A Hypothesis Test for Hydrologic Alteration

Abstract:

Hydrologic processes often undergo change due to anthropogenic influences, such as water withdrawals and regulation due to dams and reservoirs. These changes are often characterized by changes in streamflow statistics, such as the Indicators of Hydrologic Alteration. Flow regime shifts can also be observed in pre- and post-disturbance flow duration curves (FDCs), which are typically plots of daily streamflow versus the probability of exceedance. Here we develop a hypothesis of hydrologic alteration based on shifts in the expected variability of FDCs as described by their confidence intervals. Employing a database of streamflow gauges that have undergone alteration due to reservoir construction, we develop two tests: one based on deviations from confidence intervals associated with pre-disturbance FDC’s , and a second based on the magnitude of the ecodeficit and ecosurplus, which quantify the magnitude of the change in FDC. Of interest is whether hydrologic alteration can be predicted in a systematic and understandable manner, and whether one can determine an alteration to be “significant.” We also report both type I and II errors, where Type I errors correspond to concluding a river is disturbed when it is not, and type II errors correspond to concluding a river is pristine when in fact it is disturbed. An exploration of whether type II errors are related to common measures of hydrologic disturbance (e.g. water withdrawals, area of surface water, etc.) was carried out using annual FDC series.