The accurate forecast of algal blooms can provide helpful information for water resource management. However, the complex relationship between environmental variables and blooms makes the forecast challenging. In this study, we build a pipeline incorporating four commonly used machine learning models, Support Vector Regression (SVR), Random Forest Regression (RFR), Wavelet Analysis (WA)-Back Propagation Neural Network (BPNN) and WA-Long Short-Term Memory (LSTM), to predict chlorophyll-a in coastal waters. Two areas with distinct environmental features, the Neuse River Estuary, NC, USA—where machine learning models are applied for short-term algal bloom forecast at single stations for the first time—and the Scripps Pier, CA, USA, are selected. Applying the pipeline, we can easily switch from the NRE forecast to the Scripps Pier forecast with minimum model tuning. The pipeline successfully predicts the occurrence of algal blooms in both regions, with more robustness using WA-LSTM and WA-BPNN than SVR and RFR. The pipeline allows us to find the best results by trying different numbers of neuron hidden layers. The pipeline is easily adaptable to other coastal areas. Experience with the two study regions demonstrated that enrichment of the dataset by including dominant physical processes is necessary to improve chlorophyll prediction when applying it to other aquatic systems.
This paper presents a comprehensive approach to predicting short-term (for the upcoming 2 weeks) changes in estuarine dissolved oxygen concentrations via machine learning models that integrate historical water sampling, historical and upcoming 2-week meteorological data, and river discharge and discharge metrics. Dissolved oxygen is a critical indicator of ecosystem health, and this approach is implemented for the Neuse River Estuary, North Carolina, U.S.A., which has a long history of hypoxia-related habitat degradation. Through meticulous data preprocessing and feature selection, this research evaluates the predictions of dissolved oxygen concentrations by comparing a recurrent neural network with four other models, including a Multilayer Perceptron, Long Short-Term Memory, Gradient Boosting, and AutoKeras, through sensitivity experiments. The input predictors to our prediction models include water temperature, turbidity, chlorophyll-a, aggregated river discharge, and aggregated wind based on eight directions. By emphasizing the most impactful predictors, we streamlined the model-building processes and built a hindcast system from 2015 to 2019. We found that the recurrent neural network model was most effective in predicting the dissolved oxygen concentrations, with an R2 value of 0.99 at multiple stations. Different from our machine learning hindcast models that used observed upcoming meteorological and discharge data, an actual forecast system would use forecasted meteorological and discharge data. Therefore, an actual operational forecast may have lower accuracy than the hindcast, as determined by the accuracy of the predicted meteorological and discharge data. Nevertheless, our studies enhance our understanding of the factors influencing dissolved oxygen variability and set the basis for the implementation of a predictive tool for environmental monitoring and management. We also emphasized the importance of building station-specific models to improve the prediction results.
In this study, we investigate spatial and temporal variabilities of particulate organic carbon (POC) in the East China Sea (ECS) and explore the dominant influencing factors for its three subregions—coastal, continental shelf and open sea—by analyzing satellite derived POC from 2001 to 2011 after calibration against in situ samplings. The analysis reveals that the largest POC occurs in spring at the coastal region with a value of 297.5 mg m−3, and the smallest in summer at the open sea region with a value of 108.9 mg m−3. POC in the coastal region is dominated by biological activity related to river discharge of nutrients and organic matter, and is regulated by sediment discharge variation due to the Three Gorge Project and chemical fertilizer application along the Changjiang River watershed; the open sea region is dominated by water exchange with the Kuroshio current, which results in the highest sea surface temperature and the lowest chlorophyll and POC among the three subregions; POC in the continental shelf region is driven by the seasonal competition between the influence of the coastal ocean and the Kuroshio current through biological activity and water exchange. A method to estimate the organic matter reserve in the ECS was developed based on satellite data and in situ samplings. Organic carbon storage was estimated at 4.08 × 1013 g, which could be used as a baseline for future estimates. An improved estimation could be found with additional in situ data and with the use of a more sophisticated algorithm for satellite POC analysis.
27 January 2017 in Journal of Geophysical Research: Oceans
Toxic harmful algal blooms produce public health hazards in freshwater systems around the world. There is a need for forecast systems that can mitigate risk of public exposure to toxins. We improved an approach to predict the spatially and temporally resolved probability of microcystins (MCs) exceeding a threshold level (6 μg L−1) in western Lake Erie. This approach combines a 5‐day chlorophyll‐a forecast model, a weekly updated regression model predicting MCs from chlorophyll‐a, and an empirical relationship between predicted MCs and observed probability of MCs exceeding the threshold calibrated over a hindcast period. We included additional years in the database for calibration and assessment, applied an empirical bias adjustment to the Moderate Resolution Imaging Spectroradiometer for consistency with Sentinel‐3 satellite imagery, and applied a robust Siegel regression method. Cross‐validation showed reasonable skill over regions including surface water, public water system plant intake sites, and bottom waters. The forecast also presented useful skill when assessed against two intensive sampling events of Microcystis blooms in western Lake Erie in 2018 and 2019. Our results provide a comprehensive assessment of a novel method to forecast MC risk, which may be recalibrated and applied to other systems affected by toxic cyanobacterial blooms, where a similar relationship exists between chlorophyll and toxin concentrations at toxin levels relevant to advisory levels.