Closed sgrieve closed 6 years ago
Answers:
Starting to think about the removal of small areas. There are a lot of small islands on the koppen dataset as well as some very small areas inland.
In order to get big rivers and sensible results, we need to remove small patches so that we don't try to extract a river where there is not enough data. As we will be setting an area threshold of ~300 pixels (an area of ~0.3 km^2) we will need our minimum area to be significantly larger than this, so that we can capture several first order drainages per climate zone sub polygon. As a first pass at this threshold I have selected an area of 1000 Km^2, which corresponds to a square area of 1000x1000 pixels, which is generally a reasonable size of DEM to load into LSD and get sane results out.
For question 4, there is a parameter in the driver file test_drainage_boundaries
which should remove basins that intersect edges.
The final 2 parameters which need to be finalised are the minimum basin area and threshold contributing area. Both are reported in pixels. The key is to balance minimum basin size with spatial abundance of measurements, particularly for areas with fragmented climate sub zones.
I will send an email to collaborators soon to finalise their values.
Email sent to discuss parameter choices.
Final parameters chosen:
threshold_contributing_pixels: 25000
minimum_basin_size_pixels: 50000