• 대기
• /
• 제14권1호
• /
• pp.24-43
• /
• 2004

This paper describes the detailed characteristics of heavy rainfall events occurred in Chungcheong province on 15 and 16 April and from 6 to 8 August 2002 based on the analysis of raingauge rainfall rate and radar reflectivity from the METRI's X-band Weather Radar located in Cheongju. A synoptic analysis of the case is carried out, first, and then the analysis is devoted to seeing how the radar observes the case and how much information we obtain. The highly resolved radar reflectivity of horizontal and vertical resolutions of 1 km and 500 m, respectively shows a three-dimensional structure of the precipitating system, in a similar sequence with the ground rainfall rate. The radar echo classification algorithm for convective/stratiform cloud is applied. In the convectively-classified area, the radar reflectivity pattern shows a fair agreement with that of the surface rainfall rate. This kind of classification using radar reflectivity is considered to be useful for the precipitation forecasting. Another noteworthy aspect of the case includes the effect of topography on the precipitating system, following the analysis of the surface rainfall rate, topography, and precipitating system. The results from this case study offer a unique opportunity of the usefulness of weather radar for better understanding of structural and variable characteristics of flash flood-producing heavy rainfall events, in particular for their improved forecasting.

• 한국농림기상학회지
• /
• 제19권3호
• /
• pp.86-92
• /
• 2017

본 연구에서는 기상청 예보자료를 이용할 것을 전제로 서리가 발생하는 날과 서리가 발생하지 않는 날을 구분하는 모형을 구축하였다. 서리가 발생한 날과 서리가 발생하지 않은 날을 구분할 수 있는 기상인자로서 최저기온, 평균풍속, 평균상대습도, 평균이슬점온도로 나타났다. 기상인자별로 두 날을 비교한 결과 서리가 발생한 날이 서리가 발생하지 않은 날에 비해 최저기온, 이슬점온도, 평균풍속은 낮게 나타났고 상대습도는 높게 나타났다. 이러한 기상인자로 인공신경망, 랜덤포레스트, 서포트벡터 머신의 기계학습법을 이용한 모형을 연구한 결과 70%이상의 정확도를 나타내어 활용 가능성이 있을 것으로 판단된다.

• 한국지능시스템학회논문지
• /
• 제24권6호
• /
• pp.665-670
• /
• 2014

SVM은 학습 데이터를 두 개의 집단으로 분리시키는 최적의 초평면을 찾는 이진 분류기로서 우수한 성능 때문에 다양한 분야에서 귀납 추론, 이진 분류, 예측 등을 목적으로 사용되는 알고리즘이다. 또한 대표적인 블랙박스 모델 중 하나이기 때문에 학습 후 생성되는 SVM의 해석에 대한 연구도 활발히 진행되고 있다. 본 논문에서는 SVM 알고리즘을 이용하여 기상 레이더의 데이터 내에 비교적 높은 빈도로 발생하여 기상 예보의 정확도를 감소시키는 비강수에코 중 하나인 선에코를 자동으로 탐지하는 방법에 대한 연구를 수행하였다. 학습 데이터로는 평균 반사도, 크기, 발생 형태, 중심 고도 등과 같은 특성을 활용하였는데, 이는 기상 레이더 데이터에 저장된 다양한 데이터 중 반사도 값을 선택한 후 클러스터링 기법을 통해 추출한 것이다. 이와 같이 학습된 SVM 분류기를 실제 사례를 바탕으로 하여 검증하였으며, Decision Tree 알고리즘을 적용하여 생성한 분류기의 해석을 수행하였다.

• 한국방재학회 논문집
• /
• 제9권5호
• /
• pp.103-113
• /
• 2009

본 연구에서는 관측소의 지리적 위치 및 강수특성(월별, 계절별, 연평균)을 이용하여 강원도의 강수지역을 구분하였다. 강수지역 구분은 기상관측소 66개소(기상관서: 11개소, 자동기상시스템(AWS): 55개소)의 자료를 이용하였으며, 통계적 방법 중 군집 기법인 K-means 방법을 적용하였다. 지역구분 결과, 강수지역은 5개 지역(영동지방 1개 지역 및 영서지방 4개 지역)으로 구분하였다. 계절별 평균강수량은 봄에는 강원도 전체에 유사하게 발생하였으며, 여름에는 영서지방이 높게 나타났으며, 가을과 겨울에는 영동지방이 높게 발생하였다. 연평균 강수량 및 여름철 강수량의 공간분석 결과 강원도 중 일부 지역(미시령 및 대관령일원)은 산악형 강수 특성을 나타냈으나 전반적인 현상은 아닌 것으로 판단되었다. 그러나 보다 정확한 분석을 위해서는 관측소의 고도별 분포가 미흡한 것으로 나타난 관측소의 보완 및 AWS의 자료 확충이 필요할 것으로 판단된다.

• 융합정보논문지
• /
• 제11권7호
• /
• pp.14-20
• /
• 2021

겨울철 도로 결빙으로 인한 사고는 대부분 큰 사고로 이어진다. 이는 운전자가 도로의 결빙을 사전에 자각하기 어렵기 때문이다. 본 연구에서는 AutoML과 CNN의 앙상블 모델을 이용하여 도로교통 이머징 리스크를 정확하게 탐지하는 방법을 연구한다. 비정형 데이터인 이미지를 이용한 CNN 이미지 특징 추출 기반 도로교통 이머징 리스크 분류 모델과 정형 데이터인 기상 데이터를 이용한 AutoML 기반 도로교통 이머징 리스크 분류 모델을 각각 학습시킨다. 그 후 모델들에서 도출된 확률값을 입력하여 CNN 기반 분류 모델을 보완하도록 앙상블 모델을 설계한다. 이를 통해 도로교통 이머징 리스크 분류 성능을 향상하고 더 정확하고 빠르게 운전자에게 경고하여 안전한 주행이 가능하도록 한다.

• 한국환경과학회지
• /
• 제17권6호
• /
• pp.591-599
• /
• 2008

In order to support the development of wind farms in Jejudo, a wind resource database for Jejudo has been established using a meteor-statistical analysis of KIER(Korea Institute of Energy Research) met-mast measurements and KMA(Korea Meteorological Administration) weather data. The analysis included wind statistics, tower shading, an exposure category classification using satellite images, the effect of atmospheric stability on the wind profile exponent, and a correlation matrix of wind speed to gain an understanding of the meteorological correlation between long-term weather observation stations and short-term met-mast measurements. The wind resource database for Jejudo, is to be provided as an add-on to Google $Earth^{TM}$, which is expected to be utilized as a guideline for the selection of an appropriate reference site for long-term correction in the next wind farm development project.

• 천문학회보
• /
• 제38권2호
• /
• pp.33-33
• /
• 2013

We are developing empirical space weather (solar flare, solar proton event, and geomagnetic storm) forecast models based on solar data. In this talk we will review our main results and recent progress. First, we have examined solar flare (R) occurrence probability depending on sunspot McIntosh classification, its area, and its area change. We find that sunspot area and its increase (a proxy of flux emergence) greatly enhance solar flare occurrence rates for several sunspot classes. Second, a solar proton event (S) forecast model depending on flare parameters (flare strength, duration, and longitude) as well as CME parameters (speed and angular width) has been developed. We find that solar proton event probability strongly depends on these parameters and CME speed is well correlated with solar proton flux for disk events. Third, we have developed an empirical storm (G) forecast model to predict probability and strength of a storm using halo CME - Dst storm data. For this we use storm probability maps depending on CME parameters such as speed, location, and earthward direction. We are also looking for geoeffective CME parameters such as cone model parameters and magnetic field orientation. We find that all superstorms (less than -200 nT) occurred in the western hemisphere with southward field orientations. We have a plan to set up a storm forecast method with a three-stage approach, which will make a prediction within four hours after the solar coronagraph data become available. We expect that this study will enable us to forecast the onset and strength of a geomagnetic storm a few days in advance using only CME parameters and the WSA-ENLIL model. Finally, we discuss several ongoing works for space weather applications.

• 전기학회논문지
• /
• 제64권9호
• /
• pp.1337-1346
• /
• 2015

In this paper, pattern classifier is designed to classify precipitation and non-precipitation events from weather radar data. The proposed classifier is based on Fuzzy Neural Network(FNN) and consists of three FNNs which operate in parallel. In the proposed network, the connection weights of the consequent part of fuzzy rules are expressed as two polynomial types such as constant or linear polynomial function, and their coefficients are learned by using Least Square Estimation(LSE). In addition, parametric as well as structural factors of the proposed classifier are optimized through Differential Evolution(DE) algorithm. After event classification between precipitation and non-precipitation echo, non-precipitation event is to get rid of all echo, while precipitation event including non-precipitation echo is to get rid of non-precipitation echo by classifier that is also based on Fuzzy Neural Network. Weather radar data obtained from meteorological office is to analysis and discuss performance of the proposed event and echo patter classifier, result of echo pattern classifier compare to QC(Quality Control) data obtained from meteorological office.

• 천문학회보
• /
• 제36권2호
• /
• pp.90.1-90.1
• /
• 2011

We are developing empirical space weather (geomagnetic storms, solar proton events, and solar flares) forecast models based on solar information. These models have been set up with the concept of probabilistic forecast using historical events. Major findings can be summarized as follows. First, we present a concept of storm probability map depending on CME parameters (speed and location). Second, we suggested a new geoeffective CME parameter, earthward direction parameter, directly observable from coronagraph observations, and demonstrated its importance in terms of the forecast of geomagnetic storms. Third, the importance of solar magnetic field orientation for storm occurrence was examined. Fourth, the relationship among coronal hole-CIR-storm relationship has been investigated, Fifth, the CIR forecast based on coronal hole information is possible but the storm forecast is challenging. Sixth, a new solar proton event (flux, strength, and rise time) forecast method depending on flare parameters (flare strength, duration, and longitude) as well as CME parameter (speed, angular width, and longitude) has been suggested. Seventh, we are examining the rates and probability of solar flares depending on sunspot McIntosh classification and its area change (as a proxy of flux change). Our results show that flux emergence greatly enhances the flare probability, about two times for flare productive sunspot regions.

• 제어로봇시스템학회논문지
• /
• 제19권10호
• /
• pp.901-906
• /
• 2013

Chaff is a kind of matter spreading atmosphere with the purpose of preventing aircraft from detecting by radar. The chaff is commonly composed of small aluminum pieces, metallized glass fiber, or other lightweight strips which consists of reflecting materials. The chaff usually appears on the radar images as narrow bands shape of highly reflective echoes. And the chaff echo has similar characteristics to precipitation echo, and it interrupts weather forecasting process and makes forecasting accuracy low. In this paper, the chaff echo recognizing and removing method is suggested using Bayesian network. After converting coordinates from spherical to Cartesian in UF (Universal Format) radar data file, the characteristics of echoes are extracted by spatial and temporal clustering. And using the data, as a result of spatial and temporal clustering, a classification process for analyzing is performed. Finally, the inference system using Bayesian network is applied. As a result of experiments with actual radar data in real chaff echo appearing case, it is confirmed that Bayesian network can distinguish between chaff echo and non-chaff echo.