Application of the Random Forest model for chlorophyll-a forecasts in fresh and brackish water bodies in Japan, using multivariate long-term databases

Abstract

There is a growing world need for predicting algal blooms in lakes and reservoirs to better manage water quality. We applied the random forest model with a sliding window strategy, which is one of the machine learning algorithms, to forecast chlorophyll-a concentrations in the fresh water of the Urayama Reservoir and the saline water of Lake Shinji. Both water bodies are situated in Japan and have historical water records containing more than ten years of data. The Random Forest (RF) model allowed us to forecast trends in time series of chlorophyll-a in these two water bodies. In the case of the reservoir, we used the data separately from two sampling stations. We found that the best model parameters for the number of min-leaf, and with/without pre-selection of predictors, varied at different stations in the same reservoir. We also found that the best performance of lead-time and accuracy of the prediction varied between the two stations. In the case of the lake, we found the best combination of a min-leaf and pre-selection of predictors was different from that of the reservoir case. Finally, the most influential parameters for the random forest model in the two water bodies were identified as biochemical oxygen demand (BOD), chemical oxygen demand (COD), pH, and total nitrogen/total phosphorus (TN/TP).

Deep Learning Application to Time-Series Prediction of Daily Chlorophyll-a Concentration

“…The 19 environmental items are measured in the station, and input variables for models were selected based on previous studies. The nine selected indicators; water temperature, pH, dissolved oxygen (DO), electrical conductivity (EC), turbidity, total organic carbon (TOC), total nitrogen (TN), total phosphorus (TP), and chlorophyll-a concentration were found as effective predictive parameters in previous studies [1], [3]- [5], [11]- [13]. Excluded items from observation data are the concentration of volatile organic compounds (VOCs) such as trichloroethane and benzene which are measured because of their strong toxicity.…”

“…Many previous studies applied data based machine learning methods to predict the algal bloom represented by the concentration of chlorophyll-a. For example, there have been applications of the artificial neural network (ANN) [1]- [3], support vector machine (SVM) [4] and Random Forest [4], [5]. However, complexity and non-linearity among the factors associated with algal blooms make it difficult to identify the process of algal bloom occurrence.…”

A Review of Recent Machine Learning Advances for Forecasting Harmful Algal Blooms and Shellfish Contamination

“…In an RF model, each individual tree is built on a random subsampling of the input data. The final output of the model is then determined by aggregating the results generated by each tree making up the forest [35]. This method increases the robustness and generalization accuracy of individual decision trees [35,64] (Table 1).…”

“…Yajima and Derot (2018) [35] proposed an RF model to forecast algal blooms in the Urayama Reservoir (with fresh water) and the Lake Shinji (with saline water), Japan. The authors used monthly water records containing more than ten years of data to forecast chl-a concentration one to six months ahead.…”

Monthly Load Forecasting Based on Economic Data by Decomposition Integration Theory

“…Random Forest (RF) is a novel machine learning algorithm developed in recent years [35][36][37][38]. RF is robust to the input and number of samples.…”