Abstract
Indices based on optical satellite remote sensing imagery have shown to be suitable for quantifying drought-related yield losses. The Forage-Production-Index (FPI), combining meteorological observations and remote sensing-based LAI retrievals from MODIS, was adapted for the application in South Tyrol (NE Italy) in a mountainous, highly heterogeneous landscape. Yield measurements from field trials covering 39 environments (site x year) were used for validation, which was performed using mixed models describing the relationship between dry matter yield and FPI (or their variation with respect to a reference period) and accounting for the design effects treated as random factors. Following variants in computing FPI were applied: spectral unmixing of LAI, correction by means of Water Stress Coefficient (CWS) and aggregation scale. The prediction ability of the index was found to be low. Unmixing and correction by CWS resulted in a minor improvement in accuracy. Possible reasons for the low sensitivity are: i) insufficient spatial resolution of MODIS satellite data with respect to the complexity of land use; ii) lack of coincidence between yield at validation sites and surrounding grassland; iii) small number of validation sites, possibly not covering the whole yield variation over the area and period investigated.