The nitrogen-to-phosphorus ratio (TN:TP) is a key indicator influencing phytoplankton nutrient limitation and growth dynamics, directly regulating algal growth rates, abundance, and community structure, thereby affecting the process of water eutrophication. This study aims to evaluate the modeling performance of integrated machine learning approaches for lake total nitrogen to total phosphorus ratios (TN:TP), utilizing Zhuhai-1 hyperspectral satellite imagery to develop a CNN-SVR ensemble model integrating convolutional neural networks and support vector regression for remote sensing inversion of lake TN:TP ratios. Performance is evaluated against random forest (RF) and convolutional neural network (CNN) models, systematically analyzing spatial distribution patterns and primary drivers. Results indicate that the CNN-SVR model demonstrated superior performance among the tested models, with R², RMSE, MAPD, and RPD values of 0.856, 2.675, 9.516%, and 2.390, respectively. Spatially, the nitrogen-to-phosphorus ratio in lakes during the growing season exhibits an increasing trend from the western to the eastern half of the lake, progressing from northwest to southeast. When TN:TP falls below 9, algal growth becomes nitrogen-limited, indicating a higher degree of eutrophication; when TN:TP exceeds 22.6, phosphorus becomes the limiting factor, indicating lower eutrophication levels. A similar distribution pattern is observed during the non-growing season. Regarding driving mechanisms, the nitrogen-to-phosphorus ratio during the growing season is primarily influenced by TN accumulation and shows significant correlations with dissolved oxygen (DO) and pH. During the non-growing season, while still affected by TN input, its association with other water quality parameters is weaker. The results indicate that the combined use of CNN and SVR improves feature extraction and model fitting in nitrogen-to-phosphorus ratio inversion and helps clarify its ecological significance as an indicator of algal growth. This provides methodologies and evidence for precise diagnosis and ecological management of lake eutrophication.

Dissolved organic carbon (DOC) maintains the ecological balance of inland lake systems and contributes significantly to the global carbon cycle. This study aims to develop a novel deep learning algorithm to predict DOC concentrations and explore its modeling performance in nonlinear relationships. We used hyperspectral imagery (HSI) from the Chinese Ziyuan-1 satellite series alongside in situ water sample data to construct a Convolutional Multi-Head Attention Fusion Network (CMAF-Net) for prediction of DOC in Chaohu Lake, China. For comparison, we tested its performance against support vector regression (SVR), random forest (RF), and convolutional neural network (CNN) models. The spatial distribution patterns of the DOC were analyzed to explore the primary environmental drivers. The results demonstrate that CMAF-Net significantly outperforms the best-performing baseline CNN model, achieving an R² of 0.88, RMSE of 0.29 mg/L, and RPD of 2.79. Furthermore, environmental factor analysis reveals strong correlations between DOC concentrations and water temperature, total nitrogen (TN), and total phosphorus (TP), identifying them as dominant drivers of the spatial variability of DOC. Hyperspectral remote sensing integrated with CMAF-Net, under the synergistic optimization of local band feature extraction and global band-dependency modeling to screen characteristic water spectra, significantly improves DOC prediction accuracy and enhances multidimensional feature learning. The proposed approach establishes a novel pathway for the quantitative monitoring of DOC in inland aquatic lakes.