Comparative study of oil palm productivity estimation models: multilinear regression and holt-winters approach

Oil palm plays a crucial role in Indonesia’s economy through the increased production and export of oil palm. However, managing and predicting productivity in expanding oil palm plantations presents significant challenges. Current estimation methods are resource-intensive and time-consuming. This study advocates the application of remote sensing techniques, particularly through Google Earth Engine and Landsat 8 satellite imagery, for efficient oil palm productivity estimation. Using a multilinear regression model that integrates the Normalized Difference Vegetation Index (NDVI) and rainfall data from the CHIRPS dataset, historical productivity data is also incorporated using the Holt-Winter exponential smoothing approach. The research aims to enhance oil palm plantation management by harnessing remote sensing technology and weather data analysis. The research results revealed that multilinear regression for the universal model with annual data accumulation using the 12-month shifting rainfall dataset achieves an R 2 of 0.855, RMSE of 14,283 tons/ha, and MAPE of 0.975% and the AK12_M3 was the best combination to build a model for each block Meanwhile, the Holt-Winter method for universal mode as well as each block model, using a 4-month seasonal approach produces the best results. With an R 2 of 0.83, RMSE of 6,256 tons/ha, and MAPE of 7.501%.