Loading...

The Impact of Water Quality on the Accuracy of Water Depth Estimation Using Remote Sensing

Yeganeh, Yasna | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57768 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Danesh Yazdi, Mohammad
  7. Abstract:
  8. Bathymetric maps are used for monitoring water bodies such as lakes, wetlands, and coastal areas, providing valuable information regarding changes in water storage volumes in these environments. For lakes prone to drying and salinity issues, where the lakebed elevation is constantly changing, the importance of these maps becomes even more pronounced as they play a crucial role in monitoring these lakes' conditions and assessing restoration efforts. Various methods exist for generating bathymetric maps, one of which is satellite remote sensing, which is more time- and cost-effective than traditional methods. One of the most common and accessible types of remote sensing data used for bathymetric mapping is satellite hyperspectral imagery. The pixel values in satellite images represent the reflected light from the surface, which depends on factors such as water turbidity, the impact of the lakebed on reflectance, and water depth. Research based on this type of data generally assumes a dependency between pixel reflectance and the corresponding water depth, which enables the creation of bathymetric maps by establishing a relationship between these variables. However, the effect of water quality on pixel reflectance values and the increase in water depth estimation errors have not been adequately considered. This research aims to generate bathymetric maps of Lake Urmia across seven different time periods using field hydrographic data alongside Landsat 8 satellite images and machine learning methods, including Artificial Neural Networks, k-Nearest Neighbors, and Random Forest. The study incorporates water quality characteristics to reduce water depth or lakebed elevation estimation errors. For this purpose, two water quality factors—salt concentration and the NDTI (Normalized Difference Turbidity Index, representing turbidity)—were added as new inputs to the bathymetric model. Results indicate that considering salt concentration during the second to fourth data collection periods increased the mean R² for depth prediction along various transects from 0.70 to 0.88, and incorporating the turbidity index across seven data collection series increased the R² for depth prediction from 0.68 to 0.87
  9. Keywords:
  10. Remote Sensing ; Machine Learning ; Water Quality ; Lake Water Quality ; Urumieh Lake ; Optical Data ; Bathymetry

 Digital Object List

 Bookmark

No TOC