Advanced data analytical methods for citrus crop yield forecasting

Abstract

Citrus supply chains are affected by uncertainties considering farm-to-fork perspective. These uncertainties are mainly observed at both ends of the citrus supply chains: the yield at the fruit supply and the demand for products and by-products at the markets. In this chapter, we consider the farm side of citrus supply chains and examine different forecasting methods for forecasting the future crop yield accurately. The yield forecasts are crucial in the planning of supply chain activities. Accurate forecasting facilitates efficient resource allocation and proactive planning. The citrus crop yield differs depending on the geographical, topological, and climatic conditions of the fields;therefore, one model generating accurate forecasts for a particular field may not generate accurate forecasts for another field. Using statistical analysis and accuracy metrics, we evaluate the efficiency and compare the results of time series forecasting models and machine learning algorithms for each dataset. Our dataset comprises actual data from Algeria, segmented into time series citrus crop yield and a combination of citrus crop yield data with biophysical and atmospheric variables. Our preliminary results show that ARIMA is the best-fitting model for Algeria’s citrus crop yield dataset with MAPE of 1–2% for two-period (years)-ahead forecast. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.