Is there an underestimation of long-term variability of streamflow across the continental United States?

Abstract

Analysis of the long-term variability of terrestrial streamflow is important for a wide range of studies in hydroclimatology and environment. However, the common practice of detecting only the monotonic trend cannot actually reflect the long-term variability and change in streamflow, thus resulting in an insufficient or biased understanding. This study focuses on the non-monotonic trends of streamflow and its spatial variability across the continental United States (US). Annual streamflow data observed during 1951–2002 at 530 stations across the continental US are studied. The discrete wavelet spectrum (DWS) is used to detect the non-monotonic trend of streamflow and evaluate its statistical significance. The results indicate significant non-monotonic trends in annual streamflow at 172 stations in the northeast Zone and the northern part of the southeast Zone (in the regions east of 100°W). However, considering only the monotonic trends (being significant at only 38 stations) causes a biased understanding of the long-term variability of streamflow in the northeast zone. The results also reveal that the non-monotonic trend of streamflow in the continental US was mainly controlled by the climatic conditions and that the effects of landscape conditions were weak. The average aridity index (AI) of 0.86–1.35 and its DWS-based significance level >0.1 can indicate the significant non-monotonic trend of streamflow at 95% confidence level. The regions that can meet the above two conditions account for 16% of the global terrestrial surfaces. By considering non-monotonic trends, the change ratio of streamflow is found to be double of that of AI, while similar change ratios are obtained from the monotonic trend results, as the reason of significant underestimation. Due to the partly equal roles of low-frequency oscillation and monotonic trend, identification of the non-monotonic trend should be more desirable for the detection and attribution of the long-term variability and change in streamflow.