Statistical approaches for identification of low-flow drivers: Temporal aspects

Abstract

The characteristics of low-flow periods, especially regarding their low temporal dynamics, suggest that the dimensions of the metrics related to these periods may be easily related to their meteorological drivers using simplified statistical model approaches. In this study, linear statistical models based on multiple linear regressions (MLRs) are proposed. The study area chosen is the German federal state of Lower Saxony with 28 available gauges used for analysis. A number of regression approaches are evaluated. An approach using principal components of local meteorological indices as input appeared to show the best performance. In a second analysis it was assessed whether the formulated models may be eligible for application in climate change impact analysis. The models were therefore applied to a climate model ensemble based on the RCP8.5 scenario. Analyses in the baseline period revealed that some of the meteorological indices needed for model input could not be fully reproduced by the climate models. The predictions for the future show an overall increase in the lowest average 7-day flow (NM7Q), projected by the majority of ensemble members and for the majority of stations.