Cumulative impacts of multiple extreme and novel environmental changes on communities are often the result of asynchronous rather than simultaneous exposures to such stressors. Yet, the importance of temporal dynamics remains a major knowledge gap in multiple stressor ecology, lacking theory or evidence. We provide a conceptual template for predicting the ecological importance of the order in which consecutive stressors occur (i.e. an exposure order effect) based on correlated species responses. Negative correlation of species responses is hypothesized to increase, while positive correlation is expected to reduce, exposure order effects of consecutive stressors on communities. Towards a proof of concept, we experimentally exposed planktonic communities from fishless mountain lakes to different temporal sequences of two stressors, namely invasive sportfish and elevated water temperatures. Both stressors suppressed the same resident top predator and large grazers while eliciting positive responses from smaller tolerant taxa, attesting to their interchangeable effects across species based on size selection. As a result, reversal of the order of exposure to the two stressors did not alter their combined effects on community composition and function. Our findings suggest that the order in which consecutive stressors occur may not matter to a community if ecological memory of either stressor induces tolerance of the other.

Thirty 1000-L capacity cubical mesocosms each constructed of translucent high density polyethylene were deployed in a matrix arrangement on land at the Barrier Lake field station (51 01’37”N, 115 02’08”W), which is operated by the University of Calgary and the Biogeoscience Institute at the eastern edge of the Canadian Rocky Mountains. The mesocosms were filled to capacity with filtered (63-μm mesh size) snowmelt water collected from a nearby pond. Separate collections of plankton from the deepest point in each of five fishless montane lakes in Banff National Park were used to inoculate each of the six blocks of randomly positioned mesocosms with a target density of three crustacean zooplankton L^-1, which equated to the average concentration across the five lakes. Zooplankton were collected from each lake by performing vertical hauls of the entire water column one meter above the water-sediment interface using a 30-cm diameter conical net equipped with 63-μm-sized mesh. Phytoplankton were concentrated using a 30-cm diameter, 10-μm mesh-sized conical net. Planktonic communities were collected from several lakes rather than only a single lake in effort to increase the generality and applicability of our findings across individual lakes.

Based on the experimental design proposed by Giller et al. (2004), our experiment consisted of six treatment levels (warming then plus fish, fish then plus warming, simultaneous fish + warming, fish only, warming only, and control) that were each replicated five times using randomized blocking for a period of 42 days. Application of the first stressor on day 1 was followed with the addition of the second stressor on day 21 to achieve both sequential-stressor treatment levels. A warming treatment of +4°C was achieved using four 300-watt aquarium heaters (Hagen^®, Montreal, Canada) per mesocosm. The fish treatment was applied by stocking mesocosms with a single rainbow trout fingerling. Each mesocosm was thoroughly mixed prior to sampling for zooplankton and phytoplankton by draining 80 L of water via a built-in spigot. Zooplankton was concentrated by passing the entire water sample through a 63-micron mesh-sized metal sieve. Crustacean zooplankton were taxonomically enumerated and measured using a Leica MZ9s dissecting scope, and counts converted to biomass using length-mass regressions. An additional 2 L sample was filtered to concentrate phytoplankton on to a Whatman GF/F filter, which was frozen and then later freeze-dried prior to chlorophyll analysis using high performance liquid chromatography (Vinebrooke and Leavitt 1999). Additional details of the setup of the experiment and sample analyses can be found reported by MacLennan et al. (2015).