Medium-Range Forecasting for the Number of Daily Forest Fires

Abstract

In an earlier work, the authors introduced an objective forecast model for a 24-h prediction of the number of daily forest fires based on a 2-day lag autoregressive model. The meteorological inputs required for this model (temperature and geopotential height at 850 and 700 hPa and dewpoint at 850 hPa) may be predicted by a medium-range numerical weather forecast model such as that of the European Centre for Medium-Range Weather Forecasts. These predicted meteorological elements may be used to extend the range of daily forest fire forecasting. Since the forest fire forecast model is based on a categorization (type of day), an error in the meteorological predictions may not be an error in the predictive model. A meteorological error will only imply error for the model if it produces a change in the type of day (category). The forecast range for the number of forest fires per day has been extended to five days with this new model. Moreover, assuming that the weather forecast is perfect, a validation of the prediction model for forest fires is carried out.