Diagnosing and Predicting Surface Temperature in Mountainous Terrain

View More View Less
  • 1 Pacific Southwest Forest and Range Experimental Station, Forest Service, U.S. Department of Agriculture, Riverside, Calif. 92507
© Get Permissions
Restricted access

Abstract

The first two harmonics of a Fourier series temperature expansion were used to model the diurnal variation of surface temperature in mountainous terrain of southern California The temperature T at any hour t was expressed in the formwhere As is the aspect contribution to temperature and is a function of insolation, and Bt, is the bias condition function and depends on the time of day, synoptic weather class and elevation. The Fourier coefficients A0, a1, b1, a2 and b2 are all calculated independently of each other, making it possible to determine the coefficients by regression analysis. Stepwise screening regression was used to derive the Fourier coefficients by means of the “perfect prog” technique. The 17 potential predictors were valid at six times—0, 12 and 24 h in advance from both 0000 and 1200 GMT. The temperature predictions can be updated every 12 h with the input of observed surface and 850 mb data and the Limited-area Fine Mesh (LFM) model output 12 and 24 h predictions. The model then allows us to start predictions at any time, select an interval for the predictions, and predict the surface temperature out to as much as 36 h. The model was validated at four research sites in the San Bernardino Mountains of southern California with independent data. Verification results, comparing observed and predicted temperature, show root-mean-square errors ranging from 1.3 to 4.7°C. Of 48 correlation coefficients, 21 were greater than 0.90 and only one less than 0.60.

Abstract

The first two harmonics of a Fourier series temperature expansion were used to model the diurnal variation of surface temperature in mountainous terrain of southern California The temperature T at any hour t was expressed in the formwhere As is the aspect contribution to temperature and is a function of insolation, and Bt, is the bias condition function and depends on the time of day, synoptic weather class and elevation. The Fourier coefficients A0, a1, b1, a2 and b2 are all calculated independently of each other, making it possible to determine the coefficients by regression analysis. Stepwise screening regression was used to derive the Fourier coefficients by means of the “perfect prog” technique. The 17 potential predictors were valid at six times—0, 12 and 24 h in advance from both 0000 and 1200 GMT. The temperature predictions can be updated every 12 h with the input of observed surface and 850 mb data and the Limited-area Fine Mesh (LFM) model output 12 and 24 h predictions. The model then allows us to start predictions at any time, select an interval for the predictions, and predict the surface temperature out to as much as 36 h. The model was validated at four research sites in the San Bernardino Mountains of southern California with independent data. Verification results, comparing observed and predicted temperature, show root-mean-square errors ranging from 1.3 to 4.7°C. Of 48 correlation coefficients, 21 were greater than 0.90 and only one less than 0.60.

Save