[GRASS5] Re: [GRASSLIST:7895] Re: Hutchinson's Adaptive Alogrithm for sound DEMs?

Helena Mitasova hmitaso at unity.ncsu.edu
Mon Aug 15 10:58:57 EDT 2005

> That would an great leap forward for DEM interpolation in GRASS if we could
> utilise fault lines and watercourses as they are digitised from topo maps,
> i.e. lines not labelled for elevation. As well as ridgelines and
> waterbodies. I'm no programmer, so I don't know if that is hard to 
> implement  and I can't offer any help

We are looking into this right now. We have done some testing of 
watercourses and breaklines that are available for our test area to find 
out which to use to improve our DEMs - the student that
I work with tested 2 different types of USGS stream data, local 
government data (that would be very detailed cadastral maps) and stream 
data used for "improving" NC Flood mapping DEMs derived from lidar 
(these were supposed to be high quality). Surprisingly the average 
distance of these lines from the actual streams measured on ground with 
GPS was 30-70m which does not matter if you are working with 100m res. 
DEM but it is a lot if you are working with DEM that is 10m resolution 
or better. So using any of these data sets would actually introduce an 
unwanted error into the DEM. In some of our preliminary tests the 
streams derived from a DEM interpolated from lidar data without the 
streamlines were more accurate than those derived from the DEM 
interpolated with streamlines and even more accurate than the 
streamlines themselves, but more testing is needed to find out whether 
this is generally true. We are including Tpopogrid in our tests and I 
have already written to the list that we will try to implement whatever 
we will find that improves the DEM.

Anyway I will forward your suggestions to my colleagues at Duke to see 
whether they would like to implement any of the breakline/ridges to make 
the process more automated (you should keep in mind that although you 
see your watercourses drawn as lines they are actually stored as points 
so it is more the issue to hide some of the processing from the user and 
make it more automated, because you do have points and you adjust their 
elevation as you go through an iterative interpolation process)

I am totally swamped by work at this moment so I can hardly keep up with 
emails, but if somebody wants to work on this, I will be happy to assist
with the implementation - my projects focus on lidar data where the 
issues are quite different (e.g. ridges are very sharp directly from the 
data so you don't really want to mess it up with some other data that 
may be old or prone to errors), much much bigger priority for me is to 
get the vector modules support our large data sets,


More information about the grass-dev mailing list