-SMOS Level 4 Thematic SSS Research products -Product User Manual-

Transcription

1 CATDS-CECOS-L4-PUDOC Issue 2 Rev 0 - Research products -Product User Manual- Function Name Signature Date Project Manager Nicolas REUL (IFREMER C-ECOS) CATDS -CECOS Team 11 May 2015 Centre Aval de Traitement des Données SMOS (CATDS)- Expertise Center -Ocean Salinity (CEC-OS) Laboratoire d Océanographie Physique et Spatiale Z.I. Pointe du Diable-B. P. 70, Plouzané France Tél. : +33 (0) Fax : +33 (0) nreul@ifremer.fr

2 Research Products Product User Ref Level4_SSS_PUD Page 2 status Title Research Products -Products User Manual Issue Revision Date Reason for the revision /05/2015 Initial version Modification status Issue Rev Status * Modified pages Reason for the modification * I = Inserted D = deleted M = Modified IFREMER CNES 2011

3 Page 3 Table of contents 1. INTRODUCTION PRODUCTS CHARACTERISTICS, FORMATS, FILE NAMING CONVENTIONS & DATA CITATION Level 4a product content Convention for the L4a Netcdf files SSS, SST and Wind speed from SMOS data SMOS SSS SST ECMWF Wind Speed ECMWF Cummulated Evaporation OAFLUX Cummulated Precipitation TRMM-3B Cummulated Precipitation CMORPH OSCAR zonal and meridional currents Wind Stress zonal and meridional components from ASCAT Mixed-Layer Depth Salinity at the base of the Mixed Layer Time Interpolated ISAS SSS fields In Situ/SMOSL4a Match-up DataBase (MDB) files ARGO float observations Ship ThermoSalinograph Data Surface drifter Data Global tropical Moored Buoy Array data Southern Ocean SSS from Seals Match-Up Database netcdf Files Naming conventions Level 4b Product content Convention for the L4b Netcdf files Sea Surface Density Mean Sea Level Anomaly Other variables in L4b products Level 4c Product content Methodology to evaluate anomalies Level 4c product content Convention for the L4c Netcdf files DATA CITATION PRODUCT ALGORITHM, VALIDITY, COVERAGE AND KNOWN FLAWS Product Algorithm and Validity Input Data and coverage Several known flaws Remaining Major Issues in the L2OS data Solar contaminations RFI Land Contamination... 74

4 Page Major Geophysical correction issues: Sky noise... 75

5 Page 5 1. Introduction This document is the product user manual for the CATDS/CEC-OS SMOS Level 4 Thematic sea surface salinity (SSS) research products. 2.1 describes the composite L4a products major characteristics, 2.2 describes the composite L4 Match-Up databse products major characteristics, 2.3 describes the composite L4b products major characteristics, 2.4 describes the composite L4c products major characteristics, 3 explains how to aknowledge the use of these data, 4 gives an overview of the products coverage (e.g. input data features, missing output data,..) and describe several known flaws in these datasets. The scientific relevance for measuring Sea Surface Salinity (SSS) is more and more recognized in the ocean community. SSS plays an important role in the dynamics of the thermohaline overturning circulation, ENSO, and is the key tracer for the marine branch of the global hydrologic cycle, which comprises about ¾ of the global precipitation and evaporation. Ocean surface salinity is also of key importance for land-sea (river plumes), air-sea (ocean stratification, barrier layers, CO2 fluxes) and ice-sea interactions, marine biology, marine chemistry (carbonate cycle) and marine bio-optic. SSS is also essential to understanding the ocean s interior water masses, knowing that they derive their underlying temperature and salinity properties during their most recent surface interval. In the IPCC WGI Fifth Assesment Report published in October 2013, chapter 3.3 is dedicated to changes in salinity and freshwater content. Multi-decadal SSS trends have been documented in tropical and northern latitudes that are likely signatures of evaporation or precipitation trends, as predicted under global warming scenarios. As reported: Chapter 3: Observation: Ocean It has not been possible to detect robust trends in regional precipitation and evaporation over the ocean because observations over the ocean are sparse and uncertain Ocean salinity, on the other hand, naturally integrates the small difference between these two terms and has the potential to act as a rain gauge for precipitation minus evaporation over the ocean Diagnosis and understanding of ocean salinity trends is also important because salinity changes, like temperature changes, affect circulation and stratification, and therefore the ocean s capacity to store heat and carbon as well as to change biological productivity. SPM-5 It is very likely that regions of high salinity where evaporation dominates have become more saline, while regions of low salinity where precipitation dominates have become fresher since the 1950s. These regional trends in ocean salinity provide indirect evidence that evaporation and precipitation over the oceans have changed (medium confidence). In addition to the in situ observing network, two satellite missions (SMOS and Aquarius) are currently in orbit and operating with the aim of measuring salinity from space for the first time, with

6 Page 6 differing technology approaches but operating in the same spectral region (L-Band). Strong efforts have been carried out by the expert of the satellite salinity missions but also by the "first external users" to demonstrate the scientific interest of the new SSS from space products. In particular, scientific demonstrations and associated peer-reviewed publications have already revealed the strong potential and new information of these new data for the following major ocean-atmosphere processes: Ocean surface response to precipitation and evaporation fluxes New characterization of large-scale upwelling processes (fisheries) Land-ocean freshwater flux monitoring (large tropical river plumes) Ocean-atmosphere interactions (Barrier layers, Tropical cyclones, upper ocean stratification) Surface salt distrtibution links with Interanual Climatic variability (La Nina, Indian Ocean Dipole) Near-surface transport and large scale inter-gyre exchanges of salt through major oceanic currents (e.g., Gulf stream) Thermo-haline circulation (allowing first spaceborne sea water density estimates), new inputs for the Tropical instability waves monitoring Thin sea ice monitoring Surface wind speed estimation under Tropical cyclones and storms These first demonstrations of scientific applications with the new salinity satellite missions is supported by a constantly growing user community since the launch of the missions. Thus, developing scientific analyses and tools that will help to better monitor sea surface salinity changes will stimulate a growing user community and will be key to developing already identified applications but also to stimulate new ones for this ECV. In fine, such effort shall certainly contribute to our better understanding of the Earth s climate change and therefore to the IPCC initiative. Given the novelty of the Satellite SSS data, the current user and science communities being interested in SSS satellite data over the ocean are mostly connected to the mission expert team directly. The main reason of the currently limited growth of the user community for this important ECV first estimated from space is linked to the fact that this complex data are rather new and that data quality still needs to be improved or mostly better understood for applications. At the same time well informed users and experts start publishing very interesting scientific results. A non aware user starting from Level 2 or 3 data might then be surprised by the level of expertise, processing and analysis required to extract interesting scientific results from SMOS (and Aquarius) data. Hence, one of the first community and impact driven objective of the products we propose here shall therefore be to foster on a scientific level (pre-operational) the synergistic use of SSS in several parts of the ocean community. Potential science applications include: freshwater budget for climate change studies, El Nino/La Nina prediction, ocean surface circulation monitoring (coupling with altimetry and SST), operational oceanography (model assimilation, climate indicators), land-sea interactions, weather forecasts (oceanic rain gauge, extreme events), marine biology (fisheries), CO2 sequestration and ocean acidification.

7 Page 7 To reach these aims we developed 4 types of level 4 products. For this first version of the products, we chose only to produce them at fixed spatial and temporal resolution which are 1/2 and weekly composite. -Level 4 a synergistic products: These products include key geophysical variables to analyse the salinity budget in the upper ocean mixed layer. These include: -SSS from SMOS & in situ OI (ISAS), -SST from ECMWF, -wind speed modulus (ECMWF) -wind stress components (ASCAT), -ocean surface current components (OSCAR), -Evaporation (OAFLUX), -Precipitation (TRMM3B42 & CMORPH), -Mixed Layer Depth (In situ OI), -Salinity at the base of the mixed-layer These fields are averaged or cummulated in time -or interpolated- over the week of the SMOS L4a SSS and gridded at the same 1/2 spatial resolution. L4aSSS In situ Match-up products Quality controlled surface salinity and temperature measurements from Argo floats, ship TSG, surface drifters, Tropical Moorings and sea seals data provided by the Coriolis, GOSUD, SAMOS and LOCEAN data centers are collected weekly over each of the L4 composite product period and co-locatized with SMOS L4a SSS products. Supplementary data are provided with the in situ match-up database such as vertical profiles of S and T if available, co-localized ASCAT and TRMM3B42 data, etc.. A match-up database file per in situ sensor type is provided for each week of the L4a products. L4b density products L4b density products are satellite surface density fields. These surface density fields are so-called "CATDS/CECOS Ifremer SMOS Level 4c research products" and are weekly composite products of surface density at a spatial resolution of 0.50 x 0.50 deduced from the L4aSSS data and ECMWF SST data. In addition, the products include: -mean sea level anomalies from AVISO -OSCAR currents components -Wind stress components from ASCAT

8 Page 8 L4c anomaly products L4c products are Anomaly fields. These surface salinity anomaly fields are so-called "CECI-OS Ifremer SMOS Level 4c SSSA research products" and are weekly composite products at a spatial resolution of 0.50 x 0.50 deduced from the L4aSSS data. The anomalies are evaluated by removing an annual-averaged reference weekly composite field evaluated by averaging the SMOS L4aSSS data over 5 years (2010 to 2014). weekly anomalies for various thematic fields are also included such as: -ECMWF sst anomalies -Cummulated Evaporation anomalies from OAFlux -Cummulated Precipitaions anomalies from TRMM3B42 -Cummulated Precipitation anomalies from CMORPH -Mixed layer depth anomalies from APDRC -Wind stress components anomalies from ASCAT -Anomalies of the Salinity at the base of the Mixed Layer -OSCAR currents components anoamlies -ISAS temporally interpolated monthly SSS & SST fields anomalies -surface density anaomalies -MSLA

9 Page 9

10 Page Products characteristics, formats, file naming conventions & data citation 2.1 Level 4a product content The CATDS/CEC-OS SMOS Level 4a Version 1 SSS research products are weekly (7 days) composite at 50 km resolution. The products coverage is May 2010-December In addition, these products also include an ensemble of geophysical parameters derived from wellacknowledged products in the scientific communities that are useful for synergistic science applications using SMOS data. These include Sea Surface Temperature (ECMWF), surface currents (OSCAR), rain (TRMM), evaporation (OAFLUX), surface wind stresses (ASCAT), mixed-layer depth from In situ OI(APDRC), Surface salinity from in situ OI (ISAS) and salinity at the base of the mixed-layer depth estimated also from In Situ OI. Table 1: Variable Name Dimension and Description for the CECOS L4a V01 research products Variable Name Dimension Description time 1 Central date of the time period over which the SMOS data were combined to generate the composite product. Number of days since :00:00 latitude nlat Vector of the latitude of the grid nodes over which the composite product is derived. Expressed in degrees North from -90. to +90. longitude nlon Vector of the longitude of the grid nodes over which the composite product is derived. Expressed in degrees East from to sss nlat nlon Gridded Sea Surface Salinity from SMOS [Practical Salinity Scale]. Missing Values= sst nlat nlon Gridded Sea Surface Temperature colocated at SMOS pixels from ECMWF forecasts [Kelvins]. Missing Values= -

11 Page 11 Wind_Speed nlat nlon Gridded 10-m height wind speed module colocated at SMOS pixels from ECMWF forecasts [meter/seconds]. Missing Values= RFI_stat nlat nlon Gridded percentage for Radio- Frequency Interferences occurence within the brighthness temperature data set used for SSS product generation at a given pixel [%] Missing Values= Zonal_component_surface_currents nlat nlon Gridded OSCAR surface current (Ekman+Geostrophic) zonal component interpolated in space and time at SMOS pixels [m/s]. The 1/3 resolution 5 day OSCAR data were re-gridded on a 1/2 resolution grid and the 5-day currents fields were linearly interpolated in time on a daily basis. The mean current components provided into the L4a products are then the result of a time averaging over the 7-day period of the SMOS L4a product. Meridional_component_surface_currents nlat nlon Gridded OSCAR surface current (Ekman+Geostrophic) meridional component interpolated in space and time at SMOS pixels [m/s]. Missing Values= Mean_bias_In_situ_minus_SMOS nlat nlon Mean Bias between SMOS weekly L4a data and In situ observations from the Match-Up database files estimated over the period May 2010-December 2014 [pss]

12 Page 12 Standard_Deviation_In_situ_minus_SMOS nlat nlon Standard deviation of the differences between SMOS weekly L4a data and In situ observations from the Match-Up database files estimated over the period May 2010-December 2014 [pss] TRMM3B42_accumulated_rain nlat nlon Cummulated rain falls [mm] from TRMM3B42 products over the period of time of each weekly L4SSS product and avegared on the LaSSS 50 km grid. The 1/4 resolution 3-hourly TRMM-3B42 data were re-gridded on the SMOS 1/2 resolution grid and the cummulated rain [mm] was evaluated over the 7-day period corresponding to the SMOS L4a product. Data are only available for the 50 S-50 N belt and for the period until 8 august OAFlux_accumulated_Evaporation nlat nlon Cummulated evaporation [mm] from OAFLUX products over the period of time of each weekly L4SSS product and avegared on the L4aSSS 50 km grid. MLD nlat nlon Mixed-Layer depth estimated from IPC/APDRC. The 1 X1 monthly MLD orignial fields were interpolated in space and time on a 1/2 grid and daily.the MLD provided in the product is the temporal mean of the daily interpolated MLD over the 7-day period of the SMOS L4a product. surface_downward_northward_stress nlat nlon Surface wind stress meridional component included into our products are based on the Advanced

13 Page 13 SCATterometer (ASCAT) daily data produced and made available at Ifremer/cersat on a resolution (Bentamy and Croize- Fillon 2012) since November surface_downward_eastward_stress nlat nlon Surface wind stress zonal component included into our products are based on the Advanced SCATterometer (ASCAT) daily data produced and made available at Ifremer/cersat on a resolution (Bentamy and Croize- Fillon 2012) since November Salinity_MLD_base nlat nlon The salinity values at the base of the Mixed Layer Depth (MLD) [pss] correspond at each grid point to the temporally interpolated monthly ISAS value at depth p. p is chosen as the nearest standard depth levels lower than the MLD value. CMORPH_accumulated_rain nlat nlon Cummulated rain falls [mm] from CMORPH products over the period of time of each weekly L4SSS product and avegared on the LaSSS 50 km grid. CMORPH estimates cover a global belt ( 180 W to 180 E) extending from 60 S to 60 N latitude and are available for the complete period of the SMOS L4.V01 data Time_interpolated_ISAS_sss nlat nlon Optimal interpolation fields generated using delayed time quality checked in situ measurements (Argo and ship) by IFREMER/LPO, using the In Situ Analysis

14 Page 14 System (ISAS). The original files are monthly fields at 0.5 resolution. A linear interpolation in time was used to estimate the fields at the center time of the 7-day period of the SMOS L4a product. date_start 1 Start date of the time period over which the SMOS data were considered to generate the composite product date_stop 1 End date of the time period over which the SMOS data were considered to generate the composite product Convention for the L4a Netcdf files The generic filename convention for the weekly composite L4a products is given as defined below: CECOS_SMOS_L4aSSS_0.5deg_YYYY.DD_YYYY.DD_V01.nc Colored symbols indicate digital variables defined as follows: The first YYYY (4 digits) and second DD (3 digits) variables indicate the year and the ordinal number of the day in year corresponding to the start date of the time period over which the SMOS weekly data were considered to generate the composite product. The third YYYY (4 digits) and fourth DD (3 digits) variables indicate the year and the ordinal number of the day in year corresponding to the end date of the time period over which the SMOS weekly data were considered to generate the composite product. The last variable 01 (2 digits) indicate the product processing version. Example: for the weekly composite product at 0.5 degree resolution generated from 27 aug 2010 (day of year=239) to 2 of Sep 2010 (day of year=245) using the processing version 1, the file name is: CECOS_SMOS_L4aSSS_0.5deg_ _ _V01.nc In the following we describe in more detail the content of the products

15 Page SSS, SST and Wind speed from SMOS data SMOS SSS Each L4 netcdf file correspond to 7-day period composites between May 2010 and December 2014 and include the temporally & spatially averaged L4 SSS at 50 km resolution SST ECMWF Example of 7-days mean gridded Sea Surface Temperature colocated at SMOS pixels from ECMWF forecasts.

16 Page Wind Speed ECMWF Example of a 7-days averaged gridded 10-m height wind speed modulus colocated at SMOS pixels from ECMWF forecasts Cummulated Evaporation OAFLUX For ocean evaporation, we used the data from the Objectively Analyzed Air-sea Fluxes (OAFlux) project (Yu., 2007), available from the Woods Hole Oceanographic Institution (WHOI) at

17 Page 17 The daily 1-degree gridded OAFlux evaporation products were spatially interpolated on the SMOs product 1/2 degree grid and accumulated in time [mm] over the 7-day period corresponding to the SMOS L4a product. Note: at the time the version V01 of L4 products are generated, no OAFLUX Evaporation data are available after end september Yu, L Global Variations in Oceanic Evaporation ( ): The Role of the Changing Wind Speed. Journal of Climate 20: doi: JCLI Cummulated Precipitation TRMM-3B42 satellite rain rate estimates that we used in the present products are the so-called TRMM and Other Satellites (3B42) products, obtained through the NASA/Giovanni server ( The 3B42 estimates are 3-hourly rain rate at a spatial resolution of 0.25 with spatial extent covering a global belt ( 180 W to 180 E) extending from 50 S to 50 N latitude. The major inputs into the 3B42 algorithm are IR data from geostationary satellites and Passive Microwave data from the TRMM microwave imager (TMI), special sensor microwave imager (SSM/I), Advanced Microwave Sounding Unit (AMSU) and Advanced Microwave Sounding Radiometer-Earth Observing System (AMSR-E). The 1/4 resolution 3-hourly TRMM-3B42 data were re-gridded on a 1/2 resolution grid and the cummulated rain [mm} was evaluated over the 7-day period corresponding to the SMOS L4a product. Important Notice: At the time our products were generated, the 3B42 data were not available after 8 august 2014.

18 Page Cummulated Precipitation CMORPH As a complement to the TRMM3B42, we also include the cummulated rain evaluation based on the CMORPH v1.0 products derived by NOAA. CMORPH (CPC MORPHing technique) produces global precipitation analyses at very high spatial and temporal resolution. This technique uses precipitation estimates that have been derived from low orbiter satellite microwave observations exclusively, and whose features are transported via spatial propagation information that is obtained entirely from geostationary satellite IR data. At present NOAA incorporate precipitation estimates derived from the passive microwaves aboard the DMSP 13, 14 & 15 (SSM/I), the NOAA-15, 16, 17 & 18 (AMSU-B), and AMSR-E and TMI aboard NASA's Aqua and TRMM spacecraft, respectively. These estimates are generated by algorithms of Ferraro (1997) for SSM/I, Ferraro et al. (2000) for AMSU-B and Kummerow et al. (2001) for TMI. Note that this technique is not a precipitation estimation algorithm but a means by which estimates from existing microwave rainfall algorithms can be combined. Therefore, this method is extremely flexible such that any precipitation estimates from any microwave satellite source can be incorporated. With regard to spatial resolution, although the preciptation estimates are available on a grid with a spacing of 8 km (at the equator), the resolution of the individual satellite-derived estimates is coarser than that - more on the order of 12 x 15 km or so. The finer "resolution" is obtained via interpolation. In effect, IR data are used as a means to transport the microwave-derived precipitation features during periods when microwave data are not available at a location. Propagation vector matrices are produced by computing spatial lag correlations on successive images of geostationary satellite IR which are then used to propagate the microwave derived precipitation estimates. This process governs the movement of the

19 Page 19 precipitation features only. At a given location, the shape and intensity of the precipitation features in the intervening half hour periods between microwave scans are determined by performing a time-weighting interpolation between microwave-derived features that have been propagated forward in time from the previous microwave observation and those that have been propagated backward in time from the following microwave scan. NOAA refer to this latter step as "morphing" of the features. For the present CATDS products, we only considered the 3-hourly products at 1/4 degree resolution.the entire CMORPH record (December present) for 3-hourly, 1/4 degree lat/lon resolution can be found at 3-hourly, 1/4 x 1/4 degree ftp://ftp.cpc.ncep.noaa.gov/precip/cmorph_v1.0/raw/ CMORPH estimates cover a global belt ( 180 W to 180 E) extending from 60 S to 60 N latitude and are available for the complete period of the SMOS L4.V01 data (May 2010-Dec 2014). Reference: Joyce, R. J., J. E. Janowiak, P. A. Arkin, and P. Xie, 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution.. J. Hydromet., 5,

20 Page OSCAR zonal and meridional currents Here we used the 1/3 resolution global surface current products from Ocean Surface Current Analyses Realtime (OSCAR) (Bonjean and Lagerloef, 2002; directly calculated from satellite altimetry and ocean vector winds. The OSCAR data processing system calculates sea surface velocities from satellite altimetry (AVISO), vector wind fields (scatterometer winds), as well as from sea surface temperature (Reynolds-Smith) using quasisteady geostrophic, local wind-driven, and thermal wind dynamics. Near real time velocities are calculated on both a 1 x1 and 1/3 x1/3 grid on a ~5 day time base over the global ocean. Surface currents are provided on the OSCAR website ( starting from 1992 along with validations with drifters and moorings. The 1/3 resolution is available for ftp download through ftp://ftp.esr.org/pub/datasets/sfccurrents/thirddegree. The 1/3 resolution 5 day OSCAR data were re-gridded on a 1/2 resolution grid and the 5-day currents fields were linearly interpolated in time on a daily basis. The mean current components provided into the L4a products are then the result of a time averaging over the 7-day period of the SMOS L4a product.

21 Page Wind Stress zonal and meridional components from ASCAT Surface wind stress component included into our products are based on the Advanced SCATterometer (ASCAT) daily data produced and made available at Ifremer/cersat on a resolution (Bentamy and Croize-Fillon 2012) since November The daily field 1/4 fields were averaged at the SMOS L4 products 1/2 resolution and the mean wind stress components over the 7-day period of the SMOS L4a product was evaluated. Bentamy, A., and D. Croizé-Fillon Gridded Surface Wind Fields From Metop/ASCAT Measurements. International Journal Remote Sensing 33: doi: /

22 Page Mixed-Layer Depth For the Mixed Layer Depth (MLD) estimate, we used the monthly 1 X1 MLD available at the International Pacific Research Center/Asia-Pacific Data-Research Center (IPRC/APDRC): see thly_mean.info. Mixed Layer Depth (MLD) is defined here as the depth, on which density increases from 10m to the value equivalent to the temperature drop of 0.2 C. The 1 X1 monthly MLD ields were interpolated in space and time on a 1/2 grid and daily. The MLD provided in the product is the temporal mean of the dailky interpolated MLD over the 7-day period of the SMOS L4a product.

23 Page Salinity at the base of the Mixed Layer The salinity values at the base of the Mixed Layer Depth (MLD) correspond at each grid point to the monthly ISAS value at depth p. p is chosen as the nearest standard depth levels lower than the MLD value. For example, if at a specific longitude/latitude, the MLD is 52 m, we use the ISAS salinity value at depth = 55 m. Recalling that ISAS product are given at 151 standard depth levels between 0 and 2000 m ( :10: :20:2000). The obtained monthly Sd maps are linearly interpolated on a daily basis and then averaged in time to match the SMOS weekly time resolution. Original MLD data are obtained from : thly_mean.info

24 Page Time Interpolated ISAS SSS fields The global reference SSS maps we used to correct the large scale biases in our L4 products are the optimal interpolation fields generated using delayed time quality checked in situ measurements (Argo and ship) by IFREMER/LPO, Laboratoire de physique des oceans using the In Situ Analysis System (ISAS) (D7CA2S0 reanalysis product) (see a method description on Ocean- 252 T-S/Monthly-fields and in (Gaillard et al., 2009)). For the L4aSSS.V01 products, we used the monthly evolving 1/2 x1/2 fields derived from may 2010 to end 2014.

25 Page In Situ/SMOSL4a Match-up DataBase (MDB) files ARGO float observations Ensemble of ARGO float upper level measurements collected during the L4SSS product period (May Dec 2014). Salinity and Temperature measurements from Argo floats are provided by the Coriolis data centre ( We considered only delayed mode ARGO salinity and temperature float data with Quality index quality=1 and 2 observations. We collected the float data weekly over each of the L4 composite product period and performed a co-location with SMOS L4 products The upper ocean salinity and temperature values recorded between 0m and 10m depth are considered as ARGO sea surface salinities and will be referred to as Argo SSS and SST. The following variables and auxilliary data re included into each in situ match-up netcdf file: -latitude of the location where co-localized ARGO floats surfaced -longitude of the location where ARGO floats surfaced -date at which ARGO floats surfaced

26 Page 26 -sss ARGO (the upper in the upper 10 m) -sst ARGO (the upper in the upper 10 m) -depth of the SSS measurement (m) - platform number -Rain from TRMM-3B42 at ARGO float location and date - 7days -long time series of the 3-hourly rain data from TRMM-3B42 colocated atl ARGO float location and prior to and incuding ARGO data, -Wind speed components from ASCAT daily fields interpolated at ARGO float location and date, -7days -long time series of the Daily wind speed components from ASCAT daily fields colocated at ARGO float location and prior to and incuding ARGO data date -SMOS L4 SSS closest in space (within 0.25 radius) from ARGO observation and generated during the week including the float SSS observation Exemple of comparisons of weekly L4 product with ARGO observations. Table 2: Variable Name Dimension and Description for the CECOS L4a MDB ARGO V01 research products

27 Page 27 Variable Name Dimension Description time 1 Central date of the time period over which the SMOS data were combined to generate the composite product. Number of days since :00:00 latitude N_prof Vector of the latitudes of the N_prof ARGO floats that surfaced during the period over which the L4a SSS composite product is derived. Expressed in degrees North from -90. to +90. longitude N_prof Vector of the longitude of the N_prof ARGO floats that surfaced during the period over which the composite L4a SSS product is derived. Expressed in degrees East from to SSS_PRES N_prof Sea Pressure of SSS sampling at ARGO between -10m and surface [Decibar]. Missing Values= sss_at_argo_float N_prof Sea Surface salinity measured by ARGO float [practical salinity scale]. Missing Values= sst_at_argo_float N_prof Sea Surfacetemperature measured by ARGO float [degree C]. Missing Values= SMOS_sss N_prof SMOS L4a Sea Surface salinity colocalized at ARGO float location [practical salinity scale]. Missing Values= ECMWF_sst N_prof ECMWF L4a Sea Surface temperature co-localized at ARGO float location [degree C]. Missing Values= -9999

28 Page 28 latitude_of_smos_obs N_prof Vector of the latitudes of the closest L4a produtc 1/2 grid node from the ARGO float locations that surfaced during the period over which the L4a SSS composite product is derived. Expressed in degrees North from -90. to +90. longitude_of_smos_obs N_prof Vector of the longitudes of the closest L4a produtc 1/2 grid node from the ARGO float locations that surfaced during the period over which the L4a SSS composite product is derived.. Expressed in degrees East from to Date_at_argo_float N_prof Date of the time at which each argo float surfaced. Number of days since :00:00 Distance_to_coasts_at_argo_float N_prof Distance to coasts evaluated from a USGS land mask [kms] PLATEFORM_NUMBER STRING8x N_prof WMO float identifier PSAL N_LEVELSxN_prof Vertical profiles of Salinity at each ARGO float [practical salinity unit]. TEMP N_LEVELSxN_prof Vertical profiles of Temperature at each ARGO float [degree C]. PRES N_LEVELSxN_prof Sea Pressure at each level of each profile [Decibars]

29 Page 29 Ascat_daily_wind_at_ARGO N_prof Co-localized daily 1/4 x1/4 Ascat wind speed at each ARGO float [meter per seconds] TRMM3B42_3hourly_RR_at_ARGO N_prof 3-hourly rain rate from TRMM3B42 co-localized at ARGO [mm/h] Ascat_7_prior_days_wind_at_ARGO N_days_windxN_prof Preceeding 7 days time series of Ascat wind speed Co-localized at each ARGO float location and date from daily 1/4 x1/4 Ascat wind speed [meter per seconds] TRMM3B42_7_prior_days_RR_at_ARGO N_3H_RAINxN_prof Preceeding 7 days 3-hourly time series of TRMM3B42 co-localized rain rate at each ARGO float location and date from TRMM3B42 [millimeter per hour] date_start 1 Start date of the time period over which the SMOS L4a data were considered to generate the composite product date_stop 1 End date of the time period over which the SMOS L4a data were considered to generate the composite product

30 Page Ship ThermoSalinograph Data Spatial Distribution of the GOSUD V3 TSG Delayed-Mode database for May 2010-Dec 2014 Thermo-salinograph data are provided by the GOSUD V3 datasets (see We considered only Delayed Mode data, used only adjusted values and only collected TSG data that exhibit quality flags=1 & 2. We complemented the data set by adding the last processing or reprocessing of the thermosalinometer data from the French research vessels. The later are processed by IFREMER/LPO ( Note that no data from this delayed-mode database is yet available for year In addition, we considered the "research" quality data from the US Shipboard Automated Meteorological and Oceanographic System (SAMOS) initiative. Data are available at Spatial Distribution of the SAMOS TSG "research" quality database for May 2010-Dec 2014

31 Page 31 All Match-Up database files contain the following information: -latitude of the location where the co-localized SMOS/TSG data were acquired, -longitude of the location the co-localized SMOS/TSG data were acquired, -date at which TSG data were acquired, -sss measured by TSG (the upper in the upper 10 m), -sst measured by TSG (the upper in the upper 10 m), -spatially filtered TSG SSS data, using a running median filter of 25km half-width, -spatially TSG SST data, using a running median filter of 25km half-width, -depth of the TSG intake at which SSS & SST measurements are conducted, - platform number, - ship call sign, - Rain from TRMM-3B42 at float location and date, - 7days -long time series of the 3-hourly rain data from TRMM-3B42 colocated at TSG location and prior to and incuding TSG data date, -Wind speed components from ASCAT daily fields interpolated at TSG sample location and date, -7days -long time series of the Daily wind speed components from ASCAT daily fields colocated at TSG sample location and prior to and incuding TSG sample date -SMOS L4 SSS closest in space (within 0.25 radius) from each high resolution TSG observation and generated during the week including the TSG SSS observation Note: the spatially filtered SSS & SST TSG data are obtained for each HR sample along the ship track by searching for all points belonging to the track around that particular sample within a radius of 25 km. The filtered SSS is then obtained by averaging the SSS over these points. Exemple netcdf file: CECOS_MDB_TSG_L4aSSS_0.5deg_ _ _V01.nc Table 3: Variable Name Dimension and Description for the CECOS L4a MDB TSG V01 research products

32 Page 32 Variable Name Dimension Description time 1 Central date of the time period over which the SMOS data were combined to generate the composite product. Number of days since :00:00 Latitude_at_TSG DAYD, N_ship Matrix of the latitudes of the N_ship (number of distinct ships) as function of time along track (DAYD) that measured SSS during the period over which the L4a SSS composite product is derived. Expressed in degrees North from to +90. longitude_at_tsg DAYD, N_ship Matrix of the longitudes of the N_ship (number of distinct ships during the week) as function of time along track (DAYD) that measured SSS during the period over which the L4a SSS composite product is derived. Expressed in degrees East from to PRES DAYD, N_ship Matrix of the Sea Pressure of SSS intake sampling [Decibar] of the N_ship (number of distinct ships during the week) as function of time along track (DAYD) that measured SSS during the period over which the L4a SSS composite product is derived. Missing Values= sss_at_tsg DAYD, N_ship Matrix of the along-track high resolution Sea Surface salinity measured by the N_ship (number of distinct ships during the week) as function of time along track (DAYD) [practical salinity scale].

33 Page 33 Missing Values= sst_at_tsg DAYD, N_ship Matrix of the along-track high resolution Sea Surface temperature measured by the N_ship (number of distinct ships during the week) as function of time along track (DAYD) [degree C]. Missing Values= Filtered_sss_at_TSG DAYD, N_ship Matrix of the along-track spatially filtered at 50km resolution Sea Surface salinity measured by the N_ship (number of distinct ships during the week) as function of time along track (DAYD) [practical salinity scale]. Missing Values= Filtered_sst_at_TSG DAYD, N_ship Matrix of the along-track spatially filtered at 50km resolution Sea Surface temperature measured by the N_ship (number of distinct ships during the week) as function of time along track (DAYD) [degree C]. Missing Values= SMOS_sss_at_TSG DAYD, N_ship Matrix of the SMOS L4a Sea Surface salinity co-localized at each TSG location along track for N_ship (number of distinct ships during the week) as function of time along track (DAYD). [practical salinity scale]. Missing Values= ECMWF_sst_at_TSG DAYD, N_ship Matrix of the ECMWF L4a Sea Surface temperature co-localized at each TSG location along track for N_ship (number of distinct ships during the week) as function

34 Page 34 of time along track (DAYD). [degree C]. Missing Values= latitude_of_closest_smos_obs DAYD, N_ship Matrix of the latitudes of the closest L4a SSS produtc 1/2 grid node from the TSG locations for N_ship (number of distinct ships during the week) as function of time along track (DAYD). Expressed in degrees North from -90. to +90. longitude_of_closest_smos_obs DAYD, N_ship Matrix of the longitudes of the closest L4a SSS produtc 1/2 grid node from the TSG locations for N_ship (number of distinct ships during the week) as function of time along track (DAYD). Expressed in degrees East from to Date_at_TSG DAYD, N_ship Matrix of the time at which each TSG sampled was measured. This time is provided for N_ship (number of distinct ships during the week) as function of the number of time samples along track (DAYD). Number of days since :00: Distance_to_coasts_at_TSG DAYD, N_ship Distance to coasts evaluated for each TSG sample from a USGS land mask [kms]. PLATEFORM_NAME N_ship x STRING19 Ship name SHIP_CALL_SIGN N_ship x STRING6 ship call sign Ascat_daily_wind_at_TSG DAYD, N_ship Co-localized daily 1/4 x1/4 Ascat wind speed at each TSG track

35 Page 35 location & date. [meter per seconds] TRMM3B42_3hourly_RR_at_TSG DAYD, N_ship 3-hourly rain rate from TRMM3B42 co-localized at each TSG track location & date [mm/h] Ascat_7_prior_days_wind_at_TSG TRMM3B42_7_prior_days_RR_at_TSG N_ship x N_DAYS_WINDx DAYD N_ship x N_3H_RAIN xdayd Preceeding 7 days time series of Ascat wind speed Co-localized at each TSG track location and date from daily 1/4 x1/4 Ascat wind speed [meter per seconds] Preceeding 7 days 3-hourly time series of TRMM3B42 co-localized rain rate at each TSG track location & date [millimeter per hour] date_start 1 Start date of the time period over which the SMOS L4a data were considered to generate the composite product date_stop 1 End date of the time period over which the SMOS L4a data were considered to generate the composite product

36 Page Surface drifter Data Spatial Distribution of the "research" quality database for May 2010-Dec 2014 Surface Drifter data are provided by the LOCEAN datasets (see We considered only validated data. Mathc-Up databse files between SMOS and drifter observations include the following informations: -latitude of the location where the co-localized SMOS/drifter data were acquired, -longitude of the location the co-localizedsmos/drifter data were acquired, -date at which drifter data were aquired, -sss measured at drifter (the upper in the upper 10 m), -sst measured at drifter (the upper in the upper 10 m), -spatially filtered drifter SSS data, using a running median filter of 25km half-wdith, -spatially TSGdrifter data, using a running median filter of 25km half-wdith, - platform number,

37 Page 37 - instrument name, - Rain from TRMM-3B42 at float location and date, - 7days -long time series of the 3-hourly rain data from TRMM-3B42 colocated at drifter location and prior to and incuding TSG data date, -Wind speed components from ASCAT daily fields interpolated at drifter sample location and date, -7days -long time series of the Daily wind speed components from ASCAT daily fields colocated at drifter sample location and prior to and incuding drifter sample date -SMOS L4 SSS closest in space (within 0.25 radius) from drifter observation and generated during the week including the drifter SSS observation Table 4: Variable Name Dimension and Description for the CECOS L4a MDB drifter V01 research products Variable Name Dimension Description time 1 Central date of the time period over which the SMOS data were combined to generate the composite product. Number of days since :00:00 Latitude_at_drift DAYD, N_drifters Matrix of the latitudes of the N_drifters (number of distinct drifters) as function of time along track (DAYD) that measured SSS during the period over which the L4a SSS composite product is derived. Expressed in degrees North from -90. to +90. longitude_at_drift DAYD, N_drifters Matrix of the longitudes of the N_drifters (number of distinct drifters during the week) as function of time along track (DAYD) that measured SSS during the period over which the L4a SSS composite product is derived. Expressed in degrees East from to PRES DAYD, N_drifters Matrix of the Sea Pressure of SSS

38 Page 38 intake sampling [Decibar] of the N_drifters (number of distinct drifters during the week) as function of time along track (DAYD) that measured SSS during the period over which the L4a SSS composite product is derived. Missing Values= sss_at_drift DAYD, N_drifters Matrix of the along-track high resolution Sea Surface salinity measured by the N_drifters (number of distinct drifters during the week) as function of time along track (DAYD) [practical salinity scale]. Missing Values= sst_at_drift DAYD, N_drifters Matrix of the along-track high resolution Sea Surface temperature measured by the N_drifters (number of distinct drifters during the week) as function of time along track (DAYD) [degree C]. Missing Values= Filtered_sss_at_DRIFT DAYD, N_drifters Matrix of the along-track spatially filtered at 50km resolution Sea Surface salinity measured by the N_drifters (number of distinct drifters during the week) as function of time along track (DAYD) [practical salinity scale]. Missing Values= Filtered_sst_at_DRIFT DAYD, N_drifters Matrix of the along-track spatially filtered at 50km resolution Sea Surface temperature measured by the N_drifters (number of distinct drifters during the week) as function of time along track (DAYD)

39 Page 39 [degree C]. Missing Values= SMOS_sss_at_DRIFT DAYD, N_drifters Matrix of the SMOS L4a Sea Surface salinity co-localized at each DRIFT location along track for N_drifters (number of distinct drifters during the week) as function of time along track (DAYD). [practical salinity scale]. Missing Values= ECMWF_sst_at_DRIFT DAYD, N_drifters Matrix of the ECMWF L4a Sea Surface temperature co-localized at each DRIFT location along track for N_drifters (number of distinct drifters during the week) as function of time along track (DAYD). [degree C]. Missing Values= latitude_of_closest_smos_obs DAYD, N_drifters Matrix of the latitudes of the closest L4a SSS produtc 1/2 grid node from the DRIFT locations for N_drifters (number of distinct drifters during the week) as function of time along track (DAYD). Expressed in degrees North from -90. to +90. longitude_of_closest_smos_obs DAYD, N_drifters Matrix of the longitudes of the closest L4a SSS produtc 1/2 grid node from the DRIFT locations for N_drifters (number of distinct drifters during the week) as function of time along track (DAYD). Expressed in degrees East from to Date_at_DRIFT DAYD, N_drifters Matrix of the time at which each DRIFT sampled was measured. This time is provided for N_drifters

40 Page 40 (number of distinct drifters during the week) as function of the number of time samples along track (DAYD). Number of days since :00: Distance_to_coasts_at_DRIFT DAYD, N_drifters Distance to coasts evaluated for each DRIFT sample from a USGS land mask [kms]. PLATEFORM_NAME N_drifters x STRING19 Ship name SHIP_CALL_SIGN N_drifters x STRING6 ship call sign Ascat_daily_wind_at_DRIFT DAYD, N_drifters Co-localized daily 1/4 x1/4 Ascat wind speed at each DRIFT track location & date. [meter per seconds] TRMM3B42_3hourly_RR_at_DRIFT DAYD, N_drifters 3-hourly rain rate from TRMM3B42 co-localized at each DRIFT track location & date [mm/h] Ascat_7_prior_days_wind_at_DRIFT TRMM3B42_7_prior_days_RR_at_DRIFT N_drifters x N_DAYS_WINDx DAYD N_drifters x N_3H_RAIN xdayd Preceeding 7 days time series of Ascat wind speed Co-localized at each DRIFT track location and date from daily 1/4 x1/4 Ascat wind speed [meter per seconds] Preceeding 7 days 3-hourly time series of TRMM3B42 co-localized rain rate at each DRIFT track location & date [millimeter per

41 Page 41 hour] date_start 1 Start date of the time period over which the SMOS L4a data were considered to generate the composite product date_stop 1 End date of the time period over which the SMOS L4a data were considered to generate the composite product Global tropical Moored Buoy Array data Surface time series of salinity from the Global Tropical Moored Buoy Array were also collected and colocalized with SMOS L4 data. The Global Tropical Moored Buoy Array is a multi-national effort to provide data in real-time for climate research and forecasting. Major components include the TAO/TRITON array in the Pacific, PIRATA in the Atlantic, and RAMA in the Indian Ocean. The Global Tropical Moored Buoy Array is a contribution to the Global Ocean Observing System (GOOS), Global Climate Observing System (GCOS), and the Global Earth Observing System of Systems (GEOSS). Data can be accessed here: Data collected within TAO/TRITON, PIRATA and RAMA comes primarily from ATLAS and TRITON moorings. These two mooring systems are functionally equivalent in terms of sensors, sample rates, and data quality.

42 Page 42 We selected data measured at 1 meter depth and standard quality (pre-deployement calibration applied) & highest quality (pre/post calibration agree) We generated one Match-up DataBase (MDB) files per tropical basin (TAO/TRITON, PIRATA and RAMA) for the whole period May 2010-Dec Each MDB file includes for each mooring of the basin: -Mooring latitude -Mooring longitude -Date of SSS measurement -Depth of mooring salinity measurement -Depth of mooring temperature measurement -Daily mooring SSS (Quality Flag=1 (Highest quality),2 (standard)) -Daily mooring SST Quality Flag=1 (Highest quality),2 (standard)) -Daily mooring wind speed Quality Flag=1 (Highest quality),2 (standard)) -Daily mooring rain rate (Quality Flag=1 (Highest quality),2 (standard)) -Daily mooring salinity profile (Quality Flag=1 (Highest quality),2 (standard)) -Daily mooring temperature profile Quality Flag=1 (Highest quality),2 (standard)) -Weekly time averaged mooring SSS -Weekly time averaged mooring SSS quality flag" -Weekly time averaged mooring SST" -Weekly time averaged mooring wind speed -Weekly accummulated precipitation" -Weekly time averaged mooring salinity (profiles) -Weekly time averaged mooring temperature (profiles) -SMOS L4 SSS at mooring -ECMWF L4 SST at mooring -ECMWF L4 wind speed at mooring -TRMM 3B42 (v7) weekly accummulated precipitation at mooring

43 Page 43 Note that SMOS L4 SSS, ECMWF SST wind speed and TRMM rain were obtained at the moorings by bilinearly interpolating in space the 0.5 data at the mooring location Table 5 Variable Name Dimension and Description for the CECOS L4a MDB mooringsv01 research products Variable Name Dimension Description time 1 Central date of the time period over which the SMOS data were combined to generate the composite product. Number of days since :00:00 daily_date DAYD Vector of the Number of days since :00:00 at which daily SSS was measured at the moorings. Valid for all moorings. latitude N_moorings Vector of the latitudes of the N_mooring for a given basin (number of distinct TAO, or Pirata, or RAMA moorings) Expressed in degrees North from to +90. longitude N_moorings Vector of the longitudes of the N_mooring for a given basin (number of distinct moorings for TAO, or Pirata, or RAMA) Expressed in degrees East from to depth_s DEPTH_S Vector of Depths at which salinity measurements were performed. Valid for all moorings. [meter] Missing Values= depth_t DEPTH_T Vector of Depths at which temperature measurements were performed. Valid for all moorings.

44 Page 44 [meter] Missing Values= Daily_moooring_sss N_moorings, DAYD Matrix of the daily-averaged times series (DAYD) of in situ SSS as measured by the N_mooring (number of distinct moorings for a given basin [practical salinity scale]. Missing Values= Daily_moooring_sst N_moorings, DAYD Matrix of the daily-averaged times series (DAYD) of in situ SST as measured by the N_mooring (number of distinct moorings for a given basin [degrees C]. Missing Values= Daily_moooring_wind_speed N_moorings, DAYD Matrix of the daily-averaged times series (DAYD) of in situ wind speed as measured by the N_mooring (number of distinct moorings for a given basin [m/s]. Missing Values= Daily_moooring_rain_rate N_moorings, DAYD Matrix of the daily-averaged times series (DAYD) of in situ rain rate as measured by the N_mooring (number of distinct moorings for a given basin [mm/h]. Missing Values= Daily_moooring_s N_moorings, DAYD, DEPTH_S Matrix of the daily-averaged times series (DAYD) of salinity profiles as measured at the N_mooring (number of distinct moorings for a given basin) and at the depths DEPTH_S [practical salinity scale]. Missing Values= -9999

45 Page 45 Daily_moooring_t N_moorings, DAYD, DEPTH_T Matrix of the daily-averaged times series (DAYD) of tempearture profiles as measured by the N_mooring (number of distinct moorings for a given basin) and at the depth DEPTH_T [degrees C]. Missing Values= Weekly_moooring_sss N_moorings, time Matrix of the weekly-averaged times series (time) of in situ SSS as measured by the N_mooring (number of distinct moorings for a given basin [practical salinity scale]. Missing Values= Weekly_moooring_sst N_moorings, time Matrix of the weekly-averaged times series (time) of in situ SST as measured by the N_mooring (number of distinct moorings for a given basin [degrees C]. Missing Values= Weekly_moooring_wind_speed N_moorings, time Matrix of the weekly-averaged times series (time) of in situ wind speed as measured by the N_mooring (number of distinct moorings for a given basin [m/s]. Missing Values= weekly_moooring_accumulated_rain N_moorings, time Matrix of the weekly-cummulated times series (time) of rain as measured in situ by the N_mooring (number of distinct moorings for a given basin [mm]. Missing Values= Weekly_moooring_s N_moorings, time, DEPTH_S Matrix of the weeky-averaged times series (time) of salinity profiles as measured at the N_mooring (number of distinct

46 Page 46 Weekly_moooring_t N_moorings, time, DEPTH_T moorings for a given basin) and at the depths DEPTH_S [practical salinity scale]. Missing Values= Matrix of the weekly-averaged times series (time) of tempearture profiles as measured by the N_mooring (number of distinct moorings for a given basin) and at the depth DEPTH_T [degrees C]. Missing Values= Weekly_SMOS_SSS N_moorings, time Maxtrix of the weekly L4a SMOS SSS time series interpolated at each of the N_moorings locations. [practical salinity unit]. Missing Values= Weekly_ECMWF_SST N_moorings, time Maxtrix of the weekly L4a ECMWF SST time series interpolated at each of the N_moorings locations. [degree celcius]. Missing Values= Weekly_ECMWF_wind_speed N_moorings, time Maxtrix of the weekly L4a ECMWF wind speed time series interpolated at each of the N_moorings locations. [m/s]. Missing Values= Weekly_TRMM3B42_accumulated_rain N_moorings, time Matrix of the weekly-cummulated times series (time) from TRMM3B42 co-localized at the N_mooring (number of distinct moorings for a given basin [mm]. Missing Values= date_start 1 Start date of the time period over which the SMOS L4a data were considered to generate the

47 Page 47 composite product date_stop 1 End date of the time period over which the SMOS L4a data were considered to generate the composite product Example of a SSS time series at TAO/TRITTON Mooring in the east equatorial pacific (2 N, 156 E): the plot shows the mooring 1m depth SSS daily (blue), weekly (black) and the SMOS L4 SSS (red) time series Southern Ocean SSS from Seals

48 Page 48 The instrumentation of southern elephant seals with satellite-linked CTD tags has offered unique temporal and spatial coverage of the Southern Oceans since This includes extensive data from the Antarctic continental slope and shelf regions during the winter months, which is outside the conventional areas of Argo autonomous floats and ship-based studies. This landmark dataset of around 75,000 temperature and salinity profiles from E, concentrated on the sector between the Kerguelen Islands and Prydz Bay, continues to grow through the coordinated efforts of French and Australian marine research teams. The seal data (MEOP-CTD in-situ data collection) are quality controlled and calibrated using delayed-mode techniques involving comparisons with other existing profiles as well as cross-comparisons similar to established protocols within the Argo community, with a resulting accuracy of ±0.03 C in temperature and ±0.05 in salinity or better. The seal SSS dataset is acessible at the Coriolis data center ( Ocean/Marine-Mammals). The upper ocean salinity and temperature values recorded between 0m and 10m depth by the seals are considered asseal sea surface salinities and will be referred to as SEAL SSS and SST in the CEC Match-ups.

49 Page 49 The following variables and auxilliary data re included into each in situ match-up netcdf file: -latitude of the location where co-localized SEAL floats surfaced -longitude of the location where SEAL floats surfaced -date at which SEAL floats surfaced -sss SEAL (the upper in the upper 10 m) -sst SEAL (the upper in the upper 10 m) -depth of the SSS measurement (m) - platform number -Rain from CMORPH at SEAL float location and date - 7days -long time series of the 3-hourly rain data from CMORPH colocated atl SEAL float location and prior to and incuding SEAL data, -Wind speed components from ASCAT daily fields interpolated at SEAL float location and date, -7days -long time series of the Daily wind speed components from ASCAT daily fields colocated at SEAL float location and prior to and incuding SEAL data date -SMOS L4 SSS closest in space (within 0.25 radius) from SEAL observation and generated during the week including the float SSS observation Roquet, F. et al. A Southern Indian Ocean database of hydrographic profiles obtained with instrumented elephant seals. Sci. Data 1: doi: /sdata (2014). Table 6: Variable Name Dimension and Description for the CECOS L4a MDB SEAL V01 research products Variable Name Dimension Description time 1 Central date of the time period over which the SMOS data were combined to generate the composite product. Number of days since :00:00 latitude N_prof Vector of the latitudes of the N_prof SEAL floats that surfaced during the period over which the L4a SSS composite product is derived. Expressed in degrees

50 Page 50 North from -90. to +90. longitude N_prof Vector of the longitude of the N_prof SEAL floats that surfaced during the period over which the composite L4a SSS product is derived. Expressed in degrees East from to SSS_PRES N_prof Sea Pressure of SSS sampling at SEAL between -10m and surface [Decibar]. Missing Values= sss_at_seal_float N_prof Sea Surface salinity measured by SEAL float [practical salinity scale]. Missing Values= sst_at_seal_float N_prof Sea Surfacetemperature measured by SEAL float [degree C]. Missing Values= SMOS_sss N_prof SMOS L4a Sea Surface salinity colocalized at SEAL float location [practical salinity scale]. Missing Values= ECMWF_sst N_prof ECMWF L4a Sea Surface temperature co-localized at SEAL float location [degree C]. Missing Values= latitude_of_smos_obs N_prof Vector of the latitudes of the closest L4a produtc 1/2 grid node from the SEAL float locations that surfaced during the period over which the L4a SSS composite product is derived. Expressed in degrees North from -90. to +90. longitude_of_smos_obs N_prof Vector of the longitudes of the closest L4a produtc 1/2 grid node from the SEAL float locations that surfaced during the period over which the L4a SSS composite

51 Page 51 product is derived.. Expressed in degrees East from to Date_at_seal_float N_prof Date of the time at which each seal float surfaced. Number of days since :00:00 Distance_to_coasts_at_seal_float N_prof Distance to coasts evaluated from a USGS land mask [kms] PLATEFORM_NUMBER STRING8x N_prof WMO float identifier PSAL N_LEVELSxN_prof Vertical profiles of Salinity at each SEAL float [practical salinity unit]. TEMP N_LEVELSxN_prof Vertical profiles of Temperature at each SEAL float [degree C]. PRES N_LEVELSxN_prof Sea Pressure at each level of each profile [Decibars] Ascat_daily_wind_at_SEAL N_prof Co-localized daily 1/4 x1/4 Ascat wind speed at each SEAL float [meter per seconds] TRMM3B42_3hourly_RR_at_SEAL N_prof 3-hourly rain rate from TRMM3B42 co-localized at SEAL [mm/h] Ascat_7_prior_days_wind_at_SEAL N_days_windxN_prof Preceeding 7 days time series of Ascat wind speed Co-localized at each SEAL float location and date from daily 1/4 x1/4 Ascat wind

52 Page 52 speed [meter per seconds] TRMM3B42_7_prior_days_RR_at_SEAL N_3H_RAINxN_prof Preceeding 7 days 3-hourly time series of TRMM3B42 co-localized rain rate at each SEAL float location and date from TRMM3B42 [millimeter per hour] date_start 1 Start date of the time period over which the SMOS L4a data were considered to generate the composite product date_stop 1 End date of the time period over which the SMOS L4a data were considered to generate the composite product Match-Up Database netcdf Files Naming conventions The generic filename convention for the weekly MDB L4a products is given as defined below: CECOS_MDB_sensor_0.5deg_YYYY.DD_YYYY.DD_V01.nc Colored symbols indicate digital variables defined as follows: The first variable sensor indicate the in situ sensor types: Sensor="ARGO" for Coriolis DM ARGO floats Match-Ups Sensor="TSG" for GOSUDV3 TSG DM Match-Ups Sensor="TSG_SAMOS" for SAMOS TSG Match-Ups Sensor="TAO" for NOAA/AOML TAO moorings Match-Ups Sensor="PIRATA" for NOAA/AOML PIRATA moorings Match-Ups Sensor="RAMA" for NOAA/AOML RAMA moorings Match-Ups Sensor="drifter" for LOCEAN surface drifter Match-Ups Sensor="seal" for Coriolis DM sea seals CTD Match-Ups

53 Page 53 The second YYYY (4 digits) and third DD (3 digits) variables indicate the year and the ordinal number of the day in year corresponding to the start date of the time period over which the SMOS weekly data were considered to generate the composite product. The fourth YYYY (4 digits) and fifth DD (3 digits) variables indicate the year and the ordinal number of the day in year corresponding to the end date of the time period over which the SMOS weekly data were considered to generate the composite product. The last variable 01 (2 digits) indicate the product processing version. Example: for the ARGO MDB product at 0.5 degree resolution generated from 27 aug 2010 (day of year=239) to 2 of Sep 2010 (day of year=245) using the processing version 1, the file name is: CECOS_MDB_ARGO_0.5deg_ _ _V01.nc

54 Page Level 4b Product content The CATDS/CEC-OS SMOS Level 4b Version 1 Sea Surface Density (SSD) research products are weekly (7 days) composite of satellite sea surface denisty at 50 km resolution. The products coverage is May 2010-December They include some useful other variables to scientifically exploit satellite Sea surface density fields: SSS, SST, AVISO mean sea level anomaly, ekman+geostrophic OSCAR currents and wind stress components. Table 7: Variable Name Dimension and Description for the CECOS L4b V01 research products Variable Name Dimension Description time 1 Central date of the time period over which the SMOS data were combined to generate the composite product. Number of days since :00:00 latitude Nlat Vector of the latitude of the grid nodes over which the composite product is derived. Expressed in degrees North from -90. to +90. longitude nlon Vector of the longitude of the grid nodes over which the composite product is derived. Expressed in degrees East from to sss nlat nlon Gridded Sea Surface Salinity from SMOS [Practical Salinity Scale]. Missing Values= sst nlat nlon Gridded Sea Surface Temperature colocated at SMOS pixels from ECMWF forecasts [Kelvins]. Missing Values= SLA nlat nlon AVISO daily 1/4 x1/4 MSLA averaged in space and time to match the L4a SSS 1/2 Grid and weeks [meter]. Missing Values= RFI_stat nlat nlon Gridded percentage for Radio- Frequency Interferences occurence within the brighthness temperature data set used for SSS product generation at a

55 Page 55 given pixel [%]. Missing Values= Zonal_component_surface_currents nlat nlon Gridded OSCAR surface current (Ekman+Geostrophic) zonal component interpolated in space and time at SMOS pixels [m/s]. The 1/3 resolution 5 day OSCAR data were re-gridded on a 1/2 resolution grid and the 5-day currents fields were linearly interpolated in time on a daily basis. The mean current components provided into the L4a products are then the result of a time averaging over the 7-day period of the SMOS L4a product. Meridional_component_surface_currents nlat nlon Gridded OSCAR surface current (Ekman+Geostrophic) meridional component interpolated in space and time at SMOS pixels [m/s]. Missing Values= Sea_Surface_Density nlat nlon Weekly composite of Sea surface density deduced from the SMOS L4aSSS and ECMWF L4a SST[kg/m 3 ] surface_downward_northward_stress nlat nlon Surface wind stress meridional component included into our products are based on the Advanced SCATterometer (ASCAT) daily data produced and made available at Ifremer/cersat on a resolution (Bentamy and Croize- Fillon 2012) since November surface_downward_eastward_stress nlat nlon Surface wind stress zonal component included into our products are based on the Advanced SCATterometer (ASCAT)

56 Page 56 daily data produced and made available at Ifremer/cersat on a resolution (Bentamy and Croize- Fillon 2012) since November date_start 1 Start date of the time period over which the SMOS data were considered to generate the composite product date_stop 1 End date of the time period over which the SMOS data were considered to generate the composite product Convention for the L4b Netcdf files The generic filename convention for the weekly composite L4b products is given as defined below: CECOS_SMOS_L4bdens_0.5deg_YYYY.DD_YYYY.DD_V01.nc Colored symbols indicate digital variables defined as follows: The first YYYY (4 digits) and second DD (3 digits) variables indicate the year and the ordinal number of the day in year corresponding to the start date of the time period over which the SMOS weekly data were considered to generate the composite product. The third YYYY (4 digits) and fourth DD (3 digits) variables indicate the year and the ordinal number of the day in year corresponding to the end date of the time period over which the SMOS weekly data were considered to generate the composite product. The last variable 01 (2 digits) indicate the product processing version. Example: for the weekly composite product at 0.5 degree resolution generated from 27 aug 2010 (day of year=239) to 2 of Sep 2010 (day of year=245) using the processing version 1, the file name is: CECOS_SMOS_L4bdens_0.5deg_ _ _V01.nc In the following we describe in more detail the content of the products Sea Surface Density The L4b products include weekly maps of sea surface density evaluated from SMOS L4a SSS and SMOS/ECMWF L4a SST. Satellite Sea Surface Density (SSD) is calculated using the appropiate thermal expansion coefficient and the appropriate saline contraction coefficient of seawater from Absolute Salinity and Conservative Temperature. We used the computationally-efficient 48-term expression for density in terms of absolute salinity (SA), conservative temperature (CT) and sea presure p (IOC et al., 2010).

57 Page 57 IOC, SCOR and IAPSO, 2010: The international thermodynamic equation of seawater : Calculation and use of thermodynamic properties.intergovernmental Oceanographic Commission, Manuals and Guides No. 56, UNESCO (English), 196 pp. We used the matlab code available from the TEOS-10 web site. The software is available from Example of weekly L4b sea surface density SSD weekly composite Mean Sea Level Anomaly Exemple of Mean Sea Level anomaly Composite corresponding to a SMOS L4a SSS product period.

58 Page 58 We include weekly averaged Sea Level Anomalies in the L4b products. These are derived from Delayed-Time merged Global Ocean Gridded Sea Level Anomalies SSALTO/Duacs L4 products. These correspond to sea surface height above Mean Sea Surface products from multi-satellite observations over Global Ocean. The orginial products are daily at 1/4 resolution. The later were averaged in space and time to be gridded on the same grid than the L4aSSS & SST products. See Other variables in L4b products In Level 4b, we aslo reproduced some of the variables already included into the Level 4a products. These include SMOS L4a SSS, ECMWF L4a SST, OSCAR current and Ascat daily wind stress components. The reader is referred to 8.4.2, & for details.

59 Page Level 4c Product content The CATDS/CEC-OS SMOS Level 4c Version 1 anomalies research products are weekly (7 days) composite of the anomalies of an ensemble of geophysical variables with respect their mean seasonnal cycle evaluated during the complete SMOS L4a product period (May 2010-dec 2014) and at 50 km resolution. The products coverage is May 2010-December They include some useful variables to scientifically exploit satellite Sea surface salinity anomaly fields: SSSA, SSTA, AVISO mean sea level anomaly, precipitation and evaporation anaomalies, ekman+geostrophic OSCAR currents anomalies, and wind stress components anomalies Methodology to evaluate anomalies Figure: Top: example of weekly L4a SSS time series (blue) at the mouth of the amazon river (4.75 N, W) and its associated Mean Annual Cycle (black curve). Bottom: corresponding times series of the L4c SSS anomaly at that particular location. Anomalies of all variables (except for the mean sea level anomalies) included in L4c products are estimated by removing from each variable fields, the locally estimated in space and time Mean Annual Cycle contribution (MAC) :

60 Page 60 var_anomaly(lat,lon,t)=var(lat,lon,t)-mac(lat,lon,t) where (lat,lon) are geographic coordinates and t is central time of the weekly products. The Mean Annual Cycle contribution for a given variable is obtained by averaging the ensemble of observations at a given grid node (defined by lat & lon) from the same weeks of all year between 2010 and These anomalies are therefore representative of interannual variability around the mean seasonal cycle of each variable. For fields with original temporal resolution coarser than weekly (e.g. ISAS is originally provided monthly), a linear interpolation in time of the monthly anomalies was performed Level 4c product content Table 9: Variable Name Dimension and Description for the CECOS L4c V01 research products Variable Name Dimension Description time 1 Central date of the time period over which the SMOS data were combined to generate the composite product. Number of days since :00:00 latitude nlat Vector of the latitude of the grid nodes over which the composite product is derived. Expressed in degrees North from -90. to +90. longitude nlon Vector of the longitude of the grid nodes over which the composite product is derived. Expressed in degrees East from to SSSA_SMOS nlat nlon Gridded Sea Surface Salinity anomaly from SMOS [Practical Salinity Scale]. Missing Values= SSTA_ECMWF nlat nlon Gridded Sea Surface Temperature anomaly colocated at SMOS pixels from ECMWF forecasts [degree celsius]. Missing Values= SSSA_ISAS nlat nlon Gridded Sea Surface Salinity anomaly from ISAS [Practical Salinity Scale]. Missing Values= -9999

61 Page 61 SSTA_ISAS nlat nlon Gridded Sea Surface Temperature anomaly from ISAS [degree Celsius]. Missing Values= SLA nlat nlon AVISO daily 1/4 x1/4 MSLA averaged in space and time to match the L4c SSSA 1/2 Grid and weeks [meter]. Missing Values= RFI_stat nlat nlon Gridded percentage for Radio- Frequency Interferences occurence within the brighthness temperature data set used for SSS product generation at a given pixel [%]. Missing Values= OAFLux_accumulated_Evaporation_anomaly nlat nlon Gridded Weekly anomaly of cummulated Evaporation as estimated from OAFlux [mm] Missing values=-9999 Zonal_component_surface_currents_anomalies nlat nlon Gridded OSCAR surface current (Ekman+Geostrophic) zonal component anomalies interpolated in space and time at SMOS pixels [m/s]. The 1/3 resolution 5 day OSCAR data were re-gridded on a 1/2 resolution grid and the 5-day currents fields were linearly interpolated in time on a daily basis. The mean current components provided into the L4a products are then the result of a time averaging over the 7-day period of the SMOS L4a product. Meridional_component_surface_currents_anomaly nlat nlon Gridded OSCAR surface current (Ekman+Geostrophic) meridional component anomalies interpolated in space and time at SMOS pixels [m/s]. Missing Values= -9999

62 Page 62 Sea_Surface_Density_anomalies nlat nlon Weekly composite of Sea surface density anomalies deduced from the SMOS L4aSSS and ECMWF L4a SST[kg/m 3 ] surface_downward_northward_stress_anomaly nlat nlon Surface wind stress meridional component anomalies included into our products are based on the Advanced SCATterometer (ASCAT) daily data produced and made available at Ifremer/cersat on a resolution (Bentamy and Croize-Fillon 2012) since November surface_downward_eastward_stress_anomaly nlat nlon Surface wind stress zonal component anomaly included into our products are based on the Advanced SCATterometer (ASCAT) daily data produced and made available at Ifremer/cersat on a resolution (Bentamy and Croize- Fillon 2012) since November Salinity_anomaly_MLD_base nlat nlon Anomalies of the salinity values at the base of the Mixed Layer Depth (MLD) [pss] correspond at each grid point to the temporally interpolated monthly ISAS value at depth p. p is chosen as the nearest standard depth levels lower than the MLD value. CMORPH_accumulated_rain_anomaly nlat nlon Cummulated rain falls anomaly [mm] from CMORPH products over the period of time of each weekly L4SSS product and avegared on the LaSSS 50

63 Page 63 km grid. CMORPH estimates cover a global belt ( 180 W to 180 E) extending from 60 S to 60 N latitude and are available for the complete period of the SMOS L4.V01 data TRMM3B42_accumulated_rain_anomaly nlat nlon Cummulated rain falls anomaly [mm] from TRMM3B42 products over the period of time of each weekly L4SSS product and avegared on the LaSSS 50 km grid. Mixed_Layer_Depth_anomaly nlat nlon Mixed-Layer depth anomaly estimated from IPC/APDRC. The 1 X1 monthly MLD orignial fields were interpolated in space and time on a 1/2 grid and daily.the MLD provided in the product is the temporal mean of the daily interpolated MLD over the 7-day period of the SMOS L4a product. date_start 1 Start date of the time period over which the SMOS data were considered to generate the composite product date_stop 1 End date of the time period over which the SMOS data were considered to generate the composite product Convention for the L4c Netcdf files The generic filename convention for the weekly composite L4b products is given as defined below: CECOS_SMOS_L4cano_0.5deg_YYYY.DD_YYYY.DD_V01.nc Colored symbols indicate digital variables defined as follows:

64 Page 64 The first YYYY (4 digits) and second DD (3 digits) variables indicate the year and the ordinal number of the day in year corresponding to the start date of the time period over which the SMOS weekly data were considered to generate the composite product. The third YYYY (4 digits) and fourth DD (3 digits) variables indicate the year and the ordinal number of the day in year corresponding to the end date of the time period over which the SMOS weekly data were considered to generate the composite product. The last variable 01 (2 digits) indicate the product processing version. Example: for the weekly composite anomaly products at 0.5 degree resolution generated from 27 aug 2010 (day of year=239) to 2 of Sep 2010 (day of year=245) using the processing version 1, the file name is: CECOS_SMOS_L4cano_0.5deg_ _ _V01.nc

65 Page Data citation The following example show how to cite the use of these CATDS L4 reserach product data sets in a publication. List the data set title, the producing center, the year of data set release, the version number, and the dates of the data you used (for example, May to June 2014): "The SMOS L4 data were obtained from the Ocean Salinity Expertise Center (CECOS) of the CNES- IFREMER Centre Aval de Traitemenent des Donnees SMOS (CATDS), at IFREMER, Plouzane (France). V01, [list the dates of the data used]." In addition, in any publication using the other variables than SSS included into our products, all users shall aknowledge the data they used as follows: -Surface current components are based on the 1/3 resolution global surface current products from Ocean Surface Current Analyses Real time (OSCAR) (Bonjean and Lagerloef 2002; noaa.gov), as processed by CATDS/CECOS" -The global ocean evaporation products were provided by the WHOI OAFlux project ( funded by the NOAA Climate Observations and Monitoring (COM) program. -Satellite TRMM rain rate estimates that we used in the present study are based on the so-called TRMM and Other Satellites (3B42) products, obtained through the NASA/Giovanni server ( -Satellite CMORPH rain rate estimates that we used in the present study are based on National Center for Atmospheric Research Staff (Eds) datasets. "The Climate Data Guide: CMORPH (CPC MORPHing technique): High resolution precipitation (60S-60N)." Retrieved from - -The 3-D monthly fields of in situ OI temperature and salinity in NetCdf format can be found at the following DOI reference: Fabienne Gaillard (2015). ISAS-13 temperature and salinity gridded fields. Pôle Océan. -Surface wind stress component included into our products are based on the Advanced SCATterometer (ASCAT) daily data produced and made available at Ifremer/cersat on a 0.25 / 0.25 resolution (Bentamy and Croize-Fillon ) since November Bentamy, A., and D. Croizé-Fillon Gridded Surface Wind Fields From Metop/ASCAT Measurements. International Journal Remote Sensing 33: doi: / For the Mixed Layer Depth (MLD) estimate, we used the monthly 1 X1 MLD available at the International Pacific Research Center/Asia-Pacific Data-Research Center (IPRC/APDRC): _monthly_mean.info Salinity measurements from Argo floats, GOSUD TSG and Seals data are provided by the Coriolis data center ( We acknowledge the use of freely available Argo data collected by the International Argo Project and the national programs that contribute to it. We thank the GOSUD Project ( for providing free access to the TSG data.

66 Page 66 Salinity measurements from SAMOS TSG are provided by the US Shipboard Automated Meteorological and Oceanographic System (SAMOS) initiative. Data are available at The altimeter products (msla) were produced by Ssalto/Duacs and distributed by Aviso with support from Cnes.

67 Page Product Algorithm, Validity, coverage and known flaws 4.1. Product Algorithm and Validity The Algorithms used to generate these datasets are described in a dedicated document: the Algorithm Theoretical Breadboard. A preliminary assessment of the validity of these products is now being conducted: a deicated Product Validation document will be provided soon on the web site when finalized. Both documents shall be downloadable in the directory of the CATDS/CECOS ftp server Input Data and coverage The input data used to generate the present version of the CATDS CECOS-research Level 4 products are the ESA v5.50 L1B data from may 2010 to december The operational L1B data are used as input to our L3 & L4 processors for the period from january 2014 to dec The data aquired during the first four months of the commissioning phase in 2010 (from January to end of May) were not reprocessed because of reduced data quality during that period just postlaunch. Users interested in data from that period can still access the V01 products which were generated covering that period of time. Reader interested in the detailed data availability is refered to the ESA monitoring facility: Note that the last week of 2011 is missing because of electrical tests of the instrument at this period Several known flaws Remaining Major Issues in the L2OS data Some open issues remain in the level 1 algorithm which strongly impinge on the Level 2, 3 and 4 Sea Surface Salinity retrievals and their quality, but should be improved, though not fully resolved, with the next Level v620 processor now under deployement. In addition, it is clear that the L2OS algorithm, independently of the L1 data quality, has not yet reached a full maturity with remaining uncertainties in the forward models, not-yet-accounted for geophysical effects (waves, currents, diurnal effects,..), auxiliary data and non-optimal tuning of the overall algorithm(data filtering, retrieval methods,..).

68 Page 68 Figure 1:see legend below Figure 2: "First Order" remaining issues in the SMOS SSS data.

69 Page 69 The first map above is showing the difference between a SMOS Level 3 monthly-averaged product based on L2 data uncorrected for large-scale biases and an objectively analysed field of in situ observations (so-called ISAS data). The second map is showing a similar comparison but for another period of the year and reveals very strong latitudinal biases north of 20 N. The remaining issues at level 1 and 2 that need to be addressed by SMOS L1 & L2 teams in order to significantly improve the L2OS data quality can be summarized as follows: At Level 1: Strong (from the oceanographer perspective<=>±1 psu) and systematic brightness temperature data contamination over the oceans by land masses within an about 800 kmswidth band along the world coasts, Seasonal and latitudinal unexpected variations (from the oceanographer perspective) in the Brightness temperature data, Inaccuracies in the Radio Frequency Interferences (RFI) contamination filtering, Remaining noise in the Brightness temperature associated with solar radiation impacts on the reconstructed images which impact retrieved SSS data quality, Systematic spatial biases in the reconstructed TB images (partially mitigated at L2 by the OTT), Inaccuracies in the Total Electronic Content, Uncertainties in the polarization purity of the L1C data At Level 2: Non Optimal Radio Frequency Interferences (RFI) contamination filtering, Decreased sensitivity of the L-band signal to SSS in cold seas, Remaining sea water dielectric constant modelling uncertainties, Inaccuracies in the corrections for the sea surface roughness effects (wave & currents impacts), Inaccuracies in the corrections for extraterrestrial radiation glints (galactic and solar) Non yet accounted for geophysical effects in the forward models (rain impact, diurnal cycle of SST,..) Inaccuracies in the geophysical auxiliary data sets used as priors in the retrieval scheme to characterize the oceanic and meteorological conditions in the observed scenes, Non yet fully optimal iterative inversion methodology and data filtering (quality control) strategies

70 Page 70 The combination of the above listed issues in L1 & L2 products produces some well-known inhomogeneities of the retrieved L2 SSS data quality over the globe, which may prevent the use of SMOS products in certain oceanographic applications Solar contaminations Because the direct sun aliases contamines the Extended Field of View of the antenna, in our re-processing, the SSS retrievals were limited to the Alias-free Field of view domain (AF-FOV) of the instrument. Nevertheless, the direct sun aliases are sometimes located at the border of the Alias Free domain, particulary at the end of the years (November to December) in descending passes. To minimize this spurious contribution, a mask was applied around the location of the direct sun & aliases, eliminating reconstructed brightness temperatures within a radius of 0.05 in the cosine director coordinates of the antenna plane. As a very strong local source, the imaged sun disk and its aliases induce spurious brightness temperature tails after interferometric image reconstruction. While a sun correction is applied in the ESA level 1 processing, residual solar contributions that were not perfectly corrected by that processing may have produced some spurious signal in the SSS data at the end of the years. In particular, the sun disk alias propagating on the bottom left border of the AF-FOV may explain the presence of stripe-like too fresh/too salty anomalies detected in the composite SSS data that are seen to progressively amplify from october to december RFI SMOS multi-angular brightness temperature measurements at 1.4 GHz are strongly affected by radio frequency interferences (RFI) from radar networks, TV and radio links in what shall be a protected band. These intereferences are numerous over land in Europe and Asia, but can be also encountered in some other areas of Africa, America and Greenland and in some numerous islands over the world. Over the oceans, the signals emanating from land sources can extend very far away from the coasts and have dramatic consequences on the accuracy of sea surface salinity remote sensing from SMOS in some key oceanic areas like the north atlantic, north pacific and north indian oceans. The signature of RFI in SMOS data is highly variable in time and space and strongly depends on the instrument probing polarization and observation angles. Because of the interferometric principle, local strong RFI signals in the physical space can pollute a very extended area in the Fourier domain of the synthetic antenna and contaminate large portion of the SMOS reconstructed brightness temperature images. In particular, RFI sources located in the aliased regions of the image can impact the data in the (extended) alias-free field of view. Ideally, detection and Mitigation techniques of these spurious signals in SMOS data shall therefore be performed from the raw data at the visibility level prior image reconstruction and shall consider instantaneous acquisitions (snapshot information) and deal with the whole field of view images. Nevertheless, because of the strong amplitude of the RFI contamination with respect geophysical signal over the ocean, simple detection algorithm can be applied to the reconstructed brightness temperature data, identifying samples that deviate anomalously from the average of their neighbors in space, time and probing angles. Collecting acquired data over the full SMOS mission period, we were in a position to re-analyze the spatio-temporal characteristics of these signals and their varying signature as function of the instrument probing configuration (incidence angle, ascending versus descending passes). A global

71 Page 71 RFI analysis over the world ocean was performed from that data ensemble. The large number of data acquired at a given location on earth allowed us to clearly establish robust threshold detection criteria for these contaminations to best filter out the major contaminations using a multiple criteria mitigation approach. Nevertheless, it is clear that residual RFI-induced structures remain in our products, particularly in the Northern Latitudes, North Indian Ocean, along the coast of Asia and south of Madagascar. RFI density in the Mediteranean sea, Asia coastlines and Artic Sea induce a low quality retrieval in these area. We advise not to use our data for oceanographic studies in these zones. Note: We strongly recommand to only use L4 data with RFI probability equal to zero. Figure 1: these maps represent monthly averaged of the SMOS brightness temperature in terms of first stokes parameter (Th+Th)/2 at an incidence angle of 47.5 and in ascending passes. Over Sea ice the Tb is saturated because it is much higher than over the ocean. On the left: map for May 2011; right: map for May 2012 One of the largest area of contamination is the Northern Hemisphere, in particular over the North Pacific and Atlantic oceans. RFI are mostly induced here by the signals from the military radars of the Distant Early Warning (DEW) line sites. As illustrated in Figure 1, until summer 2011, the later RFI strongly contaminated SMOS data over all Canadian, Alaska waters and in the northern Atlantic (>45deg N) on ascending passes. Descending passes were less affected because of the lookangle of SMOS. Over the years, investigations of exactly where the interferences come from have been made by ESA. National authorities have collaborated with ESA to find out about the origin and how to switch these unlawful emissions off, and so RFIs have waned. Over recent years, authorities from Canada and

72 Page 72 Greenland were informed, and requested to take actions. Canada started to refurbish their equipment in autumn 2011, while Greenland switched off their transmitters in March At least 13 RFIs have now been switched off in the northern latitudes. As illustrated in Figure 2, the switch-offs have led to a significant improvement in SMOS observations at these high latitudes, which were previously so contaminated that accurate salinity measurements were not possible above 45 degrees latitude. Figure 2: (Left) Objectively analyzed in situ observations, SSS from SMOS in ascending passes in May 2011 (middle) and 2012 (right). Despite these improvements and, RFI continues to plague both salinity and soil moisture retrievals, and no solution proposed thus far can eliminate its impact in all cases. Most of the effort (including our filtering methodologies) is directed towards filtering out contaminated brightness temperature in the FOV where SSS is retrieved. However, much (but not all) of the RFI impact over the ocean is related to sources over land, and the impact in the usable portion of the field of view can be difficult to detect by simple thresholds on brightness temperatures. Therefore, inaccurate SSS retrievals are still found along the most contaminated oceanic zones. Here is one example showing intermittent contamination from radars in Alaska. The RFI induces large spatial ripples in the images far from the sources, and the impact extends into the alias-free field of view.

73 Page 73 Figure 3: RFI contaminated SMOS snapshot, showing an extremely strong source located far away from the usefull domain of the field of view but generating spurious signal where SSS retrieval is performed (the black contoured domain is the AF-FOV). The green contoured domain is the projection on earth of the fundamental hexagon delineating the Fourier component domain. Our new approach applied for the CEC V02 products is therefore to search for RFI in the entire fundamental hexagon using simple (empirically determined) thresholds on the brightness temperatures (500 K for Txx and Tyy, and 200 K for the third Stokes parameter Uxy). If one use such method to filter out RFI, a large amount of SMOS data along the world coastlines would be eliminated, including bad and good retrievals. As the decison to keep or remove an SSS retrieval in these contaminated area is not simple, we added a variable in the products (named "RFI_stat") giving the probability of remote RFI detection in the ensemble of multi-angular measurements used for the SSS retrieval. This RFI probability criterion was determined swath by swath and used to weight the SSS swath data when generating Level 3 spatio-temporal composite products. Grid points where the RFI probability flag was raised for more than 80 % of the input data were systematically eliminated. Ilustrating maps of the probability of detecting such remote RFI events over the ocean for ascending and descening passes over one month periods at 1/4 degree resolution are shown here below. As can be seen on these exemple, in some oceanic area, more than 50 % of SMOS data are contamined by RFI. As both good & bad quality data can be retrieved in such zones, our choice was to process the SSS data even in these area and to let the user perform his own filtering. A user that would like to work only with a priori free of RFI SSS SMOS data shall threrefore only select thoses data where RFI_stat=0%.

74 Page 74 The mean RFI probability over the global ocean is stable with time and reach on average 3%. It significantly decreased in the Northern Atlantic from about 10-12% in early 2010 to 4-5% end of One of the worst contaminated oceanic area is the North Indian Ocean where RFI probability systematically ranged between 10 and 35 % Land Contamination The Land Sea contamination and associated scene-dependent biases in Synthetic Aperture Radiometry were first discussed by Anterrieu (2007) two-three years before launch. While the problem was evidenced and some potential correction methods proposed (Gibbs), these were not implemented into the L1 operational chain. The L1, L2 and L3 SSS data were then further shown to be systematically highly biased on a complete band along the world coast lines, as shown at the end of the SMOS mission commissioning phase in May 2009 (see Reul presentation at the commissioning review). While methods for correcting these biases have been investigated further by L1 team members since then (e.g., Gibbs-1 to 3), no practical method has been yet implemented in the L1 processors to correct for these flaws which still strongly impact the ocean data quality, 4 years after launch. This is principally because there were other priorities to be tackled at L1 until now, because the proposed L1 solutions (e.g. Gibbs-like ) are heavy in term of computing time and therefore difficult to implement into the L1 operational processor but most evidently, because the exact source for that very complex problem and the associated optimal solution have not been yet found nor proposed as an implementable correction at L1 yet. Efforts are currently undertaken by L1 teams to find a practical correction at L1 (e.g., floor error mitigation) and if a solution rapidly emerged from the L1 efforts this would be a great progress for L2OS. It is clear for L2/ESL teams that this is an image reconstruction issue and by construction a problem that shall be solved at L1 based on a sound understanding of the MIRAS interferometer principles. Nevertheless, in parallel, L2OS teams also anticipate that empirical corrections and/or adapted data filtering should also be done at L2 while waiting for more adapted L1 solutions.

75 Page 75 While the source of the problem is certainly complex to understand (such scene-dependent biases would be present in the data even if we had a perfect knowledge of the antenna patterns, see Anterrieu 2007), it is clear on L2 and L3 SSS data that the contamination present signatures with a rather systematic character (e;g., all monthly or 10-days averaged SSS fields are contaminated with very similar patterns). SMOS is a multi-dimensional probing system: a L2 SSS retrieved on a given Earth grid point at the L2 processor output is coming from an ensemble of multi-angular data, acquired at varying position within the field of view depending on the time of passage of the satellite over that point, with varying polarization. Being a scene-dependent bias, at a given location on Earth (lat,lon), the Land Sea Contamination (LSC) is then a function of the location within the FOV (xi,eta), of the polarisation mode (XX or YY or cross pol), of the type of pass (Asc or Desc), and more importantly of the fraction of land masses and their distribution within the unit circle of the FOV (F), itself a function of the type of pass and finally, of the brightness temperature differences between land masses and ocean scenes (ΔTB LAND-OCEAN) which is also a function of time t (natural variability). An empirical correction would therefore be a complex functional of the form: LSC(lat i,lon i,pass type=asc or Desc)=func(xi,eta,p=polarisation, ΔTB LAND-OCEAN(t)) and could consist in correcting the L1C products before they are used as input to the L2 processor (socalled L1d product). This functional do not need to be evaluated for all the points of the DGG grid: it can be restricted to a band along the coast (still TBD but which could be derived from a pre-deterrmined land-fraction over the FOV metric). It is anticipated that the LSC function dependencies on ΔTB LAND- OCEAN will be a second order effect and as a first approximation might be neglected into the emprirical correction. The idea behind a potential empirical correction is based on the fact that the Level 3 observed biases are somehow apparently very stable in time over time scales 10 days. A mean bias correction could therefore be estimated from the 4 years of data and removed to the swath one Major Geophysical correction issues: Sky noise Modeling studies conducted by several teams prior to SMOS launch indicated that the downwelling celestial radiations at L-band that are scattered back by the ocean surface toward the upper hemisphere can be a source of brightness contamination affecting the quality of sea surface salinity retrieval. For sun-synchronous polar-orbiting satellite measurements of upwelling L-band radiation over the ocean, like with SMOS, this so-called sky noise depends strongly on pass direction (ascending or descending), time of year and surface roughness (wind speed). Based upon the modeling studies for SMOS sensor, the impact is expected to be strongest for descending passes in September-October and for ascending passes in March-April because for these passes the reflections of the instrument viewing directions over the field of view tend to lie along the galactic equator where L-band galactic emission is maximum.

76 Page 76 Left: model of the specularly reflected galactic signal in oct 2011 descending passes. Right: Biases between SMOS SSS retrievals and World Ocean Atlas Climatology if one correct the sky-noise contribution to the brightness temperature by assuming a perfecly flat ocean surface. As illustrated in the above figure, assuming a perfectly flat ocean surface and correcting for the sky noise using a simple specular reflection model result in significantly overpredicted SSS. To minimize the impact of that spurious signal, the sky-noise correction therefore need to account for surface roughness induced scattering impacts. Originally proposed Kirchhoff scattering model using the Kudryavtsev wave spectrum has been shown to strongly underpredicts the scattered brightness near the galactic plane and overpredicts the brightness away from the galactic plane under most surface wind speeds. Kirchhoff scattering model evaluated at surface wind speed of 3 m/s better predicts the scattered brightness under most wind conditions, but still underpredicts brightness near galactic plane at low wind speeds, and overpredicts brightness at high wind speeds. Lack of wind speed dependence is unrealistic. This was the solution used for generating the first version V01 of the CEC products. Model for scattering of celestial sky brightness has been under continual refinement and more accurate corrections for these geophysical effects are still under development in the frame of the ESA level 2 processor improvment studies. For generating the Level 3 CEC products v02, we thus applied a new semi-empirical correction algorithm for the sky noise based on the Geometric Optics (GO) scattering solution. As found, semi-empirical models based upon GO produce improved predictions relative to those based on Kirchhoff and retain wind speed dependence. Geometrical optics fits to the data were found different for ascending and descending passes, possibly due to inaccurate representation of scattering cross sections away from specular direction. Possible solution is to introduce ascending and descending lookup tables for scattered celestial sky brightness but this was not yet implemented for the V02 products. While clear improvements are expected with the new GO-correction model, particularly when strong galactic sources are scattered toward the sensor, residual erroneous correction of this effect in our alogorithm might have cause some non-geophysical variability in the Level 3 composite SSS products, particularly in the low wind speed conditions.