Probabilistic modelling of the dependence between rainfed crops and drought hazard

Abstract

Abstract. Extreme weather events, such as droughts, have been increasingly affecting the agricultural sector, causing several socio-economic consequences. The growing economy requires improved assessments of drought-related impacts in agriculture, particularly under a climate that is getting drier and warmer. This work proposes a probabilistic model that is intended to contribute to the agricultural drought risk management in rainfed cropping systems. Our methodology is based on a bivariate copula approach using elliptical and Archimedean copulas, the application of which is quite recent in agrometeorological studies. In this work we use copulas to model joint probability distributions describing the amount of dependence between drought conditions and crop yield anomalies. Afterwards, we use the established copula models to simulate pairs of yield anomalies and drought hazard, preserving their dependence structure to further estimate the probability of crop loss. In the first step, we analyse the probability of crop loss without distinguishing the class of drought, and in the second step we compare the probability of crop loss under drought and non-drought conditions. The results indicate that, in general, Archimedean copulas provide the best statistical fits of the joint probability distributions, suggesting a dependence among extreme values of rainfed cereal yield anomalies and drought indicators. Moreover, the estimated conditional probabilities suggest that when drought conditions are below moderate thresholds, the risk of crop loss increases between 32.53 % (cluster 1) and 32.6 % (cluster 2) in the case of wheat and between 31.63 % (cluster 2) and 55.55 % (cluster 2) in the case of barley. From an operational point of view, the results aim to contribute to the decision-making process in agricultural practices.

Abstract from the paper’s OpenAlex record; open copy licensed cc-by. Read at the source ↗

The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.