Frost hazard assessment from local weather and terrain data

Abstract

Minimum air temperatures were measured at 31 sites in a 170 km 2 region of undulating grazing land on 30 winter nights in 1982. The change in minimum temperature with elevation (d T/d z) could be predicted from mean night-time windspeed, total night-time all-wave radiation loss and the minimum air temperature at an exposed hill site. A multiple regression model is used which includes an interaction term for windspeed and net radiation. Excellent agreement between d T/d z values obtained with the 31 thermometer sites and a seven-site Ross recorder network was observed for 24 nights. The above regression model was then further developed with data for seven Ross recorder sites available for 86 days representing a much wider range of weather conditions. The improved model explains some 75% of the variation in minimum air temperature. Sites in convex terrain were consistently warmer and sites in concave terrain were consistently colder than predicted with the above model. A second relationship was therefore developed which predicts mean departures from the above regressions for individual sites from two simple terrain parameters: slope and catchment area. This regression explains 76% of the variation in the residual values. The combined weather—terrain model has been tested successfully with Ross recorder data from eight sites in 1980 ( n = 246), 10 sites in 1981 ( n = 189) and seven sites in 1982 ( n = 602).