Inversion of Chlorophyll-a Concentration in Donghu Lake Based on Machine Learning Algorithm

Machine learning algorithm, as an important method for numerical modeling, has been widely used for chlorophyll-a concentration inversion modeling. In this work, a variety of models were built by applying five kinds of datasets and adopting back propagation neural network (BPNN), extreme learning machine (ELM), support vector machine (SVM). The results revealed that modeling with multi-factor datasets has the possibility to improve the accuracy of inversion model, and seven band combinations are better than seven single bands when modeling, Besides, SVM is more suitable than BPNN and ELM for chlorophyll-a concentration inversion modeling of Donghu Lake. The SVM model based on seven three-band combination dataset (SVM3) is the best inversion one among all multi-factor models that the mean relative error (MRE), mean absolute error (MAE), root mean square error (RMSE) of the SVM model based on single-factor dataset (SF-SVM) are 30.82%, 9.44 μg/L and 12.66 μg/L, respectively. SF-SVM performs best in single-factor models, MRE, MAE, RMSE of SF-SVM are 28.63%, 13.69 μg/L and 16.49 μg/L, respectively. In addition, the simulation effect of SVM3 is better than that of SF-SVM. On the whole, an effective model for retrieving chlorophyll-a concentration has been built based on machine learning algorithm, and our work provides a reliable basis and promotion for exploring accurate and applicable chlorophyll-a inversion model.

Inversion of chlorophyll-a concentration in Donghu Lake based on machine learning algorithm

Machine learning algorithm, as an important method for numerical modeling, has been widely used for chlorophyll-a concentration inversion modeling. This work aims to build an effective inversion model of chlorophyll-a concentration in Lake Donghu based on machine learning algorithm. Toward this aim, a variety of models were built by applying five kinds of dataset and adopting back propagation neural network (BPNN), extreme learning machine (ELM), support vector machine (SVM). The model accuracy analysis results revealed that multi-factor dataset for modeling has the possibility to improve the accuracy of the single-factor model, and seven band combinations are better than seven single bands when modeling, Besides, SVM is more suitable than BPNN and ELM for chlorophyll-a concentration inversion modeling of Donghu Lake. SVM3 is the best inversion one among all multi-factor models that the MRE, MAE, RMSE of SF-SVM are 30.82%, 9.44 μg/L and 12.66 μg/L, respectively. SF-SVM performs a better inversion effect than SF-BPNN and SF-ELM, the MRE, MAE, RMSE of SF-SVM are 28.63%, 13.69μg/L and 16.49μg/L, respectively. In addition, the simulation effect of SVM3 is better than that of SF-SVM. On the whole, an effective model for retrieving chlorophyll-a concentration has been built based on machine learning algorithm, and our work provides a reliable basis and promotion for exploring accurate and applicable chlorophyll-a inversion model.

Temporal and Spatial Variations of Chlorophyll a Concentration and Eutrophication Assessment (1987–2018) of Donghu Lake in Wuhan Using Landsat Images

Chlorophyll a (Chl-a) concentration, which reflects the biomass and primary productivity of phytoplankton in water, is an important water quality parameter to assess the eutrophication status of water. The band combinations shown in the images of Donghu Lake (Wuhan City, China) captured by Landsat satellites from 1987 to 2018 were analyzed. The (B4 − B3)/(B4 + B3) [(Green − Red)/(Green + Red)] band combination was employed to construct linear, power, exponential, logarithmic and cubic polynomial models based on Chl-a values in Donghu Lake in April 2016. The correlation coefficient (R2), the relative error (RE) and the root mean square error (RMSE) of the cubic model were 0.859, 9.175% and 11.194 μg/L, respectively and those of the validation model were 0.831, 6.509% and 19.846μg/L, respectively. Remote sensing images from 1987 to 2018 were applied to the model and the spatial distribution of Chl-a concentrations in spring and autumn of these years was obtained. At the same time, the eutrophication status of Donghu Lake was monitored and evaluated based on the comprehensive trophic level index (TLI). The results showed that the TLI (∑) of Donghu Lake in April 2016 was 63.49 and the historical data on Chl-a concentration showed that Donghu Lake had been eutrophic. The distribution of Chl-a concentration in Donghu Lake was affected by factors such as construction of bridges and dams, commercial activities and enclosure culture in the lake. The overall distribution of Chl-a concentration in each sub-lake was higher than that in the main lake region and Chl-a concentration was highest in summer, followed by spring, autumn and winter. Based on the data of three long-term (2005–2018) monitoring points in Donghu Lake, the matching patterns between meteorological data and Chl-a concentration were analyzed. It revealed that the Chl-a concentration was relatively high in warmer years or rainy years. The long-term measured data also verified the accuracy of the cubic model for Chl-a concentration. The R2, RE and RMSE of the validation model were 0.641, 2.518% and 22.606 μg/L, respectively, which indicated that it was feasible to use Landsat images to retrieve long-term Chl-a concentrations. Based on longitudinal remote sensing data from 1987 to 2018, long-term and large-scale dynamic monitoring of Chl-a concentrations in Donghu Lake was carried out in this study, providing reference and guidance for lake water quality management in the future.

Numerical Simulation of Donghu Lake Hydrodynamics and Water Quality Based on Remote Sensing and MIKE 21

Numerical simulation is an important method used in studying the evolution mechanisms of lake water quality. At the same time, lake water quality inversion technology using the characteristics of spatial optical continuity data from remote sensing satellites is constantly improving. It is, however, a research hotspot to combine the spatial and temporal advantages of both methods, in order to develop accurate simulation and prediction technology for lake water quality. This paper takes Donghu Lake in Wuhan as its research area. The spatial data from remote sensing and water quality monitoring information was used to construct a multi-source nonlinear regression fitting model (genetic algorithm (GA)-back propagation (BP) model) to invert the water quality of the lake. Based on the meteorological and hydrological data, as well as basic water quality data, a hydrodynamic model was established by using the MIKE21 model to simulate the evolution rules of water quality in Donghu Lake. Combining the advantages of the two, the best inversion results were used to provide a data supplement for optimization of the water quality simulation process, improving the accuracy and quality of the simulation. The statistical results were compared with water quality simulation results based on the data measured. The results show that the water quality simulation of chlorophyll a and nitrate nitrogen mean square errors fell to 17% and 24%, from 19% and 31% respectively, after optimization using remote sensing spatial information. The model precision was thus improved, and this is consistent with the actual pollution situation of Donghu Lake.