Research Areas > Wetlands > Development of a GIS Method to Map Riverine Riparian Habitats
Project: Development of a GIS Method to Map Riverine Riparian Habitat
Background and Objectives
Riparian areas serve critical functions for the health of entire watersheds, and represent perhaps the most important habitat for a majority of biodiversity in the arid western United States. To make effective, rational decisions about riparian management, researchers and land managers require cost-effective, accurate, synoptic maps at sufficient spatial scales. However, many regions lack current, comprehensive maps of riparian areas. A variety of approaches have been used for delineating riparian zone boundaries. One approach is to use a distance buffer on a stream network, but distance buffers may not accurately capture the variability of floodplain width. Another approach, interpretation of aerial photography, is costly and time-consuming and may not be practical for creation of regional or statewide maps of riparian areas. In landscapes such as southern California where upland vegetation on shaded slopes is at least as vigorous, or spectrally similar to, riparian vegetation, a common strategy for riparian mapping is to first define the physical or geomorphic boundary of the riparian zone, and then to assess the vegetation characteristics within the riparian zone from aerial photography or digital remote sensing imagery.
This study explored the use of digital elevation model (DEM) to identify a topographic break that is synonymous with the riparian zone boundary. The ultimate goal was to develop and test a two-step approach to riparian mapping that can be effectively implemented at a regional scale to cost-effectively map the extent of riparian areas from headwater streams to broad floodplain valleys. The specific objectives of this study were to 1) develop a geographic information system (GIS)-based methodology to map riverine riparian zone boundaries, and 2) test the methodology in several watersheds in southern California.
Map of the Ventura River watershed illustrating nested subwatershed units generated for each stream order (left). Comparison of original predicted (pink) versus edited (green) 10-m model of the modern floodplain in the lower Ventura River (right). The original model required manual correction with ancillary data sets.
This project was conducted from 2002 to 2007.
For this project, methodology for mapping riparian areas consisted of two major steps; the first step was to map the extent of the riparian zone using DEMs from USGS. The second step was to characterize vegetation canopy cover within the mapped riparian zone using remote sensing imagery.
The first step, mapping the geomorphic boundary or extent of the riparian zone consisted of several sub-steps. These were: 1) DEM processing, 2) field data collection of valley transects for model calibration, 3) model development to predict the riparian zone extent based on field transect, 4) validation of the extent using additional field transects, and 5) editing for contemporary anthropogenic hydrologic alterations on the floodplain. This was done for each of five pilot watersheds in order to produce a customized model for each. It was also done for all pilot watersheds combined to produce a regional model. The intention was that this regional model could be applied across all southern California coastal watersheds without further data collection. In addition, models based on a 30-m DEM were compared to those of the 10-m DEM in the Ventura River watershed. This was done to test the effect of DEM spatial resolution on model accuracy, since 10-m DEMs are currently not available in many areas of California.
Next, two remote sensing imagery products were used to characterize vegetation canopy cover within the modeled riparian zone. Landsat ETM (30-m spatial resolution) and EMERGE (1-m spatial resolution) imagery were used to generate vegetation indices. The vegetation index values were compared to field-measured canopy cover values. Linear regressions were used to scale vegetation index values and to predict canopy cover values at validation locations. The accuracies of the two imagery products were compared at validation sites to assess the cost vs. benefit of using high- versus low-resolution imagery.
Lastly, the DEM-based riparian extent models were used to clip the remotely sensed canopy cover images. The clipped images served as the final riparian maps. The final riparian maps were compared to other riparian vegetation and floodplain datasets for additional validation of both riparian extent and vegetation characteristics.
Significant findings of the study are as follows:
• This study successfully developed a methodology to predict riparian geomorphic extents with the use of either a 10-m or 30-m digital elevation model (DEM). This methodology was used to predict riparian extent in 5 pilot watersheds in southern California using customized GREM models derived from fieldwork conducted in each watershed.
• In general, the Geomorphic Riparian Ecosystem Extent Model (GREM) predicts riparian habitat particularly well in areas with high topographic relief or with narrow valley walls. It does not take into account the impacts of present-day hydrology on modern riparian habitat – and thus has a tendency to overpredict the extent of habitat in wider valleys or in areas that have been altered by anthropogenic modifications to the floodplain.
• A regional model was also developed and calibrated based on the physiographic characteristics of all five pilot watersheds. While the customized models for individual watersheds understandably have lower error rates in predicting riparian geomorphic extent, the regional model can be used to predict riparian extent across southern California watersheds without further necessity for fieldwork and model development, thus making the use of this methodology more cost-effective.
• Comparison of GREM + Landsat mapped riparian habitat versus maps derived from manual interpretation of aerial photography provided an understanding tradeoffs between map accuracy, quality of information provided by the map, and cost. The regional GREM model, in combination with Landsat ETM-derived estimates of vegetative cover, represents a low-cost option for mapping riparian habitat (less than $500 per USGS 7.5 minute quad), approximately one-tenth of the cost of mapping riparian habitat using manual interpretation of aerial photography (approximately $5000 per USGS 7.5 minute quad). However, because of the definition of “riparian” employed by the GREM and how riparian extents are predicted, the delineated habitat represents “potential” or “predicted” habitat rather than actual habitat, reducing map accuracy.
• The GREM + Landsat-derived vegetative cover is more useful as a screening tool to coarsely assess riparian habitat on a regional or statewide scale. Managers who require more detailed and accurate information about the riparian habitat (i.e. riparian habitat boundaries for regulatory use planning, local land use planning, or composition of riparian vegetation), and who have adequate funding will be better served by maps derived from field-based approaches and/or manual interpretation of aerial photography.
This project was conducted in collaboration with the Conception Coast Project.
This page was last updated on: