Filling Voids: SRTM 30m DEM Africa

PeterWilson · ‎10-27-2014

I've recently downloaded SRTM 30m DEM tiles from the USGS Earth Explorer website. The current version of the SRTM 30m DEM has numerous voids that need to be filled. I'd like to hear from the GIS community on the most appropriate methods that can be used to fill the voids without adding addtional systematic errors to the DEM.

I was considering resampling the SRTM 90m DEM using natural neighbours to 30m and using the following to fill the voids within the raw SRTM 30m DEM. Would the following yeild better results than simply filling the voids from the surrounding cells. I'm concerned that some of the voids are pretty large simply filling the voids from the surrounding cells would not represent the actual topographic features correctly.

If anyone has gone through a similar exercise, please would they mind explaining the functions\tools that they have used within ArcGIS to accomplish the following?

Regards

Peter Wilson

DanPatterson_Retired · ‎10-28-2014

Peter... Several strategies are list in this link I have used the process of select by attributes to select area of a threshold area, followed by a regiongroup and then a nibble... Bill Huber provided a description in this thread but unfortunately his explanations have not been migrated as of yet but it worth bookmarking.. So in short, I have used various filters (3x3, 5x5, 7x7 etc) to smooth out the inputs and get rid of small area, or the above strategy for slightly larger areas.

To elaborate on the case of SRTM elevation, if you know the possible range of elevations within the area and you examine the histogram of the data, you will probably find anomalously large values which reflect errors, attributable to several causes. A select by attributes on those area above a comfortable threshold should mean that you have identified as errors, a regiongroup will then identify these as zones, which can then be nibbled using the values from the original surrounding elevations of the input SRTM dataset. I am doing this by memory so any errors are mine...good luck

View solution in original post

DanPatterson_Retired · ‎10-28-2014

Peter... Several strategies are list in this link I have used the process of select by attributes to select area of a threshold area, followed by a regiongroup and then a nibble... Bill Huber provided a description in this thread but unfortunately his explanations have not been migrated as of yet but it worth bookmarking.. So in short, I have used various filters (3x3, 5x5, 7x7 etc) to smooth out the inputs and get rid of small area, or the above strategy for slightly larger areas.

To elaborate on the case of SRTM elevation, if you know the possible range of elevations within the area and you examine the histogram of the data, you will probably find anomalously large values which reflect errors, attributable to several causes. A select by attributes on those area above a comfortable threshold should mean that you have identified as errors, a regiongroup will then identify these as zones, which can then be nibbled using the values from the original surrounding elevations of the input SRTM dataset. I am doing this by memory so any errors are mine...good luck

PeterWilson · ‎10-30-2014

Hi Dan

Thanks for the following help. used focal statistics for the interim, but going to develop a python script based on the NGIA Delta Surface Fill Method using the SRTM 90m DEM V4.2.

Regards