Estimation of the risk of frost occurrence applying Theory of Extreme Values

Abstract

The occurrence of frosts in temperate zones of Mexico seriously affects the volume and value of agricultural production, causing significant economic losses for producers and food shortages. Late and early frosts are highly dangerous for the development of many crops, in this sense it is important to develop sensitive risk estimation models to characterize local frost occurrence regimes to make sound decisions in production planning. This paper aims to estimate the dates of occurrence of late and early frosts, as well as the frost-free periods, by estimating the probability of occurrence from the approach of an extreme event. Daily minimum temperature data recorded in 77 meteorological stations in Mexico State between 1980 and 2010 are used, taken from CLICOM (National Meteorological Service, 2014). It is proposed to use a distribution of extreme values to model the dates of occurrence of the last spring frosts and the first autumn frosts since this family of distributions presents greater sensitivity in the tails analysis for risk estimation. The results are presented on risk maps in terms of the probability of frost occurrence. It is concluded that the Weibull distribution presented the best fit to the data according to the Akaike Information Criterion (AIC), which has enabled successfully characterize the regime of frost occurrence in Mexico State.