We have developed a prototype simulation of a localized agricultural system to test different policies for improving operations. In the simulation, we demonstrate how computational simulation could increase profits for a palm oil aggregator in the state of Sabah in Malaysian Borneo. Given long startup and shutdown times, aggregators must keep palm oil mills constantly running. Aggregators must also process palm oil the same day they collect it. Idle mill periods and expired crops lower profits. Purchasing more collection trucks or changing their routes could raise profits by delivering a greater amount of fresh crop to mills to keep them running at full capacity. At the same time, trucks cost money to operate and maintain, some plantations are closer to mills than others, and palm oil may be ready at different times depending on its growth rate and seasonality.
In the face of these complexities, aggregators may over- or under-invest in transport or dispatch trucks along suboptimal routes, causing them to lower profits or turn a loss. To assist aggregators, we constructed a computational simulation of palm oil production, collection, and processing with spatial data representing locations of plantations and mills in Sabah. In the simulation, aggregators can test different options and find which ones maximize profits. Comparing profits over time between a scenario where aggregators send trucks on suboptimal routes and buy too many trucks and the optimal scenario shows the increase in profits. In this case, computational simulation raised aggregator profits several fold.
We also adapted the simulation to help agricultural aggregators rate palm oil plantations by their risk of expanding and causing deforestation using geospatial modeling and machine learning. The simulation allows aggregators to test how avoiding high-risk plantations affects profits from palm oil. Aggregators can use the simulation to find their optimal risk tolerance balancing profits and forest loss.