Integrating piecewise and symbolic regression with remote sensing data for spatiotemporal analysis of surface water total dissolved solids in the Karun River, Iran
Monitoring water quality, including total dissolved solids (TDS), across various spatial and temporal scales is essential for comprehending water health level. The Karun River's water is a vital resource for drinking water supply and agricultural irrigation, making accurate monitoring of its qu...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-03-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123025002476 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Monitoring water quality, including total dissolved solids (TDS), across various spatial and temporal scales is essential for comprehending water health level. The Karun River's water is a vital resource for drinking water supply and agricultural irrigation, making accurate monitoring of its quality crucial for ensuring public health and sustainable resource management. This study presents a novel hybrid modeling approach that combines piecewise and symbolic regression (SR) models with Landsat imagery to predict TDS levels in river systems. Specific spectral bands, including the red and near-infrared reflectance from Landsat-9, were used as input variables. Field sampling of TDS, as the output variable, was conducted along the Karun River on three dates, synchronized with Landsat satellite images. The innovative use of a fuzzy-based uncertainty analysis, coupled with AHP weighting, allowed for a comprehensive assessment of TDS estimation accuracy and uncertainty. The Composite Uncertainty Index (CUI) approach revealed that the Multivariate Adaptive Regression Splines (MARS) and M5 models performed better than other models, with CUI values of 0.83 and 0.72, respectively. MARS demonstrated higher accuracy under low uncertainty conditions, while M5P excelled in scenarios of elevated uncertainty due to its reduced sensitivity and strong Nash-Sutcliffe coefficient. The hybrid modeling approach presented in this study offers a unique contribution to remote sensing-based water quality monitoring. By leveraging advanced regression techniques and uncertainty quantification, the findings enable more reliable predictions of TDS levels, which are crucial for sustainable river management. |
|---|---|
| ISSN: | 2590-1230 |