Testing hypotheses of hydrologic alteration beyond ecological thresholds

Abstract:

Negative impacts to riverine species and ecosystems have often been observed when the hydrologic alteration following an abrupt impact to river flow, such as the construction of a dam, exceeds percent deviation thresholds. However, reported percent deviations for different pre- and post-impact flow statistics do not state the likelihood that percent deviations in excess of any threshold are due to an actual change and not sampling variability. Such probabilistic statements about the likelihood of an actual change exceeding a given percent deviation threshold are useful for guiding river restoration decisions, but are seldom available.

We introduce a hypothesis testing approach for making probabilistic statements about changes between pre- and post-impact flow regimes that exceed of ecological thresholds. We create ensembles of annual flow duration curves (AFDCs), which provide signatures describing within-year and between-year daily flow variability during pre- and post-impact periods. Pre-impact AFDC ensembles are used to identify exceedance probabilities that are statistically independent from each other and have distinct ecological impacts, e.g. high flows and low flows. The central tendency of pre- and post-impact flows for the selected AFDC exceedance probabilities are compared with a nonparametric Mann-Whitney-Wilcoxon test. A field significance test is then employed to determine whether there has been a significant overall change in the flow regime.

Through a case study of hydropower-induced flow alteration, we demonstrate how this testing approach can screen flow restoration decisions. We weight the differences between the hydropower and ecological performance of baseline and alternative dam operation policies with Type I (over-protection) and Type II (under-protection) errors. We discuss how similar decision-oriented approaches can also be applied to sequential daily flow hydrographs, or even other ecologically relevant variables, such as water temperature.