Abstract. Modeling ecological responses to historical climate change and variability over the last 100 years requires input of temporally complete climate data at high spatial density. However, for most domains, station density during early part of the 20th century is insufficient to spatially interpolate these fields with confidence, especially in mountainous regions. For the second phase of the VEMAP (Vegetation/ Ecosystem Modeling and Analysis Project) ecological model intercomparison, we developed a 99-yr 0.5-degree latitude/longitude gridded dataset of monthly precipitation and monthly mean minimum and maximum temperature. We used a kriging model to reconstruct the 1895-1993 record for ~8500 precipitation and snotel stations and ~5500 temperature stations based on local spatial autocorrelation structure. Crossvalidation was used to evaluate errors and their time and space dependence. We then used a spatial interpolation model (PRISM) that accounts for the effects of elevation and aspect on the distribution of precipitation and temperature to grid the station data. The resulting dataset is a temporally complete and geographically realistic representation of the historical climate record.