! OVer- 
terrain 
leared 
re the 
duced 
ffsets. 
sult in 
filling 
across 
Pars to 
much 
y tree 
rences 
in the 
tween 
patchy 
esting 
of tree 
red by 
ble to 
ze and 
ıts and 
>moval 
fs and 
e river 
DEM 
' Cover 
SRTM 
1ced to 
is still 
ded by 
further 
and a 
se two 
/'e been 
logical 
ks and 
ngoing 
Fabric 
:levation (m) 
  
(AHGF,  http;//www.bom.gov.au/water/geofabric/index.shtml, 
commonly known as the ‘Geofabric’), an authoritative and rich 
representation of Australia’s hydrological features. 
The tree height offset map could be used as the basis of a high 
resolution map of forest height and biomass, provided suitably 
robust relationships can be found between height offset and 
actual tree height. Note that the tree height offset derived from 
the SRTM data is not a measure of true tree height, primarily 
because the SRTM radar signal scatters from the woody 
structure within the canopy rather than the upper margin of the 
canopy (Kellndorfer et al., 2004). Preliminary study of the 
estimated offsets suggests the estimated heights are a smaller 
fraction of true height in trees with a conical shape (typically 
conifers) than in trees with a broad canopy (such as most 
Australian eucalypts), presumably due to greater penetration of 
the radar signal into the former canopy type. 
3.2 Prospects for processing of the global SRTM DSM 
The processing method for estimating and removing tree height 
offsets in the SRTM DSM could in principle be applied to the 
entire SRTM dataset. The most significant obstacle to 
attempting this is the need for a consistent global land cover 
map compatible with the SRTM data in both resolution and 
acquisition date. Global cover maps currently exist at 
resolutions of about 300 m (GlobCover) and 500 m (MODIS 
Land Cover). Finer resolution products from Landsat might also 
be possible, and it is also possible that the raw SRTM radar 
product would contain useful information for detecting 
vegetation cover boundaries. 
Independent measures of tree height (suitably modified to 
account for SRTM penetration into the canopy) could be 
incorporated into the height offset estimation process. This 
would be particularly helpful in extensively forested areas 
where there are likely to be spatial variations in tree height that 
cannot be estimated from patch edges. Either direct 
measurements from instruments such as the GLAS laser 
altimeter aboard ICESat or indirect measurements based on 
multispectral imagery could be used. Some promising progress 
in combining those two approaches was reported by Lefsky 
(2010) using GLAS and MODIS data to produce a global tree 
height map. While this is at a relatively coarse resolution it may 
provide sufficient information to support improved offset 
estimates in extensive forest areas; further development of the 
methods could also provide improved information. Combining 
the various data sources under a model-data fusion approach 
may also be possible, yielding a tree map, tree heights and bare- 
earth DEM from a single process. 
3.3 Application to other DSM types 
While some of the aspects of the algorithm are specific to 
SRTM, notably the matching of response to patch edges, much 
of the method is in principle directly applicable to any DSM 
that responds predictably to vegetation cover. A bare-carth 
DEM is usually produced from a DSM by manually driven 
editing and hence tends to be quite expensive. A fully 
automated method, even if it is not as accurate as a manually 
driven method, is attractive where very large areas of DSM need 
to be processed, as with global or continental DSMs like 
SRTM. 
Two obvious candidates for processing using the method 
presented here are the ASTER GDEM and TANDEM-X 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B4, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
products (although the ASTER GDEM response to trees does 
not seem as consistent as the SRTM so may be less amenable to 
treatment). Both are essentially DSMs and would benefit from 
removal of offsets due to tree cover. 
4. CONCLUSIONS 
Removing artefacts from Digital Surface Models is a 
prerequisite for deriving high quality bare-earth Digital 
Elevation Models. The methods described here provide that 
capability, drawing on an appropriately fine resolution 
vegetation (tree) map as supporting data. The resulting bare- 
earth DEM for Australia is now being used for a variety of 
ecological, hydrological and geomorphological applications. 
The methods described here could be applied to the entire 
SRTM near-global DSM, provided a suitable tree cover map 
can be compiled. This would be a valuable step in realising the 
full value of this high resolution, high quality product and 
would complement other efforts such as those by CGIAR to 
provide a void-filled version of the SRTM DSM. 
The method may also be suitable for removing offsets due to 
tree cover in other remotely sensed DSMs such as ASTER 
GDEM and TANDEM-X. Some adaptation to the different 
characteristics of these DSMs would be required. 
5. REFERENCES 
Dowling, T.I, A.M. Read, M.F. Hutchinson, and J.C. Gallant, 
in prep. Drainage enforcement of the 1 second SRTM DEM for 
Australia. 
Farr, T.G., P.A. Rosen, E. Caro, R. Crippen et al. (2007), The 
shuttle radar topographic mission, Reviews of Geophysics, 45, 
RG2004. 
Gallant, J.C., 2011. Adaptive smoothing for noisy DEMs. In: 
Geomorphometry 2011, International Society for 
Geomorphometry. Redlands, California. 
http://www.geomorphometry.org/Gallant2011. 
Gallant, J.C., AM. Read, T.I. Dowling, and J.M. Austin, in 
prep. Removing vegetation offsets from the 1 second SRTM 
DEM for Australia. 
Grohman, G., G. Kroenung, and J. Strebeck, 2006. Filling 
SRTM voids: The delta surface fill method. Photogrammetric 
Engineering and Remote Sensing, 72(3), pp. 213-216. 
Hutchinson, M.F., T. Xu and J.A. Stein, 2011. Recent progress 
in the ANUDEM elevation gridding procedure. In: 
Geomorphometry | 2011, International Society for 
Geomorphometry. Redlands, California. 
http://geomorphometry.org/HutchinsonXu201 1. 
Kellndorfer, J., W. Walker, L. Pierce, C. Dobson, J.A. Fites, C. 
Hunsaker, J. Vona, and M. Clutter, 2004. Vegetation height 
estimation from Shuttle Radar Topography Mission and 
National Elevation Datasets. Remote Sensing of Environment, 
93(3), pp. 339-358. 
Kóthe, R., and M. Bock, 2009. Preprocessing of digital 
elevation models - derived from laser scanning and radar 
interferometry - for terrain analysis in geosciences. In: 
Geomorphometry 2009, Zurich. 
http://www.geomorphometry.org/KoetheBock2009.