Forecast of frost days based on monthly temperatures

Abstract

Although frost can cause considerable crop damage, and practices have been developed to mitigate forecasted frost, frost forecasting technologies have not changed for years. This paper reports on a new method based on successive application of two models to forecast the number of monthly frost days for several Community of Madrid (Spain) meteorological stations. The first is an autoregressive integrated moving average (ARIMA) stochastic model that forecasts minimum monthly absolute temperature (t min) and average monthly minimum temperature (micro t) following Box and Jenkins methodology. The second model relates monthly temperatures (t min, micro t) to the minimum daily temperature distribution during one month. Three ARIMA models were identified. They present the same seasonal behaviour (integrated moving average model) and different non-seasonal part: autoregressive model (Model 1), integrated moving average model (Model 2) and autoregressive and moving average model (Model 3). The results indicate that minimum daily temperature (t dmin) for the meteorological stations studied followed a normal distribution each month with a very similar standard deviation through out the years. This standard deviation obtained for each station and each month could be used as a risk index for cold months. The application of Model 1 to predict minimum monthly temperatures produced the best frost days forecast. This procedure provides a tool for crop managers and crop insurance companies to assess the risk of frost frequency and intensity, so that they can take steps to mitigate frost damage and estimate the damage that frost would cause.