I am working in flatlands of the Southamerican Chaco and have found that Aster DEM seems to coincide higher elevations (1 to 2 m higher) with the presence of forests, and native grasslands with lower elevations. Could it be that ASTER is showing higher altitude in these areas because of the tree canopy or is it really reflecting higher topographical areas?
ASTER DEM products have some sensitivity to tree canopy cover. Validation results (spot comparisons of ASTER DEM with canopy-penetrating methods) show this sensitivity. Globally, vertical accuracy was found to be under a meter on average. Over forested areas, vertical accuracy was about 7 m vertical on average. This may be a problem if you are dealing with areas of dense and tall tree cover.
You can find general info about the ASTER GDEM here:
http://asterweb.jpl.nasa.gov/gdem.asp
You may also find the summary of the GDEM version 2 validation results helpful:
https://lpdaacaster.cr.usgs.gov/GDEM/Summary_GDEM2_validation_report_final.pdf
ASTER digital elevation data are generated using stereo-pair based photogrammetry. The input data for the stereopairs are the nadir-looking VNIR Band 3 and the backward-looking VNIR band 3B. Since both bands are within the visible - near infrared range, they do not penetrate vegetation cover (unlike DEM's generated from radio-wave frequency sensors like SAR, SRTM, etc). Instead, they capture reflectance from the land/tree surface.
You could try checking your ASTER values in the respective regions against local elevation data. SRTM might work as a comparison. However, you would have spatial resampling issues to deal with and the vertical error of SRTM can be as much as an order of magnitude greater than ASTER GDEM.
Perhaps, this paper could be useful: http://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XXXIX-B4/275/2012/isprsarchives-XXXIX-B4-275-2012.pdf
Thank you very much György. This is helpful! and the methodology seems quite straightforward to deal with the issue. Best regards.
Thank you Subhajit. I had the same idea first. However, the article shared by György comments the opposite, and the need for correction. I would be very grateful if you could point me into more information about the ASTER DEM and its relation with forest canopy. Please remember I am not a specialist in remote sensing. Your help is very appreciated.
ASTER DEM products have some sensitivity to tree canopy cover. Validation results (spot comparisons of ASTER DEM with canopy-penetrating methods) show this sensitivity. Globally, vertical accuracy was found to be under a meter on average. Over forested areas, vertical accuracy was about 7 m vertical on average. This may be a problem if you are dealing with areas of dense and tall tree cover.
You can find general info about the ASTER GDEM here:
http://asterweb.jpl.nasa.gov/gdem.asp
You may also find the summary of the GDEM version 2 validation results helpful:
https://lpdaacaster.cr.usgs.gov/GDEM/Summary_GDEM2_validation_report_final.pdf
ASTER digital elevation data are generated using stereo-pair based photogrammetry. The input data for the stereopairs are the nadir-looking VNIR Band 3 and the backward-looking VNIR band 3B. Since both bands are within the visible - near infrared range, they do not penetrate vegetation cover (unlike DEM's generated from radio-wave frequency sensors like SAR, SRTM, etc). Instead, they capture reflectance from the land/tree surface.
You could try checking your ASTER values in the respective regions against local elevation data. SRTM might work as a comparison. However, you would have spatial resampling issues to deal with and the vertical error of SRTM can be as much as an order of magnitude greater than ASTER GDEM.
ASTER GDEM most definitely is affected by the presence of forest canopy, with elevations in these areas showing a positive bias (the same is true for built up areas, or any areas with above ground features). Please see the following paper for details:
http://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XXXIX-B4/281/2012/isprsarchives-XXXIX-B4-281-2012.pdf
The paper details validation of ASTER GDEM against 18,000 high quality geodetic control points, as well as comparison against SRTM and a bare earth elevations model of the United States (the National Elevation Dataset). The results show that ASTER GDEM is biased upward in forested areas (by about 3 meters on average), and ASTER GDEM has elevations higher in the forest canopy than SRTM does.
Dear Benjamin, Dean and Samuel. Thank you for your well documented answers to my question. I have definitely learned more about this issue with your help. Considering this, ASTER can be very misleading in identifying higher/lower areas in this extremely flat region with a landscape of native grasslands and very dense forest islets. Perhaps, as György shares, a correction is quite well justified. Thanks again!
ASTER GDEM accuracy could range in those values (1-2 m) as you mentioned. We conducted accuracy assessment of GDEM 2 (and also SRTM) for tropical montane forest by using LiDAR as the reference dataset, and we found out that the GDEM2 was higher from the canopy (LiDAR-DSM) from 0.7 to 4.7 m (SD=16.9~18.7m; RMSE= 17.0~19.3 m).
Conference Paper Accuracy assessment of global topographic data (SRTM & ASTER...
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing