Watershed Vulnerability Predictions for the Ozarks Using Landscape Models

Abstract

Forty-six broad-scale landscape metrics derived from commonly used landscape metrics were used to develop potential indicators of total phosphorus (TP) concentration, total ammonia (TA) concentration, and Escherichia coli bacteria count among 244 sub-watersheds of the Upper White River (Ozark Mountains, USA). Indicator models were developed by correlating field-based water quality measurements and contemporaneous remote-sensing-based ecological metrics using partial least squares (PLS) analyses. The TP PLS model resulted in one significant factor explaining 91% of the variability in surface water TP concentrations. Among the 18 contributing landscape model variables for the TP PLS model, the proportions of a sub-watershed that are barren and in human use were key indicators of water chemistry in the associated sub-watersheds. The increased presence and reduced fragmentation of forested areas are negatively correlated with TP concentrations in associated sub-watersheds, particularly within close proximity to rivers and streams. The TA PLS model resulted in one significant factor explaining 93% of the variability in surface water TA concentrations. The eight contributing landscape model variables for the TA PLS model were among the same forest and urban metrics for the TP model, with a similar spatial gradient trend in relationship to distance from streams and rivers within a sub-watershed. The E. coli PLS model resulted in two significant factors explaining 99.7% of the variability in E. coli cell count. The 17 contributing landscape model variables for the E. coli PLS model were similar to the TP and TA models. The integration of model results demonstrates that forest, riparian, and urban attributes of sub-watersheds affect all three models. The results provide watershed managers in the Ozark Mountains with a broad-scale vulnerability prediction tool, focusing on TP, TA, and E. coli, and are being used to prioritize and evaluate monitoring and restoration efforts in the vicinity of the White River, a major tributary to the Mississippi River and Gulf of Mexico.

Lake nutrient stoichiometry is less predictable than nutrient concentrations at regional and sub‐continental scales

“…In addition, multiple response variables can be included in the same PLS model. Generally, VIP scores above 1.0 suggest that a variable is important (Fraterrigo andDowning 2008, Mehmood et al 2012), but variables with scores slightly below 1.0 might be marginally important (Lopez et al 2008, Monk et al 2013. To assess the sensitivity of our inferences to ignoring this spatial structure, we compared PLS regression to a mixed modeling approach that explicitly accounted for the hierarchical data structure.…”

Geographically weighted regression to measure spatial variations in correlations between water pollution versus land use in a coastal watershed

“…Forest is mostly associated with low water pollution in various studies of watersheds around the world (Osborne and Kovacic, 1993;Sliva and Williams, 2001;Novotny, 2002;Bahar et al, 2008;Lopez et al, 2008;Huang et al, 2014). Based on OLS, some studies find no significant correlations between forest land and water pollution (Sponseller et al, 2001;Tran et al, 2010;Huang et al, 2013a).…”

“…Generally, built-up land and agricultural land have significant positive correlations with water pollution, which are associated with point or non-point source pollution (Johnson et al, 1997;Sliva and Williams, 2001;Fedorko et al, 2005;Mehaffey et al, 2005;Stutter et al, 2007;Tu et al, 2007;Bahar et al, 2008;Tran et al, 2010;Pratt and Chang, 2012;Yang, 2012). Woodland is significantly negatively correlated with nutrients, due to the general understanding that forests can absorb nutrients (Osborne and Kovacic, 1993;Novotny, 2002;Galbraith and Burns, 2007;Bahar et al, 2008;Lopez et al, 2008). However, the relationships between land use and water pollution can be inconsistent across time and space.…”

Sediment and total phosphorous contributors in Rock River watershed

“…Various studies have linked stream pollutants to land use variables using process-based hydrological models (Jha et al, 2010;Kirsch et al, 2002;Ullrich and Volk, 2009) or statistical methods (Lenat and Crawford, 1994;Liu et al, 2009;Lopez et al, 2008;Mehaffey et al, 2005;Nash et al, 2009). Process based hydrologic models have been successfully used to characterize watershed processes and sources of stream pollutants; however these models require detailed input data, which may not be available for some areas.…”

“…For instance, Kirsch et al (2002) showed the difficulty of calibrating a SWAT model for Rock River basin in Wisconsin, due to limited data for numerous lakes, reservoirs and dams in the basin. Using statistical regression methods, agricultural land was found to be a major contributor to nutrients in Oregon, New York, and the Missouri-Arkansas Ozark region (Lopez et al, 2008;Mehaffey et al, 2005;Nash et al, 2009). In addition, Liu et al (2009) found that urban and agricultural lands contribute many pollutants (such as TP, bacteria, metals, low dissolved oxygen, alkalinity and conductivity) to Wisconsin streams using similar statistical methods.…”