Linking process-based potato models with light reflectance data: Does model complexity enhance yield prediction accuracy?

Abstract

Data acquisition for parameterization is one of the most important limitations for the use of potato crop growth models. Non-destructive techniques such as remote sensing for gathering required data could circumvent this limitation. Our goal was to analyze the effects of incorporating ground-based spectral canopy reflectance data into two light interception models with different complexity. A dynamic- hourly scale- canopy photosynthesis model (DCPM), based on a non-rectangular hyperbola applied to sunlit and shaded leaf layers and considering carbon losses by respiration, was implemented (complex model). Parameters included the light extinction coefficient, the proportion of light transmitted by leaves, the fraction of incident diffuse photosynthetically active radiation and leaf area index. On the other hand, a simple crop growth model (CGM) based on daily scale of light interception, light use efficiency (LUE) and harvest index was parameterized using either canopy cover (CGMCC) or the weighted difference vegetation index (CGMWDVI). A spectroradiometer, a chlorophyll meter and a multispectral camera were used to derive the required parameters. CGMWDVI improved yield prediction compared to CGMCC. Both CGMWDVI and DCPM showed high degree of accuracy in the yield prediction. Since large LUE variations were detected depending on the diffuse component of radiation, the improvement of simple CGM using remotely sensed data is contingent on an appropriate LUE estimation. Our study suggests that the incorporation of remotely sensed data in models with different temporal resolution and level of complexity improves yield prediction in potato.