Monitoring soil Infiltration in Semi-Arid Regions with METEOSAT and a Coupled Model Approach using PROMET and SLC
Klug, Philipp; Bach, Heike; Migdall, Silke
VISTA Remote Sensing in Geosciences GmbH, GERMANY
Introduction
In semi-arid and arid areas, to which the Mediterranean belongs to, infiltration of the relatively sparse rainfall in the soil is a critical quantity for the water cycle. It is often assumed that sporadic rainfalls are directly evaporated without having the potential to recharge groundwater. An accurate simulation of infiltration could thus reduce the uncertainty in water balance assessment in arid and semi-arid regions.
How remote sensing data could contribute to this goal was studied in the frame of the EU FP7 project CLIMB (Climate Induced Changes on the Hydrology of Mediterranean Basins) that aims at employing integrated hydrological modeling in a new framework to reduce existing uncertainties in climate change impact analysis of the Mediterranean region. In CLIMB, remote sensing is used for monitoring purposes as well as for the hydrological parameterization. The physically based hydrological model PROMET (Processes of Radiation, Mass and Energy Transfer) models fluxes of energy and matter on the land surface of microscale (100 km2) up to mesoscale (100 000 km2) catchments and has been developed and tested in a variety of studies. With PROMET it is possible to simulate the infiltration process in detail, since 4 soil layers together with the hourly calculation time step allow simulating the vertical water transport. The adequate parameterization of the hydraulic soil properties is however crucial for a correct simulation of the infiltration and the vertical water transport. The ground water recharge in PROMET is equated with the percolation of the soil water from the 4th layer.
Changes of moisture of the soil surface are visible in measured surface reflectance, as moist soils are darker than dry soils. The idea of the research was to use this known change of reflectance with moisture conditions for a better understanding of the infiltration process. Temporal high resolution time series of surface reflectance values were used to observe the drying process of the surface. METEOSAT Second Generation (MSG) with its sensor called SEVIRI is a spinning geostationary Enhanced Vis & IR Imager with 12 spectral channels. It images every 15 minutes with a 3 km horizontal 'sampling distance' at sub-satellite point and therefore is predestined to be used for the monitoring of the soil drying process.
Methods
The chosen test site for the research was the area around the city Gafsa in Central Tunisia (34.42°N, 8.79°E) which is characterized by arid climate and yearly mean precipitation sums of fewer than 200 mm. A meteorological station is installed at Gafsa which offers continuous meteorological data on precipitation, temperature, humidity, wind speed and cloud cover to the public through the World Meteorological Organization (WMO).
After selecting a heavy rain event with clear sky conditions in the days after the rain event, the corresponding MSG data was atmospherically corrected and surface reflectance values were calculated. After correcting the effect of the bi-directional reflectance distribution function (BRDF) first investigations on the change of reflectance during the drying process could be carried out.
In the next step PROMET was run with meteorological data (precipitation, temperature, wind speed, relative humidity and cloud cover) of the Gafsa meteorological station for the selected rain event. Since the actual soil parameters are unknown and most likely sand is the predominant soil component, loamy sand (60% sand) and medium sand (95% sand) were used as soil types. Then, the model was run with a set of soil parameters for these two soil types.
The results of the model runs with PROMET, with soil moisture as exchange variable, were assimilated into the surface reflectance model SLC, which allows simulating hourly surface reflectance values during the drying process after a rain event.
Results
Investigations at the MSG BRDF corrected bottom of atmosphere reflectance values showed that the soil was darker after the rain event due to higher soil moisture and was getting brighter again in the following hours.
The simulated reflectance values derived by coupling the results from the hydrological model PROMET with the surface reflectance model SLC also showed darker reflectance values after the rain event which increase afterwards. These values were compared to the observed MSG reflectance values. The simulated reflectance values varied depending on the used hydraulic soil parameters. By calculating the RMSE between simulated and observed reflectance values, the best soil parameter set was selected.
For sensitivity analyses, PROMET was also run for a longer time period (14 months) and the impact of different soil parameter sets on the simulated evaporation and ground water recharge was investigated. Significant differences in the modeled evaporation and ground water recharge were found for the applied soil parameter sets. The total precipitation sum for the modeled 14 months period is 246 mm. Depending on the used hydraulic soil parameters the evaporation sum varies between almost 100 mm and around 65 mm. The ground water recharge sum varies between 185 mm and 135 mm. The results of the best fitting hydraulic soil parameters, as chosen by comparison of modeled surface reflectance values to MSG reflectance values, resulted that 54% of the precipitation recharged the ground water.
First investigations of the results have shown that with a coupled model approach of the hydrological model PROMET and the surface reflectance model SLC, a soil parameterization can be carried out by comparing modeled and MSG measured surface reflectance values. Using these soil parameters, a more precise monitoring of soil infiltration is expected with the hydrological model PROMET. As a result, information on ground water recharge can be retrieved und uncertainties regarding soil properties can be evaluated and minimized.
Further validations of the coupled model approach will be carried out with MSG data for a second strong rain event with over 80 mm precipitation in 2 days.