We have developed a minimum air temperature forecasting system using local meteorological
observation data in order to prevent or mitigate the frost damage. In this paper, the correlation between the meteorological factors and the decrease in air temperature during the night was shown. Then minimum air temperature was predicted by the multiple regression equation that uses air temperature and humidity at 18:00 as explanatory variables. As the result, the root mean square error (RMSE) was 2.8°C. When the difference of air temperature between 17:00 and 18:00 was used for the prediction instead of humidity, the RMSE was 3.5°C. Next, the prediction was carried out only in clear nights, then the equation that use air temperature and humidity at 18:00 showed that the RMSE was 1.3°C. Although the predictive accuracy of the equations for clear nights is low on cloudy or rainy nights, we can forecast safely on frosty nights by using those equations in combination with the equations for all nights.
Minimum air temperature
multiple regression analysis