Comparing species temperature response curves: population density versus second-hand data

Descriptions of individual species responses to temperature are required in order to assess the impact of future global warming. The response of Rhododendrom arboreum to estimated mean annual temperature was investigated in the Himalayas using General Additive Models (GAM) and Generalized Linear Models (GLM). The aim was to evaluate the consistency between the response in population density along elevation gradients versus response curves based on elevation data from herbarium specimens and vegetation surveys. The comparison was made with respect to (1) estimated temperature at the point of maximum response and (2) the shape of the response curves i.e. symmetric vs skewed. All data indicate a single optimum between 12.3 and 10.8 degreesC. The difference is only 0.4 degreesC between the optimum estimated from localities of herbarium specimens (frequency) and the population density data. The difference is larger (0.7 degreesC) when the vegetation survey data are combined with the data from the herbarium specimens. However, the differences are small when the uncertainties in temperature estimation are taken into consideration. The response curves based on herbarium specimens and vegetation survey data (frequencies) are symmetric. A sigmoid response curve was estimated from herbarium specimens (binomial data). The population density along the elevation gradients was, to some extent, asymmetric. This may reflect the underlying biological structure, but sampling bias and the numerical analyses may also influence the results.