Multi-spectral Canopy Analysis for Detection of Subcanopy Headwater Streams

Headwater streams make up a large portion of global streams by length. They play an important role in the water cycle and their related riparian zone is a diverse yet delicate ecosystem. Therefore, understanding headwater stream distribution and dynamics is of value for researchers, industry, and policy makers. There are well established methods for remote mapping of streams, but established methods are inadequate where streams are occluded by dense canopy. This work investigates the use of spring and summer leaf-on multi-spectral satellite data for detecting temperate forest canopy reflectance patterns that correlate with the presence of surface-water in mountainous forest channels. The analysis tests the use of PlanetScope and Worldview3 separately for differentiation of field-validated: 1) intermittent and perennial channels, and 2) ephemeral and non-stream channels in North Carolina, United States. Initial steps include unsupervised masking of shadow pixels and canopy segmentation into classes that align with deciduous and coniferous distribution. A dense network of flow accumulation lines is extracted to the catchment boundaries and separated into above and below field validated stream head segments. The channel network is further split into 50-meter segments to aid in understanding the potential effects of land surface aspect and elevation. Initial results show distinct average coniferous NDVI and SWIR reflectance values within 50 meters of intermittent and perennial channels when compared to the above stream head channels (two-sample t,56 degrees of freedom, p = 0.02) and hill slope regions outside of the riparian zone (two-sample t,56 degrees of freedom, p = 0.001). It is recognized that many factors contribute to reflectance variation, yet the results show promise for mapping occluded headwater streams. Work is underway to control for landscape factors in the canopy reflectance and to test a resulting channel classification model on an expanded area.