Comparison of 2D & 3D parameter-based models in urban fine dust distribution modelling

Abstract

In the present study two Land Use Regression Models for the estimation of urban fine dust distribution were established and compared. The first model used 2D parameters derived from an Open Street Map project data (OSM) and the second model used 3D parameters derived from a CityGML-based 3D city model. Both models predict fine-dust concentrations by using urban morphological (2D resp. 3D) and additional semantic parameters. The models were applied to a 2 km2 study area in Berlin, Germany. The 2D-LUR model explained 84 % of the variance of TNC for the full data set with root mean square error (RMSE) of 3284 cm−3 while the 3D-LUR explained 79 % of the variance with an RMSE of 3534 cm−3. Both models are capable to depict the spatial variation of TNC across the study area and showed relatively similar deviation from the measured TNC. The 3D-LUR needed less parameters than the 2D-LUR model. Furthermore, the semantic parameters (e.g. streets type) played a significant role in both models.