Human activities, such as climate change and pollution, are significantly altering ecosystems by modifying the strength and nature of ecological interactions, thereby impacting community composition. Ecological interactions are crucial for community dynamics and ecosystem processes, yet they are often overlooked in biodiversity change studies. Understanding the structure and dynamics of these interactions, conceptualized as information networks, is essential for comprehending how global change alters ecosystem structure and function. While the impact of human activities on ecological networks is known, effective tools are urgently needed to predict the combined effects of different stressors. Lake ecosystems are particularly vulnerable, as warming intensifies the impact of slight increases in nutrient levels, potentially triggering dramatic ecosystem shifts. Changes in network properties can lead to rapid community structural shifts and increased extinction risks. These changes can manifest as alterations in the number of interactions or the strength of existing interactions, affecting network connectance and the strength of species interactions, especially in trophic networks. These structural properties signal large-scale ecosystem changes with implications for ecosystem stability and biodiversity maintenance. However, studying these network reorganizations is challenging due to the scarcity of long-term, high-quality time series data on complete ecological networks, complex nonlinear interactions, and the need for appropriate inference methods. Most previous research focuses on single interaction types and often assumes fixed interactions over time, neglecting the complexities of interacting environmental stressors. This study aims to address these gaps by investigating the effects of two major anthropogenic stressors—warming and nutrient pollution—on plankton networks.

Existing literature highlights the reorganization of ecosystems due to human impacts, emphasizing the often-overlooked role of ecological interactions in biodiversity change. Research emphasizes the importance of studying interaction networks to understand the impact of global change on ecosystem structure and function. The need for tools to predict the combined effects of stressors, particularly warming and nutrient pollution in lakes, is highlighted. Studies have shown that changes in network properties can lead to rapid shifts in community structure and increase extinction risks. However, there is limited knowledge about how entire interaction networks reorganize in response to global change due to challenges like data scarcity, complex nonlinear interactions, and the need for specific inference methods. Much of the past research focuses solely on one type of interaction and often assumes that interactions are constant over time. Theoretical work provides only partial expectations of how natural ecosystems respond to multiple anthropogenic stressors.

This study analyzed 20–43 years of monthly plankton community data from ten peri-alpine Swiss lakes, along with measurements of water temperature and phosphate levels. The lakes experienced managed re-oligotrophication (phosphorus reduction) and a period of increased warming. Plankton species were grouped into trophic guilds based on body size, nutritional requirements, and foraging behavior, resulting in a conceptual network of up to 15 nodes. Convergent cross-mapping (CCM), a nonlinear causality test from the empirical dynamic modeling (EDM) framework, was used to identify causal associations between network nodes and quantify interaction strength. CCM quantifies how changes in one time series predict changes in another. To minimize the impact of seasonal variations, interactions were considered significant only if their strength exceeded that of a seasonal surrogate null model. Connectance (percentage of significant associations) and average interaction strength were measured using a 60-month moving window. S-maps, another EDM tool, were employed to model network properties as a function of temperature, phosphate, lake depth, and volume, disentangling the effects of warming and re-oligotrophication while accounting for lake morphometric differences. The frequency and strength of trophic, non-trophic, and hybrid links were analyzed to understand how different interaction types and guilds contribute to network changes. A sensitivity analysis was conducted to assess the influence of excluding certain guilds on the results. The data was winsorized to reduce the impact of outliers while retaining data points.

The study revealed that connectance and average interaction strengths varied over time and across lakes. Re-oligotrophication increased connectance in two lakes, while warming decreased it in six lakes. Top-down causal links were more frequent than bottom-up links in most lakes, although bottom-up links were often stronger. Re-oligotrophication slightly increased top-down links relative to bottom-up in some lakes, whereas warming generally decreased top-down links relative to bottom-up. S-map models predicted that most lakes exhibit high prevalence and strength of top-down control under low to intermediate temperatures, shifting to bottom-up control under high temperatures and phosphate levels. Hybrid links were significantly more common than trophic or non-trophic links, but were weaker in strength. Non-trophic links showed the greatest average strength. Small grazers (rotifers, ciliates, mixotrophic flagellates) and colonial cyanobacteria were found to be dominant in trophic controls and broadly connected in the network, acting as key indicators of changes in plankton network structure. Ciliate and cyanobacteria abundances were strongly influenced by long-term temperature changes.

The findings indicate that warming and nutrient fluctuations significantly disrupt plankton ecological networks in lakes. The shift from top-down to bottom-up control under warming and high phosphate levels suggests increased sensitivity to nutrient inputs. The dominance of small grazers and cyanobacteria in trophic control highlights their importance as indicators of network reorganization and ecosystem health. The frequent occurrence of weak hybrid links points to the role of intermediate consumers and generalists in structural changes and ecosystem stability. These results underscore the importance of considering multiple stressors and their non-additive effects when assessing ecosystem resilience. The study's findings are relevant for lake management and conservation strategies, highlighting the need for incorporating network dynamics into monitoring programs. The identified sensitive indicator species (small grazers and cyanobacteria) can aid in early warning systems for ecological shifts.

This research demonstrates that warming and phosphorus levels significantly affect the structure and dynamics of plankton ecological networks in lakes, leading to shifts in trophic control and increased bottom-up influence under high temperatures and phosphorus. The identification of small grazers and cyanobacteria as key indicator species offers valuable insights for monitoring and management practices. Future research should investigate the broader implications of these findings for ecosystem services and explore the effectiveness of management strategies tailored to network dynamics.

The study is limited to ten Swiss lakes within a specific geographic and climatic region, potentially restricting the generalizability of findings to other lake types or regions. The aggregation of taxa into guilds may obscure species-specific responses to environmental changes. The use of CCM and S-maps relies on the quality and completeness of the time series data, which may introduce uncertainties in the analysis.