Mapping ephemeral stream networks in desert environments using very-high-spatial-resolution multispectral remote sensing

• A new algorithm integrates knowledge about desert landscape characteristics.
• The algorithm mapped 900% more streams than those in the NHD dataset.
• The highest channel extraction accuracy was 92%.
• The algorithm captured detailed stream channels.
• The algorithm underestimated streams obscured by bright soil and sparse vegetation.

Mapping of ephemeral streams in desert environments is crucial to understanding the impacts to hydrologic and ecosystem functions caused by land-use changes. Available mapping methods at the watershed-scale typically underestimate total channel length and the size of channel networks. Although remote sensing is effective for obtaining information on large areas, conventional techniques are often ineffective or cost-prohibitive for complex stream networks in expansive desert regions. Using very-high-spatial-resolution imagery, we developed a new algorithm to map desert ephemeral streams in the southwestern U.S., where utility-scale solar energy development is altering the landscape. Knowledge about landscape features such as shrubs and desert pavement and their spatial arrangement was integrated into the algorithm using spectral transformation and spatial statistical operations. The algorithm extracted ephemeral stream lengths approximately 900% greater than those identified in the National Hydrography Dataset. The accuracy in mapping channel areas and centerlines was as high as 92% and 91%, respectively. Although the algorithm captured detailed stream channels, it often underestimated channels obscured by bright soils and sparse vegetation. Although further improvement is warranted, the algorithm provides an effective means of obtaining detailed information about ephemeral streams, which could make a significant contribution toward improving the hydrological modeling of desert environments.