Widespread Underestimation of Drought Impacts on River Ecosystems Due to Weak Experimental Designs.

Hydrologic droughts are intensifying globally due to climate change and human water demand. Although a myriad of drought impacts on river ecosystems have been described, manipulative experiments on the topic have often yielded inconclusive or inconsistent results. Here, we investigated to what extent this inconsistency could be due to uncontrolled variation in methodological choices, or to variation in how different ecological endpoints respond to drought. To this end, we developed a meta-analysis of 1284 effect sizes from 53 experiments, and estimated drought impacts on key variables of riverine biodiversity and ecosystem functioning. We evaluated study design (e.g., temporal vs. spatial focus, controls, treatment characteristics), focal ecosystem endpoint and variable type (i.e., quantity, structure, function), and level of biological organization studied. We found that methodological features strongly influenced the detectability and magnitude of the estimated drought effects, with Before-After-Control-Impact (BACI) designs reporting 2× stronger impacts than simpler designs. Effects differed among measures of quantity, structure, and function; and across ecosystem endpoints, with sensitivity to drought generally increasing with trophic position. Finally, for BACI designs, responses tended to be 1.5× weaker at higher levels of organization (ecosystem) relative to lower levels (population), suggesting an important role of compensatory dynamics and functional redundancy in absorbing drought impacts. Our findings confirm that drought degrades river ecosystem structure and functioning, but effect detectability and magnitude strongly depend on experimental choices and ecological focus. Although studies with replication over space and time (e.g., BACI) require more effort, we show that they are uniquely powerful in parsing out true drought effects from confounding effects of other time-varying processes. We contend that embracing robust experimental designs that include spatio-temporal controls and multiple levels of organization would vastly improve our ability to forecast ecological consequences of global change in river ecosystems.