Lotic Ecosystem Health Assessments Using an Integrated Analytical Approach of Physical Habitat, Chemical Water Quality, and Fish Multi-Metric Health Metrics
Sang-Jae Lee 1  
Eui-Haeng Lee 2  
Kwang-Guk An 1  
More details
Hide details
Department of Bioscience and Biotechnology, Chungnam National University, Daejeon, South Korea
Eui-Haeng Lee, Fishing Village Development Office, Korea Rural Community Corporation, Naju, South Korea
Online publish date: 2018-04-15
Publish date: 2018-05-30
Submission date: 2017-08-03
Final revision date: 2017-08-21
Acceptance date: 2017-09-27
Pol. J. Environ. Stud. 2018;27(5):2113–2131
This study evaluates integrative lotic ecosystem health using neural network modeling and principal component analysis of physical, chemical, and biological parameters in 33 streams and rivers of a large watershed. Water chemistry parameters were measured to detect chemical health, and physical habitat health was determined by a model of qualitative habitat evaluation index (QHEI). Also, biological health was determined by the multi-metric community fish model of index of biological integrity (IBI) and then analyzed trophic compositions and tolerance guilds. In addition, we analyzed fish tissues of liver, kidney, gill, vertebra, and muscle using a sentinel species of Zacco platypus. Chemical pollutions were closely associated with land-use patterns within the watershed and the locations of major point-sources. Model value of QHEI as a measure of physical habitat health averaged 144, indicating good health, and varied from 96 to 190 depending on the sampling sites. The proportion of sensitive fish species in the tolerance guilds had negative correlation with organic matter pollution (r = -0.716, p<0.001) and had positive a relationship with IBI (r = 0.683, p<0.001) and QHEI (r = 0.573, p = 0.001). The proportion of insectivore species, as a trophic composition indicator, was inversely correlated with BOD (r = -0.463, p = 0.007) and positive with IBI (r = 0.679, p<0.001). The analysis of the multi-layer perceptron (MLP) 14-5-1 model, based on the predicted IBI values in the training sites (R2 = 0.999, MSE = 0.015) and testing sites (R2 = 0.894, MSE = 27.4) showed high efficiency in the MLP model.
Kwang-Guk An   
Department of Bioscience and Biotechnology, Chungnam National University, Daejeon-34134, Department of Bioscience and Biotechnology, Chungnam National University, Daejeon-34134, 34134 Daejeon, Korea (South)