Remote sensing-based water quality index estimation using data-driven approaches: a case study of the Kali River in Uttar Pradesh, India

Abstract

The present study evaluates the water quality status of 6-km-long Kali River stretch that passes through the Aligarh district in Uttar Pradesh, India, by utilizing high-resolution IRS P6 LISS IV imagery. In situ river water samples collected at 40 random locations were analyzed for seven physicochemical and four heavy metal concentrations, and the water quality index (WQI) was computed for each sampling location. A set of 11 spectral reflectance band combinations were formulated to identify the most significant band combination that is related to the observed WQI at each sampling location. Three approaches, namely multiple linear regression (MLR), backpropagation neural network (BPNN) and gene expression programming (GEP), were employed to relate WQI as a function of most significant band combination. Comparative assessment among the three utilized approaches was performed via quantitative indicators such as R 2 , RMSE and MAE. Results revealed that WQI estimates ranged between 203.7 and 262.33 and rated as “ very poor ” status. Results further indicated that GEP performed better than BPNN and MLR approaches and predicted WQI estimates with high R 2 values (i.e., 0.94 for calibration and 0.91 for validation data), low RMSE and MAE values (i.e., 2.49 and 2.16 for calibration and 4.45 and 3.53 for validation data). Moreover, both GEP and BPNN depicted superiority over MLR approach that yielded WQI with R 2 ~ 0.81 and 0.67 for calibration and validation data, respectively. WQI maps generated from the three approaches corroborate the existing pollution levels along the river stretch. In order to examine the significant differences among WQI estimates from the three approaches, one-way ANOVA test was performed, and the results in terms of F -statistic ( F = 0.01) and p -value ( p = 0.994 > 0.05) revealed WQI estimates as “ not significant ,” reasoned to the small water sample size (i.e., N = 40). The study therefore recommends GEP as more rational and a better alternative for precise water quality monitoring of surface water bodies by producing simplified mathematical expressions.

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

“…The BP consists of input layers, implicit layers and output layers, and contains two stages: forward propagation and back propagation of the error [52]. The error is back-propagated through the implicit layer to the output layer, and apportioned to all units in each layer, until the error is eventually decreased to an acceptable level after continual training [53].…”

Spatio-Temporal Dynamics Coupling between Land Use/Cover Change and Water Quality in Dongjiang Lake Watershed Using Satellite Remote Sensing

“…In Table 5, the R 2 of TP is the lowest among the parameters. This is because remote sensing data are limited by atmospheric conditions, sensor noise, viewing angles, and other factors [32], resulting in the fact that achievement of very high R 2 values is challenging when using remote sensing to inverse water parameters. However, data provide broad and continuous observations, which offer valuable information over large areas that traditional in situ measurements cannot reach.…”

Deep learning for photovoltaic panels segmentation

“…[34] attempted to estimate and solve soil moisture using the AMSR-E sensor data as inputs to multiple back propagation neural networks. [35] evaluated the status of water quality feeding the network different band combinations to estimate the water quality index (WQI). Over all, BPNN are widely used with remotely sensed data.…”