A neural model to predict the daily minimum of air temperature

Prediction of the daily minimum temperature is very important in order to activate methods to protect plants from the effect of frost. Forecasting the minimum air temperature is complex owing to the input parameters required for the applications of analytical models and the low spatial transferability of some empirical models. This paper presents a model based on neural network technique for short-term prediction of the daily minimum temperature of air. Measurements of some atmospheric (air temperature, black body temperature, cloud amount and air humidity) and soil parameters (soil temperature) were used for training and testing the neural networks. Experimental results demonstrate the capabilities of neural techniques for forecasting the minimum air temperature. We realized seven neural models, the best one provides values of predicted minimum air temperatures with a low standard error of estimate (SEE = 1.63 °C) and a good correlation with observed values (r = 0.91). The neural network model provides information about frost risk and can be used to support decision-making in frost protection.