Earticle

As the value of Big Data was recognized as important, the government, public institutions and private companies began to pay attention to Big Data. Unlike the past, there are diverse data sources, and with the emergence of various planning and analysis techniques based on the convergence of these data, Big Data is sure to become the foundation for the creation of new advanced information and advanced decision-making. This study is to find an alternative to the atmospheric fine dust concentration that is not improved despite various measures. The purpose of this study is to exploit the public data of the Seoul Metropolitan City in order to reduce the air pollution concentration in Seoul. The study was conducted through open source R for Seoul public data analysis. As a result, policy alternatives to activate natural gas vehicles were derived.

미세먼지에 대한 연구는 실시간으로 발전하고 있으며, 예측 모델의 정확도를 향상시키기 위해 다양한 방법이 연구되고 있다. 또한 미세먼지의 정확한 원인과 영향을 파악하기 위해 이러한 다양한 요소들을 고려하는 연구들이 활발히 이루어지고 있다. 본 논문에서는 PM_2.5와 상관성이 있는 데이터를 계절을 기준으로 구분하여 학습하는 예측 모델과 특정 농도를 기준으로 저농도와 고농도를 구분하여 학습하는 모델을 통해 예측 성능의 비교 및 분석을 진행하였다. 기상데이터와 대기오염 물질 데이터를 사용하였으며 PM_2.5와 상관관계를 확인하여 학습 및 평가를 위한 데이터를 구성하였다. 계절별 예측 모델과 농도별 예측 모델은 LSTM으로 설계하였으며, 세부 파라미터는 하이퍼 파라미터 탐색을 통해 적용하였다. 예측 모델의 성능 평가는 정확도, RMSE, MAPE, 저농도와 고농도 구간에서의 정확도 그리고 AQI를 기준으로 4개의 범위에 대한 정확도로 진행하였다. 성능 평가 결과, 농도별 학습을 진행한 예측 모델이 AQI 기준 "나쁨" 구간의 정확도에서 91.02%의 정확도를 보였으며, 계절별 학습을 진행한 예측 모델보다 전반적으로 좋은 성능을 보였다.

Research on particulate matter is advancing in real-time, and various methods are being studied to improve the accuracy of prediction models. Furthermore, studies that take into account various factors to understand the precise causes and impacts of particulate matter are actively being pursued. This paper trains an LSTM model using seasonal data and another LSTM model using concentration-based data. It compares and analyzes the PM_2.5 prediction performance of the two models. To train the model, weather data and air pollutant data were collected. The collected data was then used to confirm the correlation with PM_2.5. Based on the results of the correlation analysis, the data was structured for training and evaluation. The seasonal prediction model and the concentration-specific prediction model were designed using the LSTM algorithm. The performance of the prediction model was evaluated using accuracy, RMSE, and MAPE. As a result of the performance evaluation, the prediction model learned by concentration had an accuracy of 91.02% in the "bad" range of AQI. And overall, it performed better than the prediction model trained by season.

Effluent concentration estimation equations for wetland system were developed throught statistical analysis of treatment wetland experimental data. Existin g empirical equations were reviewed for thier accuracy with experimental data, and compared with the estimatin equations. About 70 experimental data sets were used for multiple regression, and variables include influent concentration, hydraulic loading rate, average daily air temperature , and plant coverage. The estimatin equations developed for BOD5 , SS ,T-P, and T-N predicted effluent concentrations moderately well, and coefficient fo determination ($R^2$) for them was 0.74 , 0.60, 0.59 and 0.58 respectively. The equations obtained from same data but excluding plant coverage showed relatively lower $R^2$ than the former case, and it was 0.66, 0.52, 0.41 and 0.57 respectively. The EPA, WPCF , and Kadlec and Knight equations worked poorly and $R^2$ for them was significantly lower than the estimation equation developed in the study. The reason might be that the existing equations were oversimplified that they did ot include important parameters such as air temperature and plant coverage. Therefore, developing reasonable estimation equations from experiment under realistic condition is highly recommended rather than using exiting estimation equations.

Water quality samples were taken for every two hours whenever runoff occurred by rainfall to investigate concentration variations of T-N, T-P and SS during runoff process from a paddy field. The difference between the highest concentration in a runoff event and flow weighted mean concentration for T-N, T-P, SS placed between 3.07∼40.16%, 11.44∼60.80%, and 15.11∼64.5%, respectively. The difference between the lowest concentration in event and event mean concentration for T-N, T-P, SS ranged between -7.24∼-31.84%, -11.59∼-47.86%, and -5.21∼-36.20%, respectively. The relationship between runoff and mass load was derived for each storm event using observed data. The relationship between runoff and mass load showed linear relationship regardless of water quality constituents and rainfall amount. The results suggested that relationship between T-N and T-P loads and runoff should be prepared separately in considering of fertilization effect and seasonal conditions. The relationship between SS and runoff should be made to reflect seasonal conditions and tillage effect.

주요한 수질지표 중의 하나인 생물화학적 산소요구량(BOD) 농도는 호소와 하천에서 생태학적 측면에서 관측항목으로 취급하고 있다. 본 연구에서는 대한민국의 도산 및 황지지점에서 BOD 농도예측을 위하여 새로운 이단계 하이브리드 패러다임(웨이블릿 기반 게이트 순환 유닛, 웨이블릿 기반 일반화된 회귀신경망, 그리고 웨이블릿 기반 랜덤 포레스트) 을 활용하였다. 이러한 모형들은 각 대응하는 독립모형들(게이트 순환 유닛, 일반화된 회귀신경망, 그리고 랜덤 포레스트) 과 함께 평가되었다. 다양한 수질 및 수량지표들이 여러 개의 입력조합(분류1-5) 을 기본으로 하여 독립 및 이단계 하이브리드 모형을 개발하기 위하여 구현되었다. 언급한 모형들은 root mean squared error (RMSE), Nash-Sutcliffe efficiency (NSE), 그리고 correlation coefficient (CC) 를 포함한 세 개의 통계지표로서 평가되었으며, 통계결과치를 분석하면 이단계 하이브리드 모형들이 항상 대응하는 독립모형들의 예측 정도를 개선하지 않은 것으로 나타났다. 대한민국의 도산관측소에서는 DWT-RF5 (RMSE = 0.108 mg/L) 모형이 다른 최적모형과 비교하여 BOD 농도의 더 정확한 예측을 나타내었으며, 황지관측소에서는 DWT-GRNN4 (RMSE = 0.132 mg/L) 모형이 BOD 농도를 예측하는 최고의 모형이다.

Biochemical oxygen demand (BOD) concentration, one of important water quality indicators, is treated as the measuring item for the ecological chapter in lakes and rivers. This investigation employed novel two-stage hybrid paradigm (i.e., wavelet-based gated recurrent unit, wavelet-based generalized regression neural networks, and wavelet-based random forests) to predict BOD concentration in the Dosan and Hwangji stations, South Korea. These models were assessed with the corresponding independent models (i.e., gated recurrent unit, generalized regression neural networks, and random forests). Diverse water quality and quantity indicators were implemented for developing independent and two-stage hybrid models based on several input combinations (i.e., Divisions 1-5). The addressed models were evaluated using three statistical indices including the root mean square error (RMSE), Nash-Sutcliffe efficiency (NSE), and correlation coefficient (CC). It can be found from results that the two-stage hybrid models cannot always enhance the predictive precision of independent models confidently. Results showed that the DWT-RF5 (RMSE = 0.108 mg/L) model provided more accurate prediction of BOD concentration compared to other optimal models in Dosan station, and the DWT-GRNN4 (RMSE = 0.132 mg/L) model was the best for predicting BOD concentration in Hwangji station, South Korea.

Background: Paraben is a widely used substance with a preservative effect found in various materials such as food, medicine, personal care products, and cosmetics. Objectives: This study was conducted to identify the level of urinary paraben concentrations (i.e., methyl-, ethyl-, and propyl-) among Korean adults and to explore the factors related with the exposure levels. Methods: We analyzed the third period (2015~2017) of the Korean National Environmental Health Survey (KoNEHS). R statistical software (version 4.1.1) was used to estimate representative values for the whole population with weight variables to reflect sampling design. Whether urinary concentrations tended to increase as the level of paraben exposure-related characteristics increased was tested and P_trend was calculated using general linear models. Results: Urinary concentrations of all three parabens (i.e., methyl-, ethyl- and propyl-) were higher in women than in men (P_trend<0.0001, 0.008, and <0.0001), and the values of methylparaben and propylparaben tended to increase as the age of subjects increased (P_trend<0.0001, and <0.0001). Urinary concentrations of methylparaben and propylparaben were associated with intensity of exercise (P_trend<0.001, and 0.004), and that of propylparaben was higher in non-smokers (P_trend=0.01). In terms of paraben exposure-related variables, urinary concentrations of parabens (i.e., methyl-, ethyl- and propyl-) increased as the daily average frequency of teeth-brushing (P_trend<0.0001, 0.03 and 0.0001), the frequency of use of hair products (P_trend=0.005, 0.05 and 0.04), the frequency of use of makeup products (P_trend<0.001, 0.001 and <0.001), and the frequency of use of antibacterial products (P_trend=0.005, 0.02 and 0.02) increased. Conclusions: In our study, urinary concentrations of all three parabens are associated with gender, teethbrushing, hair products, make-up products, and antibacterial products. Methyl- and proyl-parabens were associated with age and intensity of exercise, and propyl-paraben was associated with smoking.

오늘날 화학 산업은 다양한 산업 및 일상 생활과 밀접하게 연관되어 그 규모가 매우 방대하다. 하지만 이 러한 규모의 유해 화학 물질 관리는 한계가 있어 신속하고 정밀한 유해 화학 물질 검출이 어려운 상황이다. 본 연 구는 이러한 동기를 바탕으로 높은 정확도와 효율성을 갖춘 농도 분류 모델을 개발하는 것을 목표로 한다. 특히, random transformation 기반의 jittering과 scaling 기법을 적용해 적은 학습 데이터로도 모델의 일반화 성능이 향상됨을 확인하였다. Jittering 기법의 경우 cosine similarity를 통해 deviation 값을 조정하여 원본 데이터의 특 성을 유지하면서 노이즈를 추가 하였고, scaling 기법의 경우 다양한 실험을 통해 최적 factor 범위를 찾음으로써 모델의 성능을 개선하였다. 이러한 증강 기법을 활용하여 제안한 시스템은 최대 17%의 성능 향상 효과를 보였다.

Today, the chemical industry is very vast in scale and deeply interconnected with various industries and daily life. However, the management of hazardous chemicals on this scale has limitations, making it difficult to quickly and accurately detect hazardous chemicals. Therefore, this study aims to develop a highly accurate and efficient concentration classification model. This study demonstrates that Random Transformation-based augmentations, jittering and scaling, improve model generalization with limited data by preserving features and optimizing factor ranges. By employing these augmentation techniques, the proposed system achieved up to a 17% improvement in performance.

한반도 미세먼지의 연도별 공간분포를 비교 분석하기 위해서 측정소 측정자료를 이용하기에는 대도시 위주로 위치해 있기 때문에 한반도 전체에 대한 미세먼지의 공간분포를 파악하기에는 어려움이 있다. 반면 화학수송모델을 이용하여 공 간분포를 분석하는 방법은 공간분포를 파악하기에는 좋으나 누실된 배출량 자료 등에 의해 미세먼지 농도가 과소 평가 되는 경향이 있다. 본 연구에서는 공간 분포 분석에 측정 자료를 이용한 방법과 화학수송모델의 장점을 결합하여 화학수 송모델에 측정소 측정자료를 자료동화방법으로 재분석하여 분석하였다. 자료동화는 Pun’s Interpolation방법으로 국내 317개소의 대기오염측정소자료를 이용하였다. 자료동화를 통해 재분석한 2006년, 2009년, 2012년에 대해 서울, 부산을 대상으로 자료동화 전·후의 모델 정합도를 분석한 결과 자료동화 전에 비해 자료동화 후에 IOA가 0.39~0.71에서 0.77~0.98로 개선되었고, MBIAS는 -32.7 ~ -17.0 μg/m3에서 -8.1 ~ -0.2 μg/m3로 개선되는 것으로 나타났다. 자료동화를 통해 재분석한 결과를 이용하여 1시간, 24시간 평균의 연간 최대농도와 연평균 지역별 공간분포를 분석하였다. PM-10 농도의 공간분포는 연평균에서는 수도권과 광역시의 미세먼지가 높은 것으로 동일하게 나타났으나 1시간 최대농도와 24 시간 최대농도의 공간분포는 연도별로 차이가 있는 것으로 나타났다. 이는 국내의 고농도 PM-10은 지역의 배출에 의한 것이 아닌 월경성 오염원의 유입으로 나타나는 양상이기 때문에 연도별로 1시간 및 24시간 최대농도 분포가 다른 것으 로 판단된다. 특히 측정소가 없는 강원도의 양양지역에서 인근 측정소 측정자료와는 다른 경향이 나타났다. 이는 자료동 화 방법을 적용한 화학수송모델을 이용한 공간분포 특징으로 본 연구를 통해 자료동화기법으로 재분석 된 화학수송모델 자료는 대기질 현황분석 등에 지역별로 상세한 자료를 제공하는 방안이 될 수 있음을 확인하였다.

To improve PM-10 spatial distribution predicted by the chemical transport model of CMAQ, re-analysis method using measured PM concentrations was developed and tested its performance. Data assimilation method for the reanalysis was Pun's interpolation method with 28 km radius of influence of measured PM-10 values and measured PM-10 data in 317 air quality monitoring stations in Korea were used in this study. Predicted PM-10 concentrations by CMAQ with the data assimilation method were compared with measured ones for Year 2006, 2009 and 2012. The results showed that CMAQ model performance with data assimilation was significantly improved in predicting PM-10 concentrations; IOA at Seoul and Busan was improved from 0.39~0.71 to 0.77~0.98 and MBIAS was also improved from -32.7~-17.0 μg/m3 to -8.1~ -0.2 μg/m3. To predict PM-10 distribution over Southern Korean Peninsula, PM-10 re-analysis for Year 2006, 2009, and 2012 was carried out. The hourly and daily maximum as well as daily and annual PM-10 concentration distribution maps were obtained using the data assimilation methods developed in this study. The spatial distribution of annual average PM- 10 concentration was relatively high in the urbanized metropolis area such as Seoul metropolitan area, Busan and Ulsan area and those trends were the same for Year 2006, 2009 and 2012. However, the spatial distributions of the hourly and the daily maximum PM-10 concentrations were different in each year. Re-analyzed PM-10 maps showed that the data assimilation method using CMAQ has an advantage in predicting spatial distributions where monitoring stations were sparsely available such as Yangyang area in Kangwon province.

Recently, according as the quality of life is improving, we are concerning the IAQ. But IAQ level is more deteriorating because people are using imprudent construction material. There are insufficient legal standard, study and research about IAQ. The purpose of this study is to apply IAQ and to develop an indoor pollutant density prediction program based on the indoor pollutant substance data.

플랫폼 비즈니스는 사람들의 생활을 풍부하게 함과 동시에 경제성장을 재촉하며 글로벌하게 전개되어 왔다. 플랫폼의 특징 즉, 네트워크효과, 쏠림현상, 고착효과 등에 의하여 플랫폼시장에 서는 독점 등이 발생하기 쉽고, 소비자나 사업자의 불이익, 이노베이션의 저해로 연결될 우려도 있다. 이에 플랫폼 비즈니스나 데이터에 대한 경쟁정책이 국제적으로 중요해지고 있다. 경쟁정책 을 둘러싼 큰 관심의 하나로 플랫폼 등 디지털 분야의 경쟁환경의 정비를 어떻게 진행시켜 나가 야 하는지가 큰 과제가 되고 있다. 신규진입자 등에 의한 새로운 도전이 저해되고 있지 않은지, 남용적인 행위에 의해서 이노베이션이 억압되고 있지 않은지에 대한 점에 초점을 두고서 경쟁당 국은 경쟁정책을 펴야 할 것으로 생각된다. 디지털 플랫폼에 의한 과점․독점이 이노베이션에 의해서 확보되는 것인 한 경쟁당국으로서 도 문제삼지 말아야 한다고 생각된다. 그러나 디지털 플랫폼이 스스로의 지배적 지위를 남용하여 소비자나 거래처 사업자에게 부당한 불이익을 부과하는 것으로 공정한 경쟁을 왜곡시키거나 스 스로의 경쟁자가 될 가능성이 있는 신규진입자를 부당하게 배제하는 등 자유로운 경쟁을 방해하 는 행위를 하는 경우에는 경쟁정책상 간과해서는 안된다. 또한, 디지털 시대에는 경쟁당국에 의 한 집행의 지연이 종래의 시장보다 장기간에 걸쳐서 경쟁을 저해하게 된다. 이 때문에 경쟁당국 은 문제가 커진 뒤 행동하는 것이 아니라 빨리 조치하는 것이 중요하다. 플랫폼에서의 데이터 집중도 경쟁법 집행에 있어서 관심을 가져야 할 현상이다. 데이터의 접근 성 확보와 관련한 법제의 마련 및 법 집행의 필요성이 존재한다. 법정책의 과제로서는 플랫폼내 의 경쟁이나 플랫폼간의 경쟁을 어떻게 평가해야 하는가 하는 문제가 있다. 데이터에 의한 이노 베이션이 중요하며, 데이터가 경쟁상 어떠한 역할을 담당하고 있는지를 검토할 필요가 있다. 경 쟁에의 영향을 검토함에 있어서 데이터의 종류를 보는 것이 중요하다. 경쟁에 영향을 주는 요인 으로서 디지털 플랫폼에서 데이터를 누구나 수집․공유할 수 있는지, 플랫폼의 멀티홈이 허용되는지, 데이터의 수집ㆍ축적에 관해서 투명성이 높은지 하는 관점을 들 수 있다. 플랫폼의 행동 자체가 새로운 가치 창출을 목표로 저렴하고 양질의 서비스를 제공하는 반면, 경쟁자의 배제나 소비자에게 불이익을 준다고 하는 경쟁제한적 효과도 존재한다. 그러므로 이와 같은 양자의 밸런스를 어떻게 판단하고 경쟁정책을 집행할 것인가 하는 것은 매우 어려운 과제이 다. 이노베이션을 저해하는 일 없이 경쟁제한행위를 배제하고, 디지털 플랫폼 이용자의 이익을 확보하고, 경제의 지속적인 성장을 실현하는 것을 경쟁당국은 목표로 해야 할 것이다. 디지털 플 랫폼에 관해서 주로 다음과 같은 5가지 대응을 추진하여야 할 것이다. 첫째, 공정거래법을 위반하 는 행위에 대해 엄정하고 신속하게 대응할 필요가 있다. 둘째, 디지털 플랫폼의 거래관행 등 실태 조사를 실시하여야 한다. 셋째, 소비자와의 관계에서 우월적 지위남용 규제의 적용을 검토할 필 요가 있다. 넷째, 데이터와 이노베이션을 고려한 기업결합심사를 할 필요가 있다. 다섯째, 각종 소프트웨어의 클라우드화에 대비한 경쟁정책의 검토 역시 필요해 보인다.

Concerns are growing over market monopolization and oligopoly due to huge platforms emerging in the Internet ecosystem. Leading platforms such as Google, Facebook, Apple and Amazon have established themselves to exercise significant monopoly power through data accumulation and use. In response to these developments, the European Union (EU) and Japan are among the first to develop legislation and rules for data protection and transparency and fairness of the platform trading environment. Platform businesses have enriched people’s lives, encouraged economic growth, and expanded globally. Platform characteristics, such as network effect, tipping effect, and lock in effect, are likely to cause monopolies in the platform market, which could lead to disadvantages for consumers and business partners, as well as impediments to innovation. Therefore, competition policies for platform businesses and data are becoming important internationally. One of the major concerns regarding competition policy is how to develop a competitive environment in the digital field such as platform. Competition authorities should develop competition policies focusing on whether new challenges by new entrants are being hampered or whether innovation is being suppressed by abuse. As long as oligopolies and monopolies by digital platforms are secured through innovation, competition authorities should not consider them problematic. However, when a digital platform abuses its dominant position and inflicts undue disadvantages on consumers and business partners, distorting fair competition or unjustly eliminating new entrants who may become its own competitors, it should not be overlooked in terms of competition policy. In the digital age, the delay in enforcement by competition authorities hinders competition for a longer period of time than conventional markets. Therefore, it is important for competition authorities not to act after the problem grows, but to take immediate action. Platform concentration of data is also a phenomenon of interest in competition law enforcement. There is a need for legislation and law enforcement to ensure access to data. Legal policy issues include how to evaluate competition within platforms and between platforms. Innovation through data is important, and we need to consider what role data plays in competition. It is important to look at the types of data when considering the impact on competition. Factors affecting competition include whether digital platforms allow everyone to collect and share data, whether multi-homing of platforms are allowed, and whether data collection and accumulation are highly transparent. While platform behavior itself is aimed at creating new value and provides low-cost, high-quality services, it also has the effect of eliminating competitors and inflicting disadvantages on consumers. Therefore, it is an urgent task to determine the balance between the two and to implement competition policies. Competition authorities should aim to eliminate competition-restricting practices without inhibiting innovation, secure the interests of digital platform users, and achieve sustainable economic growth.

본 논문에서는 인공지능 주의환기 보상전략 시스템이 발달장애인의 데이터 라벨링 작업 주의집중력에 미치는 효과를 분석하였다. 주의집중력의 척도로는 세션별 작업 정확도와 작업수행량을 사용하였다. 연구 결과, 중재가 적용된 후 연구대상자 모두 자율작업 대비 주의집중력에서 유의미한 향상이 관찰되었다. 이러한 결과는 인공지능 기술이 발달장 애인의 데이터 라벨링 작업 중 주의집중력 향상에 긍정적으로 작용할 수 있음을 의미한다. 본 연구는 인공지능 기술의 적용이 발달장애인의 데이터 라벨링 작업 정확도를 향상하여 학습데이터의 품질을 높일 수 있음을 보여주고 있으며, 발 달장애인의 데이터라벨링 관련 직업훈련 프로그램에 중요한 시사점을 제공하리라 본다.

This paper investigates the effect of an artificial intelligence attention redirection compensation strategy system on the data labeling work attention concentration by individuals with developmental disabilities. Task accuracy and task performance for each session were used as measures of attention concentration. As a result of the study, after the intervention was applied, a significant improvement in attention concentration was observed in all study subjects compared to self-serving task. These results mean that artificial intelligence technology can have a positive effect on improving the attention span of people with developmental disabilities during data labeling tasks. This study shows that the application of artificial intelligence technology can improve the quality of learning data by improving the accuracy of data labeling tasks for people with developmental disabilities, and is expected to provide important implications for vocational training programs related to data labeling for people with developmental disabilities.

본 연구는 저축은행의 소유집중도와 사업다각화 간의 관계를 분석하고자 함이 목적이다. 패널데이터를 이용하 여 저축은행의 소유집중도와 사업다각화의 관계를 분석한 것은 기존연구와의 차별성이라 할 수 있다. 분석기간은 2014년부터 2018년까지이며, 79개 저축은행을 대상으로 반기 재무자료를 이용하였다. 연구모형은 자기상관 문제를 고 려한 확률효과 GLS(Random effects Generalized Linear Square)모형을 이용하여 분석하였다. 분석결과, 저축은행의 지배주주지분율과 사업다각화의 관계는 유의한 양(+)의 값을 가지는 것으로 추정되었다. 즉 지배주주지분율이 증가할 수록 사업다각화를 대리하는 변수인 비이자수익률이 증가하는 것으로 나타났다. 이는 위험회피가설을 지지하는 결과 이며, 지배주주지분율이 높을수록 사업다각화를 통해 자신의 자산위험과 고용위험을 축소시키기 위해 적극적인 다각 화 전략을 추진하는 것으로 나타났다.

The purpose of this study is to analyze the relationship between the controlling shareholding ratio and the business diversification of savings banks The difference in this study is the analysis of the relationship between the controlling shareholding ratio of the savings bank and the business diversification using panel data. In this study, the semi-annual financial statements for the period 2014-2018 were used on the basis of a sample of 79 saving banks. The research model is analyzed using random effects generalized linear square (GLS) model considering the autocorrelation problem.As a result of the empirical analysis, it is estimated that the relationship between the controlling shareholding ratio of the savings bank and the business diversification is significant (+). This is the result of supporting the hedging hypothesis.

To produce a level-3 monthly composite image from daily level-2 Sea-viewing Wide Field-of-view Sensor (SeaWiFS) chlorophyll-a concentration data set in the East Sea, we applied four average methods such as the simple average method, the geometric mean method, the maximum likelihood average method, and the weighted averaging method. Prior to performing each averaging method, we classified all pixels into normal pixels and abnormal speckles with anomalously high chlorophyll-a concentrations to eliminate speckles from the following procedure for composite methods. As a result, all composite maps did not contain the erratic effect of speckles. The geometric mean method tended to underestimate chlorophyll-a concentration values all the time as compared with other methods. The weighted averaging method was quite similar to the simple average method, however, it had a tendency to be overestimated at high-value range of chlorophyll-a concentration. Maximum likelihood method was almost similar to the simple average method by demonstrating small variance and high correlation (r=0.9962) of the differences between the two. However, it still had the disadvantage that it was very sensitive in the presence of speckles within a bin. The geometric mean was most significantly deviated from the remaining methods regardless of the magnitude of chlorophyll-a concentration values. Its bias error tended to be large when the standard deviation within a bin increased with less uniformity. It was more biased when data uniformity became small. All the methods exhibited large errors as chlorophyll-a concentration values dominantly scatter in terms of time and space. This study emphasizes the importance of the speckle removal process and proper selection of average methods to reduce composite errors for diverse scientific applications of satellite-derived chlorophyll-a concentration data.

In this study, we analyzed how ozone pollution could be differently measured and how these different measures varied year to year and across the ten most populated cities in the United States, from 1980 to 2000. Although peak values of ozone concentration have been significantly reduced in most polluted U.S. cities for the last 20 years, the annual average values of ozone concentration have not been lowered as much as peak values. Ozone concentration data for each city shows a unique pattern of distribution, central tendency, and also there is a wide variation among different ozone measures. Two different cities with the same annual mean concentration of ozone can experience very different distributions of ozone concentration within a year. Ozone measures also show a wide margin of variability as they are estimated from different ozone monitoring sites within each city. Ozone pollution statistics can be largely varied depending on the choice of measures, monitoring sites, and averaging time period. EPA's new ozone standard of 0.08 ppm averaged over an eight-hour appears to be more stringent than the current maximum ozone standard of 0.12 ppm averaged over one hour.

Suspended sediment is an important oceanic variable for monitoring changes in coastal environment related to physical and biogeochemical processes. In order to estimate suspended sediment concentration (SSC) from satellite data, we derived SSC coefficients by fitting satellite remote sensing reflectances to in-situ suspended sediment measurements. To collect in-situ suspended sediment, we conducted ship cruises at 16 different locations three times for the periods of Sep.-November 2009 and Jul. 2010 at the passing time of Landsat $ETM_+$. Satellite data and in-situ data measured by spectroradiometers were converted to remote sensing reflectances ($R_{rs}$). Statistical approaches proved that the exponential formula using a single band of $R_{rs}$(565) was the most appropriate equation for the estimation of SSC in this study. Satellite suspended sediment using the newly-derived coefficients showed a good agreement with insitu suspended sediment with an Root Mean Square (RMS) error of 1-3 g/$m^3$. Satellite-observed SSCs tended to be overestimated at shallow depths due to bottom reflection presumably. This implies that the satellite-based SSCs should be carefully understood at the shallow coastal regions. Nevertheless, the satellite-derived SSCs based on the derived SSC coefficients, for the most cases, reasonably coincided with the pattern of in-situ suspended sediment measurements in the study region.

PM (particulate matter) is of interest to everyone because it can have adverse effects on human health by the infiltration from respiratory to internal organs. To date, many studies have made efforts for the prediction of PM10 and PM2.5 concentrations. Unlike previous studies, we conducted the prediction of tomorrow's PM10 concentration for the Air Korea stations using Chinese PM10 data in addition to the satellite AOD and weather variables. We constructed 230,639 matchups from the raw data over 3 million and built an RF (random forest) model from the matchups to cope with the complexity and nonlinearity. The validation statistics from the blind test showed excellent accuracy with the RMSE (root mean square error) of 9.905 ㎍/㎥ and the CC (correlation coefficient) of 0.918. Moreover, our prediction model showed a stable performance without the dependency on seasons or the degree of PM10 concentration. However, part of coastal areas had a relatively low accuracy, which implies that a dedicated model for coastal areas will be necessary. Additional input variables such as wind direction, precipitation, and air stability should also be incorporated into the prediction model as future work.

Since the Coastal Zone Color Scanner (CZCS)/Nimbus-7 was launched in 1978, a variety of studies have been conducted to retrieve ocean color variables from multi-satellites. Several algorithms and formulations have been suggested for estimating ocean color variables based on multi band data at different wavelengths. Chlorophyll-a (chl-a) concentration is one of the most important variables to understand low-level ecosystem in the ocean. To retrieve chl-a concentrations from the satellite observations, an appropriate algorithm depending on water properties is required for each satellite sensor. Most operational empirical algorithms in the global ocean have been developed based on the band-ratio approach, which has the disadvantage of being more adapted to the open ocean than to coastal areas. Alternative algorithms, including the semi-analytical approach, may complement the limits of band-ratio algorithms. As more sensors are planned by various space agencies to monitor the ocean surface, it is expected that continuous monitoring of oceanic ecosystems and environments should be conducted to contribute to the understanding of the oceanic biosphere and the impact of climate change. This study presents an overview of the past and present algorithms for the estimation of chl-a concentration based on multi-satellite data and also presents the prospects for ongoing and upcoming ocean color satellites.