UAV Multispectral Image-Based Urban River Water Quality Monitoring Using Stacked Ensemble Machine Learning Algorithms—A Case Study of the Zhanghe River, China

Abstract

Timely monitoring of inland water quality using unmanned aerial vehicle (UAV) remote sensing is critical for water environmental conservation and management. In this study, two UAV flights were conducted (one in February and the other in December 2021) to acquire images of the Zhanghe River (China), and a total of 45 water samples were collected concurrently with the image acquisition. Machine learning (ML) methods comprising Multiple Linear Regression, the Least Absolute Shrinkage and Selection Operator, a Backpropagation Neural Network (BP), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost) were applied to retrieve four water quality parameters: chlorophyll-a (Chl-a), total nitrogen (TN), total phosphors (TP), and permanganate index (CODMn). Then, ML models based on the stacking approach were developed. Results show that stacked ML models could achieve higher accuracy than a single ML model; the optimal methods for Chl-a, TN, TP, and CODMn were RF-XGB, BP-RF, RF, and BP-RF, respectively. For the testing dataset, the R2 values of the best inversion models for Chl-a, TN, TP, and CODMn were 0.504, 0.839, 0.432, and 0.272, the root mean square errors were 1.770 μg L−1, 0.189 mg L−1, 0.053 mg L−1, and 0.767 mg L−1, and the mean absolute errors were 1.272 μg L−1, 0.632 mg L−1, 0.045 mg L−1, and 0.674 mg L−1, respectively. This study demonstrated the great potential of combined UAV remote sensing and stacked ML algorithms for water quality monitoring.

Synergistic Use of Earth Observation Driven Techniques to Support the Implementation of Water Framework Directive in Europe: A Review

“…Interesting findings resulted for the TP (Total Phosphorous) and TN (Total Nitrogen) where five studies [7,[37][38][39][40] from the total number of 71 examined the TP while four studies [7,38,39,41] the TN. The impressive thing is that for both QEs, the UAVs' contribution was noteworthy reporting 40% for TP [38,39] and 60% for TN [38,39,41]. These findings show the great potential of the UAV platforms in the domain of water monitoring.…”

“…However, we should emphasize that research groups typically have legacy algorithms (e.g., deep learning) that are difficult to re-write and deploy on GEE or DC platforms, or researchers often find it difficult to deploy specific pre-processing routines (e.g., radiometric calibration, mosaicking). Hence, there are many software available for these needs such as the SNAP (Sentinel Application Platform) tool [110,111] and the ENVI image analysis software [32,54] that couple deep learning with all types of data including multi-spectral, hyperspectral, thermal, LiDAR and SAR and the PIX4D for UAV data processing [39,74,78] and the MEPHySto offering a hyperspectral pre-processed library [73].…”

“…Based on this review, a host of novel in situ observational platforms such as UAVs and USVs and small-sized multiple (hyperspectral, multispectral) sensors mounted on them have been shown to be rapidly maturing and becoming valuable alternatives to more expensive conventional solutions. Because of their affordability, flexibility, and susceptibility to interference from clouds, UAVs are excellent choices for surface water monitoring, especially in small water bodies [39] while they can contribute very well in shallow water applications [130]. The sensors on board could offer reliable and near-real time information having higher spectral and spatial resolution providing continuous ground feature spectral data.…”

“…ventional solutions. Because of their affordability, flexibility, and susceptibility to interference from clouds, UAVs are excellent choices for surface water monitoring, especially in small water bodies [39] while they can contribute very well in shallow water applications [130]. The sensors on board could offer reliable and near-real time information having higher spectral and spatial resolution providing continuous ground feature spectral data.…”

Remote-sensing retrieval of total phosphorus in Tiande Lake by optimizing an XGBoost machine learning model

“…Wang et al [10] proposed several regression methods and physical models based on hyperspectral satellite data and field survey to monitor the turbidity, TN, TP, and total organic carbon in alpine rivers. Taking Zhanghe River (Hubei Province, China) as the research region of interest, Xiao et al [11] established retrieval models for Chl-a, TN, TP, and chemical oxygen demand (COD) separately using the traditional regression algorithm, ML algorithm, and stacked ML algorithm to compare the applicability and accuracy of different models. The results show that the stacked ML algorithm exhibits the highest retrieval accuracy.…”