Slides for Matthias · the performances of the analysed soil moisture fields are close (within 2-3 %) to those of the open loop experiment. Analysed surface fields are used to compute

Satellite Soil Moisture Validation and Application Workshop Wien September 19-20 2017-- YHKerr

SMOS MissionLaunched in 2009 data flow started in January 2010First 6 months to be used with careTests dual pol full polOut gassing and calibration issuesMaximum RFI environment

Several re-processingsNew measurements & new instrument -> wax and

strings, trial and error approach to overcome the unexpected !Many improvements from V3 to V6, V7 underway!

Availability of L2 and L3 -L4, new products in the making


• Training on SMOS Level 2 v620 SM • Similar performances (slightly better indeed) • Much faster ! Less than 3.5 hours after sensingRodriguez-Fernandez et al. (2017, HESS)

Implemented by :

With support by :

Delivered to :

Disseminated by:

Near Real Time SM

Satellite Soil Moisture Validation and Application Workshop Wien September 19-20 2017-- YHKerrAveraged SM values over Jan/Apr/Jul/Oct from Jul-2010 to Apr-2017 (28 months)

V650 is ready ! All archive (2010 –july 2017) reprocessedSoon to start in the operational processor


First step: CalibrationComparison at TB level are conducted over various

surfaces of known temperaturesGalactic backgroundIce sheet (Antarctica)Ocean bodies

Or other similar sensorsBased on near simultaneous, iso geometryLandOcean

Ice core around Dome ConcordiaCompared TBs are ToA, without reflexion foreign source

corrections (gal, sun, moon)

SMOS’ view of the Galaxy


SMOS and DomeX

Long term stabilityOverall good agreementIncidence induced bias

between SMAP and Aquarius

Sensor Version Inc TBH TBV DTBH DTBV

Aquarius v4 28 192.90 206.19 0.59 1.16

Aquarius v4 38 189.23 210.61 0.33 -0.62

Aquarius v4 45 185.03 213.40 1.01 -0.98

SMAP R12170 40 187.67 212.46 -0.88 0.41

SMAP R13080 40 186.17 210.08 -2.38 -1.97

SMOS v620 38 188.90 211.23

SMOS v620 40 188.55 212.05

SMOS v620 28 192.31 205.03

SMOS v620 45 184.02 214.38

DOMEX-2 42 186.27 206.57 -0.015 -6.645

DOMEX-3 42 187.34 207.54 1.055 -5.675

F. Cabot


SMOS -SMAP, Successful RetrievalsMonthly Animation: 2015.04-2016.05

SMAP SMOS

SMOS-SMAPMahmoodi A.


L3 SM, SMOS Successful Rets & SMAP Recommended Rets

BIAS RMSE

Corr unB RMSEMahmoodi A.


Step 2 Validation Done over different targets

Sparse networks (Scan, SNOTel, …)

Dense networks USDA Watersheds, OziNet, Hobe, …

Specific, complicated targets and core sites Rugged terrain (Alps) Boreal areas (Sodankyla) Tropical forest (Chaco)

Satellite data different sensors (SMAP, Aquarius, AMSR, … ) Different retrieval approaches (RT model, Neural networks, simplified approaches,

….)Model outputs

Note that the retrieval approaches are all global not point to point Note that when using a large number of sites the total distribution of

quality metrics should be studied instead of just quality metrics averages


Step 2 Validation Issues to be tackledSurface area of ground measurements (representativeness)

HOBE like set upsCosmos probes

Spatio-temporal dimensionsBeatriz Molero Rodenas study

Organic soilsSpecific calibrations

Quality of measurementsGround measurements have own uncertainties and are never “truth”

Models might be “wrong” over specific ecosystems

Algorithmspoint to point or regional algorithms


Models and “proxy” sensors give erroneous estimatesA. Mialon

Very important region: • Hotspot (land feedback to atmosphere, Koster et al., Seneviratne et al.)• Very little in situ data to constrain weather models -> Remote sensing


SMOS SM used as reference for other instruments

AMSR-E retrievals using a neural network trained on SMOS L3 SM (2003-2010 -> 2010-2017)

Rodriguez-Fernandez et al. (2016. Remote Sensing)

Long time series


SMOS support to the CCI ESA SMOS/AMSR-E fusion project Neural networks are a promising approach SMOS can be used as reference for re-scaling other

instruments SMOS should be inserted into the CCI framework using LPRM

CESBIO : extraction and pre-processing of SMOS and ECMWF auxiliary data for their use by the CCI team

SMOS now taken into account in CCIv3 (Dorigo et al. 2017)


Soil Moisture 1 km Morocco

Land Surface Temperature

Optic/Thermal

Soil MoistureSMOS

1 km / 1 day

40 km / 3 days

DisPATCh-SM actual

Soil MoistureSMOS

Land Surface TemperatureMODIS (Aqua/Terra)

J. Malbeteau

A. Mahmoodi


Example of SMOS High Resolution data for irrigation monitoring

Molero et al


USING SMOS DATA IN NWP

In situOpen Loop SMOS NN SM σ x1 + T2m + RH2mSMOS NN SM σ x3 + T2m + RH2m

Assimilating SMOS data moderately improves the soil moisture analysis: On average, for more than 400 in situ sites,the performances of the analysed soil moisture fields are close (within 2-3 %) to those of the open loop experiment

Analysed surface fields are used to compute atmospheric forecasts: SMOS soil moisture (NRT, NN based product)improves the forecast in the Northern Hemisphere

Blue: positive impact

Red: negative impact

From:Rodriguez-Fernandez, de Rosnay, Albergel, et al. 2017, ECMWF ESA reportRodriguez-Fernandez et al. (in prep.)

Further work assimilating L-Band into NWP, e.g.• J. Kolassa: Merging active and passive microwave observations

in soil moisture data assimilation, RSE ,2017• G. De Lannoy: Assimilation of SMOS brightness temperatures or

soil moisture retrievals into a land surface model, Hydrology andEarth System Sciences, 2016

RMSE of 36h FC 850 hPa temperature forecastsSLV+SMOS DA (sigmao*3) minus OL SLV+SMOS DA (sigmao*9) minus OL


Summary L Band radiometry (SMOS- SMAP and Aquarius) has proven its ability to deliver reliable

accurate and absolute (non model scaled) soil moisture fields over the globe even in areas of dense vegetation thanks to its long wavelength

Independent retrieval approaches (RT, NN, SCA, SMOS-IC , DCA, ..) or completely different instruments provide very similar if not totally similar results (without fiddling our massaging such as trend correction, anomaly etc..) showing the robustness of the measurements

High temporal revisit (<14 h on average over the globe) using both SMOS and SMAP Why passive L band?

Because of its characteristics and inherent qualities The most appropriate tool as shown by all the products stemming from it Temporal stability and robustness

L band radiometry --> proof of concept demonstrated Uniqueness of the measurements hence

Many science outstanding results A very large number of operational or pre operational demonstration products

BUT … No follow on mission currently --> Data gap Need to act now


2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 20322008

L-Band radiometry missions

PrecipitationPrecipitationGravity

SMOS (ESA CNES) (40 km / 3days / L-band / global )

SMAP (NASA) (10-60 km / 3days / L-band / global)

Thermal

Color codesAltimeters

L-Band PassiveOptical

Radar

Thermal

Color codesAltimeters

L-Band PassiveOptical

Radar

Aquarius (NASA) (100km / 8days / L-Band/ global)

Kerr and Al Bitar

TODAY!


SMOS DATA PRODUCTSOver land

Data product Resolution/format Latency Available from

NRT light: Level 1 brightness temperature

30-50km (N256 Gaussian grid), swathbased; BUFR.

NRT/ 3 hours from sensing

ESAEUMETCASTWMO GTS

Level 2 soil moisture in NRT (based on Neural Network)

15 km (ISEA 4H9 grid), swath based; NETCDF.

NRT/~4 hours from sensing

Operational/NRT products / Latency < 3 hours

Data product Resolution/format Latency Available from

Level 1 brightnesstemperature

15 km (ISEA), swath EEF /NetCDF.25 km, global, EASE- NetCDF

6-8 hours after sensing1 d after sensing

ESACATDS(+ stereopolar)

Level 2 Soil moisture 15 km (ISEA), swath EEF /NetCDF.25 km, global, EASE- NetCDF

8-12 hours1 d

ESACATDS

Level 3 Brightness Temperature and Soil Moisture

15 km (ISEA 4H9) grid/ 25 km (EASE) grid depending on product. NETCDF

Daily, 3, 9 days, weekly, monthly

CATDSBEC

Level 4 fine-scale soil moisture

1 km, for Iberian Peninsula; NETCDF1 km for MODIS Tiles

2 daily maps (one asc/ one desc) in NRT

BECCATDS (2017)

Level 4 CATDS Root Zone Soil Moisture

~25 km (EASE grid version 2); NetCDF Daily, 10 days, monthly CATDS

Level 4 Drought Index 25 km EASE 2 grid netcdf Daily, 10 day, Monthly CATDS

Freeze and thaw ~25 km (EASE grid version 2); NetCDF,Northern Hemisphere

Daily Demo data set available from FMI

Surface roughness 25 km NETCDF, global Yearly CATDS

10-day global composite of L3 soil moisture. Credits CATDS/CESBIO

Root zone soil moisture in m3/m3. Credits CATDS/CESBIO

Science and composite products/ Latency > 3 hours

DATA ACCESSESA: http://smos-diss.eo.esa.int/CATDS www.catds.fr/Products/Available-products-from-CPDCBEC: http://cp34-bec.cmima.csic.es/land-datasets

SMOS swath-based L2 soil moisture product. Credits ESA

http://smos-diss.eo.esa.int/

http://www.catds.fr/Products/Available-products-from-CPDC

http://cp34-bec.cmima.csic.es/land-datasets


Acknowledgments

32nd URSI General Assembly and Scientific Symposium | Montreal, Canada | Aug 19-26, 2017 | Steven Chan et al.

This work is funded by the 2016 NASA ROSES call for the

Science Utilization of the Soil Moisture Active Passive Mission (SUSMAP)

sponsored by the NASA Terrestrial Hydrology Program


Flow Diagram

SMAPSMOS

L1 gain/offset adjustment

SMOS L1 TB

SMAP L1 TBSMOS/SMAP L1 TBfine grid resampling on

9 km EASE Grid 2.0

SMAP AlgorithmSMAPAncillary Data

SMOS/SMAP soil moisture at 9 km(2009 – present)


Part 1

Part 2

Chan et al (2017)


Part 2: Soil Moisture Retrieval


9 km soil moisture using 6 am SMOS TBs with SMAP algorithm and ancillary data

9 km soil moisture using 6 am SMAP TBs with SMAP algorithm and ancillary data

Jun 2017

Jun 2017

Good agreement

between SMOS and

SMAP

Slight code

differences from

current SMAP

operational setup

(further work

needed)

Consistent TBs,

algorithm, and

ancillary data lead

to consistent

SMOS/SMAP soil

moisture

Chan et al (2017)


Part 2: Soil Moisture Retrieval


9 km soil moisture retrieval using 6 am SMOS TBs with SMAP algorithm and ancillary data

Jun 2017

Jun 2017

9 km soil moisture retrieval using 6 am SMAP TBs with SMAP algorithm and ancillary data

Chan et al (2017)


The SMOS soil moisture validation and intercomparison approach and

results

Yann Kerr, Jean Pierre Wigneron, Beatriz Molero-Rodenas, Rajat Bindlish, Steven Chan, Roberto Fernandez-

Moran, Arnaud Mialon, Amen Al-Yaari, Ali Mahmoodi, Simone Bircher, Nemesio

Rodriguez-Fernandez, Philippe Richaume, Ahmad Al Bitarand François Cabot

CESBIO, INRA ISPA,JPL, GSFC


Part 1: SMAP TB vs. SMOS TB


Overall SMOS and SMAP in good agreement; SMAP > SMOS (ocean) but SMAP < SMOS (land)

Match SMOS to SMAP to use SMAP’s inversion setup (algorithm and ancillary data)

SMOS TBH (K)

SMAP

TB H

(K)

SMAP

TB H

(K)

SMOS TBH (K)

SMAP

TB H

(K)

SMOS TBH (K)

6:00 am + 6:00 pmMay 2015 – Jun 2017

Before Before

Before

Chan et al (2017)




SMOS TBH (K)

SMAP

TB H

(K)

SMAP

TB H

(K)

SMOS TBH (K)

SMAP

TB H

(K)

SMOS TBH (K)

After adjustments, SMOS and SMAP exhibit minimal bias over the entire TB range

Separate adjustments needed for 6:00 am TBH, 6:00 am TBV, 6:00 pm TBH, and 6:00 pm TBV

6:00 am + 6:00 pmMay 2015 – Jun 2017

After After

After

Chan et al (2017)




After adjustments, swath discontinuity between SMOS and SMAP is reduced, though not completely

eliminated due to:

▸ Different observation times away from SMOS/SMAP swath intersection region

▸ Different azimuth angles (SMOS: fore only; SMAP: fore and aft)

6:00 am SMOS TBH overlaid with 6:00 am SMAP TBH(Jun 14 – 16, 2017)

moderate swath discontinuity before adjustments minimal swath discontinuity after adjustments

SMAP6 am

SMOS6 am

SMAP6 am

SMOS6 am

Chan et al (2017)




After adjustments, SMOS and SMAP exhibit minimal bias over the entire TB range

Customized adjustments necessary for 6:00 am TBH, 6:00 am TBV, 6:00 pm TBH, and 6:00 pm TBV

Bias (SMAP minus SMOS) (K) RMSE (K)

Before Adjustments After Adjustments Before Adjustments After Adjustments

6:00 am ocean TBH 1.18 0.04 2.36 2.02

6:00 am land TBH -1.66 -0.07 3.59 3.14

6:00 am TBH 0.59 0.03 2.66 2.30

6:00 pm ocean TBH 0.57 0.01 2.20 2.10

6:00 pm land TBH -2.10 -0.02 3.91 3.27

6:00 pm TBH -0.04 0.00 2.69 2.42

6:00 am ocean TBV 0.72 -0.02 2.04 1.88

6:00 am land TBV -2.51 -0.05 3.81 2.81

6:00 am TBV 0.06 -0.03 2.50 2.11

6:00 pm ocean TBV 0.57 -0.03 2.12 2.01

6:00 pm land TBV -2.67 0.00 4.00 2.92

6:00 pm TBV -0.16 -0.03 2.66 2.25

Chan et al (2017)

Documents

Slides for Matthias · the performances of the analysed soil moisture fields are close (within 2-3 %) to those of the open loop experiment. Analysed surface fields are used to compute