Validation of a large scale hydrological model with data fields retrieved from reflective and thermal optical remote sensing data – A case study for the Upper Rhine Valley

• Published in: Physics and chemistry of the earth. Part B, Hydrology, oceans and atmosphere Vol. 33; no. 17; pp. 1061 - 1067
• Main Authors: Casper, Markus C., Vohland, Michael
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 2008
• Subjects: Evaporation; Evapotranspiration; Landsat-TM; Marine; Model validation; Remote sensing; Thermal IR; France; ANE, Germany; Europe, Rhine R; France, Vosges Mts; Switzerland, Basel; Evaporation; Model validation; Thermal IR; Remote sensing; Landsat-TM; Evapotranspiration
• ISSN: 1474-7065; 1464-1909; 1873-5193
• DOI: 10.1016/j.pce.2008.06.001

For the entire Upper Rhine Valley between Karlsruhe and Basel, a long term simulation (1985–2002) with the GWN_BW model (partly physically based 1-D water balance model) resulted in the retrieval of the following hydrological process variables: daily potential and actual evaporation, surface runoff from sealed surfaces and groundwater recharge. Meteorological data has been interpolated from all available stations in France, Germany and Switzerland including the mountainous regions of the Vosges Mountains and the Black Forest. The primary grid size of the model was 500 m; for landuse, the sub-grid variability has been taken into account additionally. In an alternative approach, Landsat-5 TM scenes from two different acquisition dates were integrated to model the daily rates of actual evaporation ( E a). To this end, data fields retrieved from both the thermal and reflective Landsat channels were combined with ancillary meteorological and digital elevation data. Assuming a single layer canopy-soil system, the daily E a rates were estimated from the modelled net radiation and the differences between maximum surface and maximum air temperatures; the final partitioning of sensible and latent heat fluxes was strongly determined by the pixel-wise derived vegetation abundances. The results of the remote sensing approach were compared quantitatively to the E a rates provided by the hydrological model by means of both correlation and geostatistical pattern analysis. Extreme differences between both approaches were detected. The low spatial variability of the simulated E a was explained by the parameterisation scheme for surface resistances. In total, the hydrological model often underestimated E a due to a non realistic representation of soil water availability under deciduous forests and a missing representation of irrigation. In addition, a static and very simple representation of capillary rise of groundwater was found to cause large overestimates of the modelled E a during periods with low groundwater table.