Planning models require accurate base-year floorspace data to properly allocate/predict activities and study the interactions between different land uses. Currently, most planning models estimate base-year floorspace data according to limited population/employment data provided by the census and in general their accuracy is unknown. In this study, building information is extracted from LiDAR data using a free LiDAR classification software. In addition, through a novel machine learning approach, the extracted buildings are further enhanced by systematically considering a set of LiDAR features in the classification process. The accuracy of building information extracted from LiDAR data and the geographic building footprint layer are then examined and validated through a field survey. It is found that LiDAR data can provide building height, footprint and, therefore, floorspace estimates with a good accuracy. Furthermore, two base-year floorspace estimation methods, one is based on the LiDAR data and the other is based on census data, are compared for sample zones and distinguished by land-use category. The results of this study show that the traditional census-based approach may be very unreliable in estimating base-year floorspace. Comparisons reveal differences as high as 37% for the Residential category. The errors are even higher for the non-residential categories, with average absolute percent errors ranging from 39% for the Office floorspace to 190% for the Accommodation and Recreation. Overall, the results obtained from this study indicate that the traditional census-based approach is very unreliable and inaccurate for modelers/planners to prepare their base-year floorspace, and therefore suggest that LiDAR data be used as a powerful add-on for planning models.
Zircon - This is a contributing Drupal Theme Design by WeebPal.
