Estimating potential illegal land development in conservation areas based on a presence-only model.

Conservation areas are facing increasing threats from anthropogenic land use activities. It is important to reasonably recognize and predict suspected illegal land development in advance. However, traditional methods easily suffer from selection bias due to the lack of accurate and reliable absence data. To tackle this problem, we have presented a novel method for estimating potential illegal land development based on the presence-only maximum entropy (MAXENT) model. The principle of MAXENT can guarantee that no additional unknown information (e.g., inaccurate pseudo-absence samples) will be introduced into the estimation procedure. This method was applied to the conservation areas in a fast-growing city, and the robustness of the MAXENT models was confirmed by the high AUC scores (over 0.80). The results indicated that the proposed method performs more effectively than the presence-absence random forest model. In addition, topographic conditions and proximity to transportation networks played dominant roles in the emergence of suspected illegal land development. Moreover, the probability map generated by MAXENT suggests that a considerable amount of forest, farmland, grassland, and water bodies will face a high degree of danger. Therefore, both superior and local governments should pay much more attention to regions with a higher potential for illegal land development. In summary, our findings are expected to support decision-making in the management and assessment of conservation areas in fast-growing regions. More importantly, the proposed method can be further applied to illegal land development estimation in many other regions.

Measuring the Non-Linear Relationship between Three-Dimensional Built Environment and Urban Vitality Based on a Random Forest Model.

Urban vitality is a major indicator used for evaluating the sustainability and attractiveness of an urban environment. Global experience indicates that urban vitality can be stimulated through a reasonable urban design. However, it remains incompletely understood in the literature which building-related indicators can substantially affect urban vitality in Asian countries. To give an insight into this question, our study took a step forward by focusing specifically on the influence of the three-dimensional built environment on urban vitality, based on which decision makers could enhance urban vitality from the perspective of vertical building design. A machine-learning-based framework was developed in this study. First, we utilized several building-related indicators to thoroughly measure the spatial characteristics of buildings at the township level. Second, the relationship between a three-dimensional built environment and urban vitality was revealed based on a combined use of the correlation method, scatter charts, and a random forest. In the random forest, both a benchmark and a new model were constructed to evaluate the importance of those building-related indicators. The results suggested that urban vitality was closely related to the three-dimensional built environment, which played an even more important role than common benchmark factors in stimulating urban vitality. The building coverage ratio, density of tall buildings, and floor area ratio were essential spatial drivers behind urban vitality. Therefore, urban designers and decision makers should not only take traditional factors into account but also carefully consider the potential influence of high-rise buildings and the outdoor thermal environment so that urban vitality can be enhanced. Our study's results can offer practical recommendations for improving urban vitality from the perspective of vertical building design. The proposed framework can also be used for measuring the potential influence of the three-dimensional built environment in other areas.