Abstract:Plenty of climate data from various sources have become available in recent years. However, to obtain climate data adequately meeting the requirement of ecological studies remains a challenge in some cases due to the difficulty of data integration and the complexity of downscaling, especially for mountainous regions. Lapse rate is one of the most important factors that influence the change of climatic variables in the mountains, and it should be incorporated into climatic models. In this study, we applied a synthetic approach combining bilinear interpolation (to produce seamless surfaces) and dynamic local regression (to obtain local lapse rates) to develop a scale-free and topography-correspondent downscaling model in R environment for Taiwan, called clim.regression. This model can generate 73 climatic variable estimates specific to the user-defined points of interest, including primary climatic variables and additional biologically relevant derivatives for historical (1960‒2009) and future periods (2016‒2035, 2046‒2065 and 2081‒2100). Results of our evaluation indicated that clim.regression reduced prediction error by 54.6%‒66.7% relative to the original gridded climate data for temperatures. In addition, we demonstrated the spatiotemporal patterns of lapse rate for different climate variables.