• Korean Journal of Agricultural and Forest Meteorology
• /
• v.18 no.4
• /
• pp.208-220
• /
• 2016

Automatic Mountain Meteorology Observation System (AMOS) is an important ingredient for several climatological and forest disaster prediction studies. In this study, we select the optimal sites for AMOS in the mountain areas of Honam and Jeju in order to prevent forest disasters such as forest fires and landslides. So, this study used spatial dataset such as national forest map, forest roads, hiking trails and 30m DEM(Digital Elevation Model) as well as forest risk map(forest fire and landslide), national AWS information to extract optimal site selection of AMOS. Technical methods for optimal site selection of the AMOS was the firstly used multifractal model, IDW interpolation, spatial redundancy for 2.5km AWS buffering analysis, and 200m buffering analysis by using ArcGIS. Secondly, optimal sites selected by spatial analysis were estimated site accessibility, observatory environment of solar power and wireless communication through field survey. The threshold score for the final selection of the sites have to be higher than 70 points in the field assessment. In the result, a total of 159 polygons in national forest map were extracted by the spatial analysis and a total of 64 secondary candidate sites were selected for the ridge and the top of the area using Google Earth. Finally, a total of 26 optimal sites were selected by quantitative assessment based on field survey. Our selection criteria will serve for the establishment of the AMOS network for the best observations of weather conditions in the national forests. The effective observation network may enhance the mountain weather observations, which leads to accurate prediction of forest disasters.

• Journal of the Korean Geographical Society
• /
• v.41 no.2 s.113
• /
• pp.168-187
• /
• 2006

Windy meteorological conditions and dried fire fuels due to higher atmospheric instability and dryness in the lower troposphere can exacerbate fire controls and result in more losses of forest resources and residential properties due to enhanced large wildland fires. Long-term (1979-2005) climatology of the Haines Index reconstructed in this study reveals that spatial patterns and intra-annual variability of the atmospheric instability and dryness in the lower troposphere affect the frequency of wildland fire incidences over the Korean Peninsula. Exponential regression models verify that daily high Haines Index and its monthly frequency has statistically significant correlations with the frequency of the wildland fire occurrences during the fire season (December-April) in South Korea. According to the climatic maps of the Haines Index created by the Geographic Information System (GIS) using the Digital Elevation Model (DEM), the lowlands below 500m from the mean sea level in the northwestern regions of the Korean Peninsula demonstrates the high frequency of the Haines Index equal to or greater than five in April and May. The annual frequency of the high Haines Index represents an increasing trend across the Korean Peninsula since the mid-1990s, particularly in Gyeongsangbuk-do and along the eastern coastal areas. The composite of synoptic weather maps at 500hPa for extreme events, in which the high Haines Index lasted for several days consecutively, illustrates that the cold low pressure system developed around the Sea of Okhotsk in the extreme event period enhances the pressure gradient and westerly wind speed over the Korean Peninsula. These results demonstrate the need for further consideration of the spatial-temporal characteristics of vertical atmospheric components, such as atmospheric instability and dryness, in the current Korean fire prediction system.

• Korean Journal of Remote Sensing
• /
• v.38 no.5_2
• /
• pp.661-667
• /
• 2022

As remote sensing and GIS have been considered to be essential technologies for disasters information production, researches on developing methods for analyzing spatial data, and developing new technologies for such purposes, have been actively conducted. Especially, it is assumed that the use of remote sensing and GIS for disaster management will continue to develop thanks to the launch of recent satellite constellations, the use of various remote sensing platforms, the improvement of acquired data processing and storage capacity, and the advancement of artificial intelligence technology. This spatial issue presents 10 research papers regarding ship detection, building information extraction, ocean environment monitoring, flood monitoring, forest fire detection, and decision making using remote sensing and GIS technologies, which can be applied at the disaster prediction, monitoring and response stages. It is anticipated that the papers published in this special issue could be a valuable reference for developing technologies for disaster management and academic advancement of related fields.

• Journal of Cadastre & Land InformatiX
• /
• v.47 no.2
• /
• pp.107-120
• /
• 2017

There are many wildfire risk indices worldwide, but objective comparisons between such various wildfire risk indices and surface dryness indices have not been conducted for the wildfire cases in Korea. This paper describes a sensitivity analysis on the wildfire risk indices and surface dryness indices for Korea using LDAPS(Local Analysis and Prediction System) meteorological dataset on a 1.5-km grid and MODIS(Moderate-resolution Imaging Spectroradiometer) satellite images on a 1-km grid. We analyzed the meteorology-based wildfire risk indices such as the Australian FFDI(forest fire danger index), the Canadian FFMC(fine fuel moisture code), the American HI(Haines index), and the academically presented MNI(modified Nesterov index). Also we examined the satellite-based surface dryness indices such as NDDI(normalized difference drought index) and TVDI(temperature vegetation dryness index). As a result of the comparisons between the six indices regarding 120 wildfire cases with the area damaged over 1ha during the period between January 2013 and May 2017, we found that the FFDI and FFMC showed a good predictability for most wildfire cases but the MNI and TVDI were not suitable for Korea. The NDDI can be used as a proxy parameter for wildfire risk because its average CDF(cumulative distribution function) scores were stably high irrespective of fire size. The indices tested in this paper should be carefully chosen and used in an integrated way so that they can contribute to wildfire forecasting in Korea.

• Korean Chemical Engineering Research
• /
• v.57 no.2
• /
• pp.274-282
• /
• 2019

In recent years, big data analysis has been expanded to include automatic control through reinforcement learning as well as prediction through modeling. Research on the utilization of image data is actively carried out in various industrial fields such as chemical, manufacturing, agriculture, and bio-industry. In this paper, we applied NASNet, which is an AutoML reinforced learning algorithm, to DeepU-Net neural network that modified U-Net to improve image semantic segmentation performance. We used BRATS2015 MRI data for performance verification. Simulation results show that DeepU-Net has more performance than the U-Net neural network. In order to improve the image segmentation performance, remove dropouts that are typically applied to neural networks, when the number of kernels and filters obtained through reinforcement learning in DeepU-Net was selected as a hyperparameter of neural network. The results show that the training accuracy is 0.5% and the verification accuracy is 0.3% better than DeepU-Net. The results of this study can be applied to various fields such as MRI brain imaging diagnosis, thermal imaging camera abnormality diagnosis, Nondestructive inspection diagnosis, chemical leakage monitoring, and monitoring forest fire through CCTV.