European High-Resolution Soil Moisture Analysis (EHRSOMA)

Transcription

1 European High-Resolution Soil Moisture Analysis (EHRSOMA) Jasmin Vural EUMETSAT Fellow Day,

2 European High-Resolution Soil Moisture Analysis (EHRSOMA) Jasmin Vural EUMETSAT Fellow Day, Goals: Error estimation and implementation Improve spatial resolution of data assimilation Performance of data assimilation

3 European High-Resolution Soil Moisture Analysis (EHRSOMA) Jasmin Vural EUMETSAT Fellow Day, Goals: Error estimation and implementation Improve spatial resolution of data assimilation Performance of data assimilation

4 Soil moisture water content of unit of soil approx % of terrestrial water

5 Soil moisture water content of unit of soil approx % of terrestrial water soil water index SWI = w root w wilt w fc w wilt

6 Wikipedia (adapted) EHRSOMA J. Vural Soil moisture Vince Migliore

7 Wikipedia (adapted) EHRSOMA J. Vural Soil moisture Vince Migliore

8 Wikipedia (adapted) EHRSOMA J. Vural Soil moisture Vince Migliore

9 Measuring soil moisture Basics Oriental polarisation of dipoles Orientation changes of applied electric field are followed below relaxation frequency f r

10 Measuring soil moisture Basics Oriental polarisation of dipoles Wikipedia Orientation changes of applied electric field are followed below relaxation frequency f r

11 Measuring soil moisture Remote sensing Backscatter coefficient measured remotely brightness temperature of backscatter T B (ε(θ)) properties of scattering objects

12 Measuring soil moisture Remote sensing Backscatter coefficient measured remotely brightness temperature of backscatter T B (ε(θ)) properties of scattering objects Ulaby, Moore, Fung 1981; adapted

13 Measuring soil moisture Remote sensing Backscatter coefficient measured remotely brightness temperature of backscatter T B (ε(θ)) properties of scattering objects Ulaby, Moore, Fung 1981; adapted

14 TU Wien EHRSOMA J. Vural Measuring soil moisture MetOp/ASCAT + Sentinel-1/SAR = SCATSAR

15 TU Wien EHRSOMA J. Vural Measuring soil moisture MetOp/ASCAT + Sentinel-1/SAR = SCATSAR TU Wien

16 TU Wien EHRSOMA J. Vural Measuring soil moisture MetOp/ASCAT + Sentinel-1/SAR = SCATSAR TU Wien

17 Data assimilation Land surface model Interaction Soil Biosphere Atmosphere (ISBA) 14 vertical levels only vertical exchange Decharme et al. 2013

18 Data assimilation Land surface model SURFEX x 133 grid points 2.5km resolution SURFEX Offline Data Assimilation (SODA) atmospheric forcing by AROME Masson et al. 2013

19 Data assimilation Land surface model SURFEX x 133 grid points 2.5km resolution SURFEX Offline Data Assimilation (SODA) atmospheric forcing by AROME Natural land surface : energy, water, carbon fluxes Hydrological and vegetation processes Masson et al. 2013

20 Data assimilation Kalman Filter Forecasting x k = F k x k 1 + B k u k Update x k = x k + K ( z k -H k x k ) Kalman gain K = P k H k T (H k P k H k T + R k ) 1

21 Data assimilation Kalman Filter Forecasting x k = F k x k 1 + B k u k Update x k = x k + K ( z k -H k x k ) Kalman gain K = P k H k T (H k P k H k T + R k ) 1 Measurement uncertainty 2 R ii = σ obs,i

22 Observation errors Systematic errors dependent on instrument Random errors noise assumed to be normally distributed

23 Observation errors Systematic errors Bias correction data set (SCATSAR) + reference (SURFEX) match cumulative density functions apply fit parameters to actual data

24 Observation errors Systematic errors Bias correction data set (SCATSAR) + reference (SURFEX) match cumulative density functions apply fit parameters to actual data

25 Observation errors Random errors Triple Collocation error model: Θ i = α i + β i Θ + ε i i: active & passive data sets & reference SCATSAR AMSR2 SURFEX

26 Observation errors Random errors Triple Collocation error model: Θ i = α i + β i Θ + ε i i: active & passive data sets & reference SCATSAR AMSR2 SURFEX assumptions: linearity of error model signal stationarity same climatology for all data sets error stationarity dependent on length of time period (seasons!) independency between Θ i and ε i (error orthogonality) negligible independency between errors (zero error cross-correlation) different type of observations

27 Observation errors Random errors Triple Collocation error model: Θ i = α i + β i Θ + ε i i: active & passive data sets & reference SCATSAR AMSR2 SURFEX assumptions: linearity of error model signal stationarity same climatology for all data sets error stationarity dependent on length of time period (seasons!) independency between Θ i and ε i (error orthogonality) negligible independency between errors (zero error cross-correlation) different type of observations Python Toolbox for the Evaluation of Soil Moisture Observations (pytesmo) of TU Wien

28 Observation errors Random errors bias corrected data time series correlation? point #? Triple Collocation error std σ εi

29 Observation errors Random errors bias corrected data time series correlation? point #? Triple Collocation error std σ εi

30 Observation errors Random errors bias corrected data time series correlation? point #? Triple Collocation error std σ εi

31 Observation errors Random errors

32 Summary SCATSAR data high spatial & temporal resolution Bias correction systematic error Triple Collocation random error Covariance matrix of observation error data assimilation

33 Summary SCATSAR data high spatial & temporal resolution Bias correction systematic error Triple Collocation random error Covariance matrix of observation error data assimilation Thanks to Stefan Schneider (ZAMG) Alexander Gruber (TU Wien)