Traditionally, snow cover has been seen as an obstacle to landcover classification and impervious surface detection based on remote sensing. However, snow cover increases the spectral contrast between impermeable surfaces and other land-use types. This study evaluated the impact of snow cover on the accuracy of impervious surface detection based on Landsat optical imagery. Data from six locations worldwide (three urban, three rural; each 30 km × 30 km) provided the input to a classification of summer (snow-free) and winter (snow-covered) data. This found that for most locations, the overall accuracy of the classification of impervious surface detection was 91.0%–94.0% (kappa 81.0%–93.0%), which was only 1.0%–3.0% less than for summer (92.0%–98.0%, kappa 86.0%–95.0%). Moreover, this difference was not statistically significant. Similarly, when summer and winter observations are processed together – as a single, bi-temporal data set – the inclusion of snow cover does not decrease the classification accuracy. We conclude that snow-covered imagery is useful for impervious surface detection and has no significant impact on classification accuracy for impervious surfaces.
Kotarba A.Z., Impact of snow cover on impervious surface detection. International Journal of Remote Sensing,