Publication Details

Ashcroft, M. B. (2006). A method for improving landscape scale temperature predictions and the implications for vegetation modelling. Ecological Modelling, 197 (3-4), 394-404.


Understanding how environmental factors influence the spatial distribution of vegetation allows environmental managers to plan for issues such as climate change, ecological restoration and intensified land use. Elevation is often used as an indirect predictor of temperature, but this limits the applicability of environmental models to other study areas and introduces errors in mountainous terrain where variations in slope, aspect, and radiation can significantly alter the relationship between elevation and temperature. Some studies have developed estimates for temperature that also consider factors such as radiation, but these usually estimate the temperature for each location without considering the surrounding environment. In this study, average summer maximum and minimum temperatures were recorded at various locations on the Illawarra Escarpment, near Sydney, Australia. It was hypothesised that wind and air movements would average out large differences in elevation and radiation in mountainous terrain and cause the temperatures to be more strongly correlated with local averages of elevation and radiation than they are with the actual elevation and radiation where the temperatures were recorded. The use of local averages improved the estimate of average summer maximum temperature from a regression r2 of 0.185 using elevation alone, to an r2 of 0.685 when using canopy cover, local average elevation and local average radiation. In contrast, average summer minimum temperatures were better predicted using the elevation of each location without averaging. The results were applied to vegetation modelling by comparing a generalised additive model using the predicted average temperatures with a model using elevation. The overall classification accuracy for vegetation communities in the study area was improved from 46.4% to 61.8%. Therefore, improved temperature estimates also improved the explanatory performance of vegetation models.



Link to publisher version (DOI)