What can we learn about rainfall from SMOS/SMAP Sea Surface Salinity ? Context and objectives Data and methods Results (learning & validation) Conclusion & perspectives ∆""" ∆""" RR RAD = 1 ##. ℎ &' RR RAD = 5 ##. ℎ &' Figure 4 : Sea surface salinity anomaly and radiometer rain rate (04/10/2012) . VALIDATION & other studies : Salinity : SMOS/ESA L2 v622 (2015-2016) & SMAP JPL/CAP L2 SSS (Summer 2016). Rain rate : REMSS, CMORPH, IMERG (μwave RR, IR RR, interpolated product (IMERG ATBD, 2015)). Salinity anomaly computation : Reference salinity is a salinity mean in a 3°x3° box estimated from pixels with a low probability of rain. ∆ = − A. Supply 1 , J. Boutin 1 , N. Martin 1 , J.L.Vergely 2 , G. Reverdin 1 , N.Viltard 3 , A. Hasson 1 , S.Marchand 1 , H. Bellenger 4 Mail : [email protected] 1 LOCEAN, Paris, France 2 ACRI-st, Guyancourt, France 3 LATMOS, Guyancourt, France 4 JAMSTEC, Yokosuka, Japan 06/01/2016 mm/h mm/h mm/h mm/h SMOS salinity SMAP salinity RR SMOS RR SMAP RR IMERG cc SMOS RR IMERG cc SMAP mm/h mm/h mm/h RR SMOS RR IMERG cc SMOS RR SMAP RR IMERG cc SMAP RR SMAP –RR SMOS RR IM. cc SMAP – RR IM. cc SMOS Figure 9 : Study case, 06/01/2015: SMOS, infrared RR and interpolated IMERG within 15mn but time lag between SMOS and nearest radiometer RR : 2h10; top, left) RR SMOS; top, right) Radiometer RR; bottom, left) RR IMERG (interpolated) ; bottom, right) RR IR. Time RAD – TIME SMOS/SMAP • Rainfall has a clear impact on Sea Surface Salinity SSS observed by L-Band radiometry (Boutin et al. 2013, 2014 ; M.E. McCulloch et al, 2012) (Figure 1). Figure 5: Determination of the relationship between SSS anomaly and Rain Rate: cumulative distribution of SSS anomaly (red) and Rain rate (black) are put in correspondence in order to get rid of time lags between SSS and RR measurements. Figure 8 : Study case, 11/08/2016: SMOS, SMAP & radiometer RR less than 15mn apart. Top: satellite SSS left) SMOS; right) SMAP; Middle: satellite RR left) SMOS; right) SMAP; Bottom: RR IMERG (Interpolated product at less than 15mn from satellite SSS) collocated with left) SMOS; right) SMAP. Figure 10 : Study case, 15/07/2016 ; SMOS, SMAP & radiometer RR at more than 15mn apart. Left: RR from satellite SSS top) SMOS; middle) SMAP; Bottom) RR displacement Right: RR from IMERG (interpolated product) collocated with top) SMOS; middle) SMAP; Bottom) RR displacement ITCZ Pacific Ocean TRAINING SET : 2012 ; 110°W-180°W ; 0°N-15°N - Salinity : SMOS/ESA L2 SSS (at 1cm depth) (v622), ascending orbits - Rain rate : REMSS (Hilburn and Wentz, 2007). • SMOS and SMAP satellite missions provide measurements at more than 30mn from rain radiometer in ~40 % of time (Figure 2). Given the strong intermittency of rain, we investigate which additional information on rain they can bring. Figure 2 : Cumulative distribution for Dt (TIME RAD – TIME SMOS/SMAP (in hours)) computed during June-July 2016. Figure 3 : Example of rain induced surface cooling observed on 14 December 2011 by a Surface Salinity Profiler (SSP; Asher et al. 2014) at -11 cm depth (black line) in the northern Pacific Ocean and simulations (in colors) of (top) salinity in the skin layer (50μm depth) blue dashed) and at 11cm depth (blue solid); (bottom) observed rain rate (solid, mmh -1 ) and wind speed (dashed, ms -1 ). (Bellenger et al. 2016) • Rain Rate derived from satellite SSS is in line with rain radiometer measurements but depends on the RR learning product. • SMOS and SMAP bring additional information on rain cells evolution that could be merged with rain radiometers data. • Validate these results using information independent from RR satellite products: e.g. in-situ rain radar data. • Take advantage of synergy between SMAP and SMOS for studying the temporal evolution of freshening after a rain event. Conclusions Perspectives Figure 1 : DSSS from SMOS versus radiometer rain rate in 2012 (1°x1° smoothing); Colorbar = Log10 of occurrences, red line = orthogonal regression, magenta points and errorbar = mean and standard deviation in rain rate classes (RR=0 and every 1 mm/h classes) ; theoretical relationship for skin layer (Schlüssel et al. 1996) . • In-situ SSS observations and surface ocean models show strong freshening after a rain event but supplementary studies are needed to improve understanding of rain/SSS interactions (Figure 3, Bellenger et al. 2016) • Asher, W. et al.(2014), Observations of rain-induced near-surface salinity anomalies, J. Geophys. Res. Oceans. • Bellenger, H. et al. (2016), Extension of the prognostic model of sea surface skin temperature to rain-induced cool and fresh lenses, submitted to JGR-Ocean. • Boutin, J. et al. (2014), Sea surface salinity under rain cells: SMOS satellite and in situ drifters observations, J. Geophys. Res. Oceans. • Boutin, J. et al. (2013), Sea surface freshening inferred from SMOS and ARGO salinity: Impact of rain, Ocean Science. • McCulloch, M. et al. (2012), Have mid-latitude ocean rain-lenses been seen by the SMOS satellite?, Ocean Model. On-going study SALINITY ANOMALY RR (mm/h) • SSM/I, SSMIS, Windsat and TMI data are produced by Remote Sensing Systems. Data are available at www.remss.com/missions/ • SMAP data are distributed by JPL OurOcean Portal and available at www.ourocean.jpl.nasa.gov • IMERG data are distributed by Goddard Space Flight Center and available at www.pmm.nasa.gov • CMORPH data are distributed by NOAA and available at http://www.cpc.noaa.gov Acknowledgements: This work is supported by the ESA/STSE SMOS+Rainfall (coordinated by ARGANS company) and by the CNES/TOSCA SMOS-OCEAN projects. We thank Cécile Mallet for her recommendations on statistical aspects. Figure 7: RR distributions for SMOS 2012, REMSS 2012, CMORPH 2012, REMSS 2015 (January to June) and SMOS 2015 (January to June). • SMOS RR, SMAP RR and radiometer RR are very consistent at <15mn lag (Figure 8). • Satellite SSS give additional information on RR when it flies over the study area at >~30mn before/after the nearest rain radiometer (Figure 9). • RR SMOS and RR SMAP allow to reconstruct the trajectory of the rain cell (Figure 10). • A DSSS-RR relationship is derived from 2012 data (Figure 5) • RR SMOS / RR RAD are in good agreement when considering the REMSS RR products both for learning and validation, with two validation methods: (Figure 6) colocations within [-15mn; 30mn] and (Figure 7) statistical distributions. A slightly different statistical distribution is obtained from a different RR product (CMORPH ; Figure 7). Figure 6: Rain rate from SMOS versus radiometer rain rate for left) 2012 (learning year) and right) 2015 (validation) ; Colorbar = Log10 of occurrences, red line = orthogonal regression. 2012 RMSE=0.29 R2=0.40 y=-0.19x-0.03 N=685944 2015 (January to June) 2012 (January to June) RMSE=0.89 Corr=0.73 y=1.06x+0.01 N=343737 RMSE=1.08 Corr=0.60 y=0.94x-0.01 N=203483 02:17 TU 02:25 TU 14:04 TU 14:25 TU