Assessing the fidelity of Landsat-based fAPAR models in two diverse sugarcane growing regions

Abstract

Sugarcane is a globally important crop used for producing sugar and for generating renewable energy. Timely and accurate forecasts of sugarcane yield and production are needed to optimize supply chain operations. Crop growth models (CGMs) are frequently used for sugarcane yield forecasting and have been shown to benefit from using remotely sensed data to force (calibrate) biophysical state variables, such as the fraction of absorbed photosynthetically active radiation (fAPAR). Little is known about the robustness of multispectral vegetation indices for modelling fAPAR in sugarcane growing regions were environmental conditions and farming practices are diverse. This study investigated how the relationships between multispectral Landsat-8 satellite imagery and in situ sugarcane fAPAR measurements vary over large heterogeneous areas. Specifically, it examined which spectral bands and indices are most appropriate for modelling fAPAR under particular production environments and assessed the robustness of the models for application in areas where sugarcane is grown under varying agro-climatic conditions. It was found that cropping and environmental conditions were the main drivers of sugarcane fAPAR modelling success. Significantly (40%) lower mean root mean squared errors (RMSEs) values were recorded in Pongola, which is attributed to the relatively homogenous conditions under which sugarcane is being grown in this area. Generally, the Sezela models were much weaker and the normalized difference vegetation index (NDVI) and soil-adjusted vegetation index (SAVI) models performed relatively poorly, with the best performing models being dominated by the SWIR bands and/or indices generated from it. The non-linear models dominated and are thus recommended for operational implementation owing to their relative simplicity and robustness. From these results we conclude that the use of remotely sensed data for estimating fAPAR throughout the growing season is highly beneficial, but that the selection of suitable variable (index) is critical, especially when the sugarcane area being considered is diverse in terms of farming practices, terrain and climate.