<p>Availability of high resolution population distribution data, independent of the administrative units in which demographic statistics are collected, is a real necessity in many fields: risk evaluation due to earthquakes, flooding or fires, to name just a few, integration between socio-demographic and environmental or geographical information collected in different formats, policy design for the provision public services, such as health, education or public transport, or mobility studies in urban areas or metropolitan regions. Because of this, the literature has explored various methods of population downscaling, collected at communality or census tract level, into smaller areas; typically urban polygons from high resolution topographic maps or land use/land cover databases, or grid cells, allowing the elaboration of raster population layers. A common feature of all these methods is that they do not incorporate building height. In this way, downscaling methods don´t distinguish between the urban sprawl type of settlement, where most of the houses are detached or semi-detached, and compact cities with high buildings. This paper examines error reduction in downscaling census tract population into 1×1 km and 1 ha grids, when we add the third dimension, building height from LIDAR remote sensing data. Algorithms used are simple, and based on areal weighting with or without auxiliary land use/land cover information, since our focus is not in fine turning algorithms, but in measuring improvements due to the missing dimension: building height. Our results indicate that improvements are noticeable. They are comparable to the ones obtained when we move from binary dasymetric methods to more general models combining densities for different land use/land cover types. Hence, adding the third dimension to population downscaling algorithms seems worth pursuing.</p>
Climate forcing reconstructions for use in PMIP simulations of the last millennium (v1.0)
