Comparing Forecasting Models for Potato Production: Evaluating T-ARMA, ARIMA-ARCH, Weibull and Score-Driven Approaches in Major Global Producers

This study examines trends in potato production across five major global producers (China, India, Russia, Ukraine and the US) using annual data from 1961 to 2022. We evaluate the T-ARMA, ARIMA-ARCH, Weibull and score-driven models to forecast production from 2023 to 2030. The results identify the T-ARMA model as optimal for China and India (validation MAE: 1,668 thousand tonnes  and 3,277 thousand tonnes, respectively), while the ARIMA-ARCH and Weibull models perform best for Russia (MAE: 1,405 thousand tonnes) and the US (MAE: 547 thousand tonnes), respectively. Projections indicate a marginal decline in China (from 96,073 to 96,066 thousand tonnes; − 0.007%) and India (from 55,701 to 55,698 thousand tonnes; − 0.005%), contrasted by robust growth in Russia (from 20,026 to 26,284 thousand tonnes; + 31.2%). Ukraine’s production remains stable at approximately 21,103 thousand tonnes annually, while the US experiences a slight decline (from 18,054 to 18,053 thousand tonnes; − 0.005%). These forecasts, derived from regionally tailored models, offer actionable insights for policymakers to strengthen food security, optimize supply chains and prioritize agricultural innovation.