Long-term trends in global river flow and the causal relationships between river flow and ocean signals

摘要

Continental flow plays a significant role in the global ecosystem. With increased global warming and a booming economy exacerbating water scarcity, global analyses of continental discharge are of great relevance to water management. In this paper, the Mann-Kendall (MK) test, with variations accounting for lag-1 autocorrelation, full autocorrelation, and long-term persistence, was used to detect the long-term flow trends in 916 of the world's largest ocean-reaching rivers over the period 1948-2004. Field significance was evaluated using Walker's test. Globally, the results revealed more decreases in streamflow than increases. With the traditional MK test, 503 rivers had decreased streamflow, with 120 showing significant decreases, whereas 408 rivers had increased streamflow, with 51 showing significant increases. Regionally, positive streamflow trends mainly occurred in high-latitude areas and negative streamflow trends mainly occurred in low-latitude areas, a pattern that can be attributed to uneven precipitation and the effects of global warming. When the full autocorrelation structure and long-term persistence behavior were taken into account in the trend analysis, there was a large reduction in the number of rivers with significant changes in streamflow. Negligible departures from the traditional MK results were observed when only the lag-1 autocorrelation was taken into consideration. For the Americas and Europe, annual integrated river flow showed a slight upward trend; for other continents, there was a slight downward trend. For all oceans except the Arctic Ocean, integrated river flow had a downward trend. We investigated the causal relationships between streamflow and ocean signals using the Granger causality test. El Nifio Southern Oscillation (ENSO) signals were significantly causal for river flow in over 36% of global rivers tested. The influence of the Arctic Oscillation (AO) and the North Atlantic Oscillation (NAO) was significant in more than a quarter