Stream community responses to stressor release are still an enigma to the science and practice of ecological restoration. Streams may show puzzling combinations of rapid early recovery (within weeks) and subsequent failure to fully reassemble, even after decades. Some streams have shifted to an alternative community state in which they remain, by processes and feedbacks that are currently not well-understood. Stream communities often show asymmetric responses to an increase and release of stressors, as indicated by new states, or slow and incomplete recovery. In restoration practice we may recognise such responses as resistance to restorative interventions. The Asymmetric Response Concept (ARC) is a theoretical framework ‘in progress’ that we designed to address these knowledge gaps. According to the ARC, asymmetries result from shifts in the relative importance of three mechanisms: tolerance, dispersal and biotic interactions, during degradation and recovery phases. Here we propose to test the core of the ARC (Main Hypothesis 1) and to develop new and more precise ARC theory. Our central aim is to mechanistically explain stream community reassembly in a food web context. Such insight is needed to enable more effective restoration of degraded stream ecosystems. In Phase I, we developed and analysed the ARC food web model. The model integrates nutrient cycling, high quality tolerance data and interactions among the major functional groups in a stream ecosystem, based on Boye catchment field data. We found that biotic interactions already strongly governed responses before stressor release. This was reflected in functional groups having different relative tolerance rankings in the dynamic food web context than would be expected from single species tolerance data. Recovery often emerged as patterns that differed between communities and ecosystem function. We further found that only very low degrees of degradation allowed symmetric recovery. As soon as degradation reached moderate to high levels, recovery was asymmetric and ecosystem function showed long return times. Under severe degradation entire functional groups were lost. Recolonisation by dispersal was then required to reach full recovery. Adequate dispersal invariably guaranteed such full recovery. This implies that the mechanisms that so strikingly give rise to recovery failure and new states in the field have not yet been included in the model. In Phase II we will include the trait-based level of ecological realism that allows recovery failure and new states to emerge. First, we include traits and processes that act below the functional group level of organisation, i.e. at the level of species or trait profile groups. This introduces redundancy (to test Main Hypothesis 3) and the possibility of priority effects. The latter are a root cause of incomplete recovery and community closure. We will expose the food web model to degradation events that vary by type and degree and evaluate recovery trajectories in relation to the five ARC outcomes of reassembly (rubber band, broken leg, partial recovery, no recovery and new state). We will then use a new diagnostic method that links underlying ecological mechanisms to outcomes of community reassembly. For each trait scenario that we run with the food web model, we will characterise the corresponding dynamical behaviour (species turnover and the time series of impact and recovery trajectories) by means of autocorrelation functions (ACFs), both in single patch and multi-patch scenarios, so that we can follow recovery in both space and time. An ACF converts the complex multivariate representation of a time series into a single equation defined by time lags. This simplifies the detection of patterns. Each ACF will be characterised to create a library of ACF fingerprints. In animals and plants, the top-five species ranked as most responsive to a degradation event often explain 70-90 % of all community level response patterns, including ACFs. These fingerprints serve to diagnose the ecological processes underlying the observed population trajectories. This represents the first part of our diagnostic method. The other part comes from characterising the same period of impact and recovery in terms of the distance to the original community (which is the measure that defines the five ARC outcome patterns). This is done by calculating the Bray-Curtis dissimilarity between the last undisturbed year or period and each of the following years, during degradation and recovery. These two parts give us (1) the diagnostic ACF fingerprints of species turnover and of the time series of the top-ranking species (and hence processes) that most strongly defined community response, and (2) the BrayCurtis pattern that defines which of the five ARC reassembly outcomes actually occurred for the community as a whole. Comparing these two enables us to link trait-based scenarios, (with f.e. dispersal-dominated, tolerance-dominated or biotic interactions-dominated dynamics), to the five ARC reassembly outcomes. Note that to mechanistically establish that very link is the ultimate goal and purpose of RESIST and ARC theory. The next level of analysis is to move from these idealised trait-based scenarios to actual cases. We will also calculate these ACFs and Bray-Curtis dissimilarities for each field-based time series analysed by A17, and for each set of patches analysed with the metacommunity model for the Boye and Kinzig catchments by A20, based on RESIST’s sampling program in these catchments. This creates three libraries of ACF fingerprint profiles: one based on process-based food web model predictions (A19), one based on advanced statistical analysis of large numbers of time series from the field with GLMs (A17), and one based on catchment-level patch occupancy metacommunity model analyses (A20). Their comparison and mapping onto the five ARC outcomes will provide a combined empirical-theoretical perspective on: which food web processes, which combinations of traits, which environmental conditions in the field and which types of spatial dynamics across the landscape give rise to the alternative outcomes of community reassembly.

Annabel Kuppels (Ruhr Universtiy Bochum)

Testing the Asymmetric Response Concept in disturbed and recovering stream ecosystems: integrating the contributions of multi-stressor tolerance, dispersal and biotic interactions to (A)symmetry of Response

Disturbance of a stream community by one or more stressors can result in the stream community being unable to recover or only partially recovering. This can even be the case when all abiotic conditions have been fully restored. This is particularly evident in the benthic invertebrate community. Using mathematical models of stream communities and food webs, the project tries to find out which ecological mechanisms lead to a complete recovery or only to a depauperate state of the system.

The occurrence of one or more stressors, individually or in combination, lead to primary local extinctions. These are predictable due to species-specific tolerance characteristics. The primary extinction of one or more species can lead to further, secondary, extinctions and changes in the relative abundance of the remaining species. These changes are dependent on the food web structure of the stream and cannot be predicted by species tolerance traits alone.

Simulations of food webs in streams and analyses of community viability make it possible to find out the number and order of secondary extinction events. These indicate the set of degraded conditions from which recovery must start. The next step is to test the contribution of ecological mechanisms such as differences in tolerances, dispersal, and biotic interactions to stunted or full recovery.

Contact: annabel.kuppels@ruhr-uni-bochum.de

First Supervisor: Prof. Dr. Matthijs Vos (Ruhr Universtiy Bochum, Theoretical and applied Biodiversity)
Second Supervisor: Prof. Dr. Ralph Tollrian (Ruhr University Bochum, Animal Ecology, Evolution and Biodiversity)
Mentor: Dr. Wouter Helmer (Rewilding Europe)