• Journal of the Korean earth science society
• /
• v.43 no.4
• /
• pp.539-556
• /
• 2022

In the coastal areas of Jeju Island, composed of volcanic rocks, saltwater intrusion occurs due to excessive pumping and geological characteristics. Groundwater level and electrical conductivity (EC) in multi-depth monitoring wells in coastal areas were characterized from 2005 to 2019. During the period of the lowest monthly precipitation, from November 2017 until February 2018, groundwater level decreased by 0.32-0.91 m. During the period of the highest monthly precipitation, from September 2019 until October 2019, groundwater level increased by 0.46-2.95 m. Groundwater level fluctuation between the dry and wet seasons ranged from 0.79 to 3.73 m (average 1.82 m) in the eastern area, from 0.47 to 6.57 m (average 2.55 m) in the western area, from 0.77 to 8.59 m (average 3.53 m) in the southern area, and from 1.06 to 12.36 m (average 5.92 m) in the northern area. In 2013, when the area experienced decreased annual precipitation, at some monitoring wells in the western area, the groundwater level decreased due to excessive groundwater pumping and saltwater intrusion. Based on EC values of 10,000 ㎲/cm or more, saltwater intrusion from the coastline was 10.2 km in the eastern area, 4.1 km in the western area, 5.8 km in the southern area, and 5.7 km in the northern area. Autocorrelation analysis of groundwater level revealed that the arithmetic mean of delay time was 0.43 months in the eastern area, 0.87 months in the northern area, 10.93 months in the southern area, and 17.02 months in the western area. Although a few monitoring wells were strongly influenced by nearby pumping wells, the cross-correlation function of the groundwater level was the highest with precipitation in most wells. The seasonal autoregressive integrated moving average model indicated that the groundwater level will decrease in most wells in the western area and decrease or increase in different wells in the eastern area.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.20 no.4
• /
• pp.382-388
• /
• 2014

This study selected five observational stations in the normal direction of Jinu-do(island) shoreline and observed water temperature, electrical conductivity and pressure from March, 2012 to January, 2013(about 11 months) and attempted to see the variation characteristics of ground watertable. This study wants to know : 1) External environment force factors(tide, climate, wave etc.) affecting ground watertable variation through time series and correlation analysis. 2) Spatial variations of ground watertable and electrical conductivity change by storm event. First, we found that the station at the intertidal zone was strongly affected by wave and tide level and the stations at sand dune and vegetation zone was affected by precipitation and tide level through time series data and correlation analysis. Second, during the storm event, we found that ground watertable and electrical conductivity are stabilized at the start line of sand dune and vegetation zone and transition zone between freshwater layer and seawater layer exists in the experiment area and is about 50~70 m from coastline of the south side of Jinu-do(island).

• Journal of Environmental Science International
• /
• v.14 no.11
• /
• pp.1015-1026
• /
• 2005

The southeastern coastal area of the Korean peninsula has a complex terrain including an irregular coastline and moderately high mountains. This implies that mesoscale circulations such as mountain-valley breeze and land-sea breeze can play an important role in wind field and ocean forcing. In this study, to improve the accuracy of complex coastal rind field(surface wind and sea surface wind), we carried out the sensitivity experiments based on PBL schemes in PSU/NCAR Mesoscale Model (MM5), which is being used in the operational system at Korea Meteorological Administration. Four widely used PBL parameterization schemes in sensitivity experiments were chosen: Medium-Range Forecast (MRF), High-resolution Blackadar, Eta, and Gayno-Seaman scheme. Thereafter, case(2004. 8. 26 - 8. 27) of weak-gradient flows was simulated, and the time series and the vertical profiles of the simulated wind speed and wind direction were compared with those of hourly surface observations (AWS, BUOY) and QuikSCAT data. In the simulated results, the strength of rind speed of all schemes was overestimated in complex coastal regions, while that of about four different schemes was underestimated in islands and over the sea. Sea surface wind using the Eta scheme showed the highest wind speed over the sea and its distribution was similar to the observational data. Horizontal distribution of the simulated wind direction was very similar to that of real observational data in case of all schemes. Simulated and observed vertical distribution of wind field was also similar under boundary layer(about 1 km), however the simulated wind speed was underestimated in upper layer.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.8
• /
• pp.311-318
• /
• 2021

Research has been ongoing to detect ships from offshore photographs for a variety of reasons, including maritime security, identifying international trends, and social scientific research. Due to the development of artificial intelligence, R-CNN models for object detection in photographs and images have emerged, and the performance of object detection has risen dramatically. Ship detection in offshore photographs using the R-CNN model has also begun to apply to satellite photography. However, satellite images project large areas, so various objects such as vehicles, landforms, and buildings are sometimes recognized as ships. In this paper, we propose a novel methodology to improve the performance of ship detection in satellite photographs using R-CNN series models. We separate land and sea via marker-based watershed algorithm and perform morphology operations to specify RoI one more time, then detect vessels using R-CNN family models on specific RoI to reduce typology. Using this method, we could reduce the misdetection rate by 80% compared to using only the Fast R-CNN.