Digital Surface Models and voxel grids for solar irradiance estimation

This dataset is supplementary material for the paper GPU-based solar irradiance estimation over Digital Surface Models using structurally lossless viewshed compression by Niko Lukač and Borut Žalik, University of Maribor. Version: 1.0 Description The dataset contains Digital Surface Model (DSM) heightmaps and the corresponding voxelized binary output with per-voxel annual solar potential estimated by the proposed method. Heightmaps format (*_heightmap.tiff) All heightmap files are single-band GeoTIFF (.tiff) rasters with 32-bit floating-point (float32) elevation values per pixel, at a spatial resolution of 1 m per pixel in both horizontal directions. Pixel values represent surface elevation in meters. Voxel binary file format (*_voxels.bin) Each voxel file is a raw little-endian binary dump with the following layout: Header (8 bytes): two 32-bit signed integers (int32): num_total - the total number of voxel records in the file, followed by num_top - the number of top-surface voxels. Body (num_total × 16 bytes): num_total consecutive voxel records, each consisting of four 32-bit floats (float32): x, y, z, value. Here x and y are the voxel's grid coordinates (column and row of the parent DSM cell, in meters at 1 m resolution), z is the voxel's elevation in meters, and value is the estimated annual solar potential (in Wh) of the voxel. The first num_top records correspond to top-surface voxels, while the remaining num_total − num_top records are the generated vertical wall voxels (which share the x, y grid position of their parent surface cell). The included voxels_vis.py Python script parses this format and renders the voxels as a 3D visualization colored by the solar potential values. Synthetic DSMs The six synthetic DSMs, data1 to data6, span an order of magnitude in voxel count. They were generated by combining low-frequency Perlin-noise terrain with a controlled population of cuboidal buildings of increasing density and height variance. File Size [m] Voxel count |V| synthetic_data1_dsm_heightmap.tiff / synthetic_data1_voxels.bin 512 × 512 389,485 synthetic_data2_dsm_heightmap.tiff / synthetic_data2_voxels.bin 512 × 512 645,573 synthetic_data3_dsm_heightmap.tiff / synthetic_data3_voxels.bin 512 × 512 878,270 synthetic_data4_dsm_heightmap.tiff / synthetic_data4_voxels.bin 1024 × 1024 1,383,851 synthetic_data5_dsm_heightmap.tiff / synthetic_data5_voxels.bin 1024 × 1024 1,687,686 synthetic_data6_dsm_heightmap.tiff / synthetic_data6_voxels.bin 1024 × 1024 1,997,288 LiDAR-derived DSMs The three LiDAR-derived DSMs cover a suburban area in Pekre, Slovenia (46.54°N, 15.59°E), a low-rise urban area in Vaihingen, Germany (48.93°N, 8.96°E), and a dense high-rise area of Lower Manhattan, New York City, USA (40.70°N, 74.01°W). The DSMs were generated using only classified LiDAR points, with gaps filled by inverse distance weighting (IDW) interpolation. File Size [m] Voxel count |V| Wall voxels [% of |V|] lidar_derived_pekre_dsm_heightmap.tiff / lidar_derived_pekre_voxels.bin 366 × 320 189,372 38% lidar_derived_vaihingen_dsm_heightmap.tiff / lidar_derived_vaihingen_voxels.bin 333 × 293 198,230 51% lidar_derived_nyc_dsm_heightmap.tiff / lidar_derived_nyc_voxels.bin 799 × 457 1,522,275 76% The LiDAR-derived DSMs are subsets of larger open datasets, and the original providers are gratefully acknowledged: Pekre, Slovenia - derived from the national airborne LiDAR survey of Slovenia, provided by the Slovenian Environmental Agency (ARSO). Vaihingen, Germany - derived from the Vaihingen airborne LiDAR dataset provided by the German Society for Photogrammetry, Remote Sensing and Geoinformation (DGPF). Lower Manhattan, NYC, USA - derived from open topographic LiDAR data provided by the State of New York (US).