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

摘要: In this study, we evaluated the aquatic ecosystem health (AEH) with five grades (A; very good to E; poor) of FAI (Fish Assessment Index), TDI (Trophic Diatom and BMI (Benthic Macroinvertebrate Index) using results SWAT (Soil Water Tool) stream water temperature (WT) quality (T-N, T-P, NH4, NO3, PO4). By applying Random Forest, one machine learning algorithms for classification analysis, each AEH index was trained graded from results. For Han river watershed (34,418 km2) in South Korea, 8 years (2008~2015) observed data Spring Fall periods at 86 locations NAEMP (National Aquatic Ecological Monitoring Program) were used. The separately (FAIs, TDIs, BMIs) (FAIa, TDIa, BMIa), Forest (WT, T-N, PO4) as input variables showed accuracy 0.42, 0.48, 0.62, 0.45, 0.4, 0.58, respectively. reason low weak strength individual trees high correlation between composing due imbalance. distribution that number Grade A total FAI, TDI, 84, 0, 158 respectively they mostly located upstream watersheds. E 4, 50, 13 shown downstream