Development of a calibration-assistance optimization framework for 3D ELCOM-CAEDYM water-quality modeling

Jo, Hyo-Yeong; Rhee, Han-Pil; Son, Jangwon; Lee, Seungjae; Jun, Jin-Young; Hwang, Soon-Jin

doi:2027.32.1.260079

Environmental Engineering Research 2027;32(1): 260079 DOI: https://doi.org/10.4491/eer.2026.079

Development of a calibration-assistance optimization framework for 3D ELCOM-CAEDYM water-quality modeling

Hyo-Yeong Jo^1,2, Han-Pil Rhee^1,2

^†, Jangwon Son², Seungjae Lee³, Jin-Young Jun², and Soon-Jin Hwang¹

^†

¹Department of Environmental and Health Sciences, Konkuk University, Seoul 05029, Republic of Korea
²ET Waters Co., Ltd., Seoul 08376, Republic of Korea
³Korea Water Resources Corporation (K-water), Daejeon 34350, Republic of Korea

Corresponding Author: Han-Pil Rhee ,Tel: +82-2-455-3931, Fax: +82-2-455-3934, Email: hprhee@etwaters.co.kr
Soon-Jin Hwang ,Tel: +82-2-450-3748, Fax: +82-2-455-3934, Email: sjhwang@konkuk.ac.kr

Received: February 5, 2026; Accepted: May 20, 2026.

ABSTRACT

Three-dimensional (3D) hydrodynamic and water-quality models can represent spatial heterogeneity in stratified reservoirs, but their routine operational application is limited because calibration requires repeated model runs. To address this limitation, this study proposes a calibration-assistance optimization framework for 3D water-quality models that reduces the number of runs needed to reach an acceptable parameter set. The framework was implemented using the Estuary, Lake and Coastal Ocean Model (ELCOM) coupled with the Computational Aquatic Ecosystem Dynamics Model (CAEDYM). Unlike conventional approaches that rely on many separate runs during calibration, the proposed method compares simulated and observed values from each run across multiple stations and sampling dates and combines those differences to guide the next parameter update. This allows each run to provide multiple pieces of calibration information while reducing repeated executions. Chlorophyll-a (Chl-a) was used as the primary calibration target for evaluating the proposed framework. Application to Daecheong Dam and Namgang Dam showed convergence within 3–7 optimization iterations (i.e., 3–7 model runs), with total turnaround times of approximately 17–24 h. These results indicate that the proposed workflow can support reproducible calibration adjustment for computationally expensive 3D models.