From Experimental to Operational Land Surface Data Assimilation for Soil Moisture Estimation using SMOS and SMAP Satellite Observations

(1)

1 / 44

From Experimental to Operational Land Surface Data Assimilation for Soil Moisture Estimation using SMOS and SMAP Satellite

Observations

Gabri¨elle De Lannoy, Rolf Reichle

Q. Liu, J. Ardizzone, A. Colliander, A. Conaty, T. Jackson, J. Kimball, R. Koster, S. Mahanama

KU Leuven, Department of Earth and Environmental Sciences, Division Soil and Water Management Global Modeling and Assimilation Office (Code 610.1), NASA/GSFC, Greenbelt, MD, USA

20 September 2016

(2)

Land Surface System

Soil Moisture Data

Model DA

SMOS DA SMAP DA Conclusion

2 / 44

Land surface:

■

interface between land and atmosphere

■

integrated system with various

compartments: soil, vegetation, snow

Processes:

■

budget of energy

■

budget of water

■

budget of carbon and other constituents

(3)

Land Surface Variables

Land Surface System

Soil Moisture Data

Model DA

3 / 44

Precipitation Soil Temperature

Evapotranspiration Runoff and drainage

Groundwater Snow

(4)

Land Surface Variables

Land Surface System

Soil Moisture Data

Model DA

3 / 44

Precipitation Soil Temperature

Evapotranspiration Runoff and drainage

Groundwater Snow

Soil Moisture

(5)

Soil Moisture

Land Surface System

Soil Moisture Data

Model DA

4 / 44

Agricultural productivity Weather and climate forecasts

■

improve flood prediction and drought monitoring capability

■

enhance weather and climate forecast skill

■

link global water, energy and carbon processes at the land surface

(6)

Earth Observing Satellite Missions

Land Surface System

Soil Moisture Data

Model DA

5 / 44

so il m o is tu re

(7)

SMOS and SMAP

Land Surface System

Soil Moisture Data

Model DA

6 / 44

L-band radiometers onboard SMOS (ESA,

Soil Moisture Ocean Salinity

)

q

1

q

2

SMAP (NASA,

Soil Moisture Active Passive

)

■

launched November 2009

■

multiple incidence angles

■

launched January 2015

■

fixed 40

^o

incidence angle SMOS and SMAP observe similar Tb at 40 km resolution, ...

... but note that the SMOS and SMAP L1 Tb data products are different (De Lannoy et al., 2015, GRSL).

(8)

SMOS and SMAP

Land Surface System

Soil Moisture Data

Model DA

7 / 44

Similar yet different ...

SMOS, 12 April 2015 SMAP, 12 April 2015

[K]

■

alias-free swath width is wider for SMAP than SMOS

■

better RFI mitigation with SMAP (e.g. Asia)

■

SMOS: ascending morning overpass; SMAP: descending morning overpass

(9)

Passive Microwave Observations over Land

Land Surface System

Soil Moisture Data

Model DA

8 / 44

L-band (1.4 GHz) brightness temperatures (Tb) are sensitive to soil moisture and temperature in the surface layer (5 cm)

NASA SMAP ESA SMOS

brightness temperature → L1 Tb

)

→ RTM (parameters)

soil temperature

soil moisture → L2 SM

vegetation water and

temperature

(10)

Land Surface Model

Land Surface System

Soil Moisture Data

Model DA

9 / 44

NASA GEOS-5 Land Surface Modeling

Land Surface Model (LSM)

Catchment land surface model (Koster et al., 2000) ■ MERRA meteorological forcings

(or GEOS-5 FP + precip corrections)

■ updated soil parameters

■ updated vegetation parameters

Little Washita (OK)

01/01/2008 07/01/20080 01/01/2009 07/01/2009 12/31/2009 0.2

0.4

sfmc [m3 /m3 ]

LW (34.88 N, 98.08 W)

in situ, MERRA, MERRA-2, revised (De Lannoy et al., 2014)

old soil

parameters revised

(11)

Observation Model

Land Surface System

Soil Moisture Data

Model DA

10 / 44

NASA GEOS-5 Land Surface Modeling

Radiative Transfer Model (RTM)

L-band tau-omega model (De Lannoy et al., 2013, 2014): transform soil moisture into Tb

0 0.1 0.2 0.3 0.4 0.5 220

240 260 280

HR = 0 0.4 0.8 1.2

1.6

Ts

dry <−−−−−−−> wet sfmc [m³/m³]

Tb H [K] Tb [K]

Global calibrated parameters, e.g.:

hmin [-] ω [-]

Spatial aggregation

From 36 km or 9 km to e.g. 40 km 3 dB footprint (antenna pattern)

(12)

Soil Moisture Observations vs. Simulations

Land Surface System

Soil Moisture Data

Model DA

11 / 44

Observation challenges:

■

raw data → land surface quantities

■

intermittent → ‘continuous’ fields in space and time

■

aggregated → ‘downscaled’ in space

■

surface layer → root-zone

■

errors (random, bias) Modeling challenges:

■

structure (LSM, RTM): simplicity ↔ reality

■

meteorological forcings

■

parameters, ancillary information, calibration

■

errors (random, bias)

obs 20150430 1200

0 0.2 0.4

Surface soil moisture

(13)

Data Assimilation

Land Surface System

Soil Moisture Data

Model DA

12 / 44

Data assimilation:

use satellite data to guide model forecasts

use model estimates to add value to satellite information

(14)

Data Assimilation

Land Surface System

Soil Moisture Data

Model DA

13 / 44

Experimental DA

■

Explorative: what is possible, what is best? Take risks, fail, succeed.

■

Develop and test algorithms

■

Optimize

■

Local validation

Example: SMOS data assimilation

(De Lannoy and Reichle, 2016a, b)

Operational DA

■

Implement best available

state-of-the-art, with gentle

incremental changes that guarantee product continuity

■

Meet set accuracy requirements

■

Meet set latency time

■

Meet user requirements

■

Product documentation, continuous monitoring

Example: SMAP L4 SM product

(Reichle et al., 2016)

(15)

SMOS Data Assimilation

Land Surface SMOS DA Products DA

Results SMAP DA Conclusion

14 / 44

(16)

SMOS L1 Tb and L2 SM Products

15 / 44

brightness temperature → L1 Tb

soil temperature

soil moisture → L2 SM

vegetation water and temperature

Tb

^H

(40

^o

)

220 230 240 250 260 270

[K]

Surface soil moisture

obs 20150430 1200

0 0.2 0.4

[m

³

/m

³

]

(30 April 2015, 12 UTC)

(17)

SMOS L1 Tb and L2 SM Products

16 / 44

3 data products:

■

L1 Tb: v620 (up to current), transformed to Tb

_BOA

◆

multi-angular SMOS Tb: H-, V-pol, [30

^o

, 35

^o

, 40

^o

, 45

^o

, 50

^o

, 55

^o

, 60

^o

]

◆

40

^o

fitted SMOS Tb: H-, V-pol (similar to SMAP)

■

L2 SM: v552 (stops May 2015)

20 30 40 50 60

210 215 220 225 230

incidence angle [^o]

multi-angular SMOS;

2nd order fit; Tb fit(40

^o

)

■

lauched in Nov 2009, multiple incidence angles

■

50% of signal within nominal resolution of 43 km (antenna pattern)

■

study period: 1 July 2010 - 1 May 2015

(De Lannoy and Reichle, 2016a, b)

(18)

Data Assimilation

17 / 44

SMOS Obs (footprint) NASA GEOS-5 Land Surface Modeling (36 km)

[K]

- Catchment land surface model - MERRA surface meteorology

——————————————–

Observation operator:

- spatial aggregation

- radiative transfer model*

only in case of Tb assimilation

Surface (0-5 cm)

“Root zone”

(0-100 cm)

(19)

Data Assimilation

17 / 44

SMOS Obs (footprint) NASA GEOS-5 Land Surface Modeling (36 km)

[K]

- Catchment land surface model - MERRA surface meteorology

——————————————–

Observation operator:

- spatial aggregation

- radiative transfer model*

only in case of Tb assimilation

Surface (0-5 cm)

“Root zone”

(0-100 cm)

Data Assimilation - 3D EnKF

- bias mitigation

^∗

- filter parameters

^∗

- Surface soil moisture ( ∼ top 5 cm)

- Root zone soil moisture ( ∼ top 1 m)

- Other consistent geophysical fields, with error estimates

⇒ * calibration using long-term SMOS record

(20)

SM Data Assimilation

18 / 44

Innovations Increments (a) O-F SM [m³.m⁻³] (b) ∆wtot [mm]

-0.02 0 0.02 -10 0 10

Analysis

(c) sfmc [m³.m⁻³] (d) rzmc [m³.m⁻³]

0 0.2 0.4 0.6 0 0.2 0.4 0.6

(30 April 2015, 12 UTC)

■

Observation-minus-forecast (O-F, innovation),

footprint-scale

■

Increment, model grid

■

Analysis, model grid

■

3D EnKF: smooth transitions,

no swath edges in analysis

(21)

Tb Data Assimilation

19 / 44

Innovations Increments

(a) O-F TbH [K] (b) O-F TbV [K] (c) ∆wtot [mm] (d) ∆tp1 [K]

-10 0 10 -10 0 10 -10 0 10 -2 0 2

Analysis

(e) sfmc [m³.m⁻³] (f) rzmc [m³.m⁻³] (g) tp1 [K]

0 0.2 0.4 0.6 0 0.2 0.4 0.6 270 280 290 300

(30 April 2015, 12 UTC)

(22)

Observation or Innovation Bias

20 / 44

■

Data assimilation system (EnKF) needs to be unbiased

■

Data assimilation only corrects for random errors

■

Climatological bias mitigation does

not assign the bias to either the observations or the forecasts.

Observation - Forecast

[O - F] → bias correction → [O - F - bias]

= [O - bias] - F

(23)

SM Observation or Innovation Bias

21 / 44

SM is relatively stationary Example: at one location,

- at any time, replace an observed SM of 0.08 m

³

/m

³

with a value of 0.10 m

³

/m

³

0 0.1 0.2 0.3

0 0.5 1

Walnut Gulch N=743

sfmc [m³/m³]

CDF [−]

SMOS model

Little River N=755

■

CDF based on 5 years, all seasons

■

separate rescaling for ascending (6 am) and descending (6 pm) times

(24)

Tb Observation or Innovation Bias

22 / 44

Tb has a strong seasonal pattern Example: at one location,

- at pentad 7, correct the observed Tb

^H

for a bias of 237-241 K - at pentad 36, correct the observed Tb

^H

for a bias of 262-260 K - at pentad ..., correct ...

model-SMOS < Tb

^H

(40

^o

) > [K], Asc, pentad 36 Little River

100 200 300 230

240 250 260 270 280

DOY

<Tb H(40o )> [K]

↓ p36 LR

- SMOS - model

■

mean-only, 5 year-average, per pentad

■

separate rescaling for ascending (6 am) and descending (6 pm), 7 angles, 2

polarizations

(25)

Innovations

Results

SMAP DA Conclusion

23 / 44

(a) O-F Tb 7ang [K] (H,V,Asc,Desc,7 angles)

-20 -10 0 10 20

(b) O-F SM [m³.m⁻³] (Asc,Desc)

-0.1 -0.05 0 0.05 0.1

Hovm¨uller plot

(longitudinal averages, 5 years)

■

unbiased O-F

■

(a) very random Tb O-F

■

(b) minimal seasonal pattern in SM O-F

(26)

Innovations

Results

SMAP DA Conclusion

24 / 44

Tb 7ang DA Tb fit DA SM DA

(a) m=0.34, s=0.08 [-] (b) m=0.25, s=0.07 [-] (c) m=0.58, s=0.17 [-]

N p e r d ay

0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

(d) m=7.42, s=2.45 [K] (e) m=7.39, s=3.16 [K] (f) m=0.03, s=0.01 [m

³

/m

³

]

st d (O -F )

0 5 10 15 0 5 10 15 0 0.05 0.1

(27)

Innovations

Results

SMAP DA Conclusion

24 / 44

Tb 7ang DA Tb fit DA SM DA

(a) m=0.34, s=0.08 [-] (b) m=0.25, s=0.07 [-] (c) m=0.58, s=0.17 [-]

N p e r d ay

0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

(d) m=7.42, s=2.45 [K] (e) m=7.39, s=3.16 [K] (f) m=0.03, s=0.01 [m

³

/m

³

]

st d (O -F )

0 5 10 15 0 5 10 15 0 0.05 0.1

Less Tb data than SM data:

Tb swaths limited to narrow, alias-free zone

(28)

Innovations

Results

SMAP DA Conclusion

24 / 44

Tb 7ang DA Tb fit DA SM DA

(a) m=0.34, s=0.08 [-] (b) m=0.25, s=0.07 [-] (c) m=0.58, s=0.17 [-]

N p e r d ay

0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

(d) m=7.42, s=2.45 [K] (e) m=7.39, s=3.16 [K] (f) m=0.03, s=0.01 [m

³

/m

³

]

st d (O -F )

0 5 10 15 0 5 10 15 0 0.05 0.1

Less Tb data than SM data:

Tb swaths limited to narrow, alias-free zone

Tb O-F larger than SM O-F

(RTM: 1.3 K / 0.01 m

³

/m

³

)

(29)

Normalized Innovations

Results

SMAP DA Conclusion

25 / 44

Tb 7ang DA Tb fit DA SM DA

(a) m=1.14, s=0.35 [K/K] (b) m=1.11, s=0.46 [K/K] (c) m=1.23, s=0.41 [-]

0.3 0.5 0.8 1.3 2.0 3.2 0.3 0.5 0.8 1.3 2.0 3.2 0.3 0.5 0.8 1.3 2.0 3.2

std(O-F/ p σ

²

F

+ σ

²

O

),

with σ

_F²

and σ

_O²

determined by DA design parameters (ensemble perturbations).

Target value = 1

< −− DA system −− >

overestimates underestimates

actual uncertainty

(30)

∆ wtot Increments

Results

SMAP DA Conclusion

26 / 44

Tb 7ang DA Tb fit DA SM DA

(a) m=0.46, s=0.11 [-] (b) m=0.36, s=0.10 [-] (c) m=0.76, s=0.19 [-]

N p e r d ay

0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

(d) m=6.86, s=3.65 [mm] (e) m=5.86, s=3.45 [mm] (f) m=4.17, s=1.93 [mm]

st d ( ∆ w to t)

0 5 10 15 0 5 10 15 0 5 10 15

(31)

∆ wtot Increments

Results

SMAP DA Conclusion

26 / 44

Tb 7ang DA Tb fit DA SM DA

(a) m=0.46, s=0.11 [-] (b) m=0.36, s=0.10 [-] (c) m=0.76, s=0.19 [-]

N p e r d ay

0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

(d) m=6.86, s=3.65 [mm] (e) m=5.86, s=3.45 [mm] (f) m=4.17, s=1.93 [mm]

st d ( ∆ w to t)

0 5 10 15 0 5 10 15 0 5 10 15

Less Tb data than SM data corresponding increments:

More increments than observations: spatial filter

(32)

∆ wtot Increments

Results

SMAP DA Conclusion

26 / 44

Tb 7ang DA Tb fit DA SM DA

(a) m=0.46, s=0.11 [-] (b) m=0.36, s=0.10 [-] (c) m=0.76, s=0.19 [-]

N p e r d ay

0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

(d) m=6.86, s=3.65 [mm] (e) m=5.86, s=3.45 [mm] (f) m=4.17, s=1.93 [mm]

st d ( ∆ w to t)

0 5 10 15 0 5 10 15 0 5 10 15

Less Tb data than SM data corresponding increments:

More increments than observations: spatial filter

stdv( ∆ wtot) for Tb DA larger than SM DA

due to relatively higher Tb O-F, more info in Tb O-F

(33)

∆ wtot Increments

Results

SMAP DA Conclusion

27 / 44

(a) R=0.72 (b) R=0.33

-50 0 50

Tb_7ang DA

-50 0 50

Tb_fit DA

wtot [mm], R=0.72

-50 0 50

Tb_7ang DA

-50 0 50

SM DA

wtot [mm], R=0.33

1

10 ² 10 ⁴

■

unbiased system

■

Tb 7ang and Tb fit correct soil moisture trajectories similarly

■

Tb DA introduces more large increments than SM DA

∼ Tb DA has larger innovations than SM DA

■

different information extracted during Tb DA and SM retrieval process?

(De Lannoy and Reichle, 2016, HESS, in review)

(34)

In Situ Evaluation

Results

SMAP DA Conclusion

28 / 44

Tb 7ang DA SM retrieval DA

(a) ∆RMSD^ub=-0.004 [m³/m³] (b) ∆RMSD^ub=-0.003 [m³/m³]

(153/187 improved) (143/187 improved)

S u rf a ce s. m .

(c) ∆RMSD^ub=-0.002 [m³/m³] (d) ∆RMSD^ub=-0.001 [m³/m³]

(125/187 improved) (121/187 improved)

R o o t- zo n e s. m .

Blue=better Red=worse

(35)

In Situ Evaluation

Results

SMAP DA Conclusion

29 / 44

a) Surface Soil Moisture

favorable non-favorable 0.4

0.5 0.6

anomR [-]

^N=98(24) ^N=83(22)

b) Root-Zone Soil Moisture

favorable non-favorable 0.4

0.5 0.6

anomR [-]

^N=98(24) ^N=83(22)

open loop, Tb 7ang DA, Tb fit DA, SM DA

■

largest soil moisture improve- ments in favorable areas

■

similar averaged skill statistics

for Tb and SM DA

(36)

SMAP Data Assimilation

Land Surface SMOS DA SMAP DA DA

Results Conclusion

30 / 44

(37)

Data Assimilation

Results Conclusion

31 / 44

SMAP L1 Tb (footprint) NASA GEOS-5 Land Surface Modeling (9 km)

[K]

- Catchment land surface model improved parameters

- radiative transfer model calibrated parameters

- GEOS-5 FP surface meteorology gage- and satellite-based

precipitation corrections

(Reichle and Liu, 2014)

Surface (0-5 cm)

“Root zone”

(0-100 cm)

(38)

Data Assimilation

Results Conclusion

31 / 44

SMAP L1 Tb (footprint) NASA GEOS-5 Land Surface Modeling (9 km)

[K]

- Catchment land surface model improved parameters

- radiative transfer model calibrated parameters

- GEOS-5 FP surface meteorology gage- and satellite-based

precipitation corrections

(Reichle and Liu, 2014)

Surface (0-5 cm)

“Root zone”

(0-100 cm)

Data Assimilation - 3D EnKF

- bias mitigation

^∗

- filter parameters

^∗

- Surface soil moisture ( ∼ top 5 cm)

- Root zone soil moisture ( ∼ top 1 m)

- Other consistent geophysical fields (e.g. land surface fluxes), with error estimates - 3-hourly, 9 km, ∼ 2.5 days latency, 3 file collections (gph, aup, lmc)

⇒ * calibration using long-term SMOS record

(39)

L4 SM Product

Results Conclusion

32 / 44

1 September 2015, 00z

Surface Soil Moisture Uncertainty

−180 −150 −120 −90 −60 −30 0 30 60 90 120 150 180

80 60 40

20

0

−20

−40

Surface Soil Moisture [m³/m³] (avg=0.210)

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65

−180 −150 −120 −90 −60 −30 0 30 60 90 120 150 180

80 60 40

20

0

−20

−40

Surface Soil Moisture Uncertainty [m³/m³] (avg=0.020)

0.00 0.01 0.02 0.03

Root-Zone Soil Moisture Uncertainty

−180 −150 −120 −90 −60 −30 0 30 60 90 120 150 180

80 60 40

20

0

−20

−40

Root−Zone Soil Moisture [m³/m³] (avg=0.211)

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65

−180 −150 −120 −90 −60 −30 0 30 60 90 120 150 180

80 60 40

20

0

−20

−40

Root−Zone Soil Moisture Uncertainty [m³/m³] (avg=0.010)

0.00 0.01 0.02 0.03

(40)

L4 SM Product

Results Conclusion

33 / 44

Soil Temperature Uncertainty

−180 −150 −120 −90 −60 −30 0 30 60 90 120 150 180

80 60 40

20

0

−20

−40

Soil Temperature [K] (avg=292.646)

260 275 290 305 320

−180 −150 −120 −90 −60 −30 0 30 60 90 120 150 180

80 60 40

20

0

−20

−40

Soil Temperature Uncertainty [K] (avg=0.733)

0.00 0.50 1.00 1.50 2.00

satellite swaths + complete model fields:

- smooth global 9-km fields

- reduction in uncertainty during assimilation

operational data publicly available https://nsidc.org/data/smap/smap-data.html

(41)

In Situ Evaluation

Results Conclusion

34 / 44

Intensively monitored CalVal watersheds

Slide withdrawn from on-line publication.

Please contact Rolf.Reichle@nasa.gov or

Gabrielle.DeLannoy@kuleuven.be for more

information.

(42)

Core Site Validation

Results Conclusion

35 / 44

Slide withdrawn from on-line publication. Please contact Rolf.Reichle@nasa.gov or

Gabrielle.DeLannoy@kuleuven.be for more information.

(43)

Core Site Validation

Results Conclusion

36 / 44

Slide withdrawn from on-line publication. Please contact Rolf.Reichle@nasa.gov or

Gabrielle.DeLannoy@kuleuven.be for more information.

(44)

Core Site Validation

Results Conclusion

37 / 44

Slide withdrawn from on-line publication. Please contact Rolf.Reichle@nasa.gov or

Gabrielle.DeLannoy@kuleuven.be for more information.

(45)

Core Site Validation

Results Conclusion

38 / 44

Slide withdrawn from on-line publication. Please contact Rolf.Reichle@nasa.gov or

Gabrielle.DeLannoy@kuleuven.be for more information.

(46)

Assimilation Diagnostics

Results Conclusion

39 / 44

High O-F in sparsely vegetated areas with large soil moisture variability

SMAP not assimilated

where SMOS was contaminated:

need bias information From brightness temperature O-F

to increments in surface and root-zone soil

moisture and soil temperature

(47)

Example

Results Conclusion

40 / 44

Assimilating 10,000 L1C obs every 3 hours

(48)

Example

Results Conclusion

41 / 44

SMAP corrects for missed precipitation!

(49)

Conclusion

Land Surface SMOS DA SMAP DA Conclusion

42 / 44

(50)

Conclusion

43 / 44

■

Experimental DA:

SMOS 36-km brightness temperature (Tb) or soil moisture (SM) retrievals?

◆

similar ‘domain-averaged’ soil moisture skills; large local skill differences

◆

most improvement in favorable areas

◆

very different assimilation diagnostics (innovations, increments)

◆

recommendation: localized optimization of observations (i.e. SM) and EnKF

parameters (e.g. obs error), keep experimenting...

(51)

Conclusion

43 / 44

■

Experimental DA:

SMOS 36-km brightness temperature (Tb) or soil moisture (SM) retrievals?

◆

similar ‘domain-averaged’ soil moisture skills; large local skill differences

◆

most improvement in favorable areas

◆

very different assimilation diagnostics (innovations, increments)

◆

recommendation: localized optimization of observations (i.e. SM) and EnKF parameters (e.g. obs error), keep experimenting...

■

Operational DA:

SMAP L4 SM: assimilation of 36-km L1 Tb

◆

global 3-hourly 9-km surface and root-zone soil moisture

◆

meets 0.04 m

³

/m

³

accuracy requirement

◆

improvements need to be considered in terms of confidence intervals

◆

operational, publicly available

All of the above research is performed with NASA funding

gabrielle.delannoy@kuleuven.be

(52)

Literature

44 / 44

■ Entekhabi, D., Yueh, S., O’Neill, P., Kellogg, K. and co-authors (2014). SMAP Handbook, JPL Publication JPL 400-1567, Jet Propulsion Laboratory, Pasadena, California, 182 pages.

■ Reichle, R. H., De Lannoy, G.J.M. , Liu, Q., Ardizzone, J.V., Chen, F., Colliander, A., Conaty, A., Crow, W., Jackson, T., Kimball, J., Koster, R.D., Smith, E.B. (2016). Soil Moisture Active Passive Mission L4 SM Data Product Assessment (Version 2 Validated Release). NASA GMAO Office Note, No. 12 (Version 1.0), National Aeronautics and Space Administration, Goddard Space Flight

Center, Greenbelt, Maryland, USA, 55pp.

■ De Lannoy, G.J.M., Reichle, R.H., Peng, J., Kerr, Y., Castro, R., Kim, E.J., Liu, Q. (2015).

Converting Between SMOS and SMAP Level-1 Brightness Temperature Observations Over Nonfrozen Land. IEEE Geoscience and Remote Sensing Letters, 12(9), 1-5,

doi:10.1109/LGRS.2015.2437612.

■ De Lannoy, G.J.M., Reichle, R.H. (2016). Global Assimilation of Multi-Angle and Multi-Polarization SMOS Brightness Temperature Observations into the GEOS-5 Catchment Land Surface Model for Soil Moisture Estimation. Journal of Hydrometeorology, 17(2), 669-691, doi:

10.1175/JHM-D-15-0037.1.

■ De Lannoy, G.J.M., Reichle, R.H. (2016). Assimilation of SMOS Brightness Temperatures or Soil Moisture Retrievals into a Land Surface Model. Hydrology and Earth System Sciences,

doi:10.5194/hess-2016-414.