CONCLUSION

An intensity ratio method developed for detecting wet snow in repeat pass SAR images of alpine areas has been tested on ERS images of a non-alpine basin, the Spey (Scotland). Intensity ratios derived from an ERS descending pass exhibit erratic behaviour when com-pared with coincident field observations including snow depth, wetness and elevation. For the intensity images of the Spey, a threshold of -2 dB was identified as consid-erably more suitable than the -3 dB previously used in alpine areas, for distinguishing areas of wet snow cover from areas of no or dry snow. The -2 dB threshold was applied to intensity ratio images of the whole basin. It detected a band of wet snow at mid elevations in the descending image and a much larger area of wet snow at mid to high elevations in an ascending image taken one and a half days later. Both results accord with field and meteorological observations.

These results demonstrate that given limited field meas-urements of snow elevation, depth, extent and wetness it is possible to identify a threshold for wet snow detection in a non-alpine basin. However, we have only consid-ered one non-alpine basin during a year of poor snow cover. The snow and weather conditions occurring at the time of image acquisition and the range of slopes cov-ered by field sites limit general applicability of the re-sults. In particular, we do not know how site and time dependent the detection threshold is.

It is not possible to answer this important question with-out independent methods of verifying wet snow cover. An alternative to further field campaigns is to compare the wet SCA detected in SAR imagery with the SCA detected in coincident cloud free optical imagery, where available. Work on this is ongoing. Similar analysis is planned for the Tjaktjajaure basin where the snow cov-ered area is greater than in Scotland and less transient. Hopefully, the method should prove more viable for providing wet snow covered area measurements for in-put to runoff models. However, no field measurements are available for this basin.

The area of missing coverage due to geometric effects was also analysed. Significant amounts of missing cov-erage were shown to occur in ERS images of the Spey and Tjaktjajaure basins (between 11 and 22% of imaged basin area). The predominant cause was foreshortening though layover was also significant. The amount of missing coverage is dependent on the range distribution of relief within the imaged scene, and affects mid to high elevations, where snow is most likely to occur, more than low elevations. This is potentially a serious problem. While missing coverage can be reduced by combining descending and ascending passes, in non-alpine areas the time lag between passes often invali-dates any wet snow cover inferences based on such a combination. As illustrated by the change in snow con-ditions between passes over the Spey basin.

It may be possible to reduce the effect of missing cover-age on wet snow mapping by extrapolating results from areas unaffected by missing coverage to areas of missing coverage with similar elevation and aspect. This is an area of future work. A major problem could be that most areas of a given elevation and aspect are affected by missing coverage. An analysis of the aspect depend-ence as well as the elevation dependence of missing coverage is needed to determine if this is the case.

For future studies a better solution to the problem of missing coverage may be to use data from a SAR with a less steep incidence angle, though the performance of the intensity ratio method for wet snow detection at such incidence angles is unknown. The choice of incidence angles and swaths provided by the SARs onboard Ra-darsat and ENVISAT provide a means of both reducing missing coverage and increasing temporal coverage. It is likely that these systems are much more suited than ERS for monitoring wet snow cover in the higher latitude basins studied here.

Last updated: 13th October, 1998 by Roger Dunham