Influence of LAI-NDVI relations on runoff modelling at basin scale

Remote sensing data are used in a semi-distributed hydrological model, aiming at the evaluation of water balance in a Mediterranean semi-arid basin. The DREAM model is applied to an intermittent stream of Southern Italy (Carapelle, basin area: 506 km2 ) at daily temporal scales considering rainfall and discharge series (period: 2006-2008). The spacetime distribution of vegetation is represented by the Leaf Area Index, derived by a simple remote sensing index (NDVI). The absence of many basic information has suggested the use of different approaches to estimate model parameters. Three main items have been investigated: potential evapotranspiration (Penman-Monteith, Thornthwaite and Hargreaves equations), vegetation state described by satellite images with different spatial resolution (NOAAAVHRR and MODIS) and different LAI-NDVI regressions (Beer’s and Caraux-Garson’s laws), soil properties (pedotransfer functions). The results, supported by various efficiency criteria, underline that a key role in runoff modelling, including detection of intermittency, is played by the empirical relationship used to obtain LAI from NDVI. In particular, the use of the logarithmic relationship, rather than the linear one, seems to be better suited to represent the spatial variation of the vegetation cover in the investigated area.