Adverse meteorological phenomenon's occurrence, such as low air temperature and frost, can cause serius damages to the agriculture of the State of São Paulo. The greater accuracy on the occurrence probability establishment for such events, based on historical data obtained from a larger agrometeorological network, appropriated probabilistic models and modern mapping technical can provides assistance for agricultural risk studies, which are very important for agricultural financing and insurance programs. The probabilistic model "normal distribution" was used to estimate the punctual risks of yearly minimum air temperatures below 0, 1, and 2 °C for 28 localities of the State of São Paulo, Brazil. A digital elevation model previously developed with remote sensing and geographical information system (GIS) techniques were used to map the estimated probabilities of frost occurrence, considering different levels of absolute minimum air temperature. Multiple regression equations were developed to estimate the regional probabilities of frost occurrence, based on altitude, latitude and longitude informations. The partial independent variables that better explained the dependent variable "probability" were, at the following importance order, altitude (R² = 0.74-0.78), latitude (R² = 0.33-0.44) and longitude (R² = 0.15-0.20). The equations generated through multiple regression analyses showed determination coefficients between 0.87 and 0.90. The estimated probabilities were spacialized using the GIS "Idrisi", based on a digital elevation model (altitude) and geographical coordinates (latitude and longitude). The multiple regression equations generated continuous pixel-to-pixel variations for different classes of probability of occurrence of absolute minimum air temperature. Two maps are presented showing the spatial variability of the frost risk occurrence with minimum temperatures below 1 and 2 °C.
Frost risk; probability; minimum air temperature; normal distribution; digital elevation; mapping
