Using Functional Assessment and Mapping Tools to Evaluate Headwater Slope Wetlands in Coastal Alabama

Abstract

Headwater slope wetlands are a ubiquitous forested wetland type located at the headwaters of coastal streams in the southeastern U.S. Coastal Plain. There is concern that past and current coastal land use / land cover (LULC) change may reduce the capacity for these wetlands to provide important functions (e.g., habitat, water quality improvement, and flood attenuation). To investigate this, 74 headwater wetlands in coastal Alabama (i.e., Mobile and Baldwin County) were assessed for important functional attributes (forest structure, soils, and hydrology) represented by various ecological measures. These data were compared to LULC data (i.e., % forest, urban and agriculture) from each wetland’s catchment over a range of surrounding landscapes typical of the Alabama coast. Wetland attributes were measured using a regionally specific rapid assessment model, the Hydrogeomorphic Approach (HGM) for the functional assessment of headwater slope wetlands in the coastal plain region of Mississippi and Alabama. Significant relationships between wetland shrub cover and agricultural and urban land use suggests LULC change may increase midstories densities. Urban land use was additionally related to increased herbaceous understory coverage and soil dewatering, as well as reductions in soil organic matter content. Despite some significant relationships and notable trends, urban and agriculture were not highly correlated with several other field measurements, suggesting other landscape factors are important for determining the functional capacity of these wetlands. Headwater wetlands can be difficult to map because of their tendency to transition gradually into uplands on the landscape. For the second part of this study, we evaluated the Wetland Intrinsic Potential (WIP) tool and its use of multi-scale topographic indices, hydrologic proxies, and random forest procedures that contribute to ‘cryptic’ wetland detection in the Bushy Creek – Dyas Creek watershed, near Bay Minette, Alabama. An initial model was trained and validated on a spatial subset of the watershed to predict headwater wetland presence, absence, and extent. The model was then applied to the remaining spatial extent of the watershed. Overall accuracy for the secondary validation dataset was 92.3%, with wetland omission and commission errors of 14.0% and 4.5%, respectively. Our statistical analyses indicated WIP reliably discerned wetlands from uplands. These findings can be used to infer the applicability and limitations of this method for wetland mapping along the northern Gulf of Mexico and support future models which explore land use/cover and hydrogeomorphic relationships with wetlands. Ultimately, information gained from this thesis study will assist in wetland monitoring efforts to better assess the environmental services provided by coastal drainages along the Gulf coast.