Ground Validation of Satellite Precipitation Estimates over Spain

1
Ground Validation of Satellite Precipitation
Estimates over Spain

• Francisco J. Tapiador
• Institute of Environmental Sciences (ICAM)
• University of Castilla-La Mancha, UCLM
• Toledo, Spain
• francisco.tapiador_at_uclm.es
• With inputs from Antonio Rodriguez and Miguel
  A.Martínez, Spanish Nal. Meteorological Institute
  (INM), Madrid, Spain

2
Intro
Val example 1
Val example 2
Conclusions
Algorithms
Val example 3

• Introduction
• A. The UCLMs Environmental Modeling Group
• GCM and NWP The PROMES model
• Remote Sensing Satellite Precipitation
• Algorithm development
• Some Validation
• B. Some examples of our validation work over
  Spain
  
• Andalusia case study (METEOSATSSM/I)
• IPWG satellite estimates over Spain (CICS,
  University of Maryland data)
  
• EUMETSAT Convective Rain Rate product (INM,
  Spain)
  
• C. Some notes on Spain as validation site

3
Intro
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions

• Merged Satellite Rainfall Algorithms
• - EURAINSAT/A algorithm (Tapiador et al.
  2004, IJRS)
  
• - PMW-calibrated IR
• Neural Networks (Tapiador et al. 2004, Met App)
• PMWIR IR spatial and temporal resolution PMW
  directness
  
• 4km/30 minutes resolution
• Used by some farmers for irrigation planning
  advised on shortcomings and limitations
  
• Cloud motion winds PMWIR estimate

4
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
SSM/I only
Neural Networks Product
Neural Net (MeteosatSSM/I)
Histogram Matching (MeteosatSSM/I)
5
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro

• Cloud Motion Winds (CMW) Scheme
• Similar to CPC Morphing
• Difference CMW are directly modeled using
  Navier-Stokes equations instead of spatial
  correlation windows more physically-direct and
  more realistic fields
• Reference Tapiador, 2004. 2nd IPWG meeting,
  Monterey, CA
  
• Used for data assimilation into GCM

6
0230
0300
0330
0430
0500
0530
ACTUAL RAIN MEASUREMENT
RAIN ESTIMATE
CMW Diffusion
ACTUAL RAIN MEASUREMENT
CMW Diffusion
RAIN ESTIMATE
Independent Validation
7
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Comparison between CMW estimate and (independent)
reference rainfall for 0230 TUC

(2 hour step, forward propagation)
8
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
What if we use the 0230 measure instead of the
0430 CMW-scheme estimate when comparing _at_
0430? So, the CMW scheme i
s actually transporting rainfall
9
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Time degradation Average for 31/OCT/2003 Usi
ng the CMW, we can maintain correlations 0.80
for up to 2.5 hours The performances of the met
hod when compared with ground rainfall at
instantaneous scale will be linked with the
performances of the rainfall to be transported
relevant perhaps for GPM
10
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
http//hermes.uclm.es
11
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
12
Intro
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions

• Validation activities
• Opportunity we needed data for algorithm
  pre-calibration
  
• Validation has a geographical component
  validation results are different in different
  places, and we need the algorithms tuned for
  Spain.
• Validation against gauge, GR comparison with
  models
  

13
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Andalusia case study
14
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro

• Half-hourly raingauge data availability
• Neural network IRPMW fusion
• Algorithm characteristics
• High temporal resolution
• High spatial resolution
• High accuracy
• Tapiador, F.J., Kidd, C., Levizzani, V., Marzano,
  F.S., 2004. A Neural Networks-Based Fusion
  Technique to Estimate Half Hourly Rainfall
  Estimates at 0.1º Resolution from Satellite
  Passive Microwave and Infrared Data. Journal of
  Applied Meteorology, 43, 576-594.

15

• Rain Gauges in Andalusia

Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Interpolation I Kriging
Interpolation II Inverse distance
SSM/I data
16
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Correlations at 0.1º (monthly)
17
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Correlations at 0.5º (monthly)
18
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Validation of IPWG Products on Spain
19
Val example 1
Algorithms
Val example 3
Conclusions
Val example 2
Intro
Geo
NOGAPS

• (CICS, University of Maryland archive)
• 00Z-00Z products
• NOGAPS
• NRL GEO
• NRL PWM
• CPC Morphing
• 3B42RT

NRL PMW
CPC
20
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Rain Gauges Location
21
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
22
Val example 1
Algorithms
Val example 3
Conclusions
Val example 2
Intro

• Geolocation error surface analysis
• Data from 01/JAN/2005 to 01/SEP/2005
• Satellite vs gauge
• Assuming 5km interval error in the nominal
  satellite data geolocation
  

23
Val example 1
Algorithms
Val example 3
Conclusions
Val example 2
Intro
CPC Morphing
3B42RT
NRL GEO
NRL PMW
NOGAPS
24
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
EUMETSATs Convective Rain Rate Product (CRR) N
owcasting Satellite Application Facility (SAF)
25
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
26
Val example 1
Val example 2
Algorithms
Conclusions
Val example 3
Intro
27
Val example 1
Val example 2
Algorithms
Conclusions
Val example 3
Intro
Validation Comparison data sources

28
Val example 1
Val example 2
Algorithms
Conclusions
Val example 3
Intro
GR Visual comparison

29
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro

• Spain as validation site

30
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro

• Pros and Cons
• Many examples of frontal, convective and
  orographic precipitation and mixed cases.
  
• Three rainfall regimes in 500,000 sq km (Texas
  696,000 sq Km)
  
• High N-S gradient. Well-calibrated, reliable
  validation net
  
• Rain gauges nets (INM, river authorities, etc.)
• Ground Radar
• TRMM coverage (South), MSG, SSM/I, AMSU, AVHRR,
  etc.
  
• Limited area
• Limited public availability of validation data
  but this could be solved for GPM
  

31
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro

• Available Validation Data
• INM gauges network
• GR
• River authorities networks
• Agrarian Meteo Nets
• Specifically-tailored nets and instrumentation

32
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro
Geography
37N
33
Val example 1
Val example 2
Algorithms
Val example 3
Conclusions
Intro

• Conclusions
• Suitability for validation site in Catalonia
  (Daniel Sempere, GRAHI)
  
• Experience in satellite rainfall estimates
  algorithms
  
• Interface with NWP modelers (NWPSatMerged
  algorithms)
  
• Data availability and support from agencies
• Geography of Spain very different from other
  validation places

34
(No Transcript)