Water quality monitoring and assessment based on cruise monitoring, remote sensing, and deep learning: A case study of Qingcaosha Reservoir

Abstract

Accurate monitoring and assessment of the environmental state, as a prerequisite for improved action, is valuable and necessary because of the growing number of environmental problems that have harmful effects on natural systems and human society. This study developed an integrated novel framework containing three modules remote sensing technology (RST), cruise monitoring technology (CMT), and deep learning to achieve a robust performance for environmental monitoring and the subsequent assessment. The deep neural network (DNN), a type of deep learning, can adapt and take advantage of the big data platform effectively provided by RST and CMT to obtain more accurate and improved monitoring results. It was proved by our case study in the Qingcaosha Reservoir (QCSR) that DNN showed a more robust performance (R2 = 0.89 for pH, R2 = 0.77 for DO, R2 = 0.86 for conductivity, and R2 = 0.95 for backscattered particles) compared to the traditional machine learning, including multiple linear regression, support vector regression, and random forest regression. Based on the monitoring results, the water quality assessment of QCSR was achieved by applying a deep learning algorithm called improved deep embedding clustering. Deep clustering analysis enables the scientific delineation of joint control regions and determines the characteristic factors of each area. This study presents the high value of the framework with a core of big data mining for environmental monitoring and follow-up assessment in a manner of high frequency, multidimensionality, and deep hierarchy.

An Intelligent Early Warning System for Harmful Algal Blooms: Harnessing the Power of Big Data and Deep Learning

“…Understandability is the degree to which the internal workings of a model can be understood. 45 The modeling logic of LSTM presupposes that data adheres to the Markov decision process, 46 considering only the relationship between two consecutive time steps. Specifically, it employs the sigmoid function in the forget gate layer to selectively inherit information from the previous time step for predicting the next one.…”

“…Understandability is the degree to which the internal workings of a model can be understood . The modeling logic of LSTM presupposes that data adheres to the Markov decision process, considering only the relationship between two consecutive time steps.…”

Water Quality Prediction Based on Machine Learning and Comprehensive Weighting Methods

“…The final evaluation result for the model’s prediction parameter was obtained by calculating the average of the values, MSE values, RMSE values, and NSE values from these 10 parallel experiments. This approach allows for a comprehensive assessment of the model’s performance across diverse experiments, reducing the potential influence of random errors associated with a single experiment [ 44 ]. Running multiple experiments yields a larger set of data points, which enhances the statistical significance of the evaluation results and provides a more-comprehensive and -accurate evaluation of the model’s performance.…”

Long-Term Monitoring of Inland Water Quality Parameters Using Landsat Time-Series and Back-Propagated ANN: Assessment and Usability in a Real-Case Scenario

“…Results show good robustness with average R 2 = 0.91. Qian et al [24] tested Multiple Linear Regression (MLR), SVM, RF and ANN for monitoring of three non-optical (pH, DO, Electrical Conductivity (EC)) and one optical parameter (Turbidity) at Qingcaosha Reservoir based on Sentinel 2 images. The results indicated that ANN showed more robust performance for all WQP (RMSE: 0.33; 0.49; 0.38; 0.26 for pH, DO, EC, and Turbidity, respectively) compared to traditional ML algorithms.…”