Three categories of digital wetland maps widely available in the United States were used to develop models relating wetlands to DOC: (1) wetlands mapped by the U.S. National Wetlands Inventory (NWI) (2) wetland vegetation cover mapped by the U.S. National Land Cover Dataset (NLCD), and (3) maps of hydric soils. Data extracted from these maps for 27 headwater catchments of the Ontonagon River in northern Michigan, USA were used with DOC concentrations measured in catchment streams to develop stepwise multiple regressions based on wetland area and type. The catchments of the 27 tributaries ranged in area from 2 to 66 km 2 and wetlands constituted 10 to 53% of their area. Although all three databases provided regressions that were highly significant ( p < 0.001), the variance explained was greater for NWI maps ( R 2 = 0.75) than for NLCD ( R 2 = 0.61) or soil maps ( R 2 = 0.60). Wetland–stream relationships were strongest during September 2002, but were significant for nine out of ten dates sampled during subsequent seasons. The individual wetland type most highly correlated ( r > 0.62) with stream DOC concentrations was conifer peatland, represented on the NWI maps as Palustrine Needle-leaved Forest, the NLCD maps as woody wetland, and the soil maps as organic soils. For the NWI dataset, DOC was negatively correlated with area of palustrine emergent wetlands (i.e., sedge meadows and graminoid fens) and bog shrubs, inferring that these wetland types may be sinks for DOC. Because of the different effects of wetland vegetation types on DOC, a GIS data source such as the NWI which depicts those wetland types is superior for predicting landscape contributions to stream DOC concentrations.
Read full abstract