THEFAM: An AI-Enhanced Framework for Assessing Thermal Ecological Flow Time Series in River Networks

Abstract This paper introduces an advanced artificial intelligence (AI)–based script called the Thermal Habitat-Based Ecological Flow Assessment Model (THEFAM) designed to assess and optimize the thermal ecological flow regime necessary for preserving thermal habitats in riverine ecosystems. At its core, the introduced model optimizes the thermal ecological flow regime by integrating the simulator of water temperature dynamics into the optimization system of ecological flow. THEFAM proposes three distinct modeling approaches for water temperature simulation: two machine learning models—namely, the long short-term memory and adaptive neuro-fuzzy inference system models—and the multiple nonlinear regression models developed using particle swarm optimization, biogeography-based optimization, and invasive weed optimization. These thermal models offer flexible pathways for users to develop a robust regional model by inputting climatic and hydrological data. Using particle swarm optimization, the model assesses the ecological flow requirements at a single measurement node or multiple measurement nodes in the catchment scale to sustain temperature-sensitive aquatic habitats. In the case study, the ideal water temperature was defined as 16°C. Moreover, the maximum and minimum tolerance were defined as 5°C and 29°C, respectively. Also, the lower bound of the ecological flow regime (minimum instream flow prescribed initially) was considered 14% of monthly flow. Case study results revealed that the closeness between the natural and optimal thermal regimes in the case study suggests a strong potential for preserving thermal habitats. Also, they show that approximately 40% of the total river flow should be allocated to meet ecological thermal flow needs.