Considering plant phenology for improving the accuracy of urban impervious surface mapping in a subtropical climate regions

Abstract

We explored the effect of plant phenology on the accuracy of urban impervious surface mapping in a subtropical climate. To compare mapping accuracies, we estimated impervious surface at the subpixel level using six Landsat Thematic Mapper (TM) images of Austin, TX, USA, obtained in different seasons. In the study area, winter images produced more accurate impervious surface maps than the images taken in other seasons. Root mean square errors (RMSEs) were 16.2% and 16.3% for February and March images, respectively. Growing season images had intermittent level accuracies in impervious surface estimation. RMSEs were 16.2%, 16.8% and 17.0% for March, June and September, respectively. An autumn image had the lowest estimation accuracy (RMSE of 17.8% for the November image). This finding is contrary to the previous studies conducted in temperate climates. We identified two major factors that improve the accuracy of impervious surface maps generated from winter images in a subtropical climate. First, many evergreen plants maintain spectral contrast between impervious surface and vegetated cover during winter in a subtropical climate. Because healthy vegetation has a distinct spectral profile, the presence of green leaves helps separate impervious surface from vegetated cover. Second, impervious surface underneath a deciduous tree canopy is more easily detectable in winter when the trees drop leaves. Based on these findings, we recommend that a winter remote-sensing image should be considered for impervious surface mapping in a subtropical climate.