Compare results computed with all stations vs. just the stations still actively reporting data (separating them out is a very easy programming exercise), and you will get nearly identical results for all stations vs. just the stations that have not been dropped.