Search

• The method combines laboratory analysis, data modelling and deep learning. • HF-DFM quantifies various water quality parameters by hyperspectral reflectance. • The method is effectively applied to monitoring water quality of urban rivers. • Prediction MAPEs range from 8.78%–12.36% for five water quality parameters. [Display omitted] Environmental protection of water resources is of critical importance to daily life of human beings. In recent years, monitoring the variation of water quality using remote sensing techniques has become prevalent. Unmanned aerial vehicle (UAV) based remote sensing techniques have been applied to quantitative retrieval of concentrations of water quality parameters including phosphorus, nitrogen, chemical oxygen demand (COD), biochemical oxygen demand (BOD), and chlorophyll a (Chl-a), successfully and efficiently. In this study, a novel method with deep factorization machine, spatial distribution pattern analysis, and probabilistic analysis engaged, named hybrid feedback deep factorization machine (HF-DFM), has been developed to quantitatively estimate concentrations of water quality parameters based on hyperspectral reflectance data on large scale effectively. Our proposed method is a unified model for quantifying concentrations of water quality parameters with an end to end structure, which integrates UAV based optical remote sensing techniques and deep learning to estimate concentrations of water quality parameters. Furthermore, our proposed model was applied to real-time quantitative monitoring the variation of water quality of Mazhou River, Shenzhen, Guangdong, China. Finally, we evaluate the performance of proposed model on a real-world dataset in terms of root of mean squared error (RMSE), mean absolute percent error (MAPE), and coefficient of determination (R 2). The experimental results show that our proposed model outperforms other state-of-the-art models with respect to RMSE, MAPE, and R 2 , where resulting MAPEs for quantifying all water quality parameters range from 8.78% to 12.36%, and resulting R 2 s range from 0.81 to 0.93. It can serve as a useful tool for decision makers in effectively monitoring water quality of urban rivers. [ABSTRACT FROM AUTHOR]

The protection of water resources is of paramount importance to human beings' practical lives. Monitoring and improving water quality nowadays has become an important topic. In this study, a novel Bayesian probabilistic neural network (BPNN) improved from ordinary Bayesian probability methods has been developed to quantitatively predict water quality parameters including phosphorus, nitrogen, chemical oxygen demand (COD), biochemical oxygen demand (BOD), and chlorophyll a. The proposed method, based on conventional Bayesian probability methods, involves feature engineering and deep neural networks. Additionally, it extracts significant information for each endmember from combinations of spectra by feature extraction, with spectral unmixing based on mathematical and statistical analysis, and calculates each of the water quality parameters. The experimental results show the great performance of the proposed model with all coefficient of determination R 2 over 0.9 greater than the values (0.6–0.8) from conventional methods, which are greater than ordinary Bayesian probability analysis. The mean percent of absolute error (MPAE) is taken into account as an important statistical criterion to evaluate model performance, and our results show that MPAE ranges from 4% (nitrogen) to 10% (COD). The root mean squared errors (RMSEs) of phosphorus, nitrogen, COD, BOD, and chlorophyll-a (Chla) are 0.03 mg/L, 0.28 mg/L, 3.28 mg/L, 0.49 mg/L, and 0.75 μg/L, respectively. In comparison with other deep learning methods, this study takes a relatively small amount of data as training data to train the proposed model and the proposed model is then tested on the same amount of testing data, achieving a greater performance. Thus, the proposed method is time-saving and more effective. This study proposes a more compatible and effective method to assist with decomposing combinations of hyperspectral signatures in order to calculate the content level of each water quality parameter. Moreover, the proposed method is practically applied to hyperspectral image data on board an unmanned aerial vehicle in order to monitor the water quality on a large scale and trace the location of pollution sources in the Maozhou River, Guangdong Province of China, obtaining well-explained and significant results. [ABSTRACT FROM AUTHOR]

Protection of water environments is an important part of overall environmental protection; hence, many people devote their efforts to monitoring and improving water quality. In this study, a self-adapting selection method of multiple artificial neural networks (ANNs) using hyperspectral remote sensing and ground-measured water quality data is proposed to quantitatively predict water quality parameters, including phosphorus, nitrogen, biochemical oxygen demand (BOD), chemical oxygen demand (COD), and chlorophyll a. Seventy-nine ground measured data samples are used as training data in the establishment of the proposed model, and 30 samples are used as testing data. The proposed method based on traditional ANNs of numerical prediction involves feature selection of bands, self-adapting selection based on multiple selection criteria, stepwise backtracking, and combined weighted correlation. Water quality parameters are estimated with coefficient of determination R 2 ranging from 0.93 (phosphorus) to 0.98 (nitrogen), which is higher than the value (0.7 to 0.8) obtained by traditional ANNs. MPAE (mean percent of absolute error) values ranging from 5% to 11% are used rather than root mean square error to evaluate the predicting precision of the proposed model because the magnitude of each water quality parameter considerably differs, thereby providing reasonable and interpretable results. Compared with other ANNs with backpropagation, this study proposes an auto-adapting method assisted by the above-mentioned methods to select the best model with all settings, such as the number of hidden layers, number of neurons in each hidden layer, choice of optimizer, and activation function. Different settings for ANNS with backpropagation are important to improve precision and compatibility for different data. Furthermore, the proposed method is applied to hyperspectral remote sensing images collected using an unmanned aerial vehicle for monitoring the water quality in the Shiqi River, Zhongshan City, Guangdong Province, China. Obtained results indicate the locations of pollution sources. [ABSTRACT FROM AUTHOR]