How Can a Changing Climate Influence the Productivity of Traditional Olive Orchards? Regression Analysis Applied to a Local Case Study in Portugal

Abstract

Nowadays, the climate is undoubtedly one of the main threats to the sustainability of olive orchards, especially in the case of rainfed traditional production systems. Local warming, droughts, and extreme weather events are some of the climatological factors responsible for environmental thresholds in relation to crops being exceeded. The main objective of this study was to investigate the influence of microclimatic variability on the productivity of traditional olive orchards in a municipality located in northeastern Portugal. For this purpose, official data on climate, expressed through agro-bioclimatic indicators, and olive productivity for a 21-year historical period (2000–2020) were used to evaluate potential correlations. In addition, a comprehensive regression analysis involving the dataset and the following modeling scenarios was carried out to develop regression models and assess the resulting predictions: (a) Random Forest (RF) with selected features; (b) Ordinary Least-Squares (OLS) with selected features; (c) OLS with correlation features; and (d) OLS with all features. For the a and b scenarios, features were selected applying the Recursive Feature Elimination with Cross-Validation (RFECV) technique. The best statistical performance was achieved considering nonlinearity among variables (a scenario, R2 = 0.95); however, it was not possible to derive any model given the underlying methodology to this scenario. In linear regression applications, the best fit between model predictions and the real olive productivity was obtained when all the analyzed agro-bioclimatic indicators were included in the regression (d scenario, R2 = 0.85). When selecting only the most relevant indicators using RFECV and correlation techniques, moderate correlations for the b and c regression scenarios were obtained (R2 of 0.54 and 0.49, respectively). Based on the research findings, especially the regression models, their adaptability to other olive territories with similar agronomic and environmental characteristics is suggested for crop management and regulatory purposes.