Forecasting the persistence of native fish in the face of drought in the American Southwest
In the American Southwest, droughts interact with an increasing human appropriation of streamflow and groundwater. As a consequence, rivers are drying with increasing frequency and intensity. Although the implications of this ongoing process for freshwater biodiversity are expected to be far reaching, the scarcity of spatially- and temporally-extensive data, and appropriate numerical tools, has hindered understanding these impacts in a rigorous framework. Here we analyzed historical discharge trends across the Southwest (285 gaging stations). We then compiled a fish abundance data-set spanning 20 years, for 17 reaches and 9 rivers in the Lower Colorado and Rio Grande river basins. We parsed out seasonal from stochastic (interannual) variation in discharge using Discrete Fast Fourier Transform (DFFT) analyses on the last 30 years of daily discharge data of these rivers. We extracted yearly strings of discharge anomalies (describing high- and low-flow events), which were subsequently coupled with the fish time series using Multivariate Autoregressive State-Space (MARSS) models. These stochastic models can quantify the effects of environmental drivers in a species-specific way, and here we used them to (i) estimate responses to discharge anomalies, and (ii) to forecast species-specific risks of hitting 80% declines within the next 10 years.
A quarter of the gaging stations in the Southwest showed significant decreasing trends over the last 30 years. In the target rivers, discharge anomalies had significant effects (both positive and negative) on the long-term community trajectories. After using these responses and the estimated process error variance to simulate trajectories 10 years into the future, we observed that, across rivers, probabilities of 80% declines were similar between native and non-native fishes. However, when comparing risks between hydrologically-unaltered and hydrologically-altered rivers (i.e. those with a significant sustained decrease in discharge relative to the long-term seasonal profile over the last 30 years) we found striking differences in how these assemblages responded: native fishes were on average 11% more likely to hit an 80% decline threshold in hydrologically-altered rivers whereas non-native fishes decreased their average risk by 16% in altered rivers. A total of 21 frequent native and 14 frequent nonnative fishes were modelled and forecasted, with the endangered spikedace and loach minnow both surpassing 50% risks of 80% decline in the future. Our study demonstrates how ongoing supraseasonal droughts undermine riverine native assemblages, decreasing their persistence and leading to communities dominated by nonnative taxa.