| Bottomland hardwoods are one of the most rapidly diminishing wetland ecosystems due to agricultural clearing, development, and reservoir construction. As society has become more aware of the values and functions of wetlands, so has the importance in conservation of these valuable resources. The objective of this study is to compare the accuracy of Remote Sensing/Geographic Information System (GIS) based functional assessment to the field based Hydrogeomorphic (HGM) approach. An accurate Remote Sensing/GIS based functional assessment can be valuable to those interested in wetland management as field work requires greater expense of labor, equipment, and time. Remote sensing models were developed for the Stephen F. Austin Experimental Forest using a combination of soil maps, soil information, QuickBird RTM multispectral satellite imagery, LiDAR derived Digital Elevation Model (DEM), and LiDAR derived Canopy Height Model. Each of the data layers was prepared in raster format and was recoded with low ratings as 1, medium ratings as 2, and high ratings as 3 in terms of its wetland function. A composite raster layer was created through pixel value addition. Then each pixel value total was divided by the highest possible value total to give the ratio pixel value as a Functional Capacity Index (FCI). The FCIs for each modeled function was then compared to the corresponding HGM field measured function FCIs for accuracy assessment.;Use of the developed models is cautioned as the statistical results are mixed. All functions and function averages have significant positive correlations except for the Cycling of Nutrients, Detain Precipitation, and Maintenance of Plant Communities functions. The Export of Organic Carbon function has the highest r value, 0.69 (p < 0.001) but a high RMSE value (0.06) and sampling error percentage (8.82%). The Detain Floodwater function had a moderately high correlation (r = 0.58, p-value < 0.001), but had the highest RMSE value (0.07) and sampling error percentage (10.61%). All functions and function averages, except for the Overall Wetland Average, have corresponding model and field based means that are significantly different. The Overall Wetland Average appears to be the most successful of all the function and function averages due to it having a moderately high correlation (r = 0.44, p-value < 0.001), model and field based means that were not significantly different (t-value = 0.47, p-value = 0.64), and the lowest RMSE value and sampling error percentage (0.01, 1.28% respectively). |
