- The coverage is patchy for many of the datasets.
- All methods produce some kind of N-S striping in the difference images.
- Some methods create very rough surfaces. This is especially evident in the SETSM images
- The differences between the LiDAR and the StereoDEM is not systematic. Though I have done the first cut analysis to look at the vertical offset, there is obviously some horizontal offset as well. This will be an avenue of continued work. Though I had hoped that each type of dataset, regardless of location, would group around a particular vertical offset, this is not the case. The fact that the distributions shown below have both different mean values and different standard deviations suggests that the lateral offset is potentially complicating things. Note that because the different datasets have different extents and resolutions (hence numbers of pixels), the heights of the peaks are variable and not any an indicator of quality or consistency.
- What is important to note is that any given dataset has a very low standard deviation for the difference between the LiDAR and the StereoDEM. This suggests that at a particular location, if you knew the true elevation you could apply a vertical offset and be roughly within a few meters of the true elevation. Not bad!
[meters]