Evaluation of Ocean Forecast within GloSea5 with Multivariable Integrated Evaluation Index
Jieun Wie1, Hyomee Lee2, Byung-Kwon Moon1, Hyojin Park3, Pil-Hun Chang4, Johan Lee5 and Yoonjae Kim5, (1)Chonbuk National University, Jeonju, South Korea, (2)National Institute of Meteorological Sciences, Global Environment System Research Division, Seogwipo, Korea, Republic of (South), (3)Gimje Girls' High School, South Korea, (4)National Institute of Meteorological Sciences, Forecast Research Department, Seogwipo, South Korea, (5)National Institute of Meteorological Sciences, Seogwipo, South Korea
Abstract:
Seasonal prediction data about the earth environment is very useful in disaster prevention or environmental policy decisions. Up until now, the assessment of long-range prediction performance generally has focused on evaluating atmosphere models, not as much surveying in ocean models. We seek to establish a system for intercomparison and diagnosis of the performance of ocean models. The long-range forecast model used for this assessment is the high-resolution Global Seasonal Forecast System version 5, which is currently operated by the Korea Meteorological Administration. Initial data and monthly hindcast forecast data that reproduced the past climate were used in the analysis. The lead time of the hindcast data ranged one to seven months, and the forecast target month was from 1992 to 2009. Variables for evaluation are sea temperature, salinity, currents, SSH, and the 13 sea areas were selected. SODA Reanalysis data were used as reference data. As the first step toward this, this study applied the multivariable integrated evaluation index (MIEI; Xu et al., 2016, 2017), which can demonstrates the overall quantitative performance of the model.
As a result of evaluating the model after eliminating the seasonal cycle of GloSea5, most shore regions showed a decreasing trend in performance as forecast lead time increased. In addition, prediction performance was determined according to target season rather than initial prediction month, for instance, when to forecast Northern hemisphere's summer season of June to September in the equatorial Pacific the prediction performance was poorer than the other seasons. Even applying MIEI to CMIP5 and CMIP6 evaluation, there were a distinct difference between the performances of the models. Through the evaluation, it was found that MIEI is an appropriate diagnostic metric to evaluate the overall ocean simulation performance.
This work was funded by the Korea Meteorological Administration Research and Development Program under Grant KMI (KMI2018-03513). This work was funded by the Korea Meteorological Administration Research and Development Program "Development of Climate Prediction System" under Grant (1365003054).
