Go through following Web pages. 

http://earthexplorer.usgs.gov/

http://glcfapp.glcf.umd.edu:8080/esdi/index.jsp

http://www2.jpl.nasa.gov/srtm/

Go through this site to get srtm data

http://srtm.csi.cgiar.org/

Hi Govindaraj .R and Balaji Kannan,

I've spent a fair bit of time on all four of the sites that you mention and they seem to point me to bare earth DEMs not DSMs (includes the tops of canopies).  Perhaps I'm missing something obvious?

Hi Thomas

Yes SRTM provide bare earth DEMs only. if you want to get DSM you to process LIDAR data for that particular area of interest and add the heights of the surface and get DSMs

To summarize I followed the advice of Dean Gesch at the USGS which is also mirrored in the Kellndorfer et al. 2004 paper.  The SRTM data is the DSM.  I tried following the Kellndorfer approach and had mixed results. I took the DSM from the SRTM and then subtracted a NED DEM to try to produce a CHM (canopy height model). The results looked promising in some areas, but it was also evident that topography had a huge effect on the accuracy and needed to be dealt with in a comprehensive manner. Miliaresis 2008 http://www.mdpi.com/1424-8220/8/5/3134/htm applied a series of aspect/slope/land cover corrections and got decent results, and I attempted some of this using slope and aspect, but never quite followed through to the full extent.  I've got some concerns that using land cover to improve the CHM prediction and then using the CHM to predict land cover may be a bit circular, however.