Urban Rural Building Network and Land Morphology Connection (URBN-LMC)

Building characteristics, such as height and footprint, are essential components of urban form models. Urban form, or the physical characteristics of urban areas, has been described thoroughly and used to predict building energy consumption, carbon footprint, and age. Although progress in characteristic prediction has been made with machine learning, open-source data, and LIDAR techniques, there are few models that include the use of land use/cover and socioeconomic data. We propose a new model, the Urban Rural Building Network and Land Morphology Connection (URBN-LMC), to derive high-resolution (30-m) building morphology statistical parameters from land cover and population data. The pilot random forest modeling efforts had relatively low performance (22% variance explained) in predicting building characteristics when constructed at granularity of the desired end product, 30-m resolution. This led us to explore a hybrid multi-scale modeling approach using machine learning prediction, followed by statistical downscaling and cellular automata approaches. URBN-LMC could be used to consider interconnections with other models, such as bridging the gap between building simulations and urban-scaling applications.