Evaluation of 14 years of SMOS and SMOS-like soil moisture against the ESA CCI dataset

Transcription

1 Evaluation of 14 years of SMOS and SMOS-like soil moisture against the ESA CCI dataset B. Cluzet, N.J. Rodriguez-Fernandez, A. Mialon, A. Al Bitar, Y. Kerr, CESBIO (CNRS, CNES, IRD, Université de Toulouse) A. Al Yaari, J.P. Wigneron INRA (Bordeaux, France) Acknowledgements: R. De Jeu, R. Van der Schalie (Vandersat)

2 Outline Introduction long times series of soil moisture the ESA AMSR-E / SMOS fusion project the AMSR-E NN dataset Results & discussion Consistency of AMSR-E NN and SMOS L3 Evaluation of the NN-based dataset with respect to CCI SM Links to climate indices Conclusions

3 Soil moisture time series SM is an essential climate variable (ECV) Remote sensing from space gives access to global SM maps but long time series are needed And merging data from different instruments is not trivial Specifically designed to measure SM What is their impact? Wagner et al 2012, Liu et al. 2012

4 AMSR-E and SMOS fusion project Best strategy to add SMOS to a long term SM record? AMSR-E TB Regressions «AMSR-E Reg» AMSR-E SM SMOS SM (Al Yaari et al. 2016) SMOS TB LPRM «SMOS LPRM» SMOS SM (van der Schalie et al. 2016, 2017) AMSR-E TB Neural Network «AMSR-E NN» AMSR-E SM SMOS L3 SM (Rodriguez-Fernandez et al. 2016)

5 The Neural Network (NN) approach Rodriguez-Fernandez et al., 2016 Promising performance! Van der Schalie et al. (2017, Submitted)

6 Objectives of the current study Evaluate consistency of AMSR-E NN and SMOS L3 SM Compare the ASMR-E NN + SMOS L3 dataset (NN) to CCI v2 SM Case study : analysis of climatic signatures comparing NAO and ENSO climatic indexes to SM anomalies NN SM ( ) CCI v2 SM ( ) ERA Interim/Land SM ( )

7 AMSR-E NN vs. SMOS L3 AMSR-E NN SMOS L3 NN SM anomaly SMOS L3 AMSR-E NN AAMSR-E NN Corr: some discrepancies BUT AMSRE-NN and SMOS L3 are mostly consistent!

8 NN / CCI climatology August CCI NN SM monthly climatologies (m3/m3)

9 NN / CCI anomalies time series Tropical Savanna Hot desert Hot desert

10 Climatic indices NAO : Northern Atlantic Oscillation Atmospheric Impact on wintertime precipitation in Europe and around Mediteranean Sea (Marshall et al., 2001, Cassou et al., 2004). Few or less observational or model evidence for SM impact NAO monthly index Source : University of Michigan ENSO : El Nino Southern Oscillation Coupled Atmospheric-Oceanic mode Worldwide impact Nino3.4 monthly index Source : MetOffice

11 NAO-SM anomalies NN CCI NAO + c) NAO - NAO + NAO - SM anomalies for NN and CCI during NAO + (a-b) and NAO- (c-d) months

12 Correlation Mediterranean Sea-Europe CCI NN Source: University of Michigan

13 ENSO NN CCI El Nino La Nina El Nino La Nina SM anomalies for NN and CCI during El Nino (a-b) and La Nina (c-d) months

14 Lagged correlation maps Increasing Time-lag (months) CCI NN ERA

15 Lagged correlation maps Increasing Time-lag (months) CCI NN ERA

16 Lagged correlation maps Increasing Time-lag (months) CCI NN ERA

17 Lagged correlation maps Increasing Time-lag (months) CCI NN ERA

18 Lagged correlation maps Increasing Time-lag (months) CCI NN ERA

19 Lagged correlation maps Increasing Time-lag (months) 0 CCI NN ERA

20 Lagged correlation maps Increasing Time-lag (months) CCI NN ERA Results (NN-CCI) Tropical/savanna : 0-3 months Arid : 4-6 months -6

21 Conclusions AMSR-E NN- SMOS L3 : some local changes NN dataset compares favorably to CCIv2 NN correlation patterns of SM anomalies and NAO are consistent with those of precipitations anomalies Cassou et al., 2008 Time lags with ENSO show two behaviours in agreement with Miralles et al., 2014, Nicolai-Shaw et al., 2016

22 Regional Max. corr and lag times Region/climate (Köppen Geiger classif.) Southern Africa (Eastern part) BWh NN CCI ERA TRMM R Lag R Lag R Lag R Lag ?? Nordeste BSh Philippines Af La Plata (Argentina)Cfa Central-North- America (texas) Bsk BWh : Arid/Desert.Hot BSh : Arid/steppe/hot Af : Tropical Rainforest Am : Tropical/Monsoon Raposa Cfa : Temperate Without dry season and with hot summer BSk : arid Steppe cold Aw: (Tropical) savanna

23 Local scale Highlight on AMZ and SAF max ( cf. corr maps) Corr : =-0.66 NN CCI El Nino Raposa : savanna (Aw) NA - La Nina Specific locations show clear evidence of ENSO-SM teleconnexions inside less significant regions

24 Feedback to the CCI project SMOS SM as reference for rescaling other instruments NN promising SMOS LPRM to CCI SMOS now in CCI (Dorigo et al. 2017, RSE in press) using LPRM not as a scaling reference

25 The ESA Near Real Time SMOS SM product Training on SMOS Level 2 SM Similar performances (slightly better indeed) Much faster! Less than 3.5 hours after sensing Rodriguez-Fernandez et al. (2017, HESS) Implemented by : With support by : Delivered to : Disseminated by:

26 AMSR-E NN SM vs SMOS L3 SM and other datasets Average from June 2010 to October 2011 Rodriguez-Fernandez et al. (2016, Remote Sensing) Triple collocation analysis as a function of NDVI Van der Schalie et al. (2017, Submitted)