Investigating the relationship between soil moisture and evapotranspiration at different surfaces: Do we improve fluxes just improving the land use in models?

C. Román-Cascón*1, 2, M. Lothon2, F. Lohou2, A. Brut3, O. Hartogensis4, O. Merlin3, N. Ojha3, C. Yagüe5, R. C. Soriguer6, R. Díaz-Delgado6, A. Andreu7 and M. P. González-Dugo7

Introduction

Final ObjectiveTo learn more about these relationships in order to improve

model forecasts

SOIL MOISTURE

SURFACE FLUXESLAND

COVER

RADIATIONATMOSPHERIC

VARIABLES(temperature,

humidity, wind)

THESE VARIABLES PRESENT BIASES

WHEN MODELLED(MAIN MOTIVATION!)

Figure 1. The land cover (type and state of vegetation) affects the net radiationthrough changes in albedo, surface temperature and emissivity. It also affects theavailable soil moisture by changing the water retention, runoff, evaporation andtranspiration. Thus, the final sensible (red arrow) and latent (blue arrow) heatfluxes are affected by the net radiation, the land cover and the water availability inthe soil. These delicate processes depend on each other and representing them innumerical models is a challenge, which affects the simulation of importantvariables such as 2-m temperature or 10-m wind speed.

THESE RELATIONSHIPS ARE NOT COMPLETELY WELL

UNDERSTOOD!!

Data & Strategy

Results

IN SITU DATA

SATELLITE DATA

MODELS

Soil Moisture

Comparison of correlations and relationships between soil moisture (SM) & latent heat flux

(Le) over different vegetation and periods.

High-resolution realistic land use (LU) from CESBIO[4] used to improve the LU

representation in the models. 1-km soil moisture (DISPATCH[5]) to evaluate

the SM spatial variability simulated by models.

Weather Research and Forecasting (WRF) model[6] & noah-mp land-surface scheme[7].

Sensitivity experiment: To what extent can we improve the modelled fluxes with

improved LU maps?

Surface Fluxes

Soil MoistureLand Use

Figure from Noah land-surface model site

Data from ICOS[1]

in France and fromEBD[2] and IFAPA[3]

in Spain

0 5 10 15 20x

0

5

10

15

20

conif-1%

wildforest-27%

conif-0%

foret feullius-3%

mixed forest-19%

closed shrub-0%open shrub-0%

savanna woody-30%

savanna-9%grassland-6%

permanent wetland-1%

croplands-1%

urban-0%cropland/natural mosaic-4%

Figure 4. a) WRF default LU map used in the reference simulation (Mod. IGBP-MODIS NOAH). The central point is the main site (60-m tower) of the BLLASTfield campaign[9] in Southern France (2011) which data are used for theevaluation. b) More realistic LU used in a new simulation, obtained fromCESBIO[4] LU data (30-m res).

x (km)

y (k

m)

0 5 10 15 20x

0

5

10

15

20

conif-2%

wildforest-0%

conif-0%

foret feullius-38%

mixed forest-0%

closed shrub-0%open shrub-0%

savanna woody-0%

savanna-0%grassland-53%

permanent wetland-0%

croplands-4%

urban-2%cropland/natural mosaic-0%

x (km)

y (k

m)

a) b)

WRF modified (new) LU in the area of study

WRF default LU in the area of study

Figure 5. a) Normalized SM RMSE calculated for a 19x19 km area aroundthe site of study. For the evaluation of the soil moisture heterogeneity,DISPATCH[5] data are used. b) Sensible heat flux (SH) mean RMSE for thesame area. c) Le mean RMSE for the same area. *Pixels where thedominant LU is forest or urban are not evaluated due to uncertainties in themeasurements. Part of these RMSE reductions are due to improvements inthe Rnet through changes in albedo (not shown).

Conclusions

0 0.1 0.2 0.3 0.40

0.5

1

0 0.1 0.2 0.3 0.40

0.5

1

0 0.1 0.2 0.3 0.40

0.2

0.4

0.6

0 0.1 0.2 0.3 0.40

0.2

0.4

0.6

0 0.1 0.2 0.3 0.40

0.2

0.4

0.6

CROPSHUMID SITE

(South France)[1]

Wheat

Le/R

net

Soil moisture (m3 m-3)

Le/R

net

Le/R

net

CROPSHUMID SITE

(South France)[1]

Corn

TREESDRIER SITE

(South Spain)[3]

Holm Oak Junipers

SHRUBDRIER SITE

(South Spain)[2]

Erica WheatFrance

CornFrance

QuercusSpain

JuniperusSpain

EricaSpain

0,3 0,36 0,56 0,50 0,39

-0,22 0,11 0,22 0,50 0,06

Different SM & Le/Rnet relationships depending on:

- Type of plant- Season / plant period- Region

- Crops already grown in sites no water-limited do NOT respond (Le/Rnet) to SM changes (dark green symbols in Fig. 2 a and b).

- Some sites do NOT respond to SM changes (e.g. Erica sp., very well adapted to extremes?) (Fig. 2e)

- Different tree species respond differently to SM changes in periods of low activity (winter) (Blue symbols in Fig. 2 c and d).

- What about models? Do they represent well these relationships?

IN SITU data

Figure 2. a-e) Scatter plots of soil moisture (x-axis, in m3 m-3) and latent heat flux (Le) divided by net radiation (Rnet) (y-axis, no units) fordifferent species in different areas (indicated in the plots). Daily values are used, calculated as the mean of each variable between 10 and 14h local time. Periods indicated in legend (right). f) Table with correlations between soil moisture and Le/Rnet for different sites and periods.

.

1st plant period

2nd plant periodBare soil (no plant)

Winter period

Summer period

SM & Le/Rnet correlations

a) b)

c) d)

e)f)

TREESDRIER SITE

(South Spain)[2]

MODELS

Sensibility experiment using more realistic LU representation:

- Improvement of albedo.

- Improvement of Rnet.

- Improvement of SM spatial heterogeneity.

- Improvement of Le and SH.

- What happens in drier areas? Is it worth to improve also the vegetation state in the models? FUTURE WORK

(1) Centre Nationale d’Études Spatiales, CNES, France; (2) Laboratoire d’Aérologie, University of Toulouse, CNRS, UPS, Toulouse, France; (3) CESBIO, France; (4) Meteorology and Air Quality Section. Wageningen University, The Netherlands; (5) Departamento de Física de la Tierra y Astrofísica. Universidad Complutense de Madrid, Spain; (6) Estación Biológica de Doñana, CSIC, Spain; (7) IFAPA, Consejería de Agricultura, Pesca y Desarrollo Rural, Spain

Summer period

1st plant period

2nd plant period

Winter period

First step:Model Sensitivity Experiment

“Improving LU data”11

0.24

0.25

0.26

0.27

0.28

0.29

0.3

0.31

0.32

0.33

0.34

Default LU

Realistic LU

0

10

20

30

40

50

60

70

80

W/m

2

Default LU

Realistic LU

0

10

20

30

40

50

60

70

W/m

2

Default LU

Realistic LU

27% of RMSE reduction

29 % of RMSE reduction

16% of RMSE reduction

Spatial evaluation

(19x19 pixels)

a) Default LU

Realistic LU

Default LURealistic LU

Default LURealistic LU

RMSE for Sensible Heat Flux

RMSE for Latent Heat Flux

RMSE for SM (normalized, values between o and 1)

0,24

0,26

0,28

0,30

0,32

0,34

8 12 16 8 12 16 8 12 16Hour (UTC) Hour (UTC) Hour (UTC)

0

20

40

60

80b)

0

20

40

60

c)

W/m

2

W/m

2

W/m

2

The relationship between soil moisture and evapotranspiration depends on the area (humid or drier), the vegetation type and the analysed period (state of vegetation).

A better representation of the vegetation (LU) map in models improves the simulation results, especially through albedo improvements.

In situ data, satellite data and models can be combined to investigate land-atmosphere interactions.

REFERENCES

[1] In situ data: ICOS data (Integrated Carbon Observation System) (https://www.icos-ri.eu/).[2] In situ data: EBD data - Doñana - Díaz-Delgado, R. (2010). In Conservation Monitoring in Freshwater Habitats (pp. 325-337). Springer, Dordrecht.[3] In situ data: IFAPA data – Sta. Clotilde site - Andreu et al. (2018). Remote Sensing, 10(4), 567.[4] Satellite data (for model experiment): CESBIO 2011 land use data from Landsat 5 (http://osr-cesbio.ups-tlse.fr/~oso/).[5] Satellite data (for model evaluation): DISPATCH data. Merlin et al. (2013). Remote Sens Environ, 130, 25–38.[6] Modelling: WRF model. Skamarock et al. (2008). NCAR Technical note-475+ STR.[7] Modelling: Noah-mp (http://www.jsg.utexas.edu/noah-mp).[8] Fig SM/LE: Seneviratne et al. (2010). Earth-Sciences Review 99. 125-161.[9] Model evaluation: BLLAST. Lothon et al (2014). Atmos Chem Phys, 14(20), 10931-10960.

Area-averaged fluxes &

satellite SM data for model evaluation

(1 km pixels)

2 areas analysed: Southern France (humid) and

Southern Spain (drier)

* Presenter, contact at carlosromancascon@ucm.es

Figure 3. Soil moisture regimes and correspondingET. Figure from Seneviratne et al. 2010[8].