We used 10-km grid data from the Finnish Bird Atlas data and high-resolution data on temperature and rainfall to estimate species richness from climate and environmental variables across spatial scales. We used an ordinary least-squares (OLS) linear-regression model with a quadratic error function to estimate the number of bird species that occur. As a baseline, we used a simple dummy model that estimated the number of species in each grid to be the average number of species over all grids. We found that the best estimator for avian species richness in Finland is the length of the growing season with R2 values from 0.5 to 0.8, depending on the scale. Our results support the energy-water hypothesis, and we suggest that the proximate control of species richness in the present case is productivity, which is in turn controlled by climate. Some of the effects conventionally attributed to scaling may have trivial causes associated with sampling, in particular the completion of missing data as primary units of observation are merged. For broad surveys of patterns, medium resolution may often be adequate and even superior to the highest nominal resolution available.