Mapping the estuarine ecosystem service of pollutant removal using empirically validated boosted regression tree models. Issue 5 (20th March 2020)

• Record Type: Journal Article
• Authors: Lohrer, Andrew M.; Stephenson, Fabrice; Douglas, Emily J.; Townsend, Michael
• Is Part Of: Ecological applications. Volume 30:Issue 5(2020)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-03-20
• Subjects: boosted regression tree -- denitrification enzyme activity -- ecosystem function -- ecosystem services -- estuary -- intertidal -- mapping -- nutrient removal -- pollution removal -- sediment -- spatial prediction; Ecology -- Periodicals; Environmental protection -- Periodicals; Biology, Economic -- Periodicals; 577.05
• DOI: 10.1002/eap.2105
• Languages: English
• ISSNs: 1051-0761

Abstract

Abstract: Humans rely on the natural environment and benefit from the goods and services provided by natural ecosystems. Quantification and mapping of ecosystem services (ES) is required to better protect valued ES benefits under pressure from anthropogenic activities. The removal of excess nitrogen, a recognized catchment‐derived pollutant, by biota in estuarine soft sediments is an important ES that potentially ameliorates the development of eutrophication symptoms. Here, we quantified estuarine benthic sediment characteristics and denitrification enzyme activity (DEA), a proxy of inorganic N removal, at 109 sites in four estuaries to develop a general ("global") model for predicting DEA. Our initial global model for linking DEA and environmental characteristics had good explanatory power, with sediment mud content having the strongest influence on DEA (60%), followed by sediment organic matter content (≈35%) and sediment chlorophyll a content (≈5%). Predicted and empirically evaluated DEA values in a fifth estuary (Whitford, n = 90 validation sites) were positively correlated ( r = 0.77), and the fit and certainty of the model (based on two types of uncertainty measures) increased further after the validation sites were incorporated into it. The model tended to underpredict DEA at the upper end of its range (at the muddier, more organically enriched sites), and the relative roles of the three environmental predictors differed in Whitford relative to the four previously sampled estuaries (reducing the explained deviance relative to the initial global model). Our detailed quantification of DEA and methodological description for producing empirically validated maps, complete with uncertainty information, represents an important first step in the construction of nutrient pollution removal ES maps for use in coastal marine spatial management. This technique can likely be adapted to map other ecosystem functions and ES proxies worldwide.