A statistical method for evaluating the effects of antecedent rainfall upon runoff: applications to the coastal plain of Georgia

Abstract

A statistical (hypothesis testing) method is described for analysing the effects of long-term (1 year) antecedent rainfall upon subsequent runoff. This method which is regional, rather than stream-specific, was applied to the southeastern US coastal plain in Georgia for the study period 1948–1994. Three sets of forty 12 month periods were used to classify antecedent rainfall into three populations: “high” (+15 to +59% normal), “low” (−11 to −42% normal), and “normal” (−9 to +8% normal). These groups were then interrelated with three groups of “subsequent” rainfall, subdivided upon a similar basis. This resulted in the creation of nine unique sub–populations consisting of combinations of antecedent and subsequent periods. Subsequent periods of 3, 6, 12, and 24 months were used. A “production ratio” (runoff/rainfall) was defined for each subsequent period and then normalized with the long-term (47 year study period) average production ratio for those months. The nonparametric Wilcoxon–Mann–Whitney test was used to evaluate differences between mean normalized production ratios for the nine subpopulations. The results of a critical hypothesis test revealed that production ratios return to normal between 1 and 2 years following a drought, given a return to normal rainfall conditions. This might represent the time required for contributions to streamflow from local ground-water flow systems to return to normal following a drought. Production ratios associated with “high” antecedent rainfall remain considerably greater than normal for a period of at least 2 years following a return to normal rainfall.