Capturing threshold responses of marine benthos along gradients of natural and anthropogenic change

摘要

1. Ecologists and managers need to understand what types of communities emerge with continued human alterations to ecosystems against a background of natural change. Both natural and anthropogenic drivers are well known to affect organisms' distributions; however, it often remains unclear where along a range of environmental and anthropogenic gradients important compositional community changes occur. 2. We used a big-data approach, including over 175,000 presence records of benthic genera for the North Sea, to identify environmental (bed shear stress, sediment grain size, temperature) and anthropogenic parameters (trawling effort) driving benthic community composition over a 21-year period. We applied a Gradient Forest analysis, based on Random Forests, to estimate the locations and importance of thresholds where small cumulative increases in the predictors drive a much greater change in genus composition than would be expected from linear effects. 3. Shear stress was the most important predictor of benthic community composition. Trawling effort, temperature gradients and sediment grain size were of intermediate importance. This corroborates that current and wave effects (typically associated with seabed substrate types) are primary determinants of benthic communities. 4. Our results suggest that a genus composition threshold for both infauna and epifaunal benthic communities is crossed when the seafloor is trawled as little as once every 4 years. Higher trawling levels corresponded with gradual compositional change without obvious thresholds, which would be consistent with chronic fishing in the North Sea over the last two centuries having caused persistent, longterm changes in ecosystem structure and functioning. This was corroborated by the large-scale spatial patterns of benthic community composition undergoing limited temporal changes during the 21-year study period. 5. Synthesis and applications. Although well established in theory, threshold effects are poorl