|ABSTRACT||Extreme weather events, such as droughts, are major sources of risk to agricultural systems, often entailing substantial crop failures. An enhanced assessment of drought-related damages in croplands is crucial, particularly in the Mediterranean regions recurrently affected by dry episodes.
This work aims to develop a multivariate probabilistic model using copulas to contribute to agricultural drought risk management and consequently attempt to prevent crop losses. The main target is to estimate the likelihood of drought risk in rainfed cropping systems. Drought impacts are assessed by wheat yield anomalies during 1986-2012 in the Iberian Peninsula (IP). Drought hazard is evaluated using the hydro-meteorological drought index SPEI (Standardized Precipitation Evapotranspiration Index) and the satellite-based indices VCI (Vegetation Condition Index), TCI (Temperature Condition Index) and VHI (Vegetation Health Index).
This study adopts a bivariate modelling approach using Elliptical (t-copula) and Archimedean (Clayton, Frank and Gumbel) copulas, and the selection of the most suitable copula function is performed based the Akaike’s Information Criteria (AIC). The copula-based approach is carried out in two steps: first the copula fits are estimated using the whole time-series and afterwards during different climatic conditions, differentiating drought and non-drought years. A good agreement is found between the bivariate copula simulations and the observations, pointing to an overall good performance of the selected copula functions and their ability in estimating the joint behaviour between yield and droughts.
The results suggest that the Archimedean copulas provide the best statistical fits of the joint probability distributions in most of the cases. These findings support that, in general, the relationship between wheat yields and drought conditions is described by an asymmetric dependence in the tails of the joint distributions. The generated probability distributions of yield using the established copula models suggest relevant risk levels of wheat under drought conditions in the major agricultural areas of the IP. From an operational point of view, the results aim to contribute to the decision-making process in agricultural practices in the IP, particularly to assist farmers in deciding whether to purchase crop insurance and managing the adequate number of employees.
Acknowledgements: This work was partially supported by Fundação para a Ciência e a Tecnologia, Portugal by the projects IMDROFLOOD funded (FCT, WaterJPI/0004/2014) and by UID/GEO/50019/2013 - Instituto Dom Luiz. Ana Russo and Andreia Ribeiro also thank FCT for grants SFRH/BPD/99757/2014 and PD/BD/114481/2016, respectively.|