Identifying yield stability and drivers of yield variability in cotton using multi-layered, whole-farm datasets

Particular areas within a cotton field often yield consistently high or low, but other areas fluctuate from season to season. Farmers have an idea of the cause of yield variability in cotton, but this is generally qualitatively assessed. The amount of spatial and temporal data that is available today is a promising opportunity to quantitatively identify areas that are constrained, and identify possible causes. This study collates yield monitor data from 16 different seasons, and publicly-available spatial data (digital soil maps, climate, etc.) from a large 13,000 ha cotton farm in Australia. The objectives were to (1) assess the spatial-temporal stability of cotton yield, and (2) use machine learning to model cotton lint yield to identify the most important driving factors of variability.