Comparison of Linear and Nonlinear Regression Models based on Machine Learning for Estimating Chlorophyll-a values in Coastal Waters of Qeshm and Hormoz Island

Chlorophyll-a, as an important indicator of algal blooms and water quality, holds significant importance in marine studies. This research aims to compare linear and nonlinear regression models based on machine learning algorithms for estimating the chlorophyll-a levels in the coastal waters of Bandar Abbas, Hormuz, and Qeshm Island. The study utilizes data from the TERRA sensor’s MODIS satellite and field measurements from various points within the study area. The examined models include linear regression, generalized linear model with Poisson distribution, random forest, and support vector machine. The performance of these models is evaluated using metrics such as root mean square error (RMSE), mean percentage error (MPE), mean absolute error (MAE), and coefficient of determination (R-squared). The results demonstrate that linear regression and generalized linear models perform poorly, while random forest and support vector machine exhibit better performance. Particularly, the random forest model shows the highest performance with an RMSE of 0.5725 and R-squared of 0.6632. This model has the capability to detect nonlinear and complex patterns and can effectively handle large datasets by employing a large number of decision trees. Overall, this research highlights the effectiveness of machine learning models, especially random forests, in accurately predicting chlorophyll-a levels as a crucial factor in managing marine ecosystems in the study area.