Improving Forecast Accuracy of Ensemble Regional Tide and Storm surge Model

The storm surge forecast model holds significant importance in minimizing coastal damages caused by strong low-pressure system (e.g., typhoon). In general, the model accuracy is largely influenced by atmospheric forcing. In the case of a deterministic forecast model, therefore, the surge forecast can vary significantly depending on the uncertainty of the atmospheric model. In the present study, we developed the Ensemble Regional Tide·Storm Surge Model (ETSM) for the sea around Korean Peninsula, utilizing the forecast results of 26 members calculated from an ensemble atmospheric model to address the uncertainty of a single forecasting model. For the assessment of ETSM, we carried out numerical experiment from July to September 2022 (ETSM_0). In addition, three sensitive experiments were conducted, considering initial bias correction (ETSM_1), the increase of ensemble members by using time-lag method (ETSM_2), and both methods (ETSM_3). When we compared ETSM_0 to additional experiments, the ensemble spread range generally widened in ETSM_2 and ETSM_3 due to the increase of ensemble members especially in the period of Typhoon HINNAMNOR, which leads to the inclusion of observations within the spread. Verification of the probability of surge heights exceeding the threshold of 10 cm for each forecast time (0, 24, 48, 72, 96, 120hr) showed that the Brier Skill Score (BSS) was 0.14 on average for ETSM_O, while it was 0.47 for ETSM_1, 0.19 for ETSM_2, and 0.47 for ETSM_3. Furthermore, the Relative Operation Characteristic (ROC) was 0.73 on average for ETSM_O, 0.79 for ETSM_1, 0.75 for ETSM_2, and 0.81 for ETSM_3. This indicates an overall improvement in probability forecast performance when applying initial bias correction and increasing the number of ensemble.