Populations proliferate unexpectedly due to disturbances prevalent in ecosystems nowadays. Recently population
size increased rapidly in a number of aquatic species in streams and rivers. However, population eruptions are a
complex manner caused by numerous environmental factors including pollutions sources. A machine learning
technique was used for data analysis in this study. Benthic macroinvertebrate communities were collected by the
Surber net and dredge across the dam areas along the main tributary in the river system in Korea including Nakdong
River, Han-River, Geumgang River and Young-san River monthly and bimonthly from March, 2018 to May, 2019.
Population densities during the survey periods were trained by Geo-Self-Organizing Map (Geo-SOM). The Geo-
SOM training illustrated geographic areas specifically presenting high population densities in Chironomids, Bryozoa,
and Oligochaetes, being concurrently associated with environmental factors pertaining to the sampled dam areas.
Some species including Tanytarsus sp. 1, Pectinatella magnifica, and Limnodrilus hoffmeisteri were locally abundant
in relation to high levels of nutrients (e.g., TN, TP), and chemical (e.g., pH, DO) and physical (e.g., water speed,
substrate size) factors. Clusters and nodes that were grouped by high levels of population densities in the Geo-SOM
were presented in an organized manner to serve as a source of prognosing population proliferations in association
with places and seasons in field conditions. Feasibility of extracting complex information from population density
changes by machine learning was further discussed regarding prediction and management of aquatic species.