Introduction   Study area   Method: Photogrammetry   Method: Water depth estimation   Final riverbed surface   Conclusions   References   Links & Acknowledgements

6: Conclusions

This paper has described the methodology used to derive high resolution, 'total' coverage DEMs for a reach of the Waimakariri River. Overall, the use of digital photogrammetry and image analysis to survey shallow gravel-bed rivers shows great promise, although some problems were encountered.

The low precision of DEM elevations as compared to the best theoretical precision (SDE = 7 cm) is likely to be caused by the relatively poor stereo-matching observed during the automated DEM collection. Where matching is unsuccessful, an elevation value is calculated by interpolating from surrounding, successful, stereo-matched points. The low relief typical of braided riverbeds means that interpolation is not expected to produce large errors, but would explain deviations from surveyed point elevations. The reason for the poor stereo-matching performance is less clear. The most likely explanation is a lack of texture in many gravel areas of the riverbed, given the scale of photographs (1:5000) and scanning resolution used. This is currently being tested using 1:4000 and 1:8000 imagery flown in February 2000.

The second issue is the positive systematic error discovered in all elevations. The maximum predictable depth obtainable using the empirial colour-depth method for estimating water depth means that is unlikely that the systematic bias in wetted areas can be reduced, and much be considered one disadvantage of the method. The positive ME in dry areas is being investigated. If it is found to be homogeneous across the riverbed then it can (in theory) simply be subtracted from all photogrammetric elevations. If the ME shows spatial variability in dry areas, this simple 'fix' cannot be applied.

Ultimately, these issues demonstrate the trade-off that is made when selecting a large-scale 'remote' survey over a conventional survey technique. Manually-surveyed cross-sections, with coarse temporal and spatial resolutions, have been shown to produce large volumetric errors (Westaway et al., 2000). Techniques such as digital photogrammetry and airborne laser survey (Hicks et al., this volume) have the potential to greatly improve the spatial and temporal resolution of riverbed monitoring programmes, yet have greater uncertainty associated with each surveyed elevation value. The debate between vertical precision and spatial extent is likely to dominate future discussions of remote survey. However, with the current pace of technological development, the difference in vertical precision between conventional and remote surveys can only reduce.

Introduction   Study area   Method: Photogrammetry   Method: Water depth estimation   Final riverbed surface   Conclusions   References   Links & Acknowledgements