Verification of X-band SRTM DEM data quality in New Zealand

The quality of 20 Shuttle Radar Topography Mis- sion DEM tiles of New Zealand supplied by DLR has been characterised in terms of the altimetric and planimetric product accuracy specifications. Using precise survey-grade elevation points and a land cover classification, we show that there is strong dependence of the X-SAR SRTM error on the land cover class, but no significant evidence of a relationship between the local estimate of error and the supplied height error map (HEM). Using a grasslands land cover class, we show that the absolute vertical accuracy of the X-SAR SRTM DEM falls within the product specifications, but the relative vertical error exceeds the specification. For the grasslands land cover class, we found no evidence of bias in the X-SAR SRTM DEM error. A regression of SRTM error against reference orthometric height and the HEM shows that there is strong evidence of a dependence of the SRTM error with altitude, and no evidence of a dependence on the supplied HEM value. We use detailed field knowledge and land form characteristics to show that there are advantages and disadvantages to using the SRTM DEM for landform analysis when compared to a conventional DEM generated from contours commonly used for this application in New Zealand.