Feeding the global population while minimizing adverse impacts on natural landscapes is integral to environmental sustainability & biodiversity conservation. Achieving zero deforestation in commodity crop supply chains requires continual monitoring of these areas.

This program estimates soybean area expansion at the continental scale in South America over a twenty year period.

The program is based on a probability sampling design that stratifies a population of 20km blocks across the continent into ‘high’, ‘medium’ and ‘low’ blocks according to soybean intensity (as primary sampling units). We employ stratified random sampling across the population for target ground truthing. Coordinating the fieldwork includes building a fieldwork data recording system that syncronises multiple field teams data together during offline/online phases, while producing live area estimates and uncertainties (SE) across the population. After experimenting with Survey123 and QField, I built a custom solution in google sheets and obtained recent satellite imagery over the target areas for accurate crop identification in the field.

During fieldwork we were able to collect waypoints to enable both a secondary sample unit area estimate and for use as training points for in-season mapping. From these training points you can create a wall to wall soybean map for South America and calculated regression based area estimates to reduce the uncertainties. Publishing this study in nature sustainability has seen hundreds of citations and thousands of views. Methodologically similar estimates were also derived for the USA for many years.

The era of big data has meant we can map crops reliably by producing area estimates with known error margins at national and continental scales using custom solutions that incorporate almost entirely free software.