Eco-hydrological dynamics and multi-temporal scale prediction models in river reservoir systems

Youkun Li¹Junqiang Lin^2*Lixiong Yu³Shangtuo Qian⁴Qidong Peng²Di Zhang²

• ¹Tianjin University, Tianjin, China
• ²China Institute of Water Resources and Hydropower Research, Beijing, China
• ³Chinese Academy of Fishery Sciences Yangtze River Fisheries Research Institute, Wuhan, China
• ⁴Hohai University, Nanjing, China

Natural reproduction is a core life process for sustaining fish populations. Exploring the response dynamics of fish spawning behavior to hydrological drivers is essential for understanding the variability and trends in fish population maintenance, thereby underpinning evidence-based formulation of conservation strategies. Leveraging over a decade of field observations (2013–2024) on the spawning abundance of the four major Chinese carps (FMCC) in the Yichang-Yidu reach downstream from the Three Gorges-Gezhouba cascade reservoirs, this study adopts statistical analysis and phase plane analysis to investigate the response patterns of FMCC spawning behavior. Addressing spawning dynamics, we develop a multi-temporal scale predictive framework, integrating imbalanced classification for long-term spawning state prediction and empirical dynamic modeling (EDM) for short-term spawning peak magnitude prediction. Long-term monitoring data reveal that the spawning processes exhibit pronounced zero-inflation characteristics on the seasonal scale, resulting in significant class imbalance among spawning states. Since 2020, the occurrence probability of small-to-medium spawning peaks has gradually declined, while spawning peaks have manifested as short-term pulse-type patterns. On the test datasets from two independent periods, the classification models successfully predict the three spawning states (no-response, low-response, and high-response) during the spawning season, achieving weighted average F1-scores of 0.66 and 0.60, respectively. The EDM models show satisfactory performance in one-step-ahead prediction of spawning peak magnitude with correlation coefficients of 0.82 and 0.71; however, nonstationary environmental regimes tend to limit the predictability of unprecedented peaks. This study establishes a regional flow-ecology relationship between hydrological processes and fish spawning dynamics under the reservoir operation regime, providing a scientific reference for the adaptive management cycle related to the ecological operation of the Three Gorges reservoir.