According to the National Oceanic and Atmospheric Administration (NOAA), the Earth's average temperature is increasing annually as global warming continues, and the temperature of the Arctic region is increasing twice as fast as those of other regions (NOAA, 2018). With Arctic glaciers melting owing to global warming, the importance of developing marine resources buried in the Arctic is emerging. According to the United States Geological Survey, undiscovered recoverable resources in the Arctic region include approximately 90 billion barrels (Bbbl) of oil, 1,669 trillion cubic feet (Tcf) of gas, and 44 Bbbl of natural gas liquid (Bird et al., 2008). It is estimated that 13% of the world's undiscovered oil volume and 30% of the world's undiscovered gas volume are buried in the Arctic region. In general, marine resources in the deep sea are developed by offshore structures, and it is important to collect and analyze the metocean data of sea areas where marine resources are buried for the operation and survival of offshore structures. Metocean data include wave, wind, current, water depth, tide, and soil conditions. To reflect metocean data in the design of offshore structures, extreme value analysis should be conducted (API, 2005). Metocean data can be collected from measurement data of actual sea areas that use buoys or ships, hindcast data that estimate the desired data of the past through numerical methods based on past metocean data, and satellite data that use satellites. According to DNVGL (2015), for the design of offshore structures, the significant wave height (Hs) and spectral peak wave period (Tp) of 100-year return values should be used for waves, whereas the 100-year return value of the one-hour average wind speed at 10 m above the mean sea level (MSL) should be used for wind speed. For ocean currents, the 10-year return value of the current speed should be reflected in the design. To derive return values through statistical methods, methods for selecting the collected data must be determined. Such methods can be primarily categorized into global and event models. Global models include the initial distribution and total sample methods that use the entire data statistically, and the event models include the peak over threshold method, which uses data above a certain threshold, and the annual maxim method, which uses only the annual maximum values (DNV, 2014). Extreme value distributions (EVDs) used for extreme value analyses include the Gumbel, Frechet, Weibull, and log-normal distributions. Methods for estimating the parameters of EVDs include

To evaluate the time-domain positioning performance of arctic marine structures, it is necessary to generate an ice load appropriate for the current position and heading of the structure. The position and orientation angle of a floating body continuously change with time. Therefore, an ice load is required for any attitude in the time-domain simulation. In this study, we present a fundamental technique for analyzing ice loads in the frequency domain based on data measured at various angles in the ice-water tank experiment. We perform spectral analysis instead of general FFT to analyze the ice load, which has the characteristics of a random signal. To generate the necessary ice load in the time domain, we must first interpolate the measured data in the frequency domain. Using the Blackman-Tukey method, we estimate the spectrum for the measured data, then process the data to generate the training set required for machine learning. Based on the results, we perform regression analysis by applying four representative techniques, including linear regression, random forest, or neural network, and compare the results with MSE. The deep neural network method performed best, but we provide further discussion for each model.