Developing Inventory of Riparian Buffer Cover in L'Anguille River Watershed using Geospatial Technologies

摘要

A watershed level analysis was conducted in East-central Arkansas along L'Anguille River and its major tributaries to assess vegetation composition in riparian buffer cover using two baselines - stream centerline and stream bank. The input data for the study comprised 1:24,000 scale National Hydrography Dataset Plus (NHDPIus) stream centerline, one meter resolution natural color digital ortho quarter quad (DOQQ) imagery, one meter resolution color infrared (CIR) DOQQ imagery, and 2006 land use-land cover (LULC) imagery. Since both the approaches required accurate stream centerline data, manual editing of NHDPIus stream centerline was performed to match streams with their actual locations. The edited stream location was verified by overlaying it on one meter color DOQQ imagery. The first approach was based on identifying the boundary of river/stream banks. The pixels identifying the boundary of river and stream banks were located on natural color and CIR DOQQ imagery by running an automated search in a direction perpendicular to both left and right side of nodal points for each segment of stream centerline. The search algorithm utilized differences in digital number (DN) values for water and land to identify pixels that marked the boundary of landarea immediately adjacent to the water. These identified pixels were joined together to create the boundary of river/stream banks. Adjacent to this boundary of river/stream banks, a 45-m (approx. 148 ft) wide land area, parallel to the banks and consisting of forests and grasslands, was considered "buffer" for mitigating water quality. This approach was compared with previously cited approach in the literature, i.e. buffer creation around stream centerline. The percentage of water pixels present in thevegetation composition analysis performed in buffer area created by using boundary of river/stream banks and stream centerline, respectively, was used as a criteria for measuring the accuracy of both the approaches. The buffer created using boundary of river/stream banks had 27.7% less water pixels compared to stream centerline approach using the natural color imagery. The automated search approach was evaluated for its use on natural and CIR imagery. Several streambank segments were manually delineatedand the error was calculated by measuring the distance between automatic and manual streambank delineations using both imagery types. It was observed that CIR imagery performed better with a mean error of 2.21 m (RMSE 3.07 m) while natural color imageryhad a mean error of 2.55 m (RMSE 3.41 m). Overall, the results show that delineation of boundary of river/stream banks using geospatial techniques and its subsequent use for buffer analysis enhances the accuracy of determining the presence of specifiedsize (width) riparian buffer along river and streams.