• Ocean and Polar Research
• /
• v.23 no.2
• /
• pp.189-194
• /
• 2001

A practical algorithm for prediction of the sea surface temperatures (SST)from the satellite remote sensing data is presented in this paper. The fluctuations of SST consist of deterministic normals and stochastic anomalies. Due to large thermal inertia of sea water, the SST anomalies can be modelled by autoregressive or Markov process, and its near future values can be predicted provided the recent values of SST are available. The actual SST is predicted by superposing the pre-known SST normals and the predicted SST anomalies. We applied this prediction algorithm to the NOAA AVHRR weekly SST data for 18 years (1981-1998) in the seas adjacent to Korea (115-$145^{\circ}E$, 20-$55^{\circ}N$). The algorithm is applicable not only for prediction of SST in near future but also for nowcast of SST in the cloud covered regions.

• Atmosphere
• /
• v.31 no.2
• /
• pp.215-227
• /
• 2021

The necessity of the prediction on the Seasonal-to-Subseasonal (S2S) timescale continues to rise. It led a series of studies on the S2S prediction models, including the Global Seasonal Forecasting System Version 5 (GloSea5) of the Korea Meteorological Administration. By extending previous studies, the present study documents sea surface temperature (SST) prediction skill around the Korean peninsula in the GloSea5 hindcast over the period of 1991~2010. The overall SST prediction skill is about a week except for the regions where SST is not well captured at the initialized date. This limited prediction skill is partly due to the model mean biases which vary substantially from season to season. When such biases are systematically removed on daily and seasonal time scales the SST prediction skill is improved to 15 days. This improvement is mostly due to the reduced error associated with internal SST variability during model integrations. This result suggests that SST around the Korean peninsula can be reliably predicted with appropriate post-processing.

• The Sea:JOURNAL OF THE KOREAN SOCIETY OF OCEANOGRAPHY
• /
• v.27 no.2
• /
• pp.49-70
• /
• 2022

Sea Surface Temperature (SST), one of the ocean features, has a significant impact on climate, marine ecosystem and human activities. Therefore, SST prediction has been always an important issue. Recently, deep learning has drawn much attentions, since it can predict SST by training past SST patterns. Compared to the numerical simulations, deep learning model is highly efficient, since it can estimate nonlinear relationships between input data. With the recent development of Graphics Processing Unit (GPU) in computer, large amounts of data can be calculated repeatedly and rapidly. In this study, Short-term SST will be predicted through Convolutional Neural Network (CNN)-based U-Net that can handle spatiotemporal data concurrently and overcome the drawbacks of previously existing deep learning-based models. The SST prediction performance depends on the seasonal and interannual SST variabilities around the southern coast of Korea. The predicted SST has a wide range of variance during spring and summer, while it has small range of variance during fall and winter. A wide range of variance also has a significant correlation with the change of the Pacific Decadal Oscillation (PDO) index. These results are found to be affected by the intensity of the seasonal and PDO-related interannual SST fronts and their intensity variations along the southern Korean seas. This study implies that the SST prediction performance using the developed deep learning model can be significantly varied by seasonal and interannual variabilities in SST.

• Proceedings of the KSRS Conference
• /
• 1998.09a
• /
• pp.236-241
• /
• 1998

The NOAA AVHRR remote sense SST data, collected by the National Fisheries Research and Development Institute (NFRDI), are analyzed in order to understand the spatial and temporal distributions of SST in the seas adjacent to Korea. Our study is based on 10-day SST images during last 7 years (1991-1997). For a time series analysis of multiple 557 images, all of images must be aligned exactly at the same position by adjusting the scales and positions of each SST image. We devised an algorithm which yields automatic detections of cloud pixels from multiple SST images. The cloud detection algorithm is based on a physical constraint that SST anomalies in the ocean do not exceed certain limits (we used $\pm$ 3$^{\circ}C$ as a criterion of SST anomalies). The remote sense SST data are tuned by comparing remote sense data with observed SST at coastal stations. Seasonal variations of SST are studied by harmonic fit of SST normals at each pixel. The SST anomalies are studied by statistical method. We found that the SST anomalies are rather persistent with time scales between 1 and 2 months. Utilizing the persistency of SST anomalies, we devised an algorithm for a prediction of future SST Model fit of SST anomalies to the Markov process model yields that autoregression coefficients of SST anomalies during a time elapse of 10 days are between 0.5 and 0.7. We plan to improve our algorithms of automatic cloud pixel detection and prediction of future SST. Our algorithm is expected to be incorporated to the operational real time service of SST around Korea.

• Atmosphere
• /
• v.16 no.2
• /
• pp.97-110
• /
• 2006

This study investigates the response of a typhoon model to the change of the sea surface temperature (SST) throughout the model integration. The SST change is parameterized as a formulae of which the magnitude is given as a function of not only the intensity and the size but the moving speed of tropical cyclone. The formulae is constructed by referring to many previous observational and numerical studies on the SST cooling with the passage of tropical cyclones. Since the parameterized cooling formulae is based on the mathematical expression, the resemblance between the prescribed SST cooling and the observed one during the period of the numerical experiment is not complete nor satisfactory. The agreements between the prescribed and the observed SST even over the swath of the typhoon passage differ from case to case. Numerical experiments are undertaken with and without prescribing the SST cooling. The results with the SST cooling do not show clear evidence in improving the track prediction compared to those of the without-experiments. SST cooling in the model shows its swath along the incomplete simulated track so that the magnitude and the distribution of the sea surface cooling does not resemble completely with the observed one. However, we have observed a little improvement in the intensity prediction in terms of the central pressure of the tropical cyclone in some cases. In case where the model without the SST treatment is not able to yield a correct prediction of the filling of the tropical cyclone especially in the decaying stage, the pulling effect given by the SST cooling alleviates the over-deepening of the model so that the central pressure approaches toward the observed value. However, the opposite case when the SST treatment makes the prediction worse may also be possible. In general when the sea surface temperature is reduced, the amount of the sensible and the latent heat from the ocean surface become also reduced, which results in the weakening of the storms comparing to the constant SST case. It turns out to be the case also in our experiments. The weakening is realized in the central pressure, maximum wind, horizontal temperature gradient, etc.

• Korean Journal of Remote Sensing
• /
• v.38 no.3
• /
• pp.265-282
• /
• 2022

Sea surface temperature (SST) is a factor that greatly influences ocean circulation and ecosystems in the Earth system. As global warming causes changes in the SST near the Korean Peninsula, abnormal water temperature phenomena (high water temperature, low water temperature) occurs, causing continuous damage to the marine ecosystem and the fishery industry. Therefore, this study proposes a methodology to predict the SST near the Korean Peninsula and prevent damage by predicting abnormal water temperature phenomena. The study area was set near the Korean Peninsula, and ERA5 data from the European Center for Medium-Range Weather Forecasts (ECMWF) was used to utilize SST data at the same time period. As a research method, Long Short-Term Memory (LSTM) algorithm specialized for time series data prediction among deep learning models was used in consideration of the time series characteristics of SST data. The prediction model predicts the SST near the Korean Peninsula after 1- to 7-days and predicts the high water temperature or low water temperature phenomenon. To evaluate the accuracy of SST prediction, Coefficient of determination (R²), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) indicators were used. The summer (JAS) 1-day prediction result of the prediction model, R²=0.996, RMSE=0.119℃, MAPE=0.352% and the winter (JFM) 1-day prediction result is R²=0.999, RMSE=0.063℃, MAPE=0.646%. Using the predicted SST, the accuracy of abnormal sea surface temperature prediction was evaluated with an F1 Score (F1 Score=0.98 for high water temperature prediction in summer (2021/08/05), F1 Score=1.0 for low water temperature prediction in winter (2021/02/19)). As the prediction period increased, the prediction model showed a tendency to underestimate the SST, which also reduced the accuracy of the abnormal water temperature prediction. Therefore, it is judged that it is necessary to analyze the cause of underestimation of the predictive model in the future and study to improve the prediction accuracy.

• Ocean and Polar Research
• /
• v.43 no.2
• /
• pp.53-63
• /
• 2021

Changes in a marine environment have a broad socioeconomic implication on fisheries and their relevant industries so that there has been a growing demand for the medium-range (months to years) prediction of the marine environment Using a medium-range ocean prediction model (Ocean Mid-range prediction System, OMIDAS) for the northwest Pacific, this study attempted to assess seasonal difference in the mid-range predictability of the sea surface temperature (SST), focusing on the Korea seas characterized as a complex marine system. A three-month re-forecast experiment was conducted for each of the four seasons in 2016 starting from January, forced with Climate Forecast System version 2 (CFSv2) forecast data. The assessment using relative root-mean-square-error was taken for the last month SST of each experiment. Compared to the CFSv2, the OMIDAS revealed a better prediction skill for the Korea seas SST, particularly in the Yellow sea mainly due to a more realistic representation of the topography and current systems. Seasonally, the OMIDAS showed better predictability in the warm seasons (spring and summer) than in the cold seasons (fall and winter), suggesting seasonal dependency in predictability of the Korea seas. In addition, the mid-range predictability for the Korea seas significantly varies depending on regions: the predictability was higher in the East Sea than in the Yellow Sea. The improvement in the seasonal predictability for the Korea seas by OMIDAS highlights the importance of a regional ocean modeling system for a medium-range marine prediction.

• Journal of the Korean Geographical Society
• /
• v.28 no.2
• /
• pp.112-122
• /
• 1993

Since more than 50${\%}$ of annual precipitation in Korea falls during Changma, the rainy season of early summer, and Late Changma, the rainy season of late summer, forcasting the onset days Changmas, and the amount related rainfalls would be necessary not only for agriculture but also for flood-control. In this study the authors attempted to build a prediction model for the forecast of the onset date of Changmas. The onset data of each Changma was derived out of daily rainfall data of 47 stations for 30 years(1961~1990) and weather maps over East Asia. Each station represent any of the 47 districts of local forecast under the Korea Meteorological Administration. The average onset dates of Changma during the period was from 21 through 26 June. The dates show a tendency to be delayed in El Ni${\~{n}}o years while they come earlier than the average in La Nina years. In 1982, the year of El Ni${\~{n}}o, the date was 9 Julu, two weeks late compared with the average. The relation of sea surface temperature(SST) over Pacific and Northern hemispheric 500mb height to the Changma onset dates was analyzed for the prediction model by polynomial regression. The onset date of Changma over Korea was correlated with SST in May(SST${_(5)}{^\circ}$C) of the district (8${^\circ}$~12${^\circ}S, 136${^\circ}~148${^\circ}W)of equatirial middle Pacific and the 500mb height in March (MB${_(3)}$"\;"m)over the district of the notrhern Hudson Bay. The relation between this two elements can be expressed by the regression: Onset=5.888SST${_5}"\;"+"\;"0.047MB${_(3)}$"\;"-251.241. This equation explains 77${\%}$ of variances at the 0.01${\%}$ singificance level. The onset dates of Late Changma come in accordance with the degeneration of the Subtro-pical High over northern Pacific. They were 18 August in average for the period showing positive correlation(r=0.71) with SST in May(SST)${_(i5)}{^\circ}$C) over district of IndiaN Ocean near west coast of Australia (24${^\circ}$~32${^\circ}$S, 104${^\circ}$~112${^\circ}$E), but negativ e with SST in May(SST${_(p5)}{^\circ}$ over district (12${^\circ}$~20${^\circ}$S,"\;"136${^\circ}$~148${^\circ}$W)of equatorial mid Pacific (r=-0.70) and with the 500mb height over district of northwestern Siberia (r=-0.62). The prediction model for Late Changma can be expressed by the regression: Onset=706.314-0.080 MB-3.972SST${_(p5)}+3.896 SST${_(i5)}, which explains 64${\%}$ of variances at the 0.01${\%}$ singificance level.

• Journal of Environmental Science International
• /
• v.23 no.8
• /
• pp.1481-1493
• /
• 2014

In order to understand the relation between the distribution of sea surface temperature and heavy snowfall over western coast of the Korean peninsula, several numerical assessments were carried out. Numerical model used in this study is WRF, and sea surface temperature data were FNL(National Center for Environment Prediction-Final operational global analysis), RTG(Real Time Global analysis), and OSTIA(Operational Sea Surface Temperature and Sea Ice Analysis). There were produced on the basis of remote sensing data, such as a variety of satellite and in situ observation. The analysis focused on the heavy snowfall over Honam districts for 2 days from 29 December 2010. In comparison with RTG and OSTIA SST data, sensible and latent heat fluexes estimated by numerical simulation with FNL data were higher than those with RTG and OSTIA SST data, due to higher sea surface temperature of FNL. General distribution of RTG and OSTIA SST showed similar, however, fine spatial differences appear in near western coast of the peninsula. Estimated snow fall amount with OSTIA SST was occurred far from the western coast because of higher SST over sea far from coast than that near coast. On the other hand, snowfall amount near coast is larger than that over distance sea in simulation with RTG SST. The difference of snowfall amount between numerical assessment with RTG and OSTIA is induced from the fine difference of SST spatial distributions over the Yellow sea. So, the prediction accuracy of snowfall amount is strongly associated with the SST distribution not only over near coast but also over far from the western coast of the Korean peninsula.

• Proceedings of the Korea Air Pollution Research Association Conference
• /
• 1999.04a
• /
• pp.187-188
• /
• 1999