Skill Assessment of Water Supply Outlooks in the Colorado River Basin

Water-supply outlooks that predict the April through July (snowmelt) runoff and assist in estimating the total water-year runoff, are very important to users that rely on the major contributing watersheds of the Colorado River. This study reviewed the skill level of April through July forecasts at 28 forecast points within the Colorado River basin. All the forecasts were made after 1950, with considerable variation in time period covered. Evaluations of the forecasts were made using summary measures, correlation measures and categorical measures. The summary measure, a skill score for mean absolute error, indicated a steady increase in forecast skill through the forecast season of January to May. The width of the distribution for each monthly forecast over the 28 locations remained similar through the forecast season. The Nash-Sutcliffe score, a correlation measure, showed similar results, with the Nash-Sutcliffe median showing an increase from 0.4 to 0.8 during the forecast season. The categorical measures used a three-section partition of the April through July runoff. The Probability of Detection for low and high flows showed an increase in skill from approx. 0.4 to 0.8 during the forecast season. The same score for mid-flow years showed limited increase in skill. The low False Alarm Rate illustrated the under forecast of high-flow years. The Bias of the mid-runoff forecasts indicated over forecast early in the forecast season (January to March), with lower Bias later in the forecast season (April and May), ending the forecast season at 1.0, indicating no Bias. Forecasts for both low and high runoff were under forecast early in the season with a Bias near 0.5, improving to nearly 1.0 by the end of the forecast season. The Hit Rate measure illustrated the difficulty of mid-flow forecasts, starting at 0.5 in January and increasing to 0.75 in May due to the forecasting assumption of normal climatology for the remaining forecast period. There was no relationship between basin elevation and forecast skill, reflecting the snow vs. rain dominance in all basins.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.