An analysis was conducted of almost 5000 operational seasonal streamflow forecast errors across the western United States. These forecasts are for 29 unregulated rivers with diversity in geography and climate. Deterministic evaluations revealed strong correspondence between observations and forecasts issued 1 April. Forecasts issued earlier in the season were more uncertain yet remained skillful. The average change in forecast performance between January and April was primarily linked to the climatological seasonal cycle of precipitation: regions with climatologically wet winters and dry springs (e.g., California) showed much more forecast improvement between January and April than did regions with dry winters and wet springs (e.g., western Great Plains, Colorado Front Range). Other climatological factors played a secondary role; for example, mixed rain-snow basins in the Pacific Northwest did not show as significant an improvement in skill versus lead time as might other wise be expected. Mixed trends in 1 April forecast skill were noted since the 1980s, with increased skill in California and Nevada, and a decline in skill in the Colorado River basin. Increased variability in streamflow was also noted across most of the western United States, although this did not appear to be the only factor responsible for trends in forecast skill.
Evaluation of Official Western U.S. Seasonal Water Supply Outlooks 1922-2002
Reference
Pagano, T., et al. Evaluation of Official Western U.S. Seasonal Water Supply Outlooks 1922-2002. 2004, pp. 896-09.
Abstract