Assimilation of hyper spectral data into crop growth models Precision farming application for maize in Catalonia

摘要

We propose a general framework for the derivation of crop information from hyperspectral airborne CASI data using physical radiative-transfer models with an assimilation link to existing crop growth models. This framework is designed for precision farming and crop monitoring applications, which take into account the agricultural and climatic conditions in Catalonia. The derivation of crop parameters is based on physical radiative-transfer models (PROSPECT and ACRM), which are numerically inverted using optimization algorithms (Simplex, Quasi-Newton, and Levenberg-Marquardt). Additionally, we propose a method for the inclusion of a priori information in order to solve the ill-posed problem. In a further processing step, existing crop growth models (ISBA and STICS) assimilate the derived crop interface parameters to benefit from the area wide and spatial high resolution information. Several measurement campaigns were conducted to calibrate and validate the physical radiative-transfer and crop growth models. In them, CASI airborne hyperspectral imagery and simultaneously in-situ measurements of crop and soil parameters were acquired on six different dates during the crop growth period in 2004. Six test sites with two maize hybrids (Oropesa and Eleonora) were selected for the in-situ measurements. The validation shows good correspondence of the derived canopy structure parameters with the in-situ measured values, whereas the derived leaf chlorophyll and nitrogen contents is not validated Further investigations will show the potential of different assimilation techniques to improve the output of the crop growth models.