Integration of in situ measured soil status and remotely sensed hyperspectral data to improve plant production system monitoring: Concept, perspectives and limitations

Abstract

A common problem in agricultural remote sensing is the sub-pixel spectral contribution of background soils, weeds and shadows which impedes the effectiveness of spectral vegetation indices to monitor site-specific variations in crop condition. To address this mixture problem, the present study combines in situ measured soil status and remotely sensed hyperspectral data in an alternative spectral unmixing algorithm. The model driven approach, referred to as Soil Modeling Mixture Analysis (SMMA), combines a general soil reflectance model and a modified spectral mixture model providing as such the opportunity to simultaneously extract the sub-pixel cover fractions and spectral characteristics of crops. The robustness of the approach was extensively tested using ray-tracer data (PBRT) from a virtual orchard, and results showed an improved monitoring of the crop's chlorophyll, water content and Leaf Area Index (LAI). A significant increase in the R2 between vegetation indices and the biophysical parameters was observed when index values were calculated from the pure vegetation signals as extracted by SMMA as opposed to index values calculated from the original (mixed) image pixels (GM1: ΔR2=0.19; MDWI: ΔR2=0.38; sLAIDI: ΔR2=0.14).