Use of Profilers in TRMM Ground Use of Profilers in TRMM Ground ValidationValidation

Kenneth GageKenneth Gage11, and Christopher Williams, and Christopher Williams1,21,2

11NOAA Aeronomy Laboratory, Tropical Dynamics and Climate GroupNOAA Aeronomy Laboratory, Tropical Dynamics and Climate Group

22University of Colorado, Cooperative Institute for Research in University of Colorado, Cooperative Institute for Research in Environmental Sciences (CIRES)Environmental Sciences (CIRES)

OutlineOutline Introduction Profiler as a Reflectivity Standard Profiler determination of vertical

reflectivity gradient structure in stratiform and convective precipitation

Profiler retrieval of DSD and precipitation parameters as an alternative to Z-R.

NASA-TRMM Ground Validation ProgramNASA-TRMM Ground Validation Program

The Aeronomy Lab was involved with these field campaigns:The Aeronomy Lab was involved with these field campaigns:Texas (April 1998) Texas (April 1998) TEFLUN-ATEFLUN-AFlorida (July-Sept 1998)Florida (July-Sept 1998) TEFLUN-BTEFLUN-BBrazil (Jan-Feb 1999)Brazil (Jan-Feb 1999) TRMM-LBATRMM-LBAKwajalein (July 1999-Present)Kwajalein (July 1999-Present) KWAJEXKWAJEXDarwin (1997-Present)Darwin (1997-Present)

TEFLUN-B, Triple-N-Ranch, Florida TEFLUN-B, Triple-N-Ranch, Florida July – September 1998July – September 1998

Joss WaldvogelDisdrometer

2-Dimensional VideoDisdrometer

2835 MHz Profiler

915 MHz Profiler

ReflectivityReflectivity(amplitude)(amplitude)

Doppler VelocityDoppler Velocity(air motion +(air motion +particle particle motion)motion)

Spectral WidthSpectral Width(turbulence (turbulence and particle and particle size variation)size variation)== 2*sqrt(var)== 2*sqrt(var)

How Can Profilers Help with How Can Profilers Help with Ground Validation?Ground Validation?

Profilers in combination with disdrometers rain gages and polarimetric scanning radar can be used to improve the quality of QPE on the ground.

Profilers in combination with other instruments can be used to provide physical validation to evaluate and improve algorithm performance.

Profilers can provide information on the variability of Precipitation parameters on a scale which lies within the pixel size of scanning radars and satellite footprints (relevant to non-uniform beam filling and background error covariance for assimilation into models).

Profilers Provide a Profilers Provide a Reflectivity Standard for otherReflectivity Standard for other

Ground-Based Instruments Ground-Based Instruments

Reflectivity Time Series in Stratiform Reflectivity Time Series in Stratiform Rain from Collocated Profilers Rain from Collocated Profilers

during TRMM LBAduring TRMM LBA

Precision StatisticsPrecision Statistics

of Two Collocatedof Two Collocated

Profilers in CentralProfilers in Central

Florida in 1998 Florida in 1998

During TEFLUN BDuring TEFLUN B

Plotted are:

a) Pdf of reflectivity difference

b) Bias and 95% confidence as a function of profiler reflectivity

c) Pdf of profiler reflectivity

Profiler and WSR-Profiler and WSR-88D Reflectivity88D ReflectivityComparison duringComparison duringTEFLUN BTEFLUN B

Top Panel:915 MHz reflectivity 21 August, 1998

Second Panel:915 MHz reflectivity degraded to 630 m & 5 minute resolution.

Third Panel:WSR-88D reflectivity over the profiler.

Bottom Panel:Profiler reflectivity variability at WSR-88D Resolution

Reflectivity Difference for all Rain Reflectivity Difference for all Rain Events (August-September 1998)Events (August-September 1998)

PDF(ZProfiler)

PDF(ΔZ) Mean +/- 95% C.I. in 2 dBZ intervals

2,062 Obs

Vertical Reflectivity GradientsVertical Reflectivity Gradients

Vertical Structure of Reflectivity Vertical Structure of Reflectivity at Legan in stratiform and at Legan in stratiform and convective precipitationconvective precipitation

Reflectivity Dependence of Vertical Structure Reflectivity Dependence of Vertical Structure of Reflectivity at TEFLUN B and LBAof Reflectivity at TEFLUN B and LBA

Top Panel: Scatter Plot of 2835 MHz profiler reflectivities at 202 m compared to calculated JWD disdrometer reflectivities excluding JWD samples with less than 64 drops.

Bottom Panel: Reflectivity differences between profiler and disdrometer as a function of disdrometer reflectivity.

Calibration of TRMM-Calibration of TRMM-LBA profiler, Ji-Pirana, LBA profiler, Ji-Pirana,

Brazil; January - Brazil; January - February 1999February 1999

Reflectivity Dependent Bias observed Reflectivity Dependent Bias observed at Wallops by Heightat Wallops by Height

An Ensemble ApproachAn Ensemble Approachto DSD Retrieval using to DSD Retrieval using

ProfilersProfilers

DSD Model DescriptionsDSD Model DescriptionsThe DSD retrieval models assume different functional forms of the DSD and are divided into two numerical classes.

AA GammaGamma N(D)=NN(D)=NooDDµµexp[(4+µ)D/Dexp[(4+µ)D/Dmm]] Ulbrich 1983Ulbrich 1983

BB Normalized Normalized GammaGamma

N(D)=NN(D)=Noo* * FFµµ(D/D(D/Dmm)) Testud et al. 2001Testud et al. 2001

CC Log-NormalLog-Normal N(D)=NN(D)=Nttexp[-lnexp[-ln22(D/D(D/Dmm))(2ln(2ln22σσ)])]

Feingold & Levin Feingold & Levin 19861986

DD Discrete DSDDiscrete DSD no mathematical functionno mathematical function

Note-For methods A,B, and C, the cost function was minimized Note-For methods A,B, and C, the cost function was minimized in Din Dkk space with k = 3, 4, 5, and 6 to produce 4 different space with k = 3, 4, 5, and 6 to produce 4 different solutions.solutions.

EE ExponentialExponential N(D)=NN(D)=Nooexp[-(4)D/Dexp[-(4)D/Dmm]] µ=0µ=0

FF GammaGamma N(D)=NN(D)=NooDD2.52.5exp[-(6.5)D/Dexp[-(6.5)D/Dmm]] µ=2.5µ=2.5

GG GammaGamma N(D)=NN(D)=NooDDµµexp[-(4+µ)D/Dexp[-(4+µ)D/Dmm]] µ=variableµ=variable

Convolution Techniques

Deconvolution Techniques

DDmm & R Statistics & R Statistics

at 3.2 kmat 3.2 km

The top panel shows the 920-MHz profiler reflectivity for the precipitation event that passed over the Darwin, Australia, profiler site on 16 February 2003.

Panel (b) shows the reflectivity at 3.2 km.

The mean and standard deviation of the ensemble Dm and R are shown for the 3.2 km altitude in panels (c) and (d).

Statistics Relative to the Ensemble MeanStatistics Relative to the Ensemble Mean

The mean mass-weighted diameter, Dm, is estimated for each DSD model and the ensemble mean is estimated for each resolution volume (each time-height sample).

The statistics relative to the ensemble mean are shown here.

)()( mmm DmeanDModelD −=Δ

Summary of Profiler Contributions to Summary of Profiler Contributions to Ground-Based QPEGround-Based QPE

Reference reflectivity in the column over the profiler

Vertical structure of the precipitation parameters above the surface

Direct estimates of precipitation parameters based on profiler retrieved drop-size distributions

GPM Front-Range Reflectivity Comparison: GPM Front-Range Reflectivity Comparison:

16-17 June 200416-17 June 2004

Calibrating Vertically Pointing ProfilersCalibrating Vertically Pointing Profilersusing a Surface disdrometerusing a Surface disdrometer

Joss-Waldvogel Disdrometer Joss-Waldvogel Disdrometer manufactured by Distromet.manufactured by Distromet.

Detects and calculates the Detects and calculates the ‘momentum’ of individual ‘momentum’ of individual rain drops hitting the head.rain drops hitting the head.

Converts the momentum of Converts the momentum of each drop to a raindrop size.each drop to a raindrop size.

Yields the number and size of Yields the number and size of raindrops hitting the 50 cmraindrops hitting the 50 cm22 head.head.

Aeronomy Lab developed Aeronomy Lab developed software to record the full software to record the full diameter resolution (127 diameter resolution (127 bins), [Manufacture’s bins), [Manufacture’s software provides 20 bin software provides 20 bin output].output].

400 MHz Doppler Spectrum400 MHz Doppler Spectrum

Source: Greg Forbes

Ground Validation for TRMM Ground Validation for TRMM and GPMand GPM

TRMM Ground Validation utilized rain maps produced by scanning radar at various locations around the globe. However it was found that the uncertainty in the ground validation rain products was large and the ability to validate satellite algorithms was limited.

GPM ground validation will place more emphasis on the physical validation of algorithms in continental and oceanic ‘supersites’. Observations will be made in a multidimensional observing volume to support the use of a Satellite Simulator Model.

Reflectivity Time Series from Reflectivity Time Series from Collocated Disdrometers Collocated Disdrometers

at Wallops Is.at Wallops Is.

Source: Ali Tokay

Use of Profilers in Ground Use of Profilers in Ground Validation for TRMMValidation for TRMM

The introduction of profilers in TRMM ground validation efforts was accomplished as a research effort and was not part of the originally planned Ground Validation activities for TRMM.

Factors Contributing to Poor QPEFactors Contributing to Poor QPE

Calibration of scanning radarCalibration of scanning radar

Reflectivity gradients in the lower Reflectivity gradients in the lower atmosphereatmosphere

Use of unrepresentative Z-R relationsUse of unrepresentative Z-R relations

Use of Disdrometers and Use of Disdrometers and Profilers to Calibrate Profilers to Calibrate

Scanning RadarsScanning Radars

Disdrometer

Profiler Reflectivities vs. Disdrometer Profiler Reflectivities vs. Disdrometer Reflectivities from TRMM LBAReflectivities from TRMM LBA

Approach to Ground Approach to Ground ValidationValidation

TRMM GV emphasized the statistical validation with surface rainfall maps at a number of locations but these products were not of sufficient accuracy to be very useful and largely ignored the physical validation that would be needed to validate algorithm assumptions and performance.

Summary of TRMM Summary of TRMM EXPERIENCEEXPERIENCE

Based on the experience of their use in TRMM ground validation field campaigns, profilers are becoming well established tools for use in multi-sensor field campaigns. A data archive of reflectivity, Doppler velocity, and spectral width was produced for each field campaign. A calibration procedure was developed utilizing co-located disdrometers and substantial progress was made toward routine retrieval of drop-size distributions from profiler measurements.

Drop Size Distributions from Drop Size Distributions from Profiler ObservationsProfiler Observations

Two profilers operating at different Two profilers operating at different frequencies to distinguish Bragg and frequencies to distinguish Bragg and Rayleigh scatterRayleigh scatter

Single profiler with Bragg and Rayleigh Single profiler with Bragg and Rayleigh components of the Doppler spectrum presentcomponents of the Doppler spectrum present

Single UHF profiler without Bragg Single UHF profiler without Bragg component: Sans Air Motion (SAM)component: Sans Air Motion (SAM)

Key is to separate air motions from hydrometeor fall velocities.

The DSD is the building block used to The DSD is the building block used to

determine precipitation parametersdetermine precipitation parameters

Examples of Precipitation Parameters:

Reflectivity = ∫ N(D) D6 dD

Liquid Water Content ~ ∫ N(D) D3 dD

Rain Rate ~ ∫v(D) N(D) D3 dD

Mean Mass-weighted Diameter, Dm

Dm = ∫ N(D) D4 dD /∫ N(D) D3 dD

Dm = (4 + µ) /

D6 Dependence – Radars are sensitive to this dimension

D3 Dependence

Dependent only on the shape of the DSD.Independent of No