international association for mathematical...
TRANSCRIPT
![Page 1: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/1.jpg)
Amilcar Soares
IAMGInternational Association for Mathematical
Geosciences
UQAM – 10 of November 2011
The use of hybrid models to integrate the main dynamic
characteristics of the physical phenomena
![Page 2: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/2.jpg)
Maximum Likelihood Classification of
Pine Forest Image ( SPOT 4)Co-located Cokriging of Pine ForestMaximum Likelihood Classification of
Pine Forest Image ( SPOT 4)Co-located Cokriging of Pine Forest
Integration of Different Data andInformation
![Page 3: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/3.jpg)
Integration of Different Data and Information to characterize geological phenomena
Geology Geophysics Well Data
Production Data
![Page 4: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/4.jpg)
“A good model is the one that starts to be not bad and , at the end, gives good results...”
![Page 5: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/5.jpg)
Deterministic (dispersion) models mime the main dynamic
characteristics of the physical phenomena
Motivation: integrate deterministic dispersion models in
stochastic charaterization of the resource by adding the
(predictive) temporal dimension, obtaining high resolution
images of the main characteritics and the uncertainty attached
Stochastic models (geostatistics) characterize the static
components of the process (polutant concentration, petrophysic
properties, etc..)
![Page 6: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/6.jpg)
Motivation: integrate deterministic dispersion
models in stochastic charaterization of natural
resources by adding the (predictive) temporal
dimension, obtaining high resolution images of the
main characteristics and the uncertainty attached
![Page 7: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/7.jpg)
In hydrogeology and petroleum applications deterministic
models, that mime the dynamic of physical phenomena,
are integrated a posteriori, through inverse models.
The idea is to calibrate/update the aquifer or reservoir
parameters – internal properties, petrophysical properties,
.. – with the dynamic data.
Hydrogeology and Petroleum applications
![Page 8: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/8.jpg)
Stochastic Simulation of
petrophysical parameters
Obtain responses from dynamic
simulator
Perturbation/Optimization/..
of initial parameters until
the desired match is
obtained
![Page 9: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/9.jpg)
Air Quality Modelling
The most of deterministic dispersion models applied for air quality
modelling faces, in general, two main limitations : the spatial
scale/resolution and the calibration with experimental data values
![Page 10: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/10.jpg)
Stochastic models (Geostatistics) produce high resolution models,
honour the experimental data but are unable to predict in time domain
Stochastic Simulation of SO2 in space-time domain
![Page 11: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/11.jpg)
Example #1 – Case study of a Costal lagoon with contaminated
sediments
Objective:
i- The use of hybrid models to integrate the main dynamic
characteristics of the phenomenon
ii- Integration of uncertainty of different conceptual models of
sediment depositional dynamics at the early stages of risk
analysis of a contaminated site
![Page 12: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/12.jpg)
Case study: Barrinha de Esmoriz, coastal lagoon
with contaminated sediments
1979
2006
1996
![Page 13: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/13.jpg)
Available data: two sampling campaigns
2001– 30 samples 2008 – 69 samples
![Page 14: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/14.jpg)
Conceptual depositional model: contaminated
sediments are driven by the fluid flow
•Morphology of meanders structure are extracted by EO –Earth Oservation data.
•Main flow characteristics (direction, velocity) are taken by a dynamic simulator
![Page 15: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/15.jpg)
Dss with local anisotropies
i) Define the random path over the entire grid of nodes xu, u=1,
Ns, to be simulated
ii) Estimate the local mean (simple kriging estimate z(xu)*) and
variance (estimation variance 2sk(xu)) conditioned to local
models of covariance anisotropy.
iii) Define the interval of Fz(z) interval to be sampled (defined by
the local mean and variance of z(x))
iv) Draw the value zs(x0) from the cdf Fz(z)
v) Loop until all Ns nodes have been visited and simulated
Proposed Methodology: Stochastic simulation of a continuous
variable with local anisotropies (Horta et al, 2009)
Continuous variable Z(x) with a distributions function Fz(z) = prob {Z(x) ≤ z}
![Page 16: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/16.jpg)
Idea: to convert the output of dynamic simulator – directions
and velocities of flow – in local anisotropy ratios and main
directions
![Page 17: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/17.jpg)
z(x0)*
local models of Cθ(h) are defined in a regular
grid node
Simple kriging with local
models of Cθ(h)
The main directions and local anisotropy ratios, are accounted in the simple
kriging system to estimate local means and variances at a given location x0,
xxcxz ,0
*
0
![Page 18: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/18.jpg)
Simple kriging system
Matrix of data covariances Vector of data-to-unknown covariances
sk(u) = Ksk-1 . ksk
Vector of simple kriging weights
![Page 19: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/19.jpg)
Data search assuming geometric anisotropy:
Pattern of spatial variability given by the
sample variogram
Elliptical diagram of ranges:
- Direction of maximum continuity (azimuth θ)
- Range of maximum continuity (major axis aθ)
-
a
aRatioAnisotropy
hy
hx
h
hФ
a
aФ
g(hФ)
g(h)
haaФ
![Page 20: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/20.jpg)
DSS
a DSSx0= a DSSxi
aФ DSSx0= aФDSSxi
Ratio DSSx0=Ratio DSSxi
DSSx0= DSSxi
1
56
2
4
x0
xi
3
![Page 21: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/21.jpg)
DSS with local anisotropy
1
56
2
4
xi
x0
3
a DSSx0= a DSSxi
aФ DSSx0≠aФDSSxi
Ratio DSSx0 ≠ Ratio DSSxi
DSSx0≠ DSSxi
![Page 22: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/22.jpg)
Depositional models of Barrinha contaminated sediments
generated by a fluid flow simulator
Winter with tide Winter without tide
![Page 23: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/23.jpg)
summer Winter (no tide)
Winter (with tide)
![Page 24: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/24.jpg)
Results
Input MOHID, Cu
Simulated image
Average
Spatial
uncertainty
![Page 25: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/25.jpg)
Cu – Scenario A
Average images
a=350m
a=500m
![Page 26: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/26.jpg)
Cu – Scenario B
a=350m
a=500m
Average images
![Page 27: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/27.jpg)
Example #2 - Downscaling of air quality dispersion
models: an hybrid multi-scale model
Objective:
. Hybrid multi-scale approach as a methodology for predicting air
quality
• Downscaling and calibration (with experimental point data) of
images obtained by deterministic dispersion models
![Page 28: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/28.jpg)
Air quality deterministic dispersion models
Advantages Disadvantages
• Incorporate meteorological
factors
• Incorporate atmospheric
photochemical reactions
• It is able to predict scenarios
• A large number of input
variables
• Difficult to calibrate for fine
grid scales
![Page 29: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/29.jpg)
Geostatistical stochastic simulation
• Prediction of pollutant dispersion in space and
time (past) at fine scales reproducing the mean
variability of data
• The ability of predicting pollutant dispersion in
time is limited: the models do not directly
incorporate meteorological data
![Page 30: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/30.jpg)
Uncertainty and Support
results from the air pollution dispersion models can
be assumed as real values of atmospheric pollutant
concentration in support v , but with associated
uncertainty
![Page 31: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/31.jpg)
Rationale – Hybrid model (proposal)
Input data:
•Meteorological
variables
•Emissions
Air quality
dispersion
model
Block data
1st Step – transformation of i input hard
data in a image of air quality
Air Quality Dispersion Model transforms the real meteorological
data (wind directions, pressure,.., emissions) in a coarse image of
a pollutant concentration – the Block data
This transformation implies an error that must be accounted in
the simulation process
![Page 32: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/32.jpg)
Rationale – Hybrid model (proposal)
Block data
Point data
• pollutant
atmospheric
concentration
measurements
Geostatistical
stochastic
simulation
2nd Step
![Page 33: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/33.jpg)
Block Sequential Simulation (BSSIM)
Simulation of point values conditionedto block and point data
Objective: to characterize high resolution images ofpollutant concentration based on block and point
data
![Page 34: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/34.jpg)
Block Direct Simulation
i. In the node x0 of a random path of a regular grid, thefollowing means and variances are calculated by block co-
kriging:
ii. Simulation of a “point” value zs(x0) by re-sampling the global cdf global of Z(x) (Direct Sequential Simulation Approach).
xzxxzxxz v.. 00
*
0
xz
xzv
xzxxzxxz v.. 00
*
0
Note that the kriging system
requires the knowledge of spatial
convariaces point-point , point-
block and block-block : C(.,.) ,
C(.,v) e C(v,v)
![Page 35: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/35.jpg)
Block Kriging
block data is a linear
average of point values
Data taken at different scales, both on block-support and
on point-support are considered in the kriging system
v
uuv
v dZLB1
![Page 36: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/36.jpg)
Block Kriging
mDmZn
t
SK
uuDuu
1
*
BP
t - Kriging weights
- data value vector
- specific datum at location u
- number of data
PBD t
uD
un
0
0
BP
PP
BB
T
PB
PBPP
C
Ck
CC
CCKkK
Kriging system:
![Page 37: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/37.jpg)
)()()( vvv RBDB
Error covariance
vv
vvvvv
0
,
2
if
ifRRCovC
R
R
Data DB is equal to the “true” value B plus an error R
Block-data error incorporation
Assuming that block errors are homoscedastic and not cross-
correlated, with zero mean and known variance, then:
![Page 38: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/38.jpg)
If errors are independent of the signal,
uncorrelated and with known variance
vvvv
vvv
,
)0( 2
ifC
ifCC
B
RB
BB
Block-data error incorporation
![Page 39: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/39.jpg)
Block-data error incorporation
)()()( vvv BDRB
Error is the diference between the data D and
the “true” value B :
The error variance can be obtained by a priori calibration
Choosing the blocks with point data
![Page 40: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/40.jpg)
)()()( vvv l
B BDR
The “true” value can be replaced by the conditional
simulated block data Bl
)( vlB
Conditioning point data:
observation from a
monitoring station
Conditional simulated
block data
Choosing the blocks with point data
![Page 41: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/41.jpg)
The variance of block error can beparametrized for different meteorological
conditions
t sN
i
i
l
i
N
lst
R tBtDNN 1
2
1
2 ,,.
1vvv
The variance of block error is obtained with the Ns realizations
of Bl and the Nt time periods.
Block-data error incorporation
![Page 42: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/42.jpg)
Block sequential simulation (BSSIM)
1. Define a random path visiting each node u of the grid. For
each location u along the path:
2. Search conditioning data (point data, block data, previously
simulated values)
3. Compute the local block-to-block, block-to-point, point-to-
block and point-to-point covariances
4. Calculate local mean and variance by solving the mixed
scale kringing system
5. Draw a value from the global cdf and add the simulated
value to data set
6. Repeat (2-6) to generate another simulated realization
![Page 43: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/43.jpg)
Air quality characterization at industrial area of Sines
![Page 44: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/44.jpg)
![Page 45: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/45.jpg)
Sines Industrial Area
TROIA’92
![Page 46: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/46.jpg)
Emissions and Meteorological Conditions
![Page 47: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/47.jpg)
Sampling the Air Quality
![Page 48: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/48.jpg)
Case study
49
• SO2 daily average
atmospheric concentrations
measurements (3 stations)
• SO2 daily average emissions
(5 stacks)
• Meteorological hourly data
• ICST3 (EPA)
• 24/07/2006 a 28/07/2006
![Page 49: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/49.jpg)
Difusive Tubes Campaign
E.D. A nalysis ,Spatial variograms, calibration of
Guassian Plume model
![Page 50: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/50.jpg)
Gaussian Plume model : Block data
51
Number of blocks: 280
Block Size: 2500 m x 2500 m
Block Discretization: 3x3
24/07/2006 25/07/2006 26/07/2006
27/07/2006 28/07/2006
![Page 51: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/51.jpg)
Results
54
24/07/2006
25/07/2006
Sim#1 Sim#2 E-type
Sim#1 Sim#2 E-type
![Page 52: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/52.jpg)
Results
26/07/2006
27/07/2006
Sim#1 Sim#2 E-type
Sim#1 Sim#2 E-type
![Page 53: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/53.jpg)
Block error effect
57
Error variance= 0.05
Error variance= 0.5 Error variance= 0.9
Block data
![Page 54: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/54.jpg)
Block Error effect
58
Error variance= 0.05
Error variance= 0.5 Error variance= 0.9
Block data
![Page 55: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/55.jpg)
Final remarks
• This hybrid multi-scale approach is a valuable methodology
for predicting air quality
• Downscaling and calibration (with experimental point data) of
maps obtained by deterministic dispersion models is achieved
by the proposed method
![Page 56: International Association for Mathematical …cosmo.mcgill.ca/wp-content/uploads/2015/06/Almicar...Amilcar Soares asoares@ist.utl.pt IAMG International Association for Mathematical](https://reader034.vdocuments.us/reader034/viewer/2022042321/5f0b7bed7e708231d430beef/html5/thumbnails/56.jpg)
Thank you