Please login first
Satellite-based Analysis of Lake Okeechobee's Surface Water: Exploring Machine Learning Classification for Change Detection.
* ,
1  FAU
Academic Editor: Riccardo Buccolieri


Water is an essential resource for the survival of living beings. Remote sensing serves as the best possible way to detect water bodies and monitor changes across time. With a surplus amount of remote sensing data, machine learning approaches have become an effective and efficient way to detect and monitor water bodies. This research focused on utilizing remote sensing and machine learning approaches to monitor changes in the surface water of Lake Okeechobee. Landsat-7 and Landsat-8 for 2002 and 2022 were used for this analysis. Further, we used Support Vector Machine (SVM) and Random Forest (RF) classification methods to compare the classification scheme and used appropriate metrics to compare their applicability for surface water detection.

For classification, all bands with three band rationing, namely Normalized Difference Water Index(NDWI), Modified Normalized Difference Water Index (MNDWI), and Normalized Difference Vegetation Index(NDVI) were used. Thus, we increased the independent variable for this model, keeping the response variable of water vs non-water areas. With an overall accuracy of 96%, SVM outperformed RF for the classification of Landsat images. The outcomes were consistent across both models, ensuring the reliability of the model along with its metrics. Both models gave an overall accuracy of more than 90% and a kappa coefficient of 0.80. Classified images were subtracted using image differencing techniques to track the change between 2002 and 2020 in Okeechobee Lake. This approach helped to assess change in lake water on a pixel-by-pixel basis. It generated images with three categories: increasing, decreasing, or no change. The SVM model suggested an increase in lake water area in 20 years by 2,1515.11 acres and decreased by 563.10 acres.

On the other hand, RF predicted an increase in lake water area by 14947.13 acres and a decrease of 2138.32 acres in the last two decades. This research can explain the changing nature of lake areas. The water management plan adopted in the southern region of Okeechobee can be a reason for this change, where they drain more water to make land available for agriculture. The insights from this research provided a foundation for further advancements in environmental assessment and sustainable water resource planning, encouraging continued exploration of innovative methodologies to address complex ecological challenges effectively.

Keywords: Remote Sensing, Change Detection, Water Extraction, Machine learning
Comments on this paper
Currently there are no comments available.