Early Within-Season Yield Prediction and Disease Detection Using Sentinel Satellite Imageries and Machine Learning Technologies in Biomass Sorghum

Machine learning models based on remote and proximal sensing as potential methods for in-season biomass yields prediction in commercial sorghum fields

“…The simple linear model was used as a benchmark to measure the relative performance of the models implemented. The other models (xgbTree, nnet, rf) evaluated in this work were selected based on the robustness they displayed in previous studies [4,5,59,60].…”

“…This crop is a staple cereal across countries under lower latitudes, but it is resurging under higher latitudes for feed, biofuel, the manufacture of specialty health-promoting foods rich in antioxidants, and for use as a substitute for traditional grains diet and meeting gluten-free needs [4][5][6][7][8]. Biomass sorghums of biofuel production interest were used in this work and were described in previous studies [4,5]. They included dual purpose (showing high grain and biomass yields), forage, sweet, and biomass per se sorghum types [7,8].…”

“…The aim of the yield forecasting is to provide scientifically sound and independent estimates of crops' yield, as precise and as early as possible within the cropping season. Crop yield forecast supports farming operations through better planning of harvest, machineries, and logistics, and helps avoid food and commodity crises [4,5,[9][10][11][12]. Crop yield forecasting can be achieved using three methods: field surveys, remote and proximal sensing models and crop simulation models [4,5,13].…”

“…Crop yield forecast supports farming operations through better planning of harvest, machineries, and logistics, and helps avoid food and commodity crises [4,5,[9][10][11][12]. Crop yield forecasting can be achieved using three methods: field surveys, remote and proximal sensing models and crop simulation models [4,5,13]. Traditionally, yield forecasting was carried out by experts through the evaluation of crop status (e.g., tiller number, spikelet number, spikelet fertility percentage, percentage of biotic adversity damage, percentage of weeds infestation) throughout the crop growing season, and using these observations and measurements to forecast yields using regression methods, or relying on the experts knowledge [13].…”

“…Crop monitoring technologies have been used to exploit this phenomenon, including satellites and handheld sensors measuring light in narrow wavebands or wavelength intervals. Plant reflectance measurements have been successfully used in several modelling instances including the quantification of canopy vigor [24][25][26], nutrient and soil moisture stresses [27,28], and to predict yields [4,5].…”