Monitoring Evapotranspiration from Remote Sensing Data for Ground Water Resources Evaluation

Evapotranspiration (ET) is a fundamental variable of the hydrological cycle which plays a major role on surface water balance and surface energy balance. At local scale ET can be estimated from detailed ground observations (eddy covariance towers), but these measurements are only representative of very limited homogeneous area. When regional information is required, e.g. for monitoring ground water resources, the flux tower measurements cannot be used directly and estimation of ET often relies on estimation from meteorological data through potential evapotranspiration formulas. At regional scale remote sensing provides spatially distributed information for mapping and monitoring ET, but this information is still rarely used for ground water assessment. Indeed, remote sensing estimation of ET suffers several drawbacks. In particular, remote sensing information by itself cannot provide a continuous monitoring of ET because of the presence of clouds and the revisit period of the sensor. Another difficulty originates in the lack of exhaustive evaluation of remote sensed ET since accurate ground measurements are scarce and representative of a limited number of homogeneous areas. This has also for consequence that a large number of methodologies to derive ET were developed with no real possibility of a consistent competitive evaluation.
We have developed the EVASPA (EVapotranspiration Assessment from SPAce) tool to monitor ET on a daily basis, together with an evaluation of the associated uncertainties, from remote sensing data in the thermal and the solar domains. This tool combines the estimation of ET from various models and various sources of data, including ground data, meteorological data and data from remote sensing sensors such as MODIS, PROBA-V, SPOT-Vegetation or sensors on board of Sentinel2 and LANDSAT platforms. The multi-model – multi-data estimations of ET are combined following an ensemble approach to generate the final monitoring of ET together with an evaluation of the ET estimation uncertainty.
Surface energy balance and evapotranspiration algorithms were fed by operational biophysical products (albedo, vegetation indices, surface temperature, Leaf Area Index, fraction cover) provided by data centers such as the Land Processes Distributed Active Archive Center (LP DAAC) in the USA, the Copernicus Global Land Service in Europe and the THEIA Land Data Centre in France. In some cases, when the required variables are not available from these data centers, we derive them using specific algorithms, in particular designed at INRA, from spectral reflectances and radiances. Meteorological variables including air temperature, air humidity, wind speed, incoming radiations and in some cases precipitations are obtained from meteorological networks or reanalysis products. For example, over France, the SAFRAN reanalysis from Météo-France provides meteorological data with an 8 km resolution. Over other parts of the world, we use the ERA5 reanalysis at a 0.25° resolution from the ECMWF European Meteorological center.
EVASPA was applied to estimate evapotranspiration over several areas in France, Spain and the Sahelian zone (Niger and Mali) to help in monitoring the water budget of different hydrosystems. The method was first evaluated against flux tower measurements of evapotranspiration over various ecosystems. RMSE ranged between 0.5 and 1 mm/day depending on the ecosystems. When integrated over watershed, ET retrievals were also compared to indirect estimates of evapotranspiration from water balance, stream flow monitoring or other modelling approaches for time period of more than a decade (these include remote sensing operational products such as MOD16 or analysis of atmospheric-hydrological modeling such as the operational Safran-Isba-Modcou application in France). The results highlight the potential use of the retrieved ET for calibrating ground water models (e.g. for estimating aquifer parameters…) or evaluating model inputs (e.g. determination of effective rainfall, identification of irrigated areas…). We also evaluated the impact of the uncertainties in the estimation of ET in the monitoring of ground water. We showed that the main sources of ET uncertainty were related to the uncertainties in incident radiations and surface temperature together with the diversity of ET models. When forced in ground water models, the uncertainties in ET had an impact almost equivalent to the impact of uncertainties in rain inputs.
Eventually, EVASPA was used for generating ET products that will be made available to the scientific community in the next future through the THEIA Land Data Centre, in particular for hydrological applications.