Evaluation of Watershed Scale Aquatic Ecosystem Health by SWAT Modeling and Random Forest Technique

Abstract

In this study, we evaluated the aquatic ecosystem health (AEH) with five grades (A; very good to E; very poor) of FAI (Fish Assessment Index), TDI (Trophic Diatom Index), and BMI (Benthic Macroinvertebrate Index) using the results of SWAT (Soil and Water Assessment Tool) stream water temperature (WT) and quality (T-N, T-P, NH 4 , NO 3 , and PO 4 ). By applying Random Forest, one of the machine learning algorithms for classification analysis, each AEH index was trained and graded from the SWAT results. For Han river watershed (34,418 km 2 ) in South Korea, the 8 years (2008~2015) observed AEH data of Spring and Fall periods at 86 locations from NAEMP (National Aquatic Ecological Monitoring Program) were used. The AEH was separately trained for Spring (FAI s , TDI s , and BMI s ) and Fall (FAI a , TDI a , and BMI a ), and the AEH results of Random Forest with SWAT (WT, T-N, T-P, NH 4 , NO 3 , and PO 4 ) as input variables showed the accuracy of 0.42, 0.48, 0.62, 0.45, 0.4, and 0.58, respectively. The reason for low accuracy was from the weak strength of the individual trees and high correlation between the trees composing the Random Forest due to the data imbalance. The AEH distribution results showed that the number of Grade A of total FAI, TDI, and BMI were 84, 0, and 158 respectively and they were mostly located at the upstream watersheds. The number of Grade E of total FAI, TDI, and BMI were 4, 50, and 13 and they were shown at downstream watersheds.

Prediction of Aquatic Ecosystem Health Indices through Machine Learning Models Using the WGAN-Based Data Augmentation Method

“…Therefore, it appears that the ML models trained using the raw training set for BMI overfit the grade A and do not correctly classify all grades in the test set. In the study of Woo et al [18], the results also showed that the predictive performance of the RF for a small number of grades (B to E) was relatively lower than that of majority grade A due to the effect of data imbalance. Additionally, although the samples by grade of the raw training set for TDI were relatively balanced, each ML model did not correctly classify the grade of TDI as a whole (Figure 11).…”

“…The data-driven Machine Learning (ML) models make effective predictions by mining the relevant information between input and output variables inherent when using a larger dataset without the physical process required by conventional numerical models [13]. In many studies, various ML models have been successfully used for the estimation of runoff [13], rainfall erosivity (R-factor) [14], dam discharge [15], sediment [16], water quality [17], and AEH [18,19], etc. However, the predictive performance of an ML model depends greatly on the quantity and quality of datasets, so it is necessary to collect sufficient amounts of data samples to build robust ML models.…”

“…However, the predictive performance of an ML model depends greatly on the quantity and quality of datasets, so it is necessary to collect sufficient amounts of data samples to build robust ML models. In S. Korea, AEH monitoring sites and time frames are irregular due to climatic characteristics, with large seasonal rainfall fluctuations [18]. Thus, continuous monitoring for BMI, TDI, and FAI could be time-consuming and labor-intensive tasks.…”

Correlation Analysis between Hydrologic Flow Metrics and Benthic Macroinvertebrates Index (BMI) in the Han River Basin, South Korea

“…However, as the extent of impervious surfaces increases, the runoff response is amplified from increasingly smaller precipitation events [9]. Although it is important to accurately determine the arrangement and proportional amounts of different functional types of impervious surface cover, it is also critical to understand how geometric characteristics can affect relationships between components of the hydrologic cycle and ecology [8,10]. Lee et al [11] examined the mediating effects of streamline geometry on the relationships between urban land use and the index of biological integrity (IBI) of the Nakdong River in Korea.…”

“…Woo et al [10] evaluated aquatic ecosystem health using water quality modeling results via the sequential wavelength assignment technique (SWAT) and random forest technique. They suggested that the aquatic ecosystem be investigated using data such as flow rate and river depth.…”