Multiple regression model for estimating vertical characteristics of built-up areas at 100 m resolution from open and global Digital Elevation Models.

Detailed and spatially consistent information on the three-dimensional characteristics of built-up areas is a key element in monitoring global urban development. We construct a regression-based method for the estimation of the vertical components of built-up areas in 100 m resolution, capturing the average height and the standard deviation of height of buildings in relation to the open spaces around them. We used open and globally available Digital Elevation Models (DEMs) fused with ancillary remote-sensing products as the model inputs, and tested a combination of radar-based and optical-based DEMs as the source of building height information. We develop and test our model on selected cities: Albuquerque, Beirut, London, Philadelphia, San Francisco and Toronto. Results of our model show realistic spatial patterns and low error for tested areas (RMSE 0.95 and R ² 0.62 for Average Gross Building Height component). We demonstrated feasibility of estimating vertical components of built-up areas in 100 m resolution using a method developed under the basic assumptions of: availability of open data, low cost of the analyses conducted, and transparency and reproducibility of the results obtained.