Modeling crop growth using multisensor analysis in remote sensing

Abstract

These last years, the use of spaceborne remote sensing and unmanned aerial vehicles (UAV) has grown exponentially in agronomy. Their abilities are theoretically complementary in terms of temporal coverage and spatial resolution. Thiswork aims to compare both approaches at the scale of a winter wheat experimental parcel during a complete growing season using green fractional cover time-series (FCOVER) and combine them to improve crop growth characterization. UAV multibands images and Sentinel2 images are analyzed on the same time interval. Eventually, the influence of landscape elements on crop growth-related variables is studied. The methodological results of this study are the processes used to transpose FCOVER time-series into a reduced amount of crop growth parameters and quantify their uncertainties. These parameters allow predicting yield using the Aquacrop Model and finally summarizing this information on a set of maps. A comparison between yield predictions to a reference yield map based on field measurement shows that yield prediction using S2 (resp.UAV) FCOVER tends to underestimate (resp. overestimate), while data combination tends to be closer to reference values.UAV provides earlier and faster growth curves, reaching higher maxima. Growth process variables are compared to covariables describing topography, the presence of historical charcoal kilns, and the ploughing date. South facing half of the parcel experiences faster growth and higher yield; an earlier ploughing date and biochar patches emphasize this trend.