Viers, JH, JL Florsheim, CM Ramirez, JF Quinn, ML Johnson, and B Kozlowicz. 2002. Detecting Patterns of Land Use Disturbance at a Watershed Scale: A Study of the Navarro River Watershed using Hyperspectral Data Analysis Techniques. EOS Trans. AGU, 83(47): B21A-0710.
Abstract: 

Analysis of hyperspectral data is a particularly novel approach to investigation of the relation between anthropogenic and natural disturbances, geomorphic responses, and ecosystem patterns at the watershed scale. During July 2000, hyperspectral imagery was collected for the Navarro basin (820km2) using the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). This NASA sensor covers the spectral wavelength range from 400nm - 2500nm, with spectral sampling of 10nm in 224 contiguous channels, and spatial resolution of 5m. These data are being analyzed for a variety of applications; however, their use for detecting patterns of disturbance within the watershed is intriguing, especially for the identification and delineation of mass wasting sites that deliver sediment to salmon bearing streams. Mass wasting sites were extracted from AVIRIS imagery using image processing techniques such as Minimum Noise Fraction and Tasseled Cap transformations, image segmentation and masking. These geospatial and spectral data were analyzed for the North Fork of the Navarro River, a sub-basin where spawning habitat for threatened coho salmon is effected by accelerated sediment delivery. Additionally, fieldwork verified the spatial position and dimensions of mass-wasting sites identified from aerial photography. A subset of 1066 identified sites was used for assessing feature extraction error from the AVIRIS imagery; the remaining sites were used for model verification. Augmenting these data within GIS, a multivariate analysis incorporated: proximity to salmon bearing streams; hillslope gradient; landslide position; and timber harvesting to identify patterns of disturbance. Preliminary results indicate that AVIRIS imagery can be segmented to identify exposed soil; furthermore, these identified areas are typically lower elevation, moderately steep hillslopes in constricted river valleys and correspond with mapped delivery sites. Hyperspectral data provide a means for the detection of sediment sources over large areas; however, the geometric and atmospheric corrections required to effectively process these data can be onerous. The current work is part of an interdisciplinary study at UC Davis intended to assist land use managers in development of TMDL guidelines in coastal watersheds in northern California. Our research in the North Fork indicates that similar approaches can be used for both inventorying and monitoring of disturbance patterns at the watershed scale.