A spline model of climate for the Western United States

Monthly climate data of average, minimum, and maximum temperature and precipitation normalized
for the period 1961 through 1990 were accumulated from approximately 3,000 weather stations in the Western United States and Southwestern Canada. About two-thirds of these observations were available from the weather services of the two countries while the remaining third were added to the normalized base from daily weather records of stations of short duration. Tests of the procedures used to normalize these supplemental data showed that estimates on average were within 0.2 °C for temperature variables and 2.7 mm for precipitation. Weather data for the 48 monthlies were fit to geographic surfaces with thin plate splines. Relationships between predicted values and observed monthlies for about 245 records withheld from the modeling process produced values of R2 that averaged about 0.95 and ranged from 0.87 to 0.99. The slope of the regression line for these relationships was essentially 1.0 for all 48 comparisons. Predictions from the climate model can then be converted to variables of demonstrated importance in plant geography, ecology, or physiology. As an illustration, algorithms are presented and justified for
estimating 18 variables derived from predicted values. These derived variables range from the straightforward such as mean annual temperature or mean temperature in the coldest month to those of degree-days >5 °C or freezing dates. Applications of the model in plant biology are illustrated for (1) generating climate estimates for locations specified by latitude, longitude, and elevation, (2) mapping climate variables, (3) separating species distributions in climatic space, and (4) relating genetic variation among populations to climatic gradients.