Forecasting of postharvest fresh produce waste at the wholesale level using time series models

Postharvest fresh produce waste at the wholesale level presents significant challenges for the food industry, as wholesale markets function as key storage and distribution points. Accurate forecasting of waste trends can enable stakeholders to implement evidence-based management and mitigation strategies. This study aims to forecast waste in wholesale produce by utilizing historical data with Autoregressive Integrated Moving Average (ARIMA) and Error, Trend, and Seasonality (ETS) time series models, targeting diverse fruit and vegetable categories. The models were applied to fruit categories (berries, citrus, melons, pome, drupes and others, stone fruits and tropical fruits) and vegetable categories (alliums, cruciferous, Cucurbitaceae, leafy vegetables, legumes, root vegetables and Solanaceae) for an 18-month waste forecast period. For fruit waste prediction, ARIMA (0,1,1)(0,1,0)[12] on citrus fruits achieved the lowest Mean Absolute Percentage Error (MAPE) values of 43.30% for ETS and 29.61% for ARIMA. In the vegetable category, ARIMA (0,1,1)(0,1,0)[12] on leafy greens showed the best performance, with MAPE values of 20.49% for ETS and 17.62% for ARIMA. These results demonstrate the models’ potential for delivering accurate, category-specific forecasts, with the ETS model yielding narrower confidence intervals than ARIMA, suggesting its greater suitability for the data used in this study. This study highlights the use of time series models for evidence-based forecasting in fresh produce waste management, providing a valuable tool for stakeholders aiming to reduce waste at the wholesale level. The findings highlight the potential of these models to support strategic decisions in food distribution and contribute to waste mitigation efforts within the fresh produce supply chain.