being taken and preliminary results will be presented.

17



AN INTEGRATED TOOL TO SUPPORT FLOOD WARNING DECISION

MAKING IN CATALUNYA (SPAIN) Carles Corral David López, Carlos Velasco, Xavier Llort, Daniel Sempere Grup de Recerca Aplicada en Hidrometeorologia. Universitat Politècnica de Catalunya. Barcelona (Spain) [email protected] ABSTRACT: A weather radar network is covering Catalunya, a region affected by Mediterranean climatological features and thus by severe flood events. An integrated flood warning system based on radar information is being developed, and it is working operationally in the regional water agency. The MOVHI tool consists in a visualisation package operating a hydrological model and a chain of rainfall processing algorithms in order to build an improved rainfall field from both radar and raingauge information. Before the merging process is done, several correction algorithms are applied to raw radar images (including attenuation and VPR). The hydrological model is a grid based model (named TOPDIST) able to provide local flow forecasts in several points of interest from user requests. Each land square of 1km2 is considered a hydrological unit, where a lumped model is applied: the SCS loss function in urban cells, and an adapted version of Topmodel in rural cells. Afterwards a routing algorithm based on the simplified drainage network is computed, providing the runoff estimates from the integration of the different cells. Currently the tool is integrated over the whole area of the country, although the hydrological model is only applied in the Besòs catchment (1000 km2, inside the important metropolitan area of Barcelona), waiting for a more reliable parameter calibration in other areas. In this presentation a general view of the tool is presented.

18



CÉVENNES-VIVARAIS MEDITERRANEAN HYDRO-METEOROLOGICAL OBSERVATORY: THE CATASTROPHIC RAIN EVENT OF 8-9 SEPTEMBER

2002 IN THE GARD AREA. Guy Delrieu Sandrine Anquetin, Hervé Andrieu, Jean-Dominique Creutin, John Nicol, Véronique Ducrocq Jacques Parent du Chatelet and E. Gaume Laboratoire d’étude des Transferts en Hydrologie et Environnement, Grenoble, France ABSTRACT: The "Cévennes-Vivarais Mediterranean Hydro-Meteorological Observatory" (OHM-CV is the French acronym) is a research initiative aimed at improving the understanding and modelling of the intense rain events that may result in devastating flash-floods in southern France. The OHM-CV observation strategy is comprised of three complementary approaches: (i) detailed, long-lasting and modern hydro-meteorological observation over part of the region of interest, i.e. the Cévennes-Vivarais region (ii) post-flood investigation following all the extreme events occurring over the entire French Mediterranean region and (iii) use of historical information available on previous floods. In parallel, several research activities, aimed in particular at a coupled hydro-meteorological modelling approach, are developed. Observational measurements over the region are based around two S-band radar systems separated by only 60 km and a network of over one hundred rain gauges, with data collected during the autumn of 2000, 2001 and 2002. This data set includes the catastrophic rain event of 8-9 September 2002 in the Gard region. In addition to the observation measurements, other aspects concerning this rain event include non-hydrostatic high-resolution modelling, results of the analog sorting technique and post-event hydrological investigation.

19


OBSTACLES TO THE GREATER USE OF WEATHER RADAR INFORMATION

F. Fabry and I. Zawadzki J.S. Marshall Radar Observatory, McGill University [email protected] ABSTRACT: The journey of the radar community towards the successful quantitative use of radar data for precipitation measurements is a complicated one, partly because of the winding nature of the road (the many steps required to convert a possibly contaminated measurement of Z aloft to the best estimate of rainfall accumulation at the ground), and mostly because of our failed attempts at trying to exploit mythical short-cuts (e.g. space-dependent instantaneous gage calibration, location-specific Z-R relationships, multiparameter approaches) to avoid tackling the problem in all its complexity. Since 1990 and the Joss and Waldvogel paper (JW90), we have had a clear road map on the steps required to make the best possible rainfall map by radar. Although different elements of the problem have been tackled with varying success, the attempts at solving all the problems for operational use have been limited at best. Furthermore, there does not exist any space-time dependent evaluation of the errors made on rainfall accumulations as a result of applying (or of not applying) all the required corrections to the radar data. As a result, potential users are left with a product of unknown quality that they can only use semi-quantitatively. And they rightfully complain. Three groups of workers, the radar engineers, the scientists, and the radar users, could try to solve the problem. Radar engineers interested in data quality issues generally work on cleaning Z data aloft, a parameter already known with a reasonable accuracy, but leave the rest to meteorologists. Scientists have mostly explored all the knowledge issues, and since we understand what to do, there is little original work left to be done except for the painstaking detailed validation of the ideas known since JW90 as well as the description of the errors, a task of limited appeal. Radar users, even in large groups like governmental organizations, generally do not have the technical expertise to handle a task of such sophistication without outside help. As a result, the situation is improving only very slowly. This presentation will summarize the road taken, the missed turns, and ends with a call for action: solving this problem requires that a few of us roll up our sleeves and attack the problem systematically. We have started to do it; anyone else wants to join?

20



CALIBRATION OF WEATHER RADARS IN SOUTHEAST QUEENSLAND G. Fields A. Seed, B. Yu, and T. Malone Griffith University, Nathan Qld 4111, Australia ABSTRACT: In Australia there is an extensive network of weather radars for all the major cities and along most of the tropical coastlines. The information obtained from this network can be used to calculate the rainfall intensities for areas within the radar range. Methodology for calibrating weather radars has been well developed for the tropics and temperate regions in Australia. The reflectivity-rainfall intensity (Z-R) relationships, however, have not been developed for the subtropical region in southeast Queensland. The purpose of this paper is to extend the calibration of the network for the weather radar stations located at Marburg and Mount Kanighan in Queensland. Data for the period from January 1999 to December 2001 has been obtained for both the rain gauge network within the radar ranges and the weather radar images. This data will be analysed on a daily and hourly basis to determine the climatological Z-R relationships for both radar stations, and their dependences on the type of rainfall events as either widespread or convective, and the spatial variation of rainfall intensities. The two radar stations in Southeast Queensland have the smallest distance between them than any other two stations within the network in Australia. The overlapping area of the radar ranges is significant. Within the overlapping area there are over forty rain gauges. Therefore an additional objective of the paper is to determine whether the combined use of the two radars and the rain gauge network in the overlapping area can provide more accurate rainfall measurements than from single radar readings.

21


GAUGE, RADAR AND SATELLITE AREAL RAINFALL ESTIMATES A. J. Pereira Filho A. Negri and P. T. Nakayama Dept. Of Atmospheric Sciences, USP, São Paulo 05508-900, Brazil ABSTRACT Gauge, weather radar and satellite data were utilized to estimate and to measure 24-hour precipitation accumulation over the São Paulo weather radar (SPWR) surveillance area. The Convective-Stratiform Technique (CST) calibrated by data from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) was compared to radar-derived and gauge-measured rainfall for the summer months of 2001 and 2002. A total of 150 days were used to compute the daily areal mean and its standard deviation in a 300-km square area centered at the SPWR. Results indicate better agreement between radar-derived and gauge-measured areal means. A scatter diagram of daily areal means and their standard deviation shows that the areal mean rainfall explains more than 87%, 88% and 96% of the gauge, radar and satellite areal rainfall variance, respectively. Radar and gauge are more dispersive due to their inherent larger errors and biases. The gauge network did not detect precipitating systems with very low areal means. Radar-derived, satellite-derived and gauge-measured total areal mean rainfall accumulation were 823 mm, 746 mm and 1224 mm, respectively. The gauge sampling area was about half of the two other systems.

22


ERRORS IN SURFACE RAINFALL RATES RETRIEVED FROM RADAR DUE TO WIND DRIFT

Neil I. Fox and Steven A. Lack Department of Atmospheric Science, University of Missouri – Columbia [email protected] ABSTRACT: A major assumption made when using radar retrieved precipitation observations in hydrological applications is that the precipitation measured aloft impacts the surface directly below the volume sampled by the radar. However, it is well known that rain may advected laterally considerable distances, implying that rainfall entered into distributed hydrological models will be inaccurate no matter how accurately the rain is measured aloft. In extreme cases rain may be observed above one catchment and actually fall in another. As moves are made toward more accurate precipitation retrievals (using dual-polarization, for example) and higher resolution (for applications such as urban hydrology) this problem has received little attention. This paper describes a series of experiments based on real data wherein the advection of the precipitation below the radar sampled volume is estimated based upon Doppler radar determined winds. The experiments show that even at standard resolutions of 2km the error can be extensive, and at higher resolutions and greater ranges / higher beam elevations the errors become very large. Errors are assessed using different Z-R relationships and resolutions as high as 0.25km. The method is robust and could easily be used to correct for wind drift.

23



PROVIDING DISTRIBUTED FORECASTS OF PRECIPITATION USING A

BAYESIAN NOWCAST SCHEME Neil I. Fox and Christopher K. Wikle University of Missouri – Columbia, Columbia, M0 65211, USA [email protected] ABSTRACT: In order to apply quantitative precipitation forecasts (QPF) in a hydrological context it is imperative to provide information regarding the uncertainty in the forecast such that a derivation of an associated uncertainty in the hydrological forecast can be estimated. A stochastic Bayesian precipitation nowcast scheme has been developed that produces an ensemble of forecasts from which can be derived a mean field along with its variance. Therefore a realistic measure of nowcast uncertainty is produced alongside the ‘best’ QPF. The methodology employs an initial series of radar reflectivity fields as its basis, but contains an integrodifference equation (IDE) formulation that allows the imposition of physically-based processes on the Bayesian structure, including the use of subsidiary observations such as Doppler radar-derived winds, that can constrain nowcasts. This paper reviews the methodology, as well as presenting examples of the nowcasts produced, and examining the distribution of nowcast QPFs. This is done by point / pixel, and by simulating small catchments to view the uncertainty characteristics at different spatial scales. The presentation to the hydrologist of the information produced by the system is discussed and the application of such distributed forecasts to hydrological problems (e.g. streamflow forecasting) described.

24



STUDY OF THE CHARACTERISTICS OF THE TROPICAL ATMOSPHERIC BOUNDARY LAYER THROUGH A DOPPLER RADAR BAND L

José Francisco de Oliveira Júnior Byung Hyuk Kwon UNEMET - Diretor de Assuntos Acadêmicos e Científico [email protected] ABSTRACT This study was accomplished using radiosonde, surface and L-Band Doppler radar data, in Serpong - Indonesia (6° 24' S - 106° 42' E), during the dry season of 1993 and rainy season of 1994. In this work the radar data were validated with the one of radiosonde through statistical methods. The structure of the CLAT (Tropical Atmospheric Boundary Layer) was interpreted through the radar echo (wind), observed 24 hours a day during the whole study period. The more evident characteristic of the CLAT is the presence of a diurnal CM (Mixed Layer) that reaches altitudes greater than 0,3 km in the morning and a reaches 3 to 5 km in the afternoon, observing a diurnal cycle of heating and cooling. The main characteristic between the dry and rainy season was that at 0300 LT (LT = local time) both presented the same behavior, but for the remaining time of the day a wide distinction exists between them. During the dry season the intensity and the direction of the wind are more variable than in the rainy season, affecting the height of the CLAT. The information obtained by the BLR (Boundary Layer Radar) shows the existence of pronounced diurnal variations in the CLAT, during clear days. It was also verified that the diurnal variation of the CLAT was affected by sea breeze circulation, confirmed by the BLR. As an example, on the 12/10/93 the breeze intensified at 1400 LT and increasing rapidly (about 1500 LT) up to a depth of 1,5 to 2 km. The validation of the radar data presented here has shown significant results, proving that the radar observations should be an important method in the study of the low equatorial atmosphere.

25



COMPARING AND IMPROVING QUANTITATIVE PRECIPITATION ESTIMATION IN THE ALPS USING VARIOUS RADAR OPERATIONAL

PRODUCTS Marco Gabella Riccardo Notarpietro, Giovanni Perona Dipartimento di Elettronica, Politecnico di Torino, Torino, Italy. [email protected] ABSTRACT A real-time application of a physically-based method, which tries to reduce the macroscopic “biases” that affect weather radar measurements in mountainous regions, will be presented. It is based on a non-linear, Weighted Multiple Regression (WMR) scheme, which separates the influences of the calibration, beam-broadening with distance, beam shielding and orography by calculating four correction coefficients that lead to the best agreement with the available in situ measurements. Once the regression coefficients have been determined, the complete map of correction factors (for the given radar site, orography and mean atmospheric refraction) can be calculated and applied to the whole radar-derived precipitation field map. The performances of the method using daily integration periods will be presented for the Mesoscale Alpine Program “Special Observation Period 2b” (September 19-21, 1999). The first day of the event is used as independent data set to train the adjustment coefficients, while the following days are used for a quantitative verification. So far, the WMR has successfully been applied and verified using 2D maximum vertical reflectivity maps (“MAX”). In this study, for the first time, we deal with more sophisticated 2D radar-derived maps (the so-called “RAIN” product, operationally generated by MeteoSwiss). These maps contain an estimate of the precipitation rate at the ground derived by extrapolating the rain rate through a real-time processing of volumetric radar observations aloft. All clutter-free reflectivity measurements along the vertical are converted into the equivalent rain rate using a single Z-R relationship (Z= 102.5 R1.5) and then weighted with weights that are inversely proportional to the reflectivity heights in order to extrapolate the rain rate on the ground. These results will be thoroughly analyzed and discussed. It will also be explained why we are confident that in more recent events the RAIN products will perform definitely better than the MAX.

26


ASSESSMENT OF THE STATISTICAL CHARACTERIZATION OF SMALL-SCALE RAINFALL VARIABILITY FROM RADAR

Mekonnen Gebremichael Witold F. Krajewski Grzegorz J. Ciach IIHR – HYDROSCIENCE & ENGINEERING, THE UNIVERSITY OF IOWA [email protected] ABSTRACT: The main objective of this study is to assess the ability of radar-derived rainfall products to characterize the small-scale spatial variability of rainfall. To perform the assessment, this study uses independent datasets from high-quality dense rain gauge networks employed during the TRMM-LBA and TEFLUN-B field experiments. A detailed comparison between the gauge- and radar-derived spatial variability estimates is carried out by means of correlation functions, variograms, and fractal characteristics. More emphasis is given to correlation function. The approach followed in the current analysis accounts for the recognized differences in the scales of observation. The performance of radar-derived correlation function is evaluated in two ways: by direct comparison with gauge-derived correlation function, and by quantifying its performance in one of its main applications – gauge sampling error estimation. Results show that at separation distances shorter than about 5 km, radar-derived correlations are lower than those obtained from gauges. Three sources of uncertainty that may have caused the discrepancy between gauge- and radar-derived correlations are identified, and their effects quantified to the extent possible. The error introduced in gauge sampling error due to the use of radar-derived correlation function is within ± 10%. Discrepancies between gauge- and radar-rainfall fields are also observed in terms of variograms and fractal characteristics.

27



PREDICTABILITY OF PRECIPITATION AND ITS DEPENDENCE ON SCALE FROM LARGE-SCALE RADAR COMPOSITES

Urs Germann Isztar Zawadzki MeteoSwiss, Locarno-Monti, Switzerland [email protected] ABSTRACT: The lifetime of precipitation patterns in Eulerian and Lagrangian space derived from large-scale radar images is used as a measure of predictability. A three-step procedure is proposed: first, the motion field of precipitation is determined by variational radar echo tracking. Second, radar reflectivity is advected by means of a modified semi-Lagrangian advection scheme assuming stationary motion. Third, the Eulerian and Lagrangian persistence forecasts are compared to observations to calculate the lifetime and other measures of predictability. The procedure is repeated with images that have been decomposed according to scales to describe the scale-dependence of predictability. Discrete cosine transform, wavelets, and simple intensity-thresholding are used for scale decomposition. The methodology has been developed and initially tested with radar composite images of relatively flat North-America. It is now applied in a more complex environment: the European Alps. Here, the steep orography influences both evolution and motion of precipitation systems and the quality of radar data. The analysis has a threefold application: i) determine the scale-dependence of predictability, ii) set a scale-dependent standard against which the skill for quantitative precipitation forecasting by numerical modeling can be evaluated, and iii) extended, partly probabilistic nowcasting by optimal extrapolation of radar precipitation patterns. Lagrangian advection on large scales was found to have significant forecast skill up to lead times of several hours. This nowcasting application was baptized MAPLE, McGill Algorithm for Precipitation nowcasting by Lagrangian Extrapolation.

28



DEVELOPMENT OF A WEATHER RADAR SIGNAL SIMULATOR TO REVISIT THE PROBLEM OF SIGNAL STATISTICS AND MEASUREMENT

ERRORS. M Gosset J Nicol, A Sanchez LTHE, BP 53, 38 041 Grenoble, France ABSTRACT: A new radar simulator was developed in our group in order to revisit the theory of radar signal statistics and its relation to drops position and movement. All the drops contained in the resolution volume are represented explicitly, their initial position is defined individually and they are allowed to move between pulses according to different assumptions concerning the wind and a given law for drop fall speed. The output of the simulator are the amplitude and phase of the signal for each pulse, which can be used for time correlation studies and the estimation of the average power for several consecutive pulses. Then appropriate statistics can be derived. The simulator was conceived as a modular tool in order to add, little by little more realism in the simulations and study step by step the sensitivity to different parameters such as : the beam pattern, receiver filters, digitalization etc…. Spectral analysis of the signal, by FFT or with a pulse pair algorithm are also being implemented as well as dual polarization treatment for oblate drops. The results for various assumptions concerning drop position and movement as well as radar characteristics will be presented.

29


X-PORT DONGA : PLANS FOR A RADAR BASED EXPERIMENT TO STUDY RANFALL VARIABILTY OVER A SMALL CATCHMENT IN BENIN,

WEST AFRICA. M Gosset F Cazenave LTHE, BP 53, 38 041 Grenoble, France ABSTRACT: A X-band polarimetric radar has been developped in our group. This radar is planned to be installed over the small Donga Catchment in North Benin, West Africa. The planned experiment is part of the ‘African Monsoon Multidisciplinary Analysis’. This project aims at studying the impact of climate variability and the role of the african monsoon on water ressources over the west african region. Several experiments are planed in the coming years as part of that project. Long term observation of the continental water cycle, at different scales have already been implemented and will be reinforced during enhanced and intensive observation periods. The Donga basin is a small (500 km2) watershed well equiped with hydrological sensors (ground water and river outflow) with the network of about 20 rain gages (tipping bucket). The X-band polarimetric radar will be installed in 2004, in order to study the rain structure in that area and the small scale variability of rain. It will also be a good opportunity to assess the limits of X-band polarimetric sensors for rain measurement and attenuation correction in such a tropical region.. This presentation will describe the planed experiment and its objectives and will also discuss the practical challenges.

30


RADAR RAINFALL ESTIMATION IN THE NEW ZEALAND CONTEXT. Warren Gray and Howard Larsen NIWA, New Zealand [email protected] ABSTRACT: The accuracy of radar measurements of precipitation is an important consideration for a number of hydrological applications. The difficulties commonly encountered in precipitation measurement by radar include errors from radar reflections from the ground or sea surface, errors in extrapolating from measurements aloft to values at the surface, errors in interpreting the radar signal as precipitation values and errors through inadequately sampling a fluctuating signal. This paper discusses these error sources, their importance in a NZ context, and solutions that have been implemented. When these solutions are incorporated, comparisons between radar estimates of rainfall over a test catchment and streamflow data from the catchment confirm a useful level of accuracy, comparable with that from a dense raingauge network.

31



NOWCASTING FOR NEW ZEALAND

Warren Gray and Howard Larsen NIWA, New Zealand [email protected] ABSTRACT: The skill with which weather radar data can be extrapolated to provide short-term rainfall forecasts is assessed for case studies over catchments near Auckland, New Zealand. The forecasts are made using a technique that decays the small spatial scale components of the precipitation pattern, as these are also the short-lived scales. At the same time the remaining, larger-scale, components are advected with a storm-scale motion. Results from applying the technique to representative frontal rain events show that the pattern of reflectivity is well forecast, with few timing errors and little bias in catchment averages. A simple estimate of skill shows improvement over catchment forecasts made assuming persistence. The larger the catchment, the more accurate the forecasts. Future work will assess the usefulness of the forecast in estimating river flow for small catchments.

32



FILTERING-OUT AREA-POINT ERRORS FROM RADAR-RAINGAUGE VERIFICATION SAMPLES

Emad Habib Grzegorz J Ciach and Witold F. Krajewski IIHR – Hydroscience & Engineering The University of Iowa [email protected] ABSTRACT: Large differences between area-averaged rainfall and point raingauge measurements pose major difficulties in conclusive verification of the radar-based quantitative precipitation estimates (QPE). We develop a nonparametric method to account for these differences in radar-raingauge comparisons. This technique, called a conditional distribution transformation (CDT) method, is a generalization of the error variance separation (EVS) method that we studied previously. The purpose of the CDT procedure is to filter-out the raingauge representativeness errors from a radar-raingauge verification sample. Similarly to the EVS, the CDT method is also based on the geostatistical apparatus and requires additional information about the small-scale spatial variability of the precipitation in the sample. However, in contrast to the EVS that was limited to the error variances only, the CDT method provides an approximation of the bivariate distribution of the radar estimates and the corresponding true area-averaged precipitation (the verification distribution). Retrieval of this distribution allows application of the distribution-oriented verification methodologies to evaluate QPE products in a systematic and comprehensive manner. We present the CDT method and the empirical tests of its validity. The tests are based on large data samples from two high-density raingauge networks. One covers the Goodwin Creek watershed in Mississippi, and the other the Little Washita watershed in Oklahoma. The tests demonstrate that the CDT method performs with satisfactory accuracy and can considerably improve on the currently used QPE verification practices.

33



COMMENSURABILITY OF OCCURRENCE OF FLOODWATER Hu Hui Han Yanben, Wang Rui Yunnan Observatory, Academia Sinica, Kunming [email protected] ABSTRACT: Professor Weng Wenbo, a famous geophysicist, firstly pointed out that the occurrence of an floodwater has the commensurability and applied the principle of commensurability which was firstly proposed in astronomy to the prediction of large natural hazards such as floodwater, droughts and earthquakes, etc., thereby developing it into the theory of prediction which have been widely used in the prediction of hazards by scholars in China. What is the commensurability? The commensurability represents the regulation of the occurrence and development of the event itself, thereby mirroring the internal causes of the development and change of things to a certain extent. The commensurability can be defined as that given that Xi and Xi+∆i are two arbitrary elements of the data set {Xi}, and Xi and Xi+∆i ∆{Xi}, if the following formula is tenable ∆X=(Xi+∆i-Xi)/K, where K=1, 2, 3,……then the data set {Xi} has the commensurability, where ∆X is the commensurable value of the data set {Xi}. If k is identically equal to 1, then ∆X is the period of the data set {Xi}. A time series with the commensurable value may be extrapolated by letting K=l according to the commensurable value. However, it is not at all certain that each extrapolated event occurs because this is only a necessary condition but not a sufficient condition. Analyzing the floodwaters that occurred in the Huaihe river of China since 1827, we discover that the occurrence of the floodwaters have the commensurability. The commensurable value is 38 years. The commensurable point and the points of golden section of the commensurable value are all the frequent occurring point of an event. The two floodwaters that occurred in Huaihe river of China in 1991 and 2003 appeared at the points of golden section of a time and two times of the commensurable value of its time axis, respectively. Therefore, the commensurability can provide scientific basis for the prediction of floodwater which will occur in the area in future. Thus, the analysis of the commensurability showed that the occurrence of the two floodwaters would be inevitable.

34



INTEGRATED CONTROL: DARSENA DRAINAGE SYSTEM CONTROL N. E. Jensen Department of Urban Water., DHI Water & Environment, Science Park, Gustav Wieds Vej 10, DK-8000 Aarhus C ABSTRACT: The paper presents the Darsena Drainage System Control (DDSC) that has been designed to form an automatic system for short-term prediction of flow and water quality development from the 8 creeks forming the sewer- and storm water drainage system for the ancient city of Genoa and for optimizing the pump control in the pumping stations along the trunk sewer taking the water to the waste Water treatment plant. A LAWR X-Band weather Radar has been installed in the catchment to provide precipitation images every 5 minutes as input to the DDSC. The system is based on the DIMS and MOUSE software packages from DHI Water and Environment.

35


HYDROLOGICAL EVALUATION OF SINGLE AND DUAL POLARISATION RADAR RAINFALL MEASUREMENT ALGORITHMS

Phillip Jordan Alan Seed, Tom Keenan, Peter May Bureau of Meteorology, Australia [email protected] ABSTRACT: The routine use of radar for rainfall measurement and short-term forecasting in flood forecasting and other hydrological applications is increasing. Dual polarisation radars have a theoretical advantage for rainfall rate estimation over conventional radars because the radiation returned by the horizontal and vertical polarised beams provide several additional parameters of the rainfall field that can be combined into an estimate of rainfall intensity. The Bureau of Meteorology Research Centre operated a 5.3 cm wavelength dual polarisation radar at Badgery’s Creek, 44 km West of Sydney Australia, for the period from September 2000 to June 2001. The aim of this paper is to analyse and optimise the performance of a C-Band dual polarisation radar for forecasting of floods in rural and urban catchments. Hourly rainfall data from a network of 239 tipping bucket raingauges, located within 110 km of the radar, has been utilised in the analysis. The Bureau of Meteorology is responsible for preparing flood warnings for the Georges River, which flows through the Western suburbs of Sydney. The URBS semi-distributed rainfall runoff routing model is used to predict flood inflows from the 350 km² catchment area upstream of the urban area. Accurate measurement of rainfall across the catchment is critical for timely and accurate flood forecasts on the Georges River. The paper compares rainfall accumulation estimates across the Georges River catchment from: 1. raingauge networks of varying density, 2. single polarisation (reflectivity only) radar, 3. reflectivity with attenuation correction calculated using dual polarisation algorithms, 4. several multi-parameter algorithms that implement measurements from the dual polarisation radar, such as reflectivity at dual polarisation, attenuation and specific differential phase, and 5. single or dual polarisation radar combined with raingauge networks of varying density. The paper then goes on to analyse the effect of each of the rainfall measurement strategies on the accuracy of prediction of the flood hydrograph from the catchment. Inferences are drawn about selection and optimisation of parameter values for the rainfall runoff routing model for use with rainfall data from each of the various sources.

36



OPERATIONAL ISSUES: REAL-TIME CORRECTION OF RADAR DATA Nicholas Kouwen Allyson Bingeman Isztar Zawadzki Department of Civil Engineering, University of Waterloo, Waterloo, Ontario, Canada, N2L 3G1 [email protected] ABSTRACT: After 30 years, the word ‘promising’ is still used too often when radar data is applied for hydrological purposes. To the long term user of radar, the words ‘erratic’ and ‘baffling’ sometimes enter the thought process. The reason is that one can not be a casual user of radar data for hydrological purposes. The list of radar errors is well known and the user must forever be on guard to recognize the conditions that affect the quality of the radar product. For the past 10 years, hourly precipitation data from the King City radar near Toronto (owned and operated by Environment Canada) have been collected at the University of Waterloo for the purpose of hydrological modelling. For the past 1.5 years, two radar products (1 and 2 km resolution hourly precipitation) have been collected from the Franktown radar near Ottawa (also owned and operated by Environment Canada) and up to seven products have been archived from the McGill radar, which is located just outside Montreal Canada. The Environment Canada radars are C-band radars while the McGill radar is an S-band radar. The ten year experience with the King Radar has shown that, along with the usual more dominant errors due to attenuation, bright band and beam filling, the periodic casual adjustment of various components of the radar data processing can affect the quality of the product from a hydrological point of view. At least, that is the only explanation for some of the problems that were encountered in using the data. A lack of maintaining the calibration of a radar can have a similar effect. For hydrological applications, thresholds are very important. For instance, a period of very light rain or snow that may go unreported can have a substantial cumulative effect on the watershed’s response to a larger event following this light rain. With this in mind, and the perception that the use of radar data in hydrology is still not nearly realized to it’s full potential, the current collaboration between University of Waterloo hydrologists and the McGill radar group was initiated. The objective is to test various algorithms to deal with the usual problems of attenuation, bright band and beam filling, as well as the removal of AP. Automatic Correction Schemes for McGill Radar Label Type of Correction Height used Resolutions C0 Clutter Mask and Anomalous Propagation 2.0 km / 1.5 km 1 km and 2 km C1 C0 + VPR correction out to 110 km 2.0 km / 1.5 km 1 km and 2 km C2 C1 + Simulated VPR out to 200 km 2.0 km / 1.5 km 1 km and 2 km C3 Best Estimate of Precipitation 1.1 km 1 km only To determine the impact of these corrections on hydrological modelling, the WATFLOOD hydrological model (http://www.watflood.ca) was set up for 72 watersheds within the McGill radar domain. Continuous simulation using the seven McGill products, as well as two from the overlapping Franktown radar are compared to show the sensitivity and improvement to the hydrological modelling.

37


http://www.watflood.ca/


CHARACTERISTICS OF RAIN DROP OBLATENESS REVEALED BY VIDEOSONDE MEASUREMENTS

T. Keenan T. Takahashi Bureau of Meteorology Research Centre, GPO Box 1289K, Melbourne, Australia [email protected] ABSTRACT: Radar rainfall estimation based on polarimetric approaches has significant advantages over traditional power-based techniques. In a physical sense it is less affected by the intrinsic variability of raindrop size distribution that impact reflectivity based rainfall estimates. In a practical sense it offers additional quality control processes and more robust techniques for the correction of power measurements affected by attenuation in rainfall. In heavy convective rainfall situations, the polarimetric measurements are particularly advantageous. The differential propagation phase shifts are easily measured and provide reliable physically based rainfall estimates i.e. radar only measurements. Typically hydrologists use engineering solutions based on rain gauges to correct rain fields estimated by radar. But in highly variable rainfall situations, these approaches are less effective and physically based approaches become more important. However, polarimetric radar techniques depend upon and are sensitive to the characteristics of the raindrop shape. Differences in the assumed model of raindrop axial ratio can have serious impact on the derived rainfall estimates. Knowledge of the raindrop shape is therefore particularly important for polarimetric radar rainfall estimation techniques. In this paper information from the balloon borne Precipitation Particle Image Sensor (PPIS) or “videosonde” developed by Takahashi (1990) are examined to investigate characteristics of raindrop axial ratio. Video images of raindrops with equivalent diameters in the range 0.5-8 mm collected from PPIS penetrations are examined. The dataset includes a variety of convective storms observed within the Asia/maritime continent region. The data are stratified as function of location, microphysical origin, height, drop size and compared to existing equilibrium shape models e.g. Pruppacher and Beard (1970) and Andsager et al. (1999). The videsonde data show considerable variability in observed raindrop axial ratio. There is clear evidence of raindrop oscillations with both prolate and oblate modes evident. Most frequently observed modes are near spherical for raindrop sizes up to 3 mm in diameter. For drops in the range 3-4.5 mm, the most frequent mode is less oblate than implied by Andsager et al. (1999) and for the larger drops, the most frequently observed mode is more oblate than indicated by the Andsager et al. (1999). In terms of microphysical origin, below the melting level, raindrops originating from frozen drops and graupel show significant oblateness whereas drops originating from snow are more prolate in character. Drops originating in warm rain develop significant oblateness near the surface. The impact of the observed characteristics of raindrop shape on polarimetric rainfall estimation relations are explored using T-matrix scattering simulations.

38



APPLICATION OF WEATHER RADARS FOR CLOUD SEEDING IN CENTRAL AREAS OF IRAN

V. Khatibi M. Seidhasani , R. Pahlavanhossieni National Cloud Seeding Research Center of IRAN P.O.Box: 89195/611, Yazd, IRAN [email protected] ABSTRACT: Regions in central Iran, especially the province of Yazd and its neighboring provinces enjoy arid and semiarid climates with an average annual precipitation of 60 mm. Strong deficiency of the water resources and decline of the aquifer created by lack of of precipitation has caused state of Iran to seek a remedy through transmission and application of cloud seeding technology. To fulfill the task, the I. R. ministry of energy for the first time called on the Russian CAO institute to set up the “Increasing precipitation through cloud seeding” project experimentally. A cloud-seeding plane and a MRL- 2 Russian weather radar were applied in 1999. Two radars were used in 2000. After a two-year interruption owing to the developement of the project, the contract with the CAO institute was renewed in 2003 and programmed to use of four weather radars in the near future. The radars will be the first weather radar network in Iran. Based on a state schedule, the research institute of water resources affiliated with the ministry of energy in collaboration with the meteorological organization, aiming to apply more equipment such as radars , planes and organizing an extensive radar network are in charge of the project throughout the country. This cooperation should examine other problems such as hail suppression and flood warning as well as research concerning scientific and practical application of radars, cloud physics and atmospheric processes, and water resources management. This article describes the condition of four installed weather radars in central Iran, their role and application for cloud seeding projects as well as the data provided by these radars.

39



THE USE OF RADAR RAINFALL DATA IN RELATION TO TOPOGRAPHY

Yuki Kimura Bertram Ostendorf University of Adelaide, Adelaide, SA 5005, Australia [email protected] ABSTRACT: Most previous studies quantifying the effects of topography on rainfall have used rain gauge network data. Although ground-based data provide long time series, their spatial distribution is generally limited. Especially in mountainous regions stations are placed with logistic constraints, which may bias a regional analysis and potentially affect the results. What seems to be lacking for these studies is a closer examination of spatial pattern of rainfall, as they can potentially be provided from radar rainfall measurements. On the other hand, studies using radar data to examine orographic rainfall have been limited to single events and therefore precluding a regional, long-term analysis. The aim of this study is to explore the utility of long-term rainfall radar data to study the topographic influence on spatial rainfall pattern. Radar data from 1997 to 2003 from the Sellicks Hills Radar station, South Australia (-35.33 latitude, 138.50 longitude) were used. The R Z relationship was calibrated using rain gauge data from selected stations along a transect from the coast to the range. Radar data showed substantial errors including clutters, backscattered signals close to the peak of Mt. Lofty Ranges and artificial patterns. The areas and time intervals where the radar data seemed to have problems were excluded from the analysis. The resulting maps were grouped using simple criteria as surrogates for the likelihood of occurrence of different mechanisms of orographic rainfall generation. Criteria for averaging were the direction of moisture advection (600mb wind direction from sounding data), rainfall magnitude and season. These averages were then compared with topographic variables using a multiple regression. We used elevation, directional slope (partial derivative of elevation in different directions), curvature and distance to the sea. Several spatial smoothing levels were applied to estimate the most appropriate spatial scale of the topographic influence on rainfall pattern. Results indicated various relationships between precipitation patterns and topographic features. Elevation showed a highest correlation with most of the precipitation patterns. Distance to sea showed the relatively higher correlation with rainfall during cold season and rainfall with the prevailing winds. Directional slopes partially accounted for the variance of rainfall patterns by wind directions. Although issues including appropriate temporal and spatial scale and the adjustment of errors still remain unclear, the results allow the conclusion that radar can be used to increase our understanding of topography-rainfall interactions.

40



TOWARDS OPERATIONAL PROBABILISTIC QUANTITATIVE PRECIPITATION ESTIMATION USING WEATHER RADAR

Witold F. Krajewski and Grzegorz J. Ciach IIHR-Hydroscience & Engineering, The University of Iowa, Iowa City, IA [email protected] ABSTRACT: Large uncertainties associated with the operational quantitative precipitation estimates produced by the U.S. national network of WSR-88D radars are well-acknowledged. These uncertainties include both systematic and random effects of numerous sources. Quantitative description of the precipitation estimation errors would help the U.S. National Weather Service (NWS) forecasters is making operational decisions on issuing forecasts and warnings on flooding potential. It would also help the public and other agencies that use the products in making decisions on their operations with likely benefits of saved resources and even lives. Currently such information is not available. The authors describe a comprehensive plan of introducing probabilistic quantitative precipitation estimation into the operational NWS environment. The plan focuses on radar-based estimates and includes research and development, experimental, and implementation components. The authors discuss three different possible approaches with their merits and problems. They recommend one of the approaches based on a parameterization of radar-rainfall errors taking into account range, seasonal, synoptic and climatic dependencies. They discuss details of the resources required and the implementation feasibility.

41



A COMPREHENSIVE DATA ANALYSIS FROM TWO VERTICALLY POINTING RADARS

Anton Kruger Witold F. Krajewski, Efthymios Nikolopoulos Christopher Williams and Kenneth Gage

IIHR-Hydroscience & Engineering, The University of Iowa, Iowa City [email protected] ABSTRACT: Two vertically pointing radars collocated high-resolution data for the period of eight months of warm season in Iowa City, Iowa. These were an X-band Doppler radar of The University of Iowa and an S-band Doppler radar of the NOAA Aeronomy Laboratory. The authors carried out a comprehensive analysis of the collected data. The analyses include power spectrum, correlation, and wavelet modeling. The authors investigated the issues of temporal scale averaging as well as the vertical scale averaging in the context of the range-dependent effects of the vertical profile of radar reflectivity. The results include characterization of rainfall type (convective and stratiform), quantification of the effect of the vertical and horizontal (time) variability on the interpretation of operational radar observations, and comparisons with an operational radar data. The authors summarize the results and their implications for quantitative rainfall estimation, radar-rainfall error structure studies, and validation of remote sensing of rainfall.

42



DEVELOPMENT OF A PRECIPITATION RESEARCH SITE IN THE NORTHERN PLAINS

Paul A. Kucera Department of Atmospheric Sciences, University of North Dakota, USA [email protected] ABSTRACT: Precipitation, especially cold season, strongly effects the hydrological cycle in the Northern Great Plains of the US. Spring and summer time flooding depends on the annual precipitation. In particular, the Red River Valley in Eastern North Dakota is susceptible to spring flooding because of the flat terrain. Quantitative precipitation estimates in the Red River Valley are generally poor especially estimation of frozen precipitation. To address this issue, we have been developing a precipitation research site in Eastern North Dakota. The University of North Dakota C-Band Doppler radar is currently being upgraded to have polarimetric capability. Also, we are deploying a 915 MhZ vertical wind profiler to examine the vertical structure of precipitation. We are also installing a dense network of rain gauges, snow gauges, and disdrometeors to examine the small-scale variability of precipitation. Are precipitation site is located in a small watershed that also has several stream gauges and cooperative network of rain gauges. We plan to use this site to improve quantitative estimates of precipitations and to examine the hydrological cycle in detail. In my talk, I will give and overview of the research site and design. Also, I plan to present some preliminary results of our initial observations.

43



ERRORS IN THE RADAR CALIBRATION BY GAGE, DISDROMETER, AND

POLARIMETRY: THEORETICAL LIMIT AND APPLICATION TO OPERATIONAL RADAR.

GyuWon Lee and Isztar Zawadzki J.S. Marshall Radar Observatory, Dept. of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada. [email protected] ABSTRACT: Radar calibration is a key source of error in quantitative precipitation estimation and forecasting. Currently, over 125 POSS (Precipitation Occurrence Sensing System) disdrometers are deployed in Canada. They can identify the type of precipitation and measure drop size distributions (DSDs) from which R and Z can be derived. In this presentation, we illustrate the potential uses of this instrument in radar calibration and a new self-consistent calibration method using radar polarimetry. A 5-year record of DSDs is used to evaluate the effect of the DSD variability on accuracy of a radar adjustment with a rain gage on a daily basis. The results show that if a single R-Z relationship is used for the entire 5-year period the relative errors in calibration of the radar as a hydrological instrument is ~26% on the average with a standard deviation of ~29%. The errors decrease to a standard deviation of 8% if the appropriate average R-Z relationship for the day is used. A calibration of reflectivity can be done if a disdrometer is available. Good correlations between radar and disdrometric reflectivities indicate that this could be an excellent way of calibrating radar on a daily basis. The information from an operational S-band polarimetric radar is also used for calibration purpose. This method is based on the fact that the specific differential phase shift ( ) or differential phase shift ( Φ ) between the horizontal and vertical polarized beams is immune to the radar calibration error whereas the reflectivity is affected by it. The error in polarimetric calibration due only to the variability of DSDs is 1 dB with a single parameter ( ) and reduces to 0.5 dB when the differential reflectivity ( ) is also added. The stability of this calibration method suggests that it can be used in real time. Its sensitivity with respect to the drop deformation has been tested. Furthermore, the consistency in the disdrometric and polarimetric calibration allows an estimation of the mean drop deformation.

DPK

DP

DP

K

DRZ

44



ERRORS IN RAIN MEASUREMENT BY RADAR DUE TO VARIABILITY OF DROP SIZE DISTRIBUTIONS

GyuWon Lee and Isztar Zawadzki J.S. Marshall Radar Observatory, Dept. of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada. [email protected] ABSTRACT: Disdrometric measurements are affected by the spurious variability due to drop sorting, small sampling volume and instrumental noise. This paper explores the importance of this variability and develops a new approach, referred to as the Sequential Intensity Filtering Technique (SIFT) that minimizes the effect of the spurious variability on disdrometric data. The time-scale dependence of the variability of drop size distributions and its effect on radar rain estimation are also quantified. A simple correction for drop sorting in stratiform rain illustrates that it generates a significant amount of spurious variability and is prominent in small drops. The SIFT filters out this spurious variability while maintaining the physical variability as evidenced by stable R-Z relationships that are independent of averaging size and by a drastic decrease of the scatter in R-Z plots. The presence of scatter causes various regression methods to yield different best-fitted R-Z equations depending on whether the errors on R or Z are minimized. The weighted total least square (WTLS) technique solves this problem by taking into account errors in both R and Z and provides the appropriate coefficient and exponent of Z=aRb. The variability of DSDs and its effect on rain estimation by radar depend on the scale of interest. The use of a climatological R-Z relationship for rainfall - affected by all of the DSDs variability - leads on the average to a random error of 41% in instantaneous rain rate estimation. This error decreases with integration time, but the decrease becomes less pronounced for integration times longer than two hours when it reaches 30%. Daily accumulations computed with the climatological R-Z relationship have a bias of 30% due to the day-to-day DSD variability. However, when daily R-Z relationships are used, a random error of 32% in instantaneous rain rate is still present due to the DSD variability within a day. This illustrates that most of the variability of DSDs has its origin within a storm or between storms within a day. The DSD variability among different physical processes is larger than the day-to-day variability. A bias of 41% in rain accumulations is due to this variability. Accurate rain rate estimation (~7%) can be achieved only after the proper underlying physical process is identified and the associated R-Z relationship is used.

45



SOURCES OF ERRORS IN PRECIPITATION MEASUREMENTS BY POLARIMETRIC RADAR

GyuWon Lee and Isztar Zawadzki J.S. Marshall Radar Observatory, Dept. of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada. [email protected] ABSTRACT: Using a set of long-term disdrometric data and of actual radar measurements from the McGill S-band operational polarimetric radar, several sources of errors in precipitation measurement with polarimetric radar are explored in order to investigate their relative importance and the feasibility of a polarimetric technique for estimating R in the context of an operational setup. The sources of errors considered are : DSD variability, uncertainty in drop deformation, observational noise, and the methods of deriving the relationships between R and polarimetric parameters. Additional polarimetric parameters reduce the effect of the DSD variability on precipitation estimates by radar, the main concern in R estimation with the traditional relationship. However, some sources of errors nullify this advantage, yielding an estimation of R that is worse than that from the conventional

hZR −

hZR − relationship. The sensitivity of polarimertic parameters to the drop deformation provides a random as well as a bias error. The random error in the estimation of R with any combination of and

is nearly insensitive to the drop deformation. However, when is included in the relationships, the random error increases for certain formulas of drop deformation and the bias becomes significantly larger.

hZ DRZ

DPK

The relationships from the literature that are widely used in the estimation of R differ from those derived from our disdrometric data set, resulting into a bias that is larger than the random error due to the DSD variability. This is due to the representation of DSDs with a specific model (gamma model) and an assumed range of its parameters. Only the relationship derived from the normalized gamma DSDs is consistent with those from measured DSDs and yields a small bias. By assuming a measurement noise as expected from a slow scanning research polarimetric radar and a 4-km range smoothing, the combined effect of the DSD variability and of measurement errors prevents any significant improvement with the

relationship compared with the conventional

),( DRh ZZR −

hZR − . The quantified measurement noise of the McGill S-band fast scanning operational radar is significantly larger than that of a slow scanning radar, implying that a spatial averaging of of at least 100 kmDRZ 2 is needed to achieve the some accuracy expected with the conventional hZR − relationship.

46



THE CHARACTERISTICS OF RAINDROP SIZE DISTRIBUTION USING FILTER PAPERS AND A PRECIPITATION OCCURRENCE SENSOR

SYSTEM DISDROMETER FROM MARCH 2001 TO FEBRUARY 2002 AT BUSAN, KOREA

Dong-In Lee Min Jang, Young-Ho Han, Chul-Hwan You and Hyo-Sang Chung Dept. of Environmental Atmospheric Sciences, Pukyong National University, Korea1

[email protected] ABSTRACT: The raindrop size distribution (DSD) is very important in microphysical modelling and estimating rainfall using radars. There are many kinds of methods for the measurement of DSD. Dyed filter papers (Marshall–Palmer, 1948), optical imaging devices (Knollenberg, 1970, Illingworth and Stevens, 1987), JWD (Joss and Waldvogel, 1967) and POSS (Precipitation Occurrence Sensor System) disdrometeters (Shepard, 1990) had been used. In this study, DSD observations by filter papers and a POSS disdrometer were accomplished from March 2001 to February 2002 at Busan, Korea to investigate the characteristics of raindrop size distributions for filter paper technique and a POSS measurement method. Filter papers were sampled at the top of building No.4 (20m a.s.l), PKNU (PuKyong National University) directly after rainfall in Busan, Korea. POSS was situated next to the filter papers. Z-R relations from many rainfall cases were obtained and compared to DWSR 88D (C band) radar reflectivity values each other. In addition, drop numbers (N), liquid water contents (M) and rain rates (R) were calculated for both methods and compared. A total of 794mm was measured by PSS on 92 rainfall days. Rain rates were from 0.1 to 19.7 mm/h and with an average of 3.7 mm/h. At each season, drop size distributions were shown as similar patterns at smaller sizes than diameter of 1 mm, however they were different from all seasons at bigger ranges than 1 mm size. The big difference of DSD was happened between spring and winter season, relatively. The DSD pattern at autumn season was very similar to winter season as compared with other seasons and its value at summer season was the largest than other seasons at Busan, Korea. Especially, DSD pattern of winter season had good accordance with spring season at smaller sizes than 1.5 mm diameter and its pattern was well followed to the autumn season distribution at bigger sizes than 1.5 mm diameter. An assessment will be made of the better method of estimating DSD.

47



RAIN AND SNOW DISCRIMINATION WITH X-BAND POLARIMETRIC RADAR

Dong-In Lee Hyo-Kyung KIM, Koyuru IWANAMI , Masayuki MAKI and Kyung-Eak KIM Department of Environmental Atmospheric Science, Pukyong National University, Busan, KOREA [email protected] ABSTRACT: Remote delineation of the transition region between rain and snow is of great importance because these two precipitation types have vastly different yet significant social and economical impacts in the regions of occurrence (Ryzhkov and Zrnic, 1998). The contour line of 0C is one of the simplest indicator to forecast the boundary between snow and rain. But this line is not good indicator at Nagaoka(138˚53′E, 37°25′N, 97m ASL), Niigata-Prefecture in Japan, where there are often heavy snowfall on condition of above 0C temperature due to East Asian winter monsoon and topographical affects. An alternative method acknowledged by radar meteorologists that polarimetric radar will play a fundamental role in measurement of hydrometeor characteristics such as size, shape, spatial orientation, and discrimination of thermodynamic phase. The purpose of this study is thus to determine the polarimetric characteristics and delineate the boundary between snow and rain region from two simple polarimetric variables, horizontal reflectivity(ZHH) and differential reflectivity(ZDR). Radar data were collected by the X-band polarimetric radar(X-POL), located at Nagaoka, Japan during winter season(Jan., 20 to Feb., 18, 2000) by NIED(National Institute for Earth Science and Disaster Prevention, Japan). The surface observation site for hydrometeor types was located on the direction of 259° azimuth and 3.7km from X-POL radar site. The discrimination of precipitation type was obtained from snow particle type sensor system developed by Dr. Tamura (1999) in Japan. Hydrometeor types in this instrument are categorized to the 6 types (rain, snow, graupel, wet snow, mixed (snow/rain) and mixed (snow/graupel)).

48



ON THE CHARACTERISTICS OF DROP SIZE DISTRIBUTIONS AND CLASSIFICATIONS OF CLOUD TYPES USING GUDUCK WEATHER

RADAR, BUSAN, KOREA Dong-In Lee Min Jang, Chul-Hwan You ,Byung-Sun Kim and Jae-Chul Nam

Dept. of Environmental Atmospheric Sciences, Pukyong National University, Korea

[email protected] ABSTRACT: A weather radar and POSS disdrometer observations were carried out to obtain a good estimation of radar rainfall and gamma distribution Z-R relationships of DSDs at Busan in Korea. The DSDs of POSS and those of M-P were compared to know the characteristics of DSDs with respect to the total number. There were significantly discrepancies between both especially for raindrops of 2mm diameter in Busan. To remove these differences between them, the least squares regression was used. The slope and intercept of DSDs obtained using least squares method are ranged from 1.9 to 2.7 and 701 to 8,044, respectively. According to the rainrate, there were classified to the six categories; very light, light, moderate, heavy, very heavy, and extreme and new gamma parameters were calculated at each one. Z-R relationships were obtained by cloud types that based on DSDs of POSS since the reflectivity and rainrate are as a function of the drop size. The constants of A and b values

in the relationship of bARZ = were also variable in Busan. In calculating Z-R relationship

with the cloud types, ST algorithm including threshold of 5.5 mm/h were added. As the result of comparison between daily rainfall amount of POSS and that of AWS rain gage, they have good agreement in the most of selected time periods even though there is a little difference. It was also found that POSS is more sensitive than that of raingage at less than 0.5 mm/h. Reflectivities obtained from POSS and Guduck weather radar were compared with each other to verify radar rainfall amount. Differences of fractional standard error were ranged from 18.4 % to 43.1 %. To reduce these fractional standard errors, new equations were obtained using linear regression relation. As a result, they were decreased to 18 %. From these results, it is considered that radar reflectivity should be changed and modified to estimate more accurate radar rainfall.

49



PRECIPITATION FORECAST BASED ON NUMERICAL WEATHER PREDICTION MODELS AND RADAR NOWCASTING METHODS

Charles A. Lin Isztar Zawadzki and Yunquing Xuan Department of Atmospheric and Oceanic Sciences, McGill University, Canada [email protected] ABSTRACT: Radar-based precipitation nowcasting methods are robust and have more skill than numerical weather prediction models over time scales of several hours. This is because the models do not generally capture well the initial precipitation distribution. Over longer time scales, the models would perform better than nowcasting methods as they resolve dynamically the large scale flow. We examine the possibility of combining precipitation nowcasts with model forecast, to yield a forecast over time scales of tens of hours. We first present results of a quantitative comparison of radar-based and model-based precipitation forecasts, using conventional and wavelet methods. We then identify the cross-over point in time where model forecasts start to have more skill than nowcasting methods, using a case study approach. The final step consists of a statistical blending of the two forecasts to produce an optimum forecast; preliminary results are presented.

50



OBSERVATION OF RAINFALL DISTRIBUTION OVER MOUNTAINOUS AND METROPOLITAN AREAS BY X-BAND MULTI-PARAMETER RADAR

Masayuki Maki Koyuru Iwanami, Sang-Goon Park, Ryohei Misumi, Hiroshi Moriwaki, Kennichi Maruyama, Mihoko Suto, Isao Watabe, Dong-In Lee, Min Jang, Hyo-Kyung Kim, V.N. Bringi National Research Institute for Earth Science and Disaster Prevention (NIED) Ibaraki, Japan [email protected] ABSTRACT: Recent studies show polarimetric radar to be advantageous for hydrological applications. In 2000, the National Research Institute for Earth Science and Disaster Prevention (NIED) developed X-band polarimetric radar MP-X and has been developing rainfall estimation algorithms. From scattering simulation using drop size distribution measured with impact type disdrometers, Maki et al. (2003) studied the effect of variations of rain drop size distribution on the estimation error of four types of rain estimator; a classic estimator R(ZH), and three polarimetric radar estimators, R(KDP), R(KDP,ZDR), R(ZH,ZDR), where R, ZH, KDP, ZDR are rain rate, reflectivity factor at horizontal polarization, differential reflectivity, and specific differential phase, respectively. Park et al. (2004) examined attenuation correction algorithms at the X-band based on information of differential phase. These studies show the potential of X-band polarimetric radar for hydrological applications. Encouraged by these results, the NIED planned three-year semi-operational MP-X observations of rainfall. The purpose of observation is to obtain data to achieve the following research goals. 1) Establishment of rain rate estimation algorithms at X-band polarimetric radar applications. 2) Retrieval of raindrop size distribution using polarimetric variables to study the microphysics of precipitations. 3) Development of extremely short-term forecasting algorithms of heavy rainfalls using vertically integrated liquid water content. 4) Development of shallow landslide risk prediction models using rainfall information provided by MP-X observations. 5) Real-time public disclosure of rainfall information using the internet. The MP-X is set up at Ebina, Kanagawa prefecture, Japan. A unique feature of the present observations is that two types of topography, mountainous and flat metropolitan areas, are found in the 80km radius radar observation area. The mountainous areas, such as Mt. Fuji and the Tanzawa mountains, are located to the west of the radar site while the flat metropolitan areas, such as Tokyo and Yokohama, are located to its east. This feature enables us to investigate rainfall distribution over different topographies concurrently. Another feature of the observation area is the availability of ground truth data from several rain-gauge networks. Pukyong National University of Korea provides two kinds of 0.1mm resolution, one-minute tipping bucket rain gauge networks. The first network comprises four rain gauges placed, in a mountainous area, on a straight line at approx. 10 km intervals from the radar site. This network is designed to investigate the effect on rain rate estimation of partial beam blocking by topography. The second network consists of four rain gauges arranged within an area of about 1km2, located at about 10 km from the MP-X. This network is designed to validate polarimetric rain rate estimators within a radar sampling volume. The local governments of both Kanagawa and Tokyo have dense networks of ten-minute rain gauges, with average intervals of about 5 km. In addition to aforementioned rain gauge networks, three impact type disdrometers (NIED and Nagoya University) and one 2D-video dsidrometer (CSU) are used in the observation. Preliminary results of the validation analysis on rain rate estimators will be presented.

51



THE MOVE TO QUANTITATIVE RADAR MEASUREMENT IN THE AUSTRALIAN BUREAU OF METEOROLOGY

Peter May Alan Seed, Tom Keenan, Frank Cummings and Ray Jones Bureau of Meteorology Research Centre ABSTRACT: The provision of weather and hydrological services associated with quantitative rainfall and severe weather are of increasing importance to the Bureau, which has recently received significant funding to upgrade the Australian Radar network of 50 radars. Service improvements in these areas will rely on the quantitative and automated applications based on the use of high quality radar data. These quantitative and severe weather products are required by a range of users including severe weather forecasters, hydrology and the aviation industry. The move away from current qualitative processes towards the quantitative use of radar data for both severe weather and rainfall estimation applications implies that new and more stringent demands will be made of the weather radar network in terms of both data quality and availability. The focus of the paper is largely on what radar signal processing is required within an end to end conventional Doppler system, but related longer term and operational issues are also discussed. In this context it is important to remember the utility of a radar network depends on a number of inter-related issues. They include the location of the radar, the type of radar and size of the antenna, scanning strategies and operational standards, the signal processing used to digitise the returned signals, and finally the algorithms used to generate products. There is a conflict between the operational requirement for a uniform radar specification and antenna programme throughout the radar network, and the need to optimize each radar given the local climate and topography. This becomes a significant issue in Australia, which has radars from the tropics to the mid-latitudes. Although dual polarisation has only been used for research application so far, it is clear from this research that dual polarisation offers considerable advantages. These include accurate rainfall estimates of convective rain, the detection of damaging hail and improved quality control. There are also potential advantages in areas with significant orographic rain. Dual polarisation places stringent demands on the quality of the antenna and radomes in particular, so current designs should allow for polarimetric upgrades without the replacement of such expensive items. Future radar products will include high quality mosaics of radar data. This in turn has implications for the radar operation and maintenance as this ideally requires volumetric scanning and calibrated radars. The new digital receiver system that has been developed in radar section offers the capability of correcting for the affects of second trip echoes, overlaid echoes, optimal estimation of both reflectivity and Doppler velocity, clutter suppression etc that will lead to high qualitative quantitative products. The approach to obtaining the best quality Doppler and reflectivity data possible will be discussed. All the new radar systems will be capable of employing dual PRT (pulse repetition time) techniques to minimise issues related to the Nyquist velocity of the radar. These methods produce some speckle in the data, but algorithms to correct this are well developed. A simple algorithm is currently in use on the Kurnell radar and a slightly more complex algorithm is used operationally in Canada. This produces high quality data suitable for both manual and objective analysis.

52


WEATHER RADAR FOR FLOOD FORECASTING: SOME UK EXPERIENCES

R J Moore A E Jones and V A Bell CEH Wallingford, Wallingford, Oxon, OX10 8BB, UK [email protected] ABSTRACT: The paper will aim to review developments in making weather radar data available for operational use in flood forecasting and warning, with particular reference to the UK experience. An integrated system configuration is used operationally that employs the HYRAD (HYdrological RADar) system to receive and store radar products from the Met Office on a spatial image database, to support animated visualisation on hundreds of PC clients, to perform further processing, and to provide an output interface to flood forecasting and modelling systems used in flood warning decision-making by the Environment Agency. One such modelling environment is the River Flow Forecasting System, or RFFS, developed by CEH and used in the Northeast Region (Yorkshire Ouse, Tees and other Northumbria rivers) and Thames Region. The paper will outline the radar products in use and the HYRAD and RFFS systems, encompassing issues such as radar data correction, radar-raingauge merging, rainfall forecasting, derivation of catchment average rainfall, and the hydrological models used for flood forecasting. Assessments of model performance for flood forecasting using radar data as input will be outlined. Future challenges for flood forecasting using weather radar will be identified, encompassing rainfall estimation, flood modelling and uncertainty aspects.

53



SPATIAL PATTERNS IN THUNDERSTORM RAINFALL EVENTS: CONCEPTUAL MODELING AND HYDROLOGICAL INSIGHTS

Efrat Morin David C. Goodrich Robert A. Maddox Xiaogang Gao Hoshin V. Gupta Soroosh Sorooshian

Department of Hydrology and Water Resources, University of Arizona, Tucson, Arizona

ABSTRACT: The characteristic spatial patterns of rainfall systems evolve from underlying processes in the systems. The important role of these rainfall patterns in watershed hydrology has long been recognized. This paper presents conceptual modeling of the rainfall spatial patterns associated with air mass thunderstorm events using observed radar data. The modeled rainfall retains a relatively simple structure while including the primary spatial features of rainfall patterns such as location and magnitude of maximum rainfall, areal extent of rain - no rain, and small-scale variability. The model was applied to radar rainfall data from Arizona and evaluated using rainfall data from a dense gauge network. Through a case study we demonstrate the use of modeled rainfall as input to a distributed hydrological model. This approach allows a comprehensive linkage between the runoff response and spatial rainfall patterns. We suggest that conceptual rainfall modeling can help in the acquisition of new insights into the behavior of rainfall systems and their interaction with the environment.

54


VERIFICATION OF NWP MODEL RAIN EVOLUTION USING RADAR DATA AND ITS APPLICATION TO THE RAINFALL NOWCASTING

Stanislaw Moszkowicz Institute of Meteorology and Water Management Researcher at Radar Operational Centre [email protected] ABSTRACT: Radar rainfall field does not contain any information about the future rain evolution. Such information is potentially included in the rain field forecasted by a NWP model. The paper deals with verification of this evolution given by different NWP models using radar data both for convective and stratiform storms. A comparison of rain field evolution predicted by the models and observed by MRL-5 radar shows there exists not very high but nevertheless significant correlation (up to 0.6) between predicted and observed evolution of area-averaged rainfall and rain field occupation (percentage of raining pixels). This correlation permit to improve predicted (area-averaged) parameters of the rain fields. Radar rainfall nowcasting, constructed as Lagrangian persistence forecast, suffers from the fact that the rain intensity or accumulation is normally assumed to be invariable though everybody knows the rain systems are changing rapidly. Introduction of the rain evolution may improve the forecast. Increasing or decreasing of area-averaged rainfall does not present any problems, but difficulties appear when one wants to kill some raining pixels or to create new ones. The question is in which location a new raining pixel should be formed or the old rain pixel annihilated. Some possible solutions are examined.

55



POLARIMETRIC RAINFALL OBSERVATION WITH COBRA IN THE RAINY

SEASON Katsuhiro Nakagawa, Hiroshi Hanado, Nobuhiro Takahashi, Toshio Iguchi, Shinsuke Satoh, Kouichi Fukutani Okinawa Subtropical Environment Remote-Sensing Center, Communications Research Laboratory, Okinawa, Japan [email protected] ABSTRACT: In recent years, many researchers have focused on polarimetric information on hydrometeors as a potential source for an improved rainfall estimation and improved understanding of physical processes within precipitating cloud systems [Bringi et al. 2001]. In order to acquire information on hydrometeors, polarimetric weather radar, such as NCAR C-2 radar, CSU-CHILL radar, BMRC C-POL radar and so on, have been developed. Various hydrometeor types, such as raindrops, ice crystals, snow, hail, or graupel, may be identified based on differences in scattered echo by the particle [Doviak et al. 1993]. Communications Research Laboratory (CRL) has developed a new C-band (5340MHz) multi-parameter Doppler radar system with a bistatic Doppler radar network to establish the next-generation technology of rainfall observation for hydrological and meteorological applications such as weather forecasts and run-off analysis in predicting floods. This new radar is named COBRA (CRL Okinawa Bistatic polarimetric Radar). Two transmitter (klystron) units are used for the polarization observation. The transmission polarization for each pulse is selected from six possible polarizations, namely, horizontal polarization, vertical polarization, ±45-degree tilt linear polarizations, and right- and left-handed circular polarizations. The return signal is measured simultaneously by two receivers, one for horizontal polarization and one for vertical polarization [Nakagawa et al. 2001]. We carried out polarimetric rainfall observation with COBRA in Okinawa at the time of the rainy season in 2003. The polarimetric observation data (ZDR, LDR, ρHV(0), KDP, ΦDP, etc.) will be validated by using the 400-MHz wind profiler and ground-based rainfall observation data. Ground-based raindrop-size distribution measurements will be made using a Joss-disdrometer and a 2-dimensional video disdrometer. An optical rain gauge (which can measure the 1-minute precipitation intensity) and a tipping-bucket rain gauge will also be used for observations of precipitation intensity.

56



THE USE OF METEOROLOGICAL RADIO-LOCATING POLARIMETERS IN HYDROLOGICAL INVESTIGATIONS IN THE REPUBLIC OF

UZBEKISTAN Z.Nazirov Central Asian Research Hydrometeorological Institute,72, K. Makhsumova Str., 700052, Tashkent, Republic of Uzbekistan [email protected] ABSTRACT: At present meteorological radio-locating polarimeters are used widely for cloud precipitation investigation and weather modification with the purpose of initiation, augmentation and redistribution of precipitation “on the fly” and on the great distance with remote methods. The methods of application of meteorological radio-locating stations for determining of storm precipitation intensity under deep cumulus clouds over local territory in mountain-valley part of Uzbekistan are discussed in the paper. The theory and method of recalculation of radio-locating characteristics of radar return from storm precipitation into hydrological characteristics of mountain rivers has been developed. The data of field-experimental investigations of hydrological characteristics of mountain rivers gained from meteorological radio-locating polarimeters and using data of mountain river hydrological stations in autumn-winter and spring-summer periods has been analyzed. The correlative function of meteorological radio-locating polarimeters data and true data have been determined. The coefficient of precipitation intensity calculation using reflecting radio-locating signal from storm precipitation taking into account the relief has been evaluated. The measurement errors of mountain river runoff caused by variations in hydrograph timing are discussed. Possibility and effectiveness of meteorological radio-locating polarimeters use for measurements of hydrological characteristics of mountain rivers are demonstrated.

57



RADAR OBSERVATION OF INTENSE RAIN EVENTS DURING THE BOLLÈNE-2002 EXPERIMENT.

John Nicol Guy Delrieu, Dominique Faure, Pierre Tabary, Hervé Andrieu Laboratoire d’étude des Transferts en Hydrologie et Environnement, Grenoble, France. ABSTRACT: The Bollène 2002 experiment was designed with Météo-France to test the relevance of a volume-scanning strategy for both hydrological and meteorological applications of weather radar. An S-band radar belonging to the French ARAMIS network was used for this purpose. Two radar data processing strategies, called the “static” and “dynamic” approaches, are developed for estimating rain-rate fields at ground level. The static approach uses “information a priori” products such as dry-weather ground clutter maps and climatological vertical profiles of reflectivity (VPR). The dynamic approach uses adaptive treatments, such as those that follow. The time series of several ground clutter targets of high intensity are used to check the radar calibration stability. Ground-cluttered areas are identified using a combination of a variability criterion derived from the pulse-to-pulse radar signal and of horizontal reflectivity gradients. Algorithms are implemented to distinguish convective and stratiform regions in the rain field. VPR corrections and specific Z-R relationships are then applied for each rain type area prior to an optimal combination of the measurements performed at the various elevation angles. The processing is based on a modular approach allowing the inclusion of various correction procedures to assess the relative improvement in rainfall estimation. The data processing techniques are evaluated with reference to the data from a dense raingauge network available in the region of interest using the autumn 2002 data set.

58


IMPROVEMENT OF MODEL OUTPUT STATISTICS PROCEDURE FOR CONSTRUCTING SIGNIFICANT RELATIONSHIP BETWEEN RAINFALL

AMOUNT AND ASSOCIATED DOMINANT INDICES Koji Nishiyama Yoshitaka I, Kenji Jinno, Akira Kawamura, Kenji Wakimizu Faculty of Engineering, Kyushu University, Fukuoka, Japan [email protected] ABSTRACT: A variety of numerical weather prediction models used in many countries provide useful information for precipitation forecast. However, the accuracy of the prediction has not reached practical level of quantitative prediction due to relatively coarse resolution and physically complicated precipitation processes in the prediction models. Therefore, the present weather prediction for providing useful precipitation guidance in many countries adopts an alternative technique represented by MOS (Model Output Statistics), which means downscaling method of prediction model outputs by relating it to observed data of a specific area using statistical techniques including multiple linear regression and neural network. However, the MOS seems to contain a fatal problem in constructing a relationship between observed rainfall and associated dominant indices because the prevailing MOS uses statistical correspondence between predicted outputs and observed rainfall in a narrow sampling area or a grid scale of prediction model. From the background, this study relates physically significant indices consisting mainly of precipitable water (PW) and convergence (CONV) of air with rainfall amount estimated from the composite of radar and rain gauge system in Japan in a narrow sampling area. The relationship between PW and rainfall at a specific point showed roughly good consistence between them in some cases of heavy rainfall excepting many cases representing relatively small amount of rainfall because of widely distributed large PW. On the other hand, the relationship between CONV and rainfall at the specific point leads to physically inconsistent result although CONV indicates sufficient narrow spatial distribution to specify the dominant area of heavy rainfall. This fact implies that a narrow sampling area do not intersect a narrow CONV area. Therefore, the improvement of physically and statistically meaningless relationship into significant one will require the extension of a sampling area. This method will lead to the construction of significant MOS, in other words, significant rainfall prediction. Actual detailed analysis and some results of the improvement will be shown.

59



MICROPHYSICAL ANALYSIS OF ARTIFICIALLY INDUCED RADAR ECHOES BY LIQUID CARBON DIOXIDE SEEDING

Misato Nagata Kenji Wakimizu, Koji Nishiyama, and Norihiko Fukuta Faculty of Agriculture, Kyushu University, Fukuoka ,Japan. [email protected] ABSTRACT: In order to effectively convert large amount of cloud water into precipitation, it is important that artificially formed ice particles attains sufficient size to be detected by meteorological radar at the cloud top and large fall velocity to arrive at the ground within the period of seeding operation. To achieve the goal, Fukuta (1999) suggested LOLEPSHIN (Low Level Penetration Seeding of Homogeneous Ice Nucleant) using liquid carbon dioxide (LC), which causes strong evaporative cooling reaching -90 °C and the subsequent generation of approximately 1013 ice particles per gram of LC by homogeneous nucleation. The method is designed to induce a large amount of precipitation through two basic processes. The first process is called RETHIT, Roll-up Expansion of Twin Horizontal Ice Crystal Thermal. In the process, artificially formed ice particles can grow into a size large enough to fall out and to be detected by meteorological radar with relatively little competition among growing ice particles until the thermal arrives at the cloud top. The second process is called FILAS, Falling-growth Induced Lateral Air Spreading. In the process, the volume of artificial ice particles expands through microphysical and dynamical interactions in the cloud. Consequently, it is expected that falling ice particles continue to grow further by consuming a large amount of cloud water during the horizontal expansion and fall. This seeding method was applied to supercooled convective clouds under the outbreak of cold air mass from the Siberia in Northern Kyushu, Japan, on February 2, 1999. Under the condition, tracing of radar images confirmed the formation of two artificial radar echoes. The echoes expanded horizontally and reached to approximately 17 km in the upper width and took a mushroom shape with the echo symmetrically spread on both sides of the upper portion of the artificially induced ice crystal thermal. In this paper, focusing on simple RETHIT process, dynamical and microphysical processes linking to artificial radar reflectivity induced by LOLEPTHIN were investigated using a simple time-dependent microphysical numerical model. The results indicate that the reflectivity of seeded radar echo depends on the available ice water content and the number concentration of ice particles. This relationship shows that inferred properties of artificially induced radar echo in FILAS process are consistent with the features of the seeded radar echoes. Application of the radar-based microphysical evaluation to cloud seeding studies and operation will provide useful information to judge whether seeding operation leads to a significant result.

60



THE USE OF HIGH RESOLUTION RADAR-RAINFALL ESTIMATES TO IDENTIFY RUNOFF PROCESSES, THRESHOLDS, AND SPACE-TIME

VARIABILITY AT THE CONTINENTAL SCALE. Fred Ogden Ryan Knox, and John Zahner Dept. of Civil and Environmental Engineering, University of Connecticut, Storrs, CT 06269 USA ABSTRACT: Weather radar rainfall estimates (2x2 km, 15 minute) are used to estimate storm total rainfall, fractional coverage of rain, and to model soil moisture content as a surrogate for antecedent soil moistures. 15-minute discharge data from a significant number of catchments are analyzed. Catchments considered are distributed throughout the Mississippi River basin, have areas less than 1000 km2, and are within 100 km of a WSR-88D radar to avoid significant range effects. Only warm season storms are considered. Runoff production thresholds are identified using a combination of runoff efficiency, recession curve analysis, and comparisons with hydrologic simulations driven by radar-rainfall. This methodology provides a physically-consistent methodology for identifying runoff production mechanism and key thresholds. Thresholds that are strongly dependent on rainfall volume or rainfall rate are indicative of the mechanism. Results have important implications for appropriately formulating land-surface schemes and in prediction of floods, as well as in understanding the space-time variability of runoff production at the continental scale.

61


COMPARING RIVER BUREAU’S RADAR AND JAPAN METEOROLOGICAL AGENCY’S AND ESTABLISHING THE BEST COMBINATION OF DATA IN

KISARATSU RIVER BASIN AND KATSURA RIVER BASIN Masayuki Ote Hydrologic Engineering Team, Hydraulic Engineering Research Group, Public Works Research Institute [email protected] ABSTRACT: Currently, there are two systems of radar rain gauges operated in Japan. River Bureau of Land, Infrastructure and Transportation Ministry has a type 26 and Japan Meteorological Agency has a type 20. The ones owned by River Bureau are used for the purpose of disaster prevention and collecting rainfall data in certain area for short period while rain gauges belong to Japan Meteorological Agency are for weather forecasting. This paper compares the accuracy of rainfall estimates from the River Bureau and Meteorological Agency radars. Comparisons are performed with and without calibration using ground-based rain gauges. Optimal combinations of rainfall measurements from the two radars are also examined.

62



ATMOSPHERIC RAINFALL INTENSITY IN ALBANIA Niko Pano Bardhyl Avdyli Hydrometeorological Institute, Hydrology Department, Rr."Durresit", 219, Tirana, Albania. ABSTRACT: In this paper it is attempted to present a general evaluation of the atmospheric rainfall intensity in the Albanian territory. Albanian territory is one of the most complicated natural area in the Europe because of its physiographical specific conditions: mountaineous regions ,with average altitude 785m above the sea level, complicated morphometrical territory, typical Mediterranian climate etc. Atmospheric rainfall intensity evaluation was based on the multi annual data of the of the Albanian hydrometeorological Institute of the Academy of Sciences.Meteorological monitoring network consist of more then 125 pluviometric and 36 pluviographic stations with observed period of 15-45 years. These stations are located all over the territory. The division scheme of the Albanian territory in homogeneous areas, based on the evaluation and determination of the natural factors influencing in the rainfall intensity is presented.One of the important representative indicators to estimate the integral impact of these factors is the 24 hours precipitation with p=1% probability H24

p=1% on the Albanian territoty is realized according to the respective values of the gradient PM=H24

P=1% / h, where:h—is altitude of the territory.Region is accepted as the smallest tacsionometric unit.There are three particular homogeneous regions,the first region:75-120mm;the second one:121-200mm;and the third one:201-450mm. The reduction lines of the rainfall stratum together with their respective coordinations are also presented.The rainfalls of different duration т and values of different percents security are calculated for each mteorological station.The calculated duration of rainfalls are т = 10,20,30, one hour,2,3,4,6 and 12 hours. For each one zone the rainfalls intensity and regional parameters such as reduction of the rainfall depth-Ψp(т) mean rainfalls intensity ITP,the coordinates of the subsidiary curves S(т) and E(т),precipitation reduction coefficent ŋ, are evaluated.For the rainfalls of different duration т =10,20,30,one hour,2,3,4,6 and12 hours with p=1%probability are composed of geographical distribution maps. Evaluation of the atmospheric precipitation intensity is used for hydrological forecasting,calculation in hydrotechnic projects such as bridges, artificial channels to discharge maximal flow,erosion process etc. Further modernization of the meteorological monitoring network with up-to-date apparatus such as radars will help us with meteorological accuracy and as a consequence to improve our hydrological studies and applications by profiting from utilization of weather radar according to your experience in this field.

63


CORRECTION OF RADAR REFLECTIVITY AND DIFFERENTIAL REFLECTIVITY FOR RAIN ATTENUATION AND ESTIMATION OF

RAINFALL AT X-BAND WAVELENGTH S.-G. Park M. Maki, K. Iwanami, R. Misumi, V. N. Bringi National Research Institute for Earth Science and Disaster Prevention, Tsukuba, Japan [email protected] ABSTRACT: The effect of attenuation of radar intensity measurements (radar reflectivity ZHH and differential reflectivity ZDR) at X-band wavelength on rainfall estimation is examined. The attenuation of ZHH and ZDR are corrected by the self-consistent method proposed by Bringi et al. (2001), which was evaluated with C-band polarimetric radar data. In this study, the self-consistent method is modified for the MP-X radar of the National Research Institute for Earth Science and Disaster Prevention (NIED), Japan. The MP-X radar is a linear orthogonal polarized radar operating at 9.375 GHz. Since 2001, the MP-X radar has observed various precipitation systems in Japan, such as typhoons, frontal systems, and thunderstorms. In addition to the radar observations, a few disdrometers have also collected surface drop size information within the radar observable area. The method of attenuation correction has been evaluated for various precipitation systems, based on usage of the relations mentioned above, and the effect of attenuation on rainfall estimations has been examined by comparison to ground-based disdrometer data. These results will be presented.

64



IDEALIZED SIMULATION OF MOUNTAIN FLOW WITH SPECIAL EMPHASIS ON RAINFALL GENERATION

Assela Pathirana Tadashi Yamada Department of Civil Engineering, Faculty of Science and Engineering, Chuo University, Tokyo [email protected] ABSTRACT: Explicit physical modeling of atmospheric processes in numerical models as a means of predicting rainfall and related phenomena has largely been limited to larger spatial scales in the past. However, with the increasing availability of cheap computing power, the resolutions at which the atmospheric models can be reasonably operated are rapidly increasing. This situation makes it possible to use physical models of the atmosphere to revisit ‘classical’ atmospheric problems whose treatment were was traditionally limited to simplified linear treatment. One such problem is the analysis of a steady wind field with a mountain ridge. This interaction is one of the most important rainfall generation mechanisms in many climatic regions of the world. Idealized numerical experiments help to understand the effects on various otopographical, static and dynamical parameters on the behavior and rainfall response of mountain flow. Scope of most of the past studies on idealized simulation of mountain flow has been limited to the dynamical behavior of atmosphere, thus neglecting the microphysical aspects and thus the rainfall generation. In the present research, we have modified the MM5 modeling system version 3 of Pennsylvania State University, to give it the ability to perform idealized simulations. The modified modeling system was used to simulate the mountain flow phenomena under a large number of different atmospheric and topographical conditions, with special focus on quantitative distribution of rainfall. One of the objectives of this modeling study was to understand the rainfall generation of the interaction of summer monsoon winds with the central mountains of Sri Lanka. It was found that the rainfall patterns like that of monsoon Sri Lanka are results of delicate balance of number of factors and a change of a single factor can result in quite different rainfall distributions. One of the notable factors is the critical importance of the easterly jet mechanism to restrict rainfall only to wind-side slopes. Further, vertical distribution of humidity, the effect of land heating during the day time and the interaction of sea-breeze with monsoon wind fields also seems to affect the rainfall quantity.

65



NATIONAL FLOOD NOWCASTING SYSTEM TOWARDS AN INTEGRATED MITIGATION STRATEGY IN SOUTH AFRICA

Geoff Pegram and Scott Sinclair Civil Engineering, University of Natal, Durban, 4041, South Africa. [email protected] ABSTRACT: Based on a successfully completed pilot project funded by the Water Research Commission (WRC) for the city of Durban and surrounds: “Umgeni river nowcasting using radar – an integrated pilot study”, the WRC has awarded a contract to the University of Natal (UN) to go national. This brings together, under one umbrella led UN, the latently disparate Government departments: South African Weather Services (SAWS), Department of Water Affairs and Forestry (DWAF) and the newly promulgated National Disaster Management Centre (NDMC). SAWS is responsible for the dissemination of accurate rainfall estimates and forecasts based on gauges, radar and satellite data; DWAF is responsible for large rivers, primarily the Orange and the Vaal and for providing streamflow gauging data countrywide. The catchments prone to flash floods, especially those in urban areas are the responsibility of the newly formed Metro Disaster Management Centres. They need to have flows converted to images of levels of inundation assisted by their Metro engineers. The aims of the project are:

1. To put an effective, efficient, available national flood-forecasting system in place. 2. Use this system to forecast flood inundation levels routinely. 3. Use these forecast inundation levels to alert vulnerable people, industry, Disaster

Managers in order to mitigate the effects of floods. 4. Have recent information (satellite, radar and gauge estimation of rainfall) distributed

from the NDMC/DWAF:PSU (National Disaster Management Centre/Department of Water Affairs & Forestry: Public Safety Unit) to the regions.

5. Provide flood nowcasts/forecasts (with horizons of 1/2, 1, 2, 4, 8 hour etc) to sensitive regions in as much detail as required

6. Interact and work with local Disaster Managers and Local Authorities to convert flows to inundation levels.

7. Provide training initiatives (annual courses, presentations and software) for local Disaster Managers using simulated weather systems to augment training on historical events.

The presentation will focus on the overall organisation of the project and will highlight those aspects of the technical input which have so far been successfully completed. The latter include the powerful combination of linear catchment transfer function models combined with Kalman Filtering in providing optimal ensemble streamflow forecasts, the stochastic nowcasting of advecting rainfields measured by radar using the “String of Beads” model and the merging of satellite, radar and raingauge fields to provide a “best” estimate of areal rainfall country-wide, important as input to the catchment modelling process.

66



USE OF A STOCHASTIC PRECIPITATION NOWCASTING SCHEME FOR FLUVIAL FLOOD FORECASTING AND WARNING.

Clive Pierce Neill Bowler, Alan Seed, Anne Jones Met Office, Joint Centre for Hydro-Meteorological Research, Maclean Building, Crowmarsh Gifford, Wallingford, OXON, OX10 8BB, UK [email protected] ABSTRACT: At present, operational Quantitative Precipitation Nowcasting (QPN) typically involves the generation of a single, best guess precipitation nowcast produced by extrapolating the motion (and sometimes the intensity) of radar inferred precipitation. The quality of these best guesses can vary greatly with the nature and extent of the precipitation. Furthermore, on their own, they do not provide the end-user with information on the associated forecast error. In collaboration with the Bureau of Meteorology (Melbourne, Australia), the Met Office (Joint Centre for Hydro-Meteorological Research, UK) has developed a stochastic precipitation nowcast scheme, designed to model and predict the PDF of surface rain rate and rain accumulation in space and time. Here we demonstrate the range of probabilistic products generated by the scheme, and their potential applications for fluvial flood forecasting and warning. With the aid of a hydrological model (the PDM), we consider the use of ensembles of predicted catchment rain accumulation in evaluating the range of possible river flow responses from a given catchment. When employed in conjunction with a catchment specific, cost-loss, decision-making model, we demonstrate the value of PDFs of catchment rainfall accumulation and river flow as an aid to objective decision making within the flood warning process.

67



RADAR UNCERTAINTIES IN CELL-BASED TITAN ANALYSES Arquimedes Ruiz-Columbie Orlando Nuňez-Russis, Dale Lee Bates Active Influence Assoc. [email protected] ABSTRACT: Radar uncertainties are analyzed through the use of error analysis in numerical radar data obtained by TITAN (Thunderstorm Identification Tracking And Nowcasting) in clouds and rainfall over Texas. The estimation of uncertainties in repeatable measurements is discussed due to the fact that repeated measurements cannot be obtained when dealing with clouds. Every cloud is a unique and unrepeatable process, and therefore, the actual estimation involves radar uncertainties plus natural variability. Although usually meteorologists and hydrologists do not bother to consider uncertainties in every day measurements, these uncertainties are very important to validate the radar as a tool in the measure of hydro-meteorological quantities and in the evaluation of potential increases in applied weather modification projects.

68



ASSESSING ERRORS IN URBAN DRAINAGE SYSTEM MODEL OUTPUT FLOWS DRIVEN BY RADAR AND OTHER RAINFALL DATA

G. L. Robbins C. G. Collier Telford Institute of Environmental Systems, School of Environment & Life Sciences, University of Salford [email protected] ABSTRACT: Over the last 15 years or so procedures have been developed to assess the errors involved when assimilating meteorological data into numerical weather prediction models. One approach, based upon Baysian statistics, enables the combination of the model error probability density function (p.d.f.) (background error) with the error p.d.f. of the observations to provide an anlaysed p.d.f. for the model output. In this paper we describe a procedure based upon this approach which has been developed to assimilate rainfall measurements made by different hydrometeorological networks into an urban drainage system (UDS) in order to assess the likely error to be associated with flow predictions. The technique has been developed using data collected within the Bolton town centre UDS in N.W. England. The UDS is modelled using a version of the commercially available hydraulic model, HydroworksTM. The assimilation technique, which has its origins in Generalised Likelihood Uncertainty Estimation (GLUE) originally developed to optimise hydrological model parameters, uses an objective measure of the performance of the model in predicting flows at the UDS outlet. The Baysian statistics approach is then used to combine the model, derived from model validation measurement campaigns, and observational p.d.f. enabling an estimate to be made of the error in the model output flow predictions. We describe how a calibration dataset using data from raingauges, dual wavelength microwave links and the Hameldon Hill C-band weather radar sited some 25km from the UDS, is constructed. Work is ongoing to investigate the performance of a model of a large fluvial catchment. It is noted that the methodology offers a way to associate an error to the model predicted output flows in real-time dependent upon knowledge of the model error p.d.f., rainfall variability and the radar error p.d.f. Whilst the measured radar error p.d.f. may be derived from the calibration dataset, an added degree of sophistication could involve a classification scheme based upon the measured radar vertical reflectivity profile rather than upon rainfall variability. This would enable radar errors to be taken account of in the error derivation. An attempt is made in this paper to demonstrate how this might work.

69



MEASUREMENTS OF RAIN WITH A POLARIMETRIC WSR-88D RADAR. OPERATIONAL DEMONSTRATION. Alexander Ryzhkov Scott Giangrande Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma, USA. [email protected] ABSTRACT: In the past year, the National Severe Storms Laboratory has been conducting an operational demonstration of the polarimetric utility of the KOUN WSR-88D radar. The capability of the KOUN radar to estimate rainfall was tested on a large data set containing more than 30 rain events representing different seasons and different types of rain. Two gage networks, the Oklahoma Mesonet and the Agricultural Research Service (ARS) Micronet, are used to test different polarimetric algorithms for rainfall estimation. One-hour rain totals are estimated from the KOUN radar using different polarimetirc algorithms and compared with hourly accumulations measured by the gages. Both point and areal rain estimates are examined. The performance of different algorithms as a function of range is analyzed up to 230 km from the radar. At relatively close distances from the radar (< 125 km) where bright band contamination is negligible, the quality of radar rainfall estimates is mostly determined by DSD variations and the possible presence of hail. As our analysis shows, these two problems are best addressed by the synthetic R(Z,KDP,ZDR) algorithm. It combines merits of the Z-ZDR pair for light rain, the KDP-ZDR combination for moderate-to-heavy rain, and capitalizes on relative insensitivity of KDP to the presence of hail. All polarimetric methods outperform the conventional R(Z) algorithm in terms of RMS error. In the range interval 125 – 200 km, the bright band becomes a leading factor affecting the performance of all algorithms during cold season, when rain is predominantly stratiform and the melting level is quite low. At these distances, the synthetic algorithm is no longer superior because Z and ZDR are substantially affected by melting hydrometeors. Surprisingly, the R(KDP) algorithm is more immune to the bright band contamination than the others. It performs best of all, both in terms of bias and RMS error. The situation is very different in the warm season when rain is mostly associated with strong localized convection, rain fields are very non-uniform, and bright band contamination is not a key factor. Rain estimates based on KDP rapidly degrade with distance. The reasons for such degradation are discussed. At the ranges beyond 200 km, all radar algorithms for rainfall estimation perform equally poor due to overshooting, beam broadening, and loss of sensitivity.

70



HAILSTORMS AND THE ESTIMATION OF THEIR IMPACT ON INSURED PROPERTIES USING RADAR

Sandra Schuster Risk Frontiers, Macquarie University, Sydney, Australia [email protected] As property values inevitably increase, especially in major metropolitan areas, so does the exposure to storm and hail losses. The main variables responsible for hail damage to property are:

• Frequency of hailstorms • Intensity of hailstorms, dependant on number, size and kinetic energy of hailstones • Wind speed • Vulnerability and exposure of buildings to hailstorms

This presentation focuses on comparisons between radar-derived variables (radar reflectivity & hail kinetic energy), hailstone size data, and insurance claims. A methodology for establishing a relationship to estimate hail damage using radar technology is introduced. Initial results are presented using data from the metropolitan Sydney area. The data comprise recent thunder and hailstorms including the 1999 Sydney hailstorm, the most damaging natural disaster event in Australian insurance history. The Bureau of Meteorology Weather Watch Radar, located at Letterbox, 60km southwest of Sydney, provides radar data. This “WSR 74 S Band” non-Doppler radar operates with a 1.9deg bandwidth, 10min time steps and a 1km range resolution. CAPPI (Constant Altitude Plan Position Indicators) reflectivities are derived from sector-volume scans at a height of 1.5km (low storm level and close to the surface where hailstones occur). The hail kinetic energy per radar pixel Ekinp [J/m2] is calculated using the CAPPI radar reflectivity integrated over the duration and area. This represents the entire volume of hail received per surface unit over time, a quantity that is related to hail damage. This empirical relationship is based on 175 hailstone spectra and therefore avoids hailstone sizes. Consequently Ekinp is the same for a volume with few large or many small hailstones. Radar reflectivities greater than 55 dBZ are assumed to contain a significant amount of hail. The introduced radar methodology is likely to be a very useful tool in order to estimate extent of hail damage on buildings. Initial results indicate a preferred area for hail damage to the left side of the storm path. Merging cells tend to cause more damage than single cells. Insurance claims data are required in order to calculate loss curves (damage functions), which relate hail Ekinp to loss ratio (damage in relation to total sums insured) and mean damage & cross-correlations. Insurance companies could use damage functions resulting from this research to estimate the extent of hail damage to their property portfolio (the number and costs of insurance claims) immediately after the event and could derive probable maximum losses (PML) to building portfolios. A nowcasting warning system, assisting State Emergency Services to manage demand in certain areas (e.g. manpower and tarpaulins to protect damaged roofs) could prove beneficial.

71



USE OF RADAR RAINFALL ESTIMATES TO EVALUATE MESOSCALE MODEL FORECASTS OF CONVECTIVE RAINFALL

Alan Seed Elizabeth Ebert Bureau of Meteorology Research Centre, Melbourne, Australia [email protected] ABSTRACT: High resolution mesoscale models are capable of producing realistic precipitation fields in many convective systems. While it is impossible to accurately predict the precise evolution of convective rainfall at all points in time and space, a model forecast will be useful in a forecast situation if it can predict:

• the general region and timing of convection, • the mean rain accumulation over the lifetime of the event (i.e., is unbiased), • an appropriate range of rain intensities (accounting for model resolution dependency), • rain structures of realistic size.

If radar data are available, then one can visually compare model rain forecasts with radar observed rainfall on time scales of several minutes to several hours. Real-time monitoring allows the forecaster to assess whether the model is on the right track, and if not, whether mental adjustments can be made to the model output to improve near-term rainfall predictions. Post-event analysis of a larger number of cases using a variety of diagnostic verification strategies can reveal systematic errors in the model's representation of convection. This talk will describe the use of radar observations over southeastern Australia to evaluate rainfall forecasts from the Bureau's 5-km mesoLAPS model on a variety of space and time scales.

72



COULD WE PLAY HYDROLOGY WITH RADARS? Daniel Sempere-Torres and Rafael Sánchez-Diezma Grup de Recerca Aplicada en Hidrometeorologia (GRAHI). Universitat Politècnica de Catalunya, Barcelona, Spain [email protected] ABSTRACT: Since these first days, radar has been seen as a powerful tool to describe and understand the meteorological phenomena related to precipitation, and a promising technology for hydrology. Even nowadays, meteorological radar is still the unique instrument able to provide a description of the precipitation field with the resolution required in hydrological modeling (spatial resolution about 1 km2 and temporal resolution of 5 to 10 minutes). Requirements that lies far out of the conventional raingauge networks densities (usually about 1 raingauge per 50-200 km2 (Andrieu et al., 1992; Sempere Torres, 2000). This important role radars could play has been widely understood, and almost all countries around the world have carried out significant efforts to install radar networks. Although these operational networks have usually been designed from purely meteorological criteria and requirements, commonly they have been seen as tools to improve hydrological models and forecasting systems. Nevertheless, up to now they have essentially been used for qualitative applications and the quantitative applications are still rare, out of a few research developments. The present situation is that this promising tool has not been able to fulfill the hopes of the hydrologists as potential users, and radars are still seldom used on operational hydrology. The main reason for this situation is that radar measures rainfall in an indirect way, which needs a sequence of very complex correction procedures. Procedures that only have been clearly established since the early 1990s, when hydrologists working on the problem of radar rainfall estimation have begun to take a physical approach to study the principle of radar measurements from a hydrological perspective, leading to what it is now known as ‘radar hydrology’. Thus the advancements of the last 20 years (the work of Zawadzki, 1984 can be considered as the milestone showing this change of mentality) have allowed us to state that the hydrological applications of weather radar can be achieved if a systematic correction of the errors and inaccuracies, inherent to radar measurement in itself, are corrected. This presentation will show through selected case studies several illustrations of the importance of taking into account these corrections and the need we still have to improve the quality of estimates. The key question about if we could play hydrology with meteorological radars, and why radar still remains a “revolutionary promise” for hydrologists will be thus addressed.

73



ADJUSTMENT AND SPECIAL PROCESSING OF THE RADAR PRECIPITATION DATA AS APPLIED TO ENGINEERING HYDROLOGY

Vladimir A. Shutov Valday Branch of the State Hydrological Institute. [email protected] ABSTRACT: Problems are addressed of inaccuracies and adjustment of the radar data with special attention to the winter precipitation measurements for a hilly area in northwestern European Russia. Calibration of the Valday radar has been done by using a special rain-gauge network over the Polomet' river watershed. The ground network consisted of 38 gauged sites aggregated in a number of nested groups. Hourly (3-hourly) rates were interpolated into grid cells. The ratio gauge/radar (G/R) was found varied from 0,5 to 3,8 for 3-hourly rates, and from 0,7 to 2,2 for daily amounts. Of particular importance is that the G/R ratios are not randomly distributed. By analyzing their maps, one can infer about specific inaccuracies which are correlated with the terrain features. On winter the radar can underestimate those snow fallen onto the hills, there is a considerably scattered snow water equivalent (SWE) and accumulated radar estimates. Based upon experience, the problem of how to improve that estimate seems to be clarified. Methods are proffered to process the radar-based precipitation data as convenient to the engineering hydrology applications. Three main properties of precipitation fields have been examined: (1) spatial coverage, (2) spatial variograms and (3) empirical conditional frequency of at-cell rain rates by specified spatial averages. Examples of the data processing are given to expound upon the ideas of how to describe precipitation patterns statistically. For instance, a procedure is proffered to reveal rain rates exceeding that threshold correlated with antecedent wetness of a drainage basin. Classification was done of the radar images using the approach based on cluster analysis. Two types of rainfall fields with three subtypes each were identified as depending on intensity and localization of the most heavy rainfall in the area examined. The procedure has been developed for hydrological calculations with the use of radar precipitation and its spatial characteristics. It is based on the so-called “extreme intensity” method accepted in Russia for engineering when the maximal (peak) flows and their frequency are to be determined. The peak flow conditions are simulated and represented as spatial frequency curves for runoff depth on conventional watersheds 10x10 km in area composed of different soils and produced, respectively, different runoff rates. Presented allows us to guess where there would be runoff depth exceeding some specified value or, in opposite, to evaluate what a flow may be expected by given rain rates fallen onto the given area. An expert outlook is finally given: how to deal with the precipitation space-time series when extended studies will focus on the climate variability and water resources assessment for any water project.

74



COMBINING RADAR AND RAIN GAUGE RAINFALL ESTIMATES FOR FLOOD FORECASTING IN SOUTH AFRICA

Scott Sinclair and Geoff Pegram

Civil Engineering University of Natal Durban [email protected]

ABSTRACT: The Hydrologist’s traditional tool for measuring rainfall is the rain gauge. Rain gauges are relatively cheap, easy to maintain and provide a direct and suitably accurate estimate of rainfall at a point. What rain gauges fail to capture well is the spatial variability of rainfall with time, an important aspect for the credible modelling of a catchment’s response to rainfall. This spatial variability is particularly evident at short timescales of up to several days. As the period of accumulation increases to months and years the expected spatial variability in rainfall is reduced and rain gauges provide an improved estimate of the spatial rainfall fields. Due to the fractal variability of rainfall in space, simple interpolation between rain gauges does not provide an accurate estimate of the true spatial rainfall field, at short time scales. Weather radar provides (with a single instrument) a highly detailed representation of the spatial structure and temporal evolution of rainfall over a large area. Estimated rainfall rates are derived indirectly from measurements of reflectivity and are therefore subject to errors that need to be corrected by some form of ground truthing. Rain gauges are commonly used as such a ground truth but their point measurements are not directly comparable to the volume-averaged estimates of weather radar. This paper describes a recently proposed merging technique and presents an operational application for flood forecasting in South Africa. The technique employed is a conditional merging technique (Ehret, 2002), which makes use of Kriging to extract the optimal information content from the observed data. Adopting a mean field based on the Kriged rain gauge data, while the spatial detail from the radar is retained reduces bias in the rainfall estimate. The variance of the estimate is reduced in the vicinity of the gauges where they are able to provide good information on the true rainfall field. At greater distances from the gauges the radar gives the best available estimate and the variance associated with the radar estimate is retained. Real-time rain gauge measurements are combined with data from the South African Weather Services (SAWS) network of weather radars, as an input component of a flood forecasting system.

75



SUSTAINING INVESTMENT IN RADAR DATA PROCESSING Alison Smith Malcolm Kitchen Met Office, Exeter, UK In the UK, it is perceived that the benefit to operational hydrology derived from the radar network is limited by the quantitative accuracy of the rainfall estimates. Steady improvement in radar data quality over recent years has resulted in the data being used as input into flow forecasting models in a limited number of catchments. If flood warning service delivery targets are to be met in a larger number of catchments whilst avoiding heavy investment in new gauge networks, further improvement in radar accuracy is essential. Unfortunately, all other things being equal, the benefit to cost ratio of investment in radar data processing is likely to decline over time. It becomes increasingly difficult to make further gains in data quality (because all the simple things have been done) and the more fundamental limitations of the radar technique start to assume a greater relative importance. To respond to this challenge in the UK, the Met Office and the Environment Agency are investing in a project to re-engineer the radar data processing chain. The objective is not only to bring about an incremental improvement in the accuracy, availability and timeliness of radar rainfall products, but also to design a system that should improve the benefit to cost ratio from future development work. The design fundamentals are:- Concentration of all quality control and correction procedures on a central system, with minimal processing at radar sites. Deferment of categorical quality control decisions to the point in the processing chain where the decisions can be made in the light of all available meteorological evidence. Retention of maximum data resolution as close as possible to the point where products for end-users are generated. Use of best-practice correction algorithms and collaborative development to minimise costs. Common processing of UK and other radar data from Europe. These features should provide a much higher level of efficiency and flexibility in the processing than has previously been available in the UK. This in turn, should enable cost-effective development to continue. Change on this scale is difficult because existing radar systems are involved and implementation has to avoid interruption to the existing services of operational products. The task is made easier by the flexibility of TCP/IP connections to both the radar sites and the principle user systems. The paper will describe in more detail the system architecture and the algorithm selection/testing process.

76


ANALYSIS OF RAINFALL-TOPOGRAPHY RELATIONSHIPS IN JAPAN WITH REGARD TO THE SPATIAL SCALE OF MOUNTAIN SLOPES

Yoshiharu SuzukI Eiichi Nakakita, Masahiko Hasebe and Shuichi Ikebuchi Department of Architecture and Civil Engineering, Faculty of Engineering, Utsunomiya University [email protected] ABSTRACT: One of the important subjects on hydrology is determining the properties of spatial-temporal rainfall distribution with regard to topographic features in a target region. In the current study, for the purpose of making a stochastic model of rainfall distribution in mountainous regions, an attempt was made to determine the rainfall-topography relationships through the analysis of rainfall observed by the Miyama radar for a period of about ten years. The radar is located in the Kinki region of Japan and has a quantitative observation range of 120 km radius, a spatial resolution of 3 km and a time resolution of 5 minutes. The current study investigates the mechanism of topographic effects on rainfall distribution with regard to the spatial scale of mountain slopes and some topographic factors; topographic elevation, incline and direction of mountain slopes. For a quantitative estimation of topographic effects, the authors have focused much attention on “the Dependence Line on Topographic Elevation (DLTE)” of rainfall distribution, which represents the correlation between rainfall distribution and topographic elevation. Through the analysis of various cases of rainfall distribution which have various time scales of accumulation, DLTE was found to be one of universal rainfall-topography relationships in previous studies by the authors. It has not been, however, determined what scale of mountain slopes DLTE generally holds in and what spatial scale rainfall-topography relationships should be modeled in to represent the properties of rainfall distribution in an entire region. Thus, investigations were carried out on rainfall-topography relationships by partitioning mountains into some slopes which have almost uniform topographic factors respectively after computing topographic factors which represent the average features of topography in various scales. Using digital elevation data with a resolution of 1 km, incline and direction of a plane which approximates to the form of mountains within a certain radius of each grid were computed as the incline and direction of topography at each grid with changing the radius of influence variously. It was found through the investigations that DLTE of rainfall distribution could hold on relatively small mountain slopes if mountains were partitioned appropriately. It was investigated additionally what spatial scale was most appropriate for the partitioning to represent rainfall-topography relationships in an entire region using DLTE of rainfall distribution. In the investigation, AIC (Akaike Information Criterion) was used as a criterion for estimation.

77



QUANTITATIVE PRECIPITATION ESTIMATION AND FORECASTS FOR FLOOD DEFENCE BASED ON POLISH WEATHER RADAR NETWORK

Jan Szturc Institute of Meteorology and Water Management, Branch of Katowice, PL 40-045 Katowice, ul. Bratkow 10, Poland [email protected] ABSTRACT: A modern weather radar network POLRAD is been developed in Poland. Finally the network operated by Institute of Meteorology and Water Management will consists of eight Gematronik C-Band Doppler, magnetron or klystron radars. At present fifth radars work in frame of the network. One of the most important tasks of the POLRAD network will be support of flood protection since weather radar provides a high-resolution input to hydrological models. Research activities are focused on their application to flood forecasting system. So there are two main investigated areas: quantitative precipitation estimation (QPE) and quantitative precipitation forecasts (QPF). Problem of application of radar precipitation data is a problem of estimation of precipitation field which should be as close as possible to “true” precipitation. It is done by combination with raingauge network data. Different approaches are assumed to merging like weighted mean method or using of Bayesian approach. Radar data is especially useful for precipitation forecasting for very short time periods ahead, what is so called nowcasting. Used methods are mostly based on extrapolation technique, but results are not quite satisfactory. The other approach using artificial neural network (NN) concept was investigated.

78



THE DEVELOPMENT OF SHORT-TERM RAINFALL PREDICTION SYSTEM IN MOUNTAINOUS REGION BY THE COMBINATION OF EXTRAPOLATION

MODEL AND MESO-SCALE ATMOSPHERIC MODEL Nozomu Takada Japan Weather Association [email protected] ABSTRACT: In steep mountainous regions, it is very important for runoff prediction to develop the rainfall prediction model which predicts the short-term rainfall with high spatial resolution. In this study, short-term rainfall prediction system over Kurobe-valley, where is formed by several steep high mountains, has been developed based on local weather characteristics. The longest prediction lead-time is six hours, and the grid interval is 2.5km, and the update interval of prediction is every 10 minutes. The size of the target area is 461km2.The system is composed of two types of prediction model. One is extrapolation model using radar data, and the other is meso-scale atmospheric model with data assimilation. The extrapolation model is based on linear extrapolation technique, but it is not simple translation method. To understand rainfall and meteorological characteristics of Kurobe-valley and around the region, we observed several rainfall events with a volume scanning Doppler radar and analyzed three-dimensional structure of rain echo and wind field. Through the observation, rainfall characteristics around Kurobe-valley were found and the characteristics have been considered in the extrapolation model. We also analyzed surface rainfall data and atmospheric GPV data (output of numerical weather prediction model of Japan Meteorological Agency). As a result, the relationships between cloud development and the characteristics of atmospheric data found by the analysis have been also considered in the extrapolation model. The meso-scale atmospheric model is modified using radar data and the results of extrapolation model. The calibrated radar data and the results of extrapolation model (until two hours ahead) are assimilated into the atmospheric model every 10 minutes in calculating process. In the assimilation process, predicted rainfall data of meso-scale atmospheric model and the calibrated radar data (or the results of extrapolation model) are compared, then water vapor mixing ratio in the atmospheric model is modified. The developed short-time rainfall prediction system provides more accurate rainfall prediction data until six hours ahead with the spatial resolution of 2.5km than other operationally available prediction services for the valley area.

79



MULTI-SENSOR MEASUREMENTS OF PRECIPITATION Ali Tokay Paul Bashor and K. R. Wolff NASA/Goddard Space Flight Center ABSTRACT: A unique surface precipitation measurement site has been operating at NASA Wallops Flight Facility since May 2002. The site hosts various types of disdrometers and rain gauges in addition to several Doppler radars and a Microwave Link. The site has recently incorporated into a newly developed ground validation site at the mid-Atlantic Coast as part of the NASA’s Tropical Rainfall Measuring Mission (TRMM). Efforts have been made to generate instantaneous and monthly radar rainfall maps utilizing WSR-88D at Wakefield, Virginia, and a rain gauge and disdrometer network. It is anticipated that Greater Wallops Island will serve as a long-term rainfall verification site for the NASA Precipitation Program that includes satellites equipped with microwave sensors. This study focuses on the small-scale of variability of radar reflectivity and rainfall within ground-based and space-borne radar pixel size. A set of three impact disdrometers has been operated over a year at Wallops Island. It is expected that the network of disdrometers will be expanded in August 2003. As of now, multiple units of the impact disdrometers have been collocated at the site to determine their calibration bias. The measurement uncertainty of the disdrometer and gauges will be incorporated to this study.

80


MERGING RADAR - RAINGAUGES DATA TO ESTIMATE RAINFALL FIELDS: AN IMPROVED GEOSTATISTICAL APPROACH USING NON

PARAMETRIC SPATIAL MODELS Carlos A. Velasco-Forero Eduardo F. Cassiraga, Daniel Sempere-Torres, Rafael Sánchez-Diezma, and J. Jaime Gómez-Hernández Grup de Recerca Aplicada en Hidrometeorologia. Universitat Politècnica de Catalunya. Barcelona (SPAIN) [email protected] ABSTRACT: Among the rainfall measuring sensors, raingauges and ground radars are probably the two most important in rainfall estimation. Therefore, the development of improved methodologies to estimate rainfall fields merging radar and raingauges data has been an objective from the initial studies of hydrological applications of meteorological radar. Thus, some previous works have reported that rainfall estimated by radar – raingauges cokriging improved flood estimates (e.g. Sun et. al., 2000). However, spatial and temporal high variability of rainfall fields introduces problems to select and use a valid spatial model (covariance or variogram) in the cokriging system. Additionally, real time hydrological models have very restrictive requirements about maximum computing time that usually cokriging does not fulfil. All these problems mean that cokriging has not been applied in real time applications. The aim of this work is to provide an improved geostatistical approach able to be applied operationally. First, to avoid the classical selection of theoretical covariance models, we obtained valid covariance tables at each time step using a nonparametric methodology based on FFT. Then, four geostatistical alternatives are analyzed to test its efficiency in terms of time computing. The selected approaches used are Ordinary Kriging (OK), Kriging with External Drift (KED), CoKriging (COK), and Collocated CoKriging (ColCOK). 10-minute radar scans of a 22-hour rainfall case study in Catalunya is used to test the performances of these methodologies. For this event, results show that rainfall field estimated by ColCOK and KED give the best results in a statistical and qualitative way, and that these methodologies can be used in a real time scheme application.

81



MULTI RADAR INTRANET SUPPORTED FLOOD DETECTION AND MONITORING SYSTEM FOR POLAND

André M. Weipert AMS-Gematronik, Germany [email protected] ABSTRACT: Due to severe flooding and disastrous natural phenomena in Poland between the years 1997 and 1998 the Institute of Meteorology and Water Management (IMGW Poland) initiated the modernization of the existing hydrological and meteorological management system. The project is called "Emergency Flood Recovery Project" mainly financed by the World Bank. This paper describes the current status and latest results of the integration and extension of the Polish weather radar network POLRAD consisting of 8 METEOR® Gematronik Doppler weather radars and the radar data processing software Rainbow®. In order to provide long-term rainfall forecast capabilities the integration includes additionally the adaptation to the Met Office (UK) rainfall forecast system Nimrod.

82



QUANTITATIVE PRECIPITATION NOWCASTING

James W. Wilson NCAR, Boulder CO USA [email protected] ABSTRACT: The 0-6 hr forecasting (nowcasting) of thunderstorms and convective precipitation has been historically based on the extrapolation of radar echoes or satellite cloud images. During the last few decades this has led to major improvements in severe storm warnings and to a lesser extent quantitative precipitation nowcasting (QPN). However, except for highly organized supercells or squall lines, the utility of extrapolation techniques decreases at a very rapid rate particularly during the first 60 min. Thus precision nowcasting of convective precipitation by extrapolation techniques is generally unreliable beyond 30-60 min because there is no accounting for storm initiation, growth and dissipation. Until recently there has been essentially no improvement in convective precipitation nowcasting beyond these simple extrapolation techniques. In the last decade some capability to anticipate storm initiation has been demonstrated by monitoring the location of boundary layer convergence lines, such as gust fronts, by radar and satellite. Physically based computerized nowcasting systems that assimilate knowledge of these convergence lines are now showing the ability to improve on extrapolation only techniques. It is believed that further improvement will require high resolution specification of the static stability and boundary layer winds; knowledge that is not routinely available. Recent exciting developments may be rectifying this situation. For example during the past four years high resolution boundary layer winds are being retrieved from single Doppler radar data. During the last year high resolution near surface water vapor measurements have been retrieved from radar data. A number of other very promising water vapor measurement techniques were also tested during the 2002 International H2O Project. In addition numerical modeling techniques which assimilate very high resolution data have shown promise. A sampling of new instrumentation and nowcasting techniques will be shown. Commentary on the future direction and accuracy of QPN will be presented. This will include the role of numerical models, automation, data fusion systems and the forecaster.

83



PROGRESS IN RADAR DATA QUALITY CONTROL AND ASSIMILATION Qin Xu Leilei Wang, Pengfei Zhang, Shun Liu, Li Bi and Qingyun Zhao, Keith Sashegyi, Jason E. Nachamkin ABSTRACT: Previously, a 3.5-dimensional variational (3.5dVar) package of radar wind analysis was developed and incorporated into the Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS). The 3.5dVar package contains three major components: (i) data quality control, (ii) 3D-wind analysis, and (iii) thermodynamic analysis. This paper reports progress in upgrading the first two components of the 3.5dVar. First, built upon the recent successes of the Collaborative Radar Acquisition Field Test (CRAFT) project, an automated analysis system is developed to detect and record data quality problems in level-II wind data received real-time from KTLX in Oklahoma and from eight radars in New England by since June 2002. Based on the detected data quality problems, new data quality control algorithms are developed for radar data assimilation. These algorithms are being used to upgrade the data quality control in 3.5dVar. Secondly, a new method of statistical analysis of innovation vector (discrete arrays of observation minus independent background values at observation points) is developed to estimate radar observation and background error covariances from the real-time level-II data. The 3D-wind analysis in 3.5dVar is being upgraded with multivariate scheme to use the estimated error covariances. The upgraded 3.5dVar will be tested with COAMPS on selected cases of severe storms occurred in central and eastern US during 7-10 May 2003. The results will be reported at the conference.

84


APPLICATION OF WEATHER RADAR OBSERVATION FOR FLOOD FORECAST AND CONTROL IN THE UPPER TONE RIVER OF JAPAN

Dawen Yang Toshio Koike Department of Civil Engineering, University of Tokyo, Bunkyo-ku, Tokyo 113-8656, Japan [email protected] ABSTRACT: Weather radar measurement is expected to improve the accuracy of flood forecasting together with a distributed hydrological model, and then provide a tool for flood control through dam operation in real time. This research includes a distributed hydrological model with reservoir operation to make a combined system for exploring efficiency of reservoir operation in flood control. Tone River basin usually suffers heavy typhoon rainfall in late summer. The reservoirs located in the upper Tone River play an important role on flood regulation. Application to the upper Tone River basin by simulating flood event in August 2001 shows potential applicability of radar rainfall for real-time flood control.

85



WEATHER RADAR CALIBRATION IN CENTRAL IRAN APPLYING GROUND BASED RAIN GAUGES DATA

A. A. Semsar Yazdi F. Amiryazdani, M. khalili, M. Kalantarzade, F. Golkar National Cloud Seeding Research Center of IRAN [email protected] ABSTRACT: The first weather radar was installed and put into practice in Iran with the callaboration of the Russian CAO institute in Jun. 1999. This coincided with the inanguration of weather modification project. The radar was situated in the mountain regions of central Iran. According to the anticipated programs, the second radar was installed 200 Km farther in 2001 which is a Russian radar (MRL-2) with 3.2 Cm Wave length in x band radar and pulse frequency of 400 Hz and ultimate range of 200 Km. Since precipitation enhancement is targeted through weather modification in Iran, evaluation and measurement of the rain rate as well as the increace or decrease of precipitation are of great importance in this project. To apply radars for more accurate estimation of precipitation rate within the region of their vision , the radar calibration project using assessment factor method has picked out which enjoys the data gained from 55 ground rain gauge stations and separate radar estimation data from radial distance of 50, 100, 150 and 200 Km.

86



A STUDY ON THE RAIN CHARACTERISTICS OF THE MESO-ΒSCALE USING DOPPLER RADAR.

Abe Yoichi Yamada Tadashi Chuo University [email protected] ABSTRACT: Concentrated heavy rainfalls cause serious damage to human life. However, forecasting of Meso-β scale rainfall using numerical models is still a difficult task due to accuracy and resolution of the models are not sufficient to recreate Meso-β scale rainfall phenomena, and due to the difficulty in obtaining detailed information for initial conditions of the numerical models. To forecast with accuracy it is important to understand the mechanisms of the rainfall phenomenon. Hence, observational data on the Meso-βscale rainfall phenomena is very important information. We have observed the characteristics of Meso-β scale rainfall using an x-band Doppler radar since 1995. The radar is located at the center in metropolis Tokyo and can observe almost the entire area of the Kanto Plain. Based on a large number of time-series of radar images, the Meso-βscale rainfall occurring in the Kanto Plain is classified into the following three types: 1) front formation-type, 2) cells advection-type and 3) isolated cell-type. Occurrence of these three types is mainly decided by surface wind and upper-level winds, and the time of rainfall continuation differs for each rainfall type. In this research, we investigated the initiation point of rainfall, and the initiation point of strong rainfall (radar reflectivity factor is over 32 dBZ), the speed of precipitation movement, and duration of rainfall. Especially we analyzed the spatial and temporal characteristics of radar accumulated precipitation and instantaneous strong intensity of rainfall observed by the radar. This precipitation is compared with the AMeDAS (Automated Meteorological Data Acquisition System) data of the Japan Meteorological Agency for quantitative comparison. Moreover, we examined the possibility of the forecasting each classified rainfall type. It was found that the rainfall amount is greater in mountain regions compared with surrounding plains, but regions with strong rainfall frequency is much more frequent in plains than on mountains in Meso-βscale rainfall of summer season. Further, it became clear that these local characteristics differ for each rainfall type. Furthermore, front formation-category can be further classified into three types depending on rainfall progression patterns, and forecast of the location of precipitation is possible to some extent based on surface wind direction.

87



THE STUDY ON KOREAN HYDROLOGICAL RADAR NETWORK DESIGN FOR FLOOD WARNING AND MONITORING SYSTEM

Dr. Kanghoon Yoon Dr. Taegyun Kim Korean Institute of Construction Technology Korea Ministry of Construction and Transportation ABSTRACT: Research Purposes

• To develop new flood hydrological monitoring system including hydrological radar networks

• To study the feasibility of Korea hydrological radar networks considering and satisfying hydrological, meteorological, geophysical, and flood management aspects

• To investigate the radar rainfall estimation methodologies including areal rainfall analysis and flood rainfall forecasting using radar rainfall

• To study the hardware and software linkage of the proposed radar network to recent flood warning system and suggest prospective system in future

Future Research

• Propose optimal hydrological radar network in Korea • Expand the study of radar rainfall estimation and forecasting methodology • Propose appropriate radar-based flood monitoring and forecasting system in Korea • Suggest future operational strategy for hydrological radars

88


INTEGRATION OF WEATHER RADAR DATA INTO A RASTER GIS

FRAMEWORK FOR IMPROVED FLOOD ESTIMATION B. Yu A. Seed, L. Pu and T. Malone Griffith University, Nathan Qld 4111, Australia ABSTRACT: Rainfall shows high variability in space and time. Weather radars have the potential to capture this variability at an acceptable cost. Direct use of weather radar data for flood estimation and forecasting, however, has not been tested in Australia. Traditional approach to runoff routing for flood estimation and forecasting relies on an arbitrary sub-division of the catchment. URBS, a runoff routing model using this approach, has been widely used for flood estimation and forecasting. In URBS and other similar runoff-routing models, the spatial variation in rainfall is represented at the sub-catchment scales. Typical resolution of rainfall totals for a flood event is of the order of 100 km2 per sub-catchment for medium to large catchments, which is broadly compatible with the density of rain gauges in the catchment for flood estimation. With radar-based rainfall data at 1 km resolution, the question is whether this significant improvement in the spatial resolution of rainfall can lead to an improvement of the quality of estimated hydrographs and forecast skills. This paper presents a trial study of flood estimation using the radar data for a medium-sizes catchment in southeast Queensland. Mary River at Bellbird Creek has a catchment of 480 km2 with a long-term mean annual runoff depth of 360 mm. One of the largest floods on record occurred on 9 February 1999. A 1 km raster-based GIS framework was developed for flood estimation. DEM (digital elevation model) was used to generate flow directions and accumulations, and flood routing sequences. A spatially variable infiltration model indexed on vegetation types was used to produce rainfall excess for routing purposes. Flow routing at the hill slope scale and along the drainage network was considered. Radar-based rain field at the same spatial resolution of 1 km is dynamically loaded at 10-min intervals to estimate hydrographs at the catchment outlet as well as for each of the 480 cells. Simulated hydrograph was compared with the observed hydrograph at the site and with that generated with URBS.

89


THE ERROR STRUCTURE OF RADAR PRECIPITATION MAPS Isztar Zawadzki J. S. Marshall Radar Observatory, McGill University ABSTRACT: 1-Radar gives information on precipitation at scales of a few kilometers, every five to ten minutes and over an area of a radius of hundred to two hundred kilometers. 2-Radar is a semi-quantitative instrument. The two above statements were made repeatedly in the past. Which of the two is true? In fact, stated in this manner the problem is ambiguous and can be misleading. Since radar does not measure precipitation, but rather the backscattered power from whatever targets the radar beam encounters, it is clear that precipitation is a derived quantity and consequently is affected by a number of sources of uncertainty discussed by the author in the past. Thus, perhaps the more appropriate formulation of the question pertaining radar as a hydrological tool is: What is the error or radar derived precipitation amount for a given time-space resolution and area coverage? Even posed in this manner the question is difficult to answer. Although gages are considered the ultimate validation for radar precipitation the difference in scales sampled by the two instruments coupled with the variability of precipitation at all scales makes the radar gage comparison a science in itself. Perhaps the best first step is to estimate the errors in QPE from each source independently. In this presentation a summary of a quantitative evaluation of errors in radar derived precipitation fields will be given. A series of other papers in the conference will describe the details of the work.

90


ASSIMILATING RADAR-OBSERVED WINDS INTO A HIGH-RESOLUTION NUMERICAL WEATHER ANALYSIS AND PREDICTION SYSTEM AT NRL

Qingyun Zhao John Cook Qin Xu

Paul Harasti Mike Frost

Naval Research Laboratory, Monterey, California, USA [email protected] ABSTRACT: A high-resolution radar data assimilation system is under development at the US Naval Research Laboratory for assimilating radar-observed winds into the Navy’s Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPSTM). The data assimilation system uses a variational approach to retrieve three-dimensional wind and cloud information from radial-velocity and reflectivity fields observed by multiple radars over the analysis domain. Extensive tests of the system have been conducted with NEXRAD level-II data and the results showed positive impacts of radar observations on wind and cloud analyses. The products from this system include the mesoscale dynamical and hydrological fields required for initializing the high-resolution COAMPS model. These products will be cycled into the high-resolution COAMPS model to improve the model capability and accuracy in predicting storms and other various types of hazardous weather. COAMPSTM is a registered trade mark of Naval Research Laboratory

91



METEOROLOGICAL ANALYSIS OF THE HAIL EVENT ON FEB. 5, 2003 IN

NORTHERN KYUSHU, JAPAN Kenji Wakimizu Taichi Maki, Misato Nagata and Koji Nishiyama Faculty of Agriculture, Kyushu University, Hakozaki, Higashi-ku, Fukuoka 812-8581, Japan. [email protected] ABSTRACT: A hail storm occurred in Northern Kyushu from 2130JST to 2200JST on February 5, 2003. The behavior of hail precipitation could be observed by Kyushu University (KU) radar. In this day, the cold air mass broke into Northern Kyushu from northwest to southeast. The KU radar echo (PPI) was like the shape of tongue. The radar echo had the width of about 10 km and the length of about 30km. The diameter of hailstone was about 1.0 to 1.5 cm. The depth of fallen hail was about 3.0 to 5.0 cm. On this day, the damage of crops was less than expected from the strength and the scale of hail precipitation. One of the reasons is that it occurred in the off-season of cropping. In this study, meteorological factors on this hail event and past hail events were analyzed, paying attention to the forecast of hail precipitation, because the strength and the scale were so large.

92



93

ERROR STATISTICS OF VPR CORRECTIONS IN STRATIFORM PRECIPITATION

Isztar Zawadzki and Aldo Bellon J. S Marshall Weather Radar Observatory, Dept. of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada [email protected] ABSTRACT: Errors in surface rainfall estimates caused by ignoring the vertical profile of reflectivity, (VPR), have been assessed by simulating how fine resolution 3-D reflectivity measurements at close ranges are sampled by the radar at various ranges and heights. This approach has the advantage of isolating errors due solely to the VPR while ignoring all other sources of errors that would be present when comparing radar measurements with gauges. About 200 hours of stratiform precipitation distributed among 21 events and with a distinct melting layer have been used to derive uncorrected and corrected 1-hour accumulations using various procedures. The latter include (a): “local” VPR obtained by a space-time averaging procedure over 30 minutes, (b): intensity dependent “climatological” VPR derived from the entire data set and (c) “event” VPR. The rainfall estimates, 0.2 km apart in height up to 5km and 40 km apart in range up to 210 km, are compared with the “ground truth” of the original near-range data at the lowest height of 1 km. The RMS error structure has thus been derived as a function of height and range and for verification areas ranging from (2x2) km2 to (20x20) km2. However, it is the errors at 1.5 km in height up to 90 km and along the height of the lowest elevation angle afterwards that are most relevant since this corresponds to the height of rainfall estimates of the Canadian radars. The stratification of the results in terms of the height of the bright band is essential in order to understand the influence of the bright band with range. The largest errors (> 100% at near ranges without correction) are encountered with lower bright bands that are both stronger and occurring at heights potentially used for surface estimates. The “local” VPR correction reduces these errors to 30 or 40% at 2 km resolution and by an additional 10% at 10 km resolution. This result is considered as a “best-case” scenario achievable only under conditions of homogeneity of the bright band at all ranges. Surprisingly, the “climatological” and “event” based VPR performed equally well, with errors about 10% higher than those from the “local” VPR. Errors for accumulations ranging from 5 minutes two 2 hours are also obtained.



94

DISTRIBUTED COLLABORATIVE ADAPTIVE SENSING , AN END TO END ENGINEERED SYSTEM FOR FLOOD MONITORING AND FORECASTING

V.Chandrasekar (assisted by the other collaborators in the large scale project ). ABSTRACT: Recently the National Science Foundation has Established an Engineering research Center, a collaborative effort between Colorado State University, University of Massachusetts, University of Oklahoma, and University of Puerto Rico, with partnership between government agencies such as NOAA and a consortium of private industries, to develop the technology and science for a " Distributed collaborative adaptive sensor network" to revolutionize the way we sample the atmosphere where the needs are the greatest. As a part of this a flood forecast research test bed is being developed in Houston , to implement the DCAS paradigm in precipitation sensing where the engineered system will have the capability to be driven according to the needs of the end user. This paper will present the description of the program along with the scientific and operational plan for the DCAS paradigm.

the effects of raingage network density on weather … · 6th international symposium on...

Documents