NIFS announced that it is going to provide water temperature prediction service in full swings from May of this year with the algorithms that it developed by combining artificial intelligence and sea surface temperature data obtained from satellite observations for improving the accuracy of water temperature prediction technology.

Earlier, NIFS advanced the technology for the weekly oceanic condition data service, considering the result of public opinion survey conducted online (e-People) last year.

With the newly developed water temperature prediction algorithms, it is possible to predict water temperature in wide sea areas with high resolution(1㎞) going beyond the existing way of qualitative water temperature prediction based on statistical indicators. 

The algorithms, where a ConvLSTM* was applied, has continued to improve the prediction accuracy through AI learning reflecting region-specific characteristics including East Sea, West Sea, South Sea, East China Sea, etc. In practice, the prediction accuracy has been improved from 90% in 2024 to 94% up to the present time through the algorithm improvement.

*A Convolutional LSTM (ConvLSTM)- a deep learning method for time series prediction- is a type of neural network that combines Convolution Neural Network(CNN) and Long-Short Term Memory(LSTM). 

Since November of last year, NIFS has been provided the pilot service for the prediction in Korean waters. Building on that, NIFS is going to expand the prediction areas to Northwest Pacific from this May*.

*You can find the weekly oceanic condition data on https://www.nifs.go.kr.

The water temperature prediction algorithms this time are expected to enable preemptive response to the fisheries disasters caused by abnormal ocean conditions such as high water temperature, low water temperature, cold water mass, etc. as well as to expand scientific bases for reducing fisheries damages. In addition, the related technologies are also expected to contribute to upgrading the overall level of ocean science and technology.