Eliminating Station Location Bias from Long-Term Climate Division Data

Previous research has identified biases in long-term temperature trends in climate division data in New England due to changes in station locations with respect to elevation. A similar issue affects historic precipitation trends in Texas. We show that the spread of stations from east to west across climate divisions has led to masking of a long-term positive precipitation trend through climate-division averaging.

To eliminate this effect, we implemented a technique to extrapolate limited duration COOP data to fill the entire 1895-2007 period. This technique uses the inverse weighting of squared difference approach with the four nearby available stations with the highest correlations to the target station during the data overlap period. The result is a serially-complete set of actual and estimated station data, which may then be averaged within climate divisions without any time-dependent station location bias. The effect on long-term trends in climate-division precipitation is shown for the continental United States.