• Journal of Ecology and Environment
• /
• v.45 no.4
• /
• pp.313-327
• /
• 2021

Background: Citizen science is becoming a mainstream approach of baseline data collection to monitor biodiversity and climate change. Dragonflies (Odonata) have been ranked as the highest priority group in biodiversity monitoring for global warming. Ischnura senegalensis Rambur has been designated a biological indicator of climate change and is being monitored by the citizen science project "Korean Biodiversity Observation Network." This study has been performed to understand changes in the distribution range of I. senegalensis in response to climate change using citizen science data in South Korea. Results: We constructed a dataset of 397 distribution records for I. senegalensis, ranging from 1980 to 2020. The number of records sharply increased over time and space, and in particular, citizen science monitoring data accounted for the greatest proportion (58.7%) and covered the widest geographical range. This species was only distributed in the southern provinces until 2010 but was recorded in the higher latitudes such as Gangwon-do, Incheon, Seoul, and Gyeonggi-do (max. Paju-si, 37.70° latitude) by 2020. A species distribution model showed that the annual mean temperature (Bio1; 63.2%) and the maximum temperature of the warmest month (Bio5; 16.7%) were the most critical factors influencing its distribution. Future climate change scenarios have predicted an increase in suitable habitats for this species. Conclusions: This study is the first to show the northward expansion in the distribution range of I. senegalensis in response to climate warming in South Korea over the past 40 years. In particular, citizen science was crucial in supplying critical baseline data to detect the distribution change toward higher latitudes. Our results provide new insights on the value of citizen science as a tool for detecting the impact of climate change on ecosystems in South Korea.

• Journal of Ocean Engineering and Technology
• /
• v.36 no.5
• /
• pp.340-352
• /
• 2022

This article presents a modeling method for the uncorrelated measurement error of the ultra-short baseline (USBL) acoustic positioning system for aiding navigation of underwater vehicles. The Mahalanobis distance (MD) and principal component analysis are applied to decorrelate the errors of USBL measurements, which are correlated in the x- and y-directions and vary according to the relative direction and distance between a reference station and the underwater vehicles. The proposed method can decouple the radial-direction error and angular direction error from each USBL measurement, where the former and latter are independent and dependent, respectively, of the distance between the reference station and the vehicle. With the decorrelation of the USBL errors along the trajectory of the vehicles in every time step, the proposed method can reduce the threshold of the outlier decision level. To demonstrate the effectiveness of the proposed method, simulation studies were performed with motion data obtained from a field experiment involving an autonomous underwater vehicle and USBL signals generated numerically by matching the specifications of a specific USBL with the data of a global positioning system. The simulations indicated that the navigation system is more robust in rejecting outliers of the USBL measurements than conventional ones. In addition, it was shown that the erroneous estimation of the navigation system after a long USBL blackout can converge to the true states using the MD of the USBL measurements. The navigation systems using the uncorrelated error model of the USBL, therefore, can effectively eliminate USBL outliers without loss of uncontaminated signals.

• Journal of Ecology and Environment
• /
• v.47 no.2
• /
• pp.49-62
• /
• 2023

Background: Over the past three decades, gradual eustatic sea-level rise has been considered a primary exogenous factor in the increased frequency of flooding and biological changes in several salt marshes. Under this paradigm, the potential importance of short-term events, such as ocean storminess, in coastal hydrology and ecology is underrepresented in the literature. In this study, a simulation was developed to evaluate the influence of wind waves driven by atmospheric oscillations on sedimentary and vegetation dynamics at the Skallingen salt marsh in southwestern Denmark. The model was built based on long-term data of mean sea level, sediment accretion, and plant species composition collected at the Skallingen salt marsh from 1933-2006. In the model, the submergence frequency (number yr^-1) was estimated as a combined function of wind-driven high water level (HWL) events (> 80 cm Danish Ordnance Datum) affected by the North Atlantic Oscillation (NAO) and changes in surface elevation (cm yr^-1). Vegetation dynamics were represented as transitions between successional stages controlled by flooding effects. Two types of simulations were performed: (1) baseline modeling, which assumed no effect of wind-driven sea-level change, and (2) experimental modeling, which considered both normal tidal activity and wind-driven sea-level change. Results: Experimental modeling successfully represented the patterns of vegetation change observed in the field. It realistically simulated a retarded or retrogressive successional state dominated by early- to mid-successional species, despite a continuous increase in surface elevation at Skallingen. This situation is believed to be caused by an increase in extreme HWL events that cannot occur without meteorological ocean storms. In contrast, baseline modeling showed progressive succession towards the predominance of late-successional species, which was not the then-current state in the marsh. Conclusions: These findings support the hypothesis that variations in the NAO index toward its positive phase have increased storminess and wind tides on the North Sea surface (especially since the 1980s). This led to an increased frequency and duration of submergence and delayed ecological succession. Researchers should therefore employ a multitemporal perspective, recognizing the importance of short-term sea-level changes nested within long-term gradual trends.

• International journal of advanced smart convergence
• /
• v.12 no.4
• /
• pp.434-442
• /
• 2023

The relationship between acute kidney injury (AKI) prediction and nephrotoxic drugs, or drugs that adversely affect kidney function, is one that has yet to be explored in the critical care setting. One contributing factor to this gap in research is the limited investigation of drug modalities in the intensive care unit (ICU) context, due to the challenges of processing prescription data into the corresponding drug representations and a lack in the comprehensive understanding of these drug representations. This study addresses this gap by proposing a novel approach that leverages patient prescription data as a modality to improve existing models for AKI prediction. We base our research on Electronic Health Record (EHR) data, extracting the relevant patient prescription information and converting it into the selected drug representation for our research, the extended-connectivity fingerprint (ECFP). Furthermore, we adopt a unique multimodal approach, developing machine learning models and 1D Convolutional Neural Networks (CNN) applied to clinical drug representations, establishing a procedure which has not been used by any previous studies predicting AKI. The findings showcase a notable improvement in AKI prediction through the integration of drug embeddings and other patient cohort features. By using drug features represented as ECFP molecular fingerprints along with common cohort features such as demographics and lab test values, we achieved a considerable improvement in model performance for the AKI prediction task over the baseline model which does not include the drug representations as features, indicating that our distinct approach enhances existing baseline techniques and highlights the relevance of drug data in predicting AKI in the ICU setting.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.3
• /
• pp.32-43
• /
• 2010

This paper presents deblocking filter for low-complexity video decoder. Baseline profile of the H.264/AVC used for mobile devices such as mobile phones has two times higher compression performance than the MPEG-4 Visual but it has a problem of serious complexity as using 1/4-pel interpolation filter, adaptive entropy model and deblocking filter. This paper presents low-complexity deblocking filter for decreasing complexity of decoder with preserving the coding efficiency of the H.264/AVC. In this paper, the proposed low-complexity deblocking filter decreased 49% of branch instruction than conventional approach as calculating value of BS by using the CBP. In addition, a range of filtering of strong filter applied in intra macroblock boundaries was limited to two pixels. According to the experimental results, the proposed low-complexity deblocking filter decreased -0.02% of the BDBitrate comparison with baseline profile of the H.264/AVC, decreased 42% of the complexity of deblocking filter, and decreased 8.96% of the complexity of decoder.

• Korean Journal of Remote Sensing
• /
• v.28 no.1
• /
• pp.121-128
• /
• 2012

High resolution intertidal DEM is a basic material for science research like sedimentation/erosion by ocean current, and is invaluable in a monitoring of environmental changes and practical management of coastal wetland. Since the intertidal zone changes rapidly by the inflow of fluvial debris and tide condition, remote sensing is an effective tool for observing large areas in short time. Although radar interferometry is one of the well-known techniques for generating high resolution DEM, conventional repeat-pass interferometry has difficulty on acquiring enough coherence over tidal flat due to the limited exposure time and the rapid changes in surface condition. In order to overcome these constraints, we tested the feasibility of radar interferometry using Cosmo-SkyMed tandem-like one-day data and ERS-ENVISAT cross tandem data with very short revisit period compared to the conventional repeat pass data. Small temporal baseline combined with long perpendicular baseline allowed high coherence over most of the exposed tidal flat surface in both observations. However the interferometric phases acquired from Cosmo-SkyMed data suffer from atmospheric delay and changes in soil moisture contents. The ERS-ENVISAT pair, on the other hand, provides nice phase which agree well with the real topography, because the atmospheric effect in 30-minute gap is almost same to both images so that they are cancelled out in the interferometric process. Thus, the cross interferometry with very small temporal baseline and large perpendicular baseline is one of the most reliable solutions for the intertidal DEM construction which requires very accurate mapping of the elevation.

• Journal of Preventive Medicine and Public Health
• /
• v.37 no.4
• /
• pp.381-389
• /
• 2004

Objective : To compare the efficacy between SKI306X and Diclofenac by using generalized estimating equations (GEE) methodology in the analysis of correlated bivariate binary outcome data in Osteoarthritis (OA) diseases. Methods : A randomized, double-blind, active comparator-controlled, non-inferiority clinical trial was conducted at 5 institutions in Korea with the random assignment of 248 patients aged 35 to 75 years old with OA of the knee and clinical evidence of OA. Patients were enrolled in this study if they had at least moderate pain in the affected knee joint and a score larger than 35mm as assessed by VAS (Visual Analog Scale). The main exposure variable was treatment (SKI 306X vs. Diclofenac) and other covariates were age, sex, BMI, baseline VAS, center, operation history (Yes/No), NSAIDS (Y/N), acupuncture (Y/N), herbal medicine (Y/N), past history of musculoskeletal disease (Y/N), and previous therapy related with OA (Y/N). The main study outcome was the change of VAS pain scores from baseline to the 2nd and 4th weeks after treatment. Pain scores were obtained as baseline, 2nd and 4th weeks after treatment. We applied GEE approach with empirical covariance matrix and independent(or exchangeable) working correlation matrix to evaluate the relation of several risk factors to the change of VAS pain scores with correlated binary bivariate outcomes. Results : While baseline VAS, age, and acupuncture variables had protective effects for reducing the OA pain, its treatment (Joins/Diclofenac) was not statistically significant through GEE methodology (ITT:aOR=1.37, 95% CI=(0.8200, 2.26), PP:aOR=1.47, 95% CI=(0.73, 2.95)). The goodness-of-fit statistic for GEE (6.55, p=0.68) was computed to assess the adequacy of the fitted final model. Conclusions : Both ANCOVA and GEE methods yielded non statistical significance in the evaluation of non-inferiority of the efficacy between SKI306X and Diclofenac. While VAS outcome for each visit was applied in GEE, only VAS outcome for the fourth visit was applied in ANCOVA. So the GEE methodology is more accurate for the analysis of correlated outcomes.

• Nutrition Research and Practice
• /
• v.9 no.4
• /
• pp.379-384
• /
• 2015

BACKGROUND/OBJECTIVE: Apolipoprotein A5 gene promoter region T-1131C polymorphism (APOA5 T-1131C) is known to be associated with elevated plasma TG levels, although little is known of the influence of the interaction between APOA5 T-1131C and lifestyle modification on TG levels. To investigate this matter, we studied APOA5 T-1131C and plasma TG levels of subjects participating in a three-month lifestyle modification program. SUBJECTS/METHODS: A three-month lifestyle modification program was conducted with 297 participants (Age: $57{\pm}8years$) in Izumo City, Japan, from 2001-2007. Changes in energy balance (the difference between energy intake and energy expenditure) and BMI were used to evaluate the participants' responses to the lifestyle modification. RESULTS: Even after adjusting for confounding factors, plasma TG levels were significantly different at baseline among three genotype subgroups: TT, $126{\pm}68mg/dl$; TC, $134{\pm}74mg/dl$; and CC, $172{\pm}101mg/dl$. Lifestyle modification resulted in significant reductions in plasma TG levels in the TT, TC, and CC genotype subgroups: $-21.9{\pm}61.0mg/dl$, $-20.9{\pm}51.0mg/dl$, and $-42.6{\pm}78.5mg/dl$, respectively, with no significant differences between them. In a stepwise regression analysis, age, APOA5 T-1131C, body mass index (BMI), homeostasis model assessment-insulin resistance (HOMA-IR), and the 18:1/18:0 ratio showed independent association with plasma TG levels at baseline. In a general linear model analysis, APOA5 T-1131C C-allele carriers showed significantly greater TG reduction with decreased energy balance than wild type carriers after adjustment for age, gender, and baseline plasma TG levels. CONCLUSIONS: The genetic effects of APOA5 T-1131C independently affected plasma TG levels. However, lifestyle modification was effective in significantly reducing plasma TG levels despite the APOA5 T-1131C genotype background.

• Journal of Information Technology Services
• /
• v.19 no.5
• /
• pp.83-91
• /
• 2020

Language Models such as BERT has been an important factor of deep learning-based natural language processing. Pre-training the transformer-based language models would be computationally expensive since they are consist of deep and broad architecture and layers using an attention mechanism and also require huge amount of data to train. Hence, it became mandatory to do fine-tuning large pre-trained language models which are trained by Google or some companies can afford the resources and cost. There are various techniques for fine tuning the language models and this paper examines three techniques, which are data augmentation, tuning the hyper paramters and partly re-constructing the neural networks. For data augmentation, we use no-answer augmentation and back-translation method. Also, some useful combinations of hyper parameters are observed by conducting a number of experiments. Finally, we have GRU, LSTM networks to boost our model performance with adding those networks to BERT pre-trained model. We do fine-tuning the pre-trained korean-based language model through the methods mentioned above and push the F1 score from baseline up to 89.66. Moreover, some failure attempts give us important lessons and tell us the further direction in a good way.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.12
• /
• pp.565-572
• /
• 2017

Traditional method for time series analysis, autoregressive integrated moving average (ARIMA) allows to mine significant patterns from the past observations using autocorrelation and to forecast future sequences. However, Korean baseball games do not have regular intervals to analyze relationship among the past attendance observations. To address this issue, we propose artificial neural network (ANN) based attendance prediction model using various measures including performance, team characteristics and social influences. We optimized ANNs using grid search to construct optimal model for regression problem. The evaluation shows that the optimal and ensemble model outperform the baseline model, linear regression model.