Multispectral Remote Sensing for Estimating Water Quality Parameters: A Comparative Study of Inversion Methods Using Unmanned Aerial Vehicles (UAVs)

Abstract

Multispectral remote sensing technology using unmanned aerial vehicles (UAVs) is able to provide fast, large-scale, and dynamic monitoring and management of water environments. We here select multiple water-body indices based on their spectral reflection characteristics, analyze correlations between the reflectance values of water body indices and the water quality parameters of synchronous measured sampling points, and obtain an optimal water body index. A representative selection, such as statistical analysis methods, neural networks, random forest, XGBoost and other schemes are then used to build water-quality parameter inversion models. Results show that the XGBoost model has the highest accuracy for dissolved oxygen parameters (R2 = 0.812, RMSE = 0.414 mg L−1, MRE = 0.057) and the random forest model has the highest accuracy for turbidity parameters (R2 = 0.753, RMSE = 0.732 NTU, MRE = 0.065). Finally, spatial distribution maps of dissolved oxygen and turbidity of water bodies in the experimental domain are drawn to visualize water-quality parameters. This study provides a detailed comparative analysis of multiple inversion methods, including parameter quantity, processing speed, algorithm rigor, solution accuracy, robustness, and generalization, and further evaluates the technical characteristics and applicability of several inversion methods. Our results can provide guidance for improved small- and medium-sized surface-water quality monitoring, and provide an intuitive data analysis basis for urban water environment management.

Water Quality Inversion of a Typical Rural Small River in Southeastern China Based on UAV Multispectral Imagery: A Comparison of Multiple Machine Learning Algorithms

“…In a study by Huangfu et al [75], the inversion of several small/mediumsized rivers (with a minimum width of 10 m) in the Xinyang section of the Huaihe River was achieved based on Sentinel-2 satellite images, yielding R 2 values ranging from 0.60 to 0.67. Yan et al [76] successfully inverted dissolved oxygen (DO) and TUB in the Biyu River (with a water area of 10 5 m 2 and a length of 2.5 × 10 3 m) using UAV multispectral imagery and XGBoost and RF models, with inversion accuracies (R 2 ) ranging from 0.75 to 0.81. Hou et al [16] conducted a study around the Fuyang River section (with a water area covering about 1.5 × 10 5 m 2 and a length of 700 m), achieving the inversion of six water quality indicators based on UAV images and employing partial least squares (PLS), RF, and Lasso models, with R 2 reaching up to 0.90.…”

Predictive Modeling of Urban Lake Water Quality Using Machine Learning: A 20-Year Study

“…For instance, increasing the treatment efficiency of sewage before it is released can reduce BOD. Similarly, implementing green infrastructure like rain gardens or permeable pavements can help control runoff and reduce turbidity [57][58][59].…”

A Feature Selection Method Based on Relief Feature Ranking with Recursive Feature Elimination for the Inversion of Urban River Water Quality Parameters Using Multispectral Imagery from an Unmanned Aerial Vehicle

“…Xiang et al used Temporal Convolutional Networks (TCNs), LightGBM, and four single features to make a quadratic decomposition-based water quality prediction model, demonstrating that LightGBM is suitable for handling the low-frequency components in information, making the model more flexible [22]. Yan et al conducted a comparative analysis of the technical characteristics and accuracy of several inversion models, including RF, and constructed an optimal water quality parameter inversion model, discussing the influence of different inversion methods on the prediction of water quality parameters [23].…”