The ability of various statistical techniques to forecast the July-August-September (JAS) total rainfall and monthly streamflow in the Sirba watershed (West Africa) was tested. First, multiple linear regression was used to link predictors derived from the Atlantic and Pacific sea-surface temperatures (SST) to JAS rainfall in the watershed up to 18 months ahead; then, daily precipitation was generated using temporal disaggregation; and finally, a rainfall-runoff model was used to generate future hydrographs. Different combinations of lag times and time windows on which SSTs were averaged were considered. Model performance was assessed using the Nash-Sutcliffe coefficient (E f), the coefficient of determination (R 2) and a three-category hit score (H). The best results were achieved using the Pacific Ocean SST averaged over the March-June period of the year, before the rainy season, and led to a performance of R 2 = 0.458, E f = 0.387 and H = 66.67% for JAS total rainfall, and R 2 = 0.552, E f = 0.487 and H = 73.28% for monthly streamflow.
Abstract:In most studies, trend detection is performed under the assumption of a monotonic trend. However, natural processes and, in particular, hydro-climatic variables may not conform to this assumption. This study performs a simultaneous evaluation of gradual and abrupt changes in Canadian low streamflows using a modified Mann-Kendall (MK) trend test and a Bayesian multiple change-point detection model. Statistical analysis, using the whole record of observation (under a monotonic trend assumption), shows that winter and summer low flows are dominated by upward and downward trends, respectively. Overall, about 20% of low flows are characterized by significant trends, where ¾80% of detected significant trends are upward (downward) for winter (summer) season. Change-point analysis shows that over 50% of low-flow time series experienced at least one abrupt change in mean or in direction of trend, of which ¾50% occurred in 1980s with a mode in 1987. Analysis of segmented time series based on a common change-point date indicates a reduced number of significant trends, which is attributed to first, the change in nonstationarity behaviour of low flows leading to less trend-type changes in the last few decades; and second, the false detection of trends when the sample data are characterized by shifts in mean. Depending on whether the monotonic trend assumption holds, the on-site and regional interpretation of results may vary (e.g. winter low flow) or even lead to contradictory conclusions (e.g. summer low flow). Trend analysis of last two decades of streamflows shows that (1) winter low flows are increasing in eastern Canada and southern British Columbia, whereas they are decreasing in western Canada; (2) summer low flows are increasing in central Canada, southern British Columbia and Newfoundland, whereas they are decreasing in Yukon and northern British Columbia and also in eastern Ontario and Quebec.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.