National parks serve as protected areas that ensure the sustainable use of representative natural ecosystems. In South Korea, approximately 46% of the nation’s total species and 65% of its endangered species inhabit national parks. Comprehensive resource surveys have been conducted every 5 years in accordance with the Natural Parks Act to safeguard these invaluable biological resources. However, species identification in these surveys still largely relies on traditional methods, which are time-consuming and labor-intensive, and often subject to investigator bias. This study investigated freshwater fish communities in stream ecosystems at four national parks-Seoraksan, Woraksan, Deogyusan, and Hallyeohaesang-using environmental DNA (eDNA) metabarcoding to evaluate the consistency of its results with those of conventional surveys. A total of 33 fish species (five orders, 11 families) were recorded in both methods. eDNA detected 31 species (detection rate: 93.9%), whereas conventional methods identified 22 species (66.7%). However, species richness values were significantly correlated between the two methods. eDNA exhibited higher detection efficiency (88.7-100.0%) across all parks than conventional methods (64.3-72.2%). Six endangered species were identified in total, four of which (66.7%) were detected by eDNA. The results indicate that eDNA metabarcoding provides a more sensitive, efficient, and reliable tool for assessing freshwater fish diversity in national parks than conventional methods. Overall, eDNA offers a valuable complement or potential alternative to conventional survey methods for the long-term monitoring of freshwater fish biodiversity.