Optimizing the spatiotemporal configuration of water quality monitoring networks in reservoirs using anisotropic Bayesian maximum entropy method

Variations in water quality along the length and depth of a reservoir reveal anisotropic conditions, which pose significant challenges when designing effective monitoring networks. Geostatistical techniques like Bayesian maximum entropy (BME) have proven effective in designing monitoring systems, but they fall short when it comes to planning water quality monitoring in the depth and length of reservoirs. This paper introduces a novel approach for designing long-term, routine water quality monitoring networks specifically tailored for deep reservoirs. Due to the considerable anisotropy in the data and the large length-to-depth ratio of the reservoir, we modeled the anisotropies by scaling the longitudinal distances and rotating the coordinate axes. To examine long-term variations in water quality within reservoirs, a calibrated CE-QUAL-W2 hydrodynamic and water quality simulation model was employed, along with a regular hexagonal grid pattern to determine potential locations for monitoring stations. The proposed methodology outlined the ideal configuration for a reservoir water quality monitoring network, specifying the number of monitoring stations needed and the sampling frequency. The quality monitoring network was designed based on two crucial criteria: the variance of estimation error of the BME method and the sampling cost. The BME method, which can integrate information from various sources, including both hard (deterministic) and soft (stochastic) data, reduces the variance of the estimation error compared to traditional geostatistical methods, leading to more accurate estimates. Using the evidential reasoning (ER) method based on the criteria mentioned earlier, we ranked various alternatives for the locations of monitoring stations and their sampling frequencies.

We applied the proposed methodology to the Karkheh Dam reservoir, the largest reservoir in Iran, which faces notable challenges related to thermal stratification and water quality. The results suggest a monitoring network of 10 sampling stations with a 75-day sampling interval for effective water quality management. This approach offers a robust framework for water quality monitoring and resource management in large reservoirs by helping decision-makers balance accuracy, cost, and uncertainty to design resilient and cost-effective monitoring networks.