In basins, information on long-term air temperature is useful for farming. In the present study, air temperatures at 1.5m were measured every 10 minutes at various elevations on the slopes of a caldera basin (Akaigawa Basin, Hokkaido) during 2 periods (July to October 1987 and May to October 1988). The relationships between elevation and daily minimum and maximum air temperatures at 1.5m were anälyzed. Since the relationships had small regional differences, and were approximately linear relationship, the daily minimum and maximum air temperatures at any elevation in the basin could be estimated using gradients of air temperature to elevation (air temperature gradient) and the daily minimum and maximum air temperatures at a particular site. In order to estimate the gradients, multiple regression models were tested which include various meteorological elements as independent variables. The best model for the minimum air temperature gradient (An, °C/100m) is as follows, where the independent variables are daily minimum air temperature (Tn, °C) and mean water vapor pressure (VPn, mb), accumulated radiation heat loss (Rn, kWh/m2) and wind speed (Wn m/s) during night time, as measured at the bottom of the basin. [formulla omitted] This regression model explains 77% of the Variation in the daily minimum air temperature gradient, and its Standard estimate error (0.5°C/100m) is equivalent to 10% of the gradient range. For the daily maximum air temperature gradient, an adequate regression model was not found by the independent variables tested. Other results suggested that the daily minimum air temperature gradient at the Akaigawa Basin can also be predicted from the air temperature gradient of the lower atmosphere in Sapporo (about 70 km distance from Akaigawa Basin) which is calculated by the aerological data reported at 21:00 (observed at about 20:30) of the previous night. © 1992, The Society of Agricultural Meteorology of Japan. All rights reserved.