Enhancing reservoir water quality simulation through machine learning-driven remote sensing integration with EFDC: a...

Area of Science:

• Environmental Science
• Hydrology
• Remote Sensing

Background:

• Reservoir water quality management is crucial but often hindered by a lack of in-situ data.
• Accurate simulation of water quality parameters like total nitrogen (TN), total phosphorus (TP), and chlorophyll-a (Chl-a) is essential for effective eutrophication control.

Purpose of the Study:

• To develop and validate a coupled modeling framework integrating machine learning-based remote sensing retrievals with the Environmental Fluid Dynamics Code (EFDC).
• To improve the accuracy of reservoir water quality simulations, particularly in data-scarce environments.
• To assess the effectiveness of different management strategies for eutrophication control.

Main Methods:

• Utilized Landsat imagery (2013-2023) and machine learning algorithms (Random Forest, Gradient Boosting, AdaBoost) to derive spatiotemporal distributions of TN, TP, and Chl-a.
• Incorporated these remote sensing-derived parameters as dynamic boundary conditions into the EFDC model.
• Conducted seasonal analysis and scenario simulations to evaluate management strategies, including external load reduction and outflow regulation.

Main Results:

• The coupled model demonstrated reduced simulation errors (0.13-5.28%) and increased mean R² from 0.70 to 0.81 compared to standalone EFDC.
• Retrieval-based estimates showed lower mean relative errors for TN (20.61%), TP (28.95%), and Chl-a (26.08%).
• An optimal strategy combining 30% external load reduction and 1.0% outflow reduction achieved significant concurrent reductions in TN (27.4%), TP (23.7%), and Chl-a (13.2%).

Conclusions:

• Integrating remote sensing data significantly enhances hydrological model accuracy for reservoir water quality simulation.
• Reducing external nutrient loads alone is insufficient for effective algal biomass control; combined strategies with hydrodynamic outflow regulation are necessary.
• The developed framework provides a valuable tool for data-scarce reservoir management and informs operational decision-making for eutrophication control.