Network indices show heat wave-induced fragility in experimental food webs: Operational insights for ecosystem...

Emerging global change imposes challenges for marine ecosystems placing them at the verge of collapse. The identification of early signs of ecosystem collapse is crucial to assist adaptive management. Shifts in the ecosystems are preceded by rewiring of ecological networks. Environmental assessment still lacks indicators that quantify the state of ecological processes (e.g., trophic interactions). Thus, evaluating ecosystems using network indices is appropriate, although their responses to stress factors are not fully resolved. Network indices have been proposed to identify the good environmental status, but experimental evidence of their behavior under disturbance remains limited. We used data from a mesocosm experiment simulating different frequencies of sublethal heat waves (three sequential events across summer-3HW, and a single abrupt event at the end of summer-1HW) to construct carbon flow networks and calculate information theory indices, testing their responsiveness to disturbance. A comparison with ecological attributes was conducted to interpret the trends of network indices and assess whether these indices provide complementary insights to the ecological attributes. Our results show that, even without ecosystem collapse, carbon flow networks representing communities exposed to heat waves exhibited reduced link-weighted connectance and lower internal carbon-use efficiency compared to the control. These changes were associated with increased detritivory under both heat wave treatments, along with higher respiration losses and shorter average residence time of carbon after 1HW. Information theory indices proved responsive to disturbance, detecting early signs of stress that were not captured by changes in ecological attributes alone. Our findings show that information theory indices respond consistently to heat waves. Grounded in robust theoretical foundations, our study advances the operationalization of these indices for ecosystem-based management.