Downsample Grids by Averaging

Average to new projection Example from El Hiero, Canary Islands, Spain

• Create high resolution (2.5 m) UTM DEM from lidar point cloud (Figure 1), or use the DTM and DSM from the national mapping agency

• Resample UTM to 1"x1" geographic, by averaging all points in the 2.5 m DEM that fall within a single 1" pixel.  Do this as both pixel-is-point and pixel-is-area.  On average about 135 points are averaged.

• Perform elevation and slope moments on the grids.  The slopes will be greatly different for the 1" DEM compared to the 2.5 m or 5 m averaged grids (Table 2).

• These results have not changed the vertical datums, which does not effect the slope results.

Figure 1.  2.5 m UTM, created from lidar point cloud.  This is a DSM, taking the highest return in each pixel.  There are some data gaps in the point cloud.  This is relatively low resolution lidar, and 2.5 m might be the best resolution

Difference map between two 1" resamplings of the 2.5 m DEM.  Each point in the 2.5 m UTM DSM was mapped to the correct pixel in the 1x1" grid, one using the pixel-is-point (which matches Copernicus, ASTER, and SRTM), and the other pixel-is-area (which matches ALOS).  Because there is only 1/4 overall in the pixel areas between the two mappings, the DEMs are not the same.  In order to create the difference map, one of the DEMs has to use bilinear interpolation to get the corresponding value in the other.  The difference map reflects slope and aspect; in steep areas, the 1/2 pixel offsets have significantly different elevations.

Point density for the number of 2.5 m pixels that went into each 1" pixel.  Except along the edges and data voids in the lidar data, each pixel in the 1" DEM has about 135 points in the 2.5 m DSM. The differences come from voids (mostly water), and the rotation of the UTM grid compared to the geographic grid.

Four DEMs of El Hiero, all at 1"x1". Lidar DSM created directly from the point cloud; Copernicus, and two arc second grids created from a 2.5 m UTM DEM created from the point cloud

Table 1.  Moment report for the 4 DEMs above

Table 2.  Slope moments for El Hiero.  The 2.5 m and 5 m UTM DEMs have significantly different slopes than the 1" grids.  The lidar DSM is slightly different than the one used in Table 1 (masking the coast, and whether to take the highest point in the pixel, or the 95th percentile to deal with potential noise)