The 1-km-mesh nocturnal minimum temperatures in winter were estimated for all of Kyushu by using variational analysis. The 1-km-mesh is not sufficient for prediction of frost occurrence and prediction of road freezing. So, a method to convert the nocturnal minimum temperatures from 1-km-mesh to 50-m-mesh was examined. The estimated deviation for 1-km-mesh nocturnal minimum temperature (DTr-m) was computed by subtracting 1-km-mesh nocturnal minimum temperature from observed nocturnal minimum temperature at JH (Japan Highway) weather stations. Each day from December 1, 1997 to February 28, 1998 was classified into one of five kinds of weather condition days. The properties of DTr-m distribution for each weather condition were investigated by rank sum test and by using GIS with a DEM (GTOPO30). The values of DTr-m tended to become nearly zero on rainy days, with large positive or negative values on fine days. But, in areas where flat topography predominates, the values of DTr-m were distributed around zero regardless of the weather conditions. On fine days, the values of DTr-m were large and positive in mountainous regions, and were large and negative in the West sea coastal area. The differences in the degree of localized cooling between JH weather and AMeDAS stations which depend on geographical and topographical features may cause the variation in the DTr-m values.
The difference of the degree of basin (BDJ-A) was calculated from 50-m-mesh elevation data for the JH weather stations and the nearest AMeDAS stations. The values of BDJ-A have a negative relationship to the median of DTr-m frequency. These results show that 1-km-mesh temperature may be convertible to 50-m-mesh temperature by using BDJ-A.