COMPARISON BETWEEN THE LINEAR MODEL AND K-NEAREST NEIGHBOR METHOD FOR PREDICTING MACROINVERTEBRATE ASSEMBLES IN A CITY RIVER IN BEIJING, CHINA

L. YANG
2018 Applied Ecology and Environmental Research  
Yang et al.: Comparison between the linear model and k-nearest neighbor method for predicting macroinvertebrate assembles in a city river in Beijing, China -387 -APPLIED ECOLOGY AND ENVIRONMENTAL RESEARCH 16(1):387-406. Abstract. Benthic macroinvertebrates play an important role in materials and energy flow in river ecosystems. In this paper, we built models, a linear model and k-nearest neighbor method, for predicting biodiversity of macroinvertebrates in a city river using the data from Wenyu
more » ... River. Both Shannon-Wiener index and Simpson index were considered for measuring the biodiversity of macroinvertebrates. The observed data of macroinvertebrates and 12 water quality indicators in Wenyu River, from 2010 to 2012, were applied in building and validating the predicted models. The results indicated that 1) The validity of the linear model was, though not perfect, better for predicting macroinvertebrates diversity using water quality indicators than k-nearest neighbor method in a city river; 2) Simpson index was more robust and accurate than the other biodiversity index to act as the variable of predicting benthic macroinvertebrates in a city river. There were 89.47% observations within the 99% confidence intervals. The developed predictive model was a useful tool for assessing river health, especially city river health, without taking into account the abundances of invertebrates.
doi:10.15666/aeer/1601_387406 fatcat:grrgm2t6fbhqjiduchboftg6k4