Landscape and Ecological Engineering Экология

Data-driven typology of Japanese dam reservoirs: a machine learning approach

Doi, Hideyuki, Ichiyanagi, Hidetaka, Urabe, Jotaro
Springer Nature · 2026-11-16 · DOI 10.1007/s11355-026-00776-8

Аннотация

Dams are fundamental components of modern river systems in Japan. However, the current classification system for these reservoirs is based primarily on administrative definitions rather than functional characteristics. This study establishes a data-driven, structural-morphological typology with functional implications for 110 major Japanese reservoirs using random forest (RF) and principal component analysis (PCA) based on their structural specifications together with hydrological and operational characteristics compiled from the national dam database. Using bRF, with PCA as a linear benchmark, we classified the reservoirs into six distinct types, ranging from old, large-catchment reservoirs to compact modern water-supply reservoirs. We discuss the utility of this typology as a framework for stratifying sites for studying dam-reservoir issues. Dam functional grouping provides a scientific basis for differentiating management strategies across the diverse landscape of Japanese reservoirs.

Читать у издателя →