Search

While numerous studies have employed deep learning and high-resolution remote sensing to predict future land use and land cover (LULC) changes, no study has integrated these predictive tools with the current urban planning context to find a potential issues for sustainability. This study addresses this gap by examining the planning context of Busan Metropolitan City (BMC) and analyzing the paradoxical objectives within the city's 2040 Master Plan and the subordinate 2030 Busan Master Plan for Parks and Greenbelts. Although the plans advocate for increased green areas to enhance urban sustainability and social wellbeing, they simultaneously support policies that may lead to a reduction in these areas due to urban development. Using the CA-ANN model in the MOLUSCE plugin, a deep learning-based LULC change analysis, we forecast further urban expansion and continued shrinkage of natural green areas. During 1980–2010, Busan Metropolitan City (BMC) underwent high-speed urban expansion, wherein the urbanized areas almost doubled and agricultural lands and green areas, including forests and grassland, reduced considerably. Forecasts for the years 2010–2040 show continued further expansion of urban areas at the expense of areas for agriculture and green areas, including forest and grasslands. Given the master plans, these highlight a critical tension between urban growth and sustainability. Despite the push for more green spaces, the replacement of natural landscapes with artificial parks and green areas may threaten long-term sustainability. In view of these apparently conflicting goals, the urban planning framework for BMC would have to take up increasingly stronger conservation policies and adaptive planning practices that consider environmental preservation on a par with economic development in the light of the planning context and trajectory of urbanization. [ABSTRACT FROM AUTHOR]

The concept of social infrastructure and its role in building disaster resilience has garnered significant attention in recent years. Defined as physical spaces that foster social connections and vibrant communities, social infrastructure has been recognized as a crucial element in providing support and resources during disasters. The coronavirus disease 2019 pandemic has further emphasized the importance of enhancing social infrastructure. However, divergent perspectives exist regarding the relationship between social infrastructure and socioeconomic factors in terms of distribution. Some studies suggest that the disproportionate distribution of social infrastructure may disadvantage certain racial and socioeconomic groups. Conversely, other research highlights instances where marginalized communities with robust social infrastructure exhibit resilience despite facing social and economic challenges. This discrepancy underscores the need for further investigation to reconcile these differing descriptions. Addressing this gap, this study aims to examine the relationship between social infrastructure, socioeconomic factors, and environmental justice. Specifically, the research question guiding this study is: Is the distribution of social infrastructure in a city related to specific racial or class factors from an environmental justice perspective? Utilizing the ordinary least squares regression, this study reveals that communities with a higher proportion of the non-White population have a greater number of social infrastructures, challenging expectations based on the environmental justice literature. [ABSTRACT FROM AUTHOR]

With the recent increase in food demand, urban agriculture has gained attention as a way of increasing food self-sufficiency and providing recreational spaces in cities. In this study, the suitability of rooftop greenhouses (RGs), a type of urban agriculture, was analyzed by combining the analytical hierarchy process (AHP) and geographic information systems (GIS) in Seoul, the capital city of Korea. To achieve this, we derived location suitability factors through expert consultations and calculated the weights of each factor through AHP. After building the spatial data according to these factors, they were weighted and summed then scaled to a score of 0–100. The highest weight of the RG location factors was for benefit (0.1782), followed by officially assessed land prices (0.0913) and supermarket density (0.0802). The weights of supermarket density and accessibility were high because they are considered the main distribution channels. When analyzing the location of RGs by linking these results with the spatial data according to factor, we revealed that Gangseo-gu (a district of Seoul) had relatively high location suitability scores. This trend was determined to be caused by the rather low officially assessed land price, high supermarket density, and productive population. This result could prove useful when selecting the approximate locations of RGs in Seoul and for promoting food self-sufficiency in cities. [ABSTRACT FROM AUTHOR]

Deep learning-based medical image analysis technology has been developed to the extent that it shows an accuracy surpassing the ability of a human radiologist in some tasks. However, data labeling on medical images requires human experts and a great deal of time and expense. Moreover, medical image data usually have an imbalanced distribution for each disease. In particular, in multilabel classification, learning with a small number of labeled data causes overfitting problems. The model easily overfits the limited number of labeled data, while it still underfits the large amount of unlabeled data. In this study, we propose a method that combines entropy-based Mixup and self-training to improve the performance of data-imbalanced chest X-ray classification. The proposed method is to apply the Mixup algorithm to limited labeled data to alleviate the data imbalance problem and perform self-training that effectively utilizes the unlabeled data while iterating this process by replacing the teacher model with the student model. Experimental results in an environment with a limited number of labeled data and a large number of unlabeled data showed that the classification performance was improved by combining entropy-based Mixup and self-training. [ABSTRACT FROM AUTHOR]