Increased precipitation as the main driver of increased streamflow in tile-drained watersheds of the upper midwestern U.S.

Abstract

Recent increased streamflow (Q) and its associated impacts on water quality have frequently been linked to land use and land cover (LULC) changes such as increased tile drainage, cultivation of prairies, and adoption of soybean (Glycine max) in modern-day cropping systems. However, many previous studies have assumed minimal to no change in precipitation during their study period. A recent analysis of streamflow records from 29 HUC 8 (Hydrologic Unit Code 008) watersheds in Iowa and Minnesota showed that increased precipitation instead of LULC change was the main driver of increased streamflow. The analysis was done through hierarchical regression of annual streamflow as a function of annual precipitation for the periods prior to 1975 (pre-change period) and after 1976 (post-change period). A statistical shift in annual relationship from the pre- to post-change period was assumed to be an indication of LULC changes, whereas a lack of statistical shift suggested no change in the relationship and higher flows were mainly driven by increased precipitation. In this article, we further show that annual streamflow and annual baseflow were influenced not only by the current year's precipitation but also by precipitation in the preceding one to two years, and this effect was manifested through increased or decreased stored soil water. The present analysis was done using backward stepwise hierarchical regression with the natural log of annual streamflow as the predictor variable and three to five years of precipitation, the area under soybean production, a group variable simulating pre- and post-change periods, and its interaction terms with precipitation and soybean area as the explanatory variables. This analysis also showed that precipitation was the main driver of annual streamflow or baseflow; however, for some rivers, the area under soybean production and group differences were also significant variables, although at a much smaller confidence level. Annual streamflow testing was done for the Blue Earth River, Redwood River, Cottonwood River, and Whetstone River watersheds in Minnesota and the Maquoketa River and Raccoon River watersheds in Iowa. Annual baseflow testing was done only on the Redwood River and Raccoon River watersheds. Using similar backward stepwise regressions, the analysis showed that changes in Ln(monthly streamflow) were linked to stored soil water through the preceding months' and years' precipitation. On a daily scale, comparison of slopes of the hydrograph's rising limb for two large precipitation events in 1957 (pre-change period) and 1993 (post-change period) showed less watershed connectivity due to LULC changes such as drainage, a finding that is contrary to what has been suggested in the literature. A similar comparison of the falling limb slopes for a given streamflow condition showed similar slope values in pre-change (1947) and post-change (1991) periods, thus suggesting no changes in storage capacity of the watershed as a result of LULC changes. Further comparisons of high-level streamflows (Q > 100 m3 s-1) as a function of average daily storm precipitation showed no effect of drainage between the pre- and post-change periods. A statistical analysis of the relationship between the falling limb slopes as a function of streamflow for low (dQ/dt