Estimating minimum temperature and climatological freeze risk in hilly terrain

Abstract Differences between weather station and on-farm minimum temperatures in hilly terrain were related to general weather parameters, using stepwise multiple linear regression and correlation techniques. Opacity (cloud cover) and wind speed were the most significant parameters, explaining up to 79% of the variation in night-time temperature differences between base station and field sites 20–40 km away. Considering dew point depression, days since 0.25 cm rainfall, absolute minimum temperature, cloud base height, opacity as quadratic and an exponential relationship did not improve the regression estimates significantly. Standard error of estimates ranged from 0.8°C for sites on hilltops to 2.3°C for valley bottoms. In independent data tests using 19 cases, the average minimum temperature difference between the base station and a high risk site was estimated within 0.4°C of the measured value. Significance of the regression relationships increased proportionally with distance down the slopes. Base station-site minimum temperature differences on relatively clear calm nights averaged near 1°C on hilltops, 2° to 3°C midway down slopes and 5° to 6°C in the hollows. Regression equations developed from short-term temperature measurements at more than 50 sites were used to estimate daily site minimum temperatures for a 21-year period. Spring and fall freeze probabilities for critical temperatures of 32°, 30° and 28°F (0.0°, −1.1° and −2.2°C) were then calculated for each site using standard climatological techniques. Estimated average freeze dates in hollows were up to 34 days later in spring and 39 days earlier in fall than on hilltops. Average dates of spring and fall freeze occurrences were delayed and advanced respectively about 6 days for each degree (C) drop in critical temperature. In independent data tests the average spring and fall ground frost dates at the base station were estimated within 2 days of measured values, using similar techniques. Minimum temperatures measured by traverses on a clear, calm night were highly correlated with estimated frost dates (R = 0.99).