A literature review of recent trends in environmental flow assessment and applications in water-limited rivers

Environmental flows (e-flows) are defined as the quantity, timing, and quality of water flows required to support aquatic ecosystem health while supporting human needs and well-being. This review paper explores the development and application of e-flows assessment methods over the last decade and the growing implementation of machine learning models in e-flows management. The e-flows methods identified in this review were classified into four groups: hydrological, habitat simulation, multivariate approaches and holistic frameworks. While hydrological methods remain the most common approach, a significant tendency towards multimetric hydrological methods, vs single minimum flow value prescription is evident, considering the natural variability of the hydrological regime and the impact of hydrological flow alteration on ecological processes. Combined multivariate approaches are increasingly applied in e-flows studies by linking several methods (hydrological, hydraulic, habitat) or including additional variables other than flow. For example, water temperature is a key water quality variable, increasingly included to strengthen proposed e-flows and to support fish habitat quality and recruitment. Recent literature also indicates that water quality parameters (e.g., dissolved oxygen, turbidity, pH) can be more easily implemented as part of e-flows frameworks than in past decades, given the advances in remote sensing techniques and open data accessibility. This review highlights recent research related to the growing concerns of climate change and groundwater pumping impacts on e-flows depletion and aquatic ecosystem degradation, emphasizing their urgent inclusion in e-flows assessment frameworks. The results also underscore the recent advancements in remote sensing, machine learning models, and statistical modelling related to assessment and monitoring, showing the potential of these techniques in improving e-flows estimation. The increasing application of machine learning models in low-flow prediction demonstrates their reliability as tools in e-flows management, and hybrid models that combine machine learning and physically based approaches represent promising developments in e-flows studies.