PubMed · 2026-06-10
Researchers built a new forecasting system that combines climate modeling and machine learning to predict oil palm yields in southern Thailand with much greater accuracy. The results suggest oil palm production there is broadly resilient to climate change through the end of the century, with only a modest mid-century dip.
The hybrid APSIM + Random Forest model reduced yield prediction error from 15.51 t/ha (model alone) to 2.74 t/ha (hybrid averaged across sites) — roughly a 6-fold improvement.
Seasonal yield forecasts driven by downscaled data matched the accuracy of those using observed reanalysis data, enabling reliable predictions up to eight months ahead.
CMIP6 climate projections show oil palm yields in southern Thailand remain stable or slightly higher early this century, dip modestly at mid-century, then stabilize again by 2100 — suggesting relative climate resilience.
