Despite the consensus that warming will affect biodiversity, alter physicochemical environments, and disrupt biological interactions, the relative importance of these key processes and how they interact to determine overall ecosystem function is poorly understood. Here, we analyze long-term (16~39 years) time series data from ten aquatic ecosystems and use convergent cross mapping (CCM) to quantify the hidden causal network linking species diversity, ecosystem function, and physicochemical factors. We observe that aquatic ecosystems subject to stronger warming exhibit decreased stability (larger fluctuations in phytoplankton biomass). We further show that this effect can be attributed to a weakening of stabilizing causal pathways between biodiversity, nutrient cycling, and phytoplankton biomass. Thus, rather than thinking in terms of separate factors, a more holistic view, that causally links biodiversity and the other ecosystem components, is required to understand and predict climate impacts on the temporal stability of aquatic ecosystems.only data from freshwater systems by excluding the two marine datasets (Wc and Ng). We, however, do not have sufficient data to examine exclusively marine systems.
ComputationAll analyses were done with R (ver. 3.1.2). The CCM analyses and structural equation modeling were implemented using the rEDM (Hao Ye et al. 2013) and lavaan (Rosseel 2012) packages, respectively. Documentation of all the analytical procedures mentioned above is provided in the Supplementary Information R script.