Does consideration of human population density on land classified as ‘forest’ improve predictions of water clarity in lakes across the northeastern United States?

Kathryn L. Cottingham and Ian McGrory, Biological Sciences, Dartmouth College, Hanover, NH, Kenneth M. Johnson and Barbara D. Cook, Carsey School of Public Policy, University of New Hampshire, Durham, NH, Bethel Steele and Kathleen C. Weathers, Cary Institute of Ecosystem Studies, Millbrook, NY, Mark J. Ducey, Natural Resources and the Environment, University of New Hampshire, Durham, NH, Jennifer A. Brentrup, Biology, St. Olaf College, Northfield, MN, Jessica V Trout-Haney, Ecology, Evolution, Ecosystems, and Society, Dartmouth College, Hanover, NH, Michael W. Palace, Christina Herrick and Franklin Sullivan, Earth Sciences, University of New Hampshire, Durham, NH, Michael C. Thompson, Natural Resources & the Environment, University of New Hampshire, Durham, NH, David A. Lutz, Environmental Studies, Dartmouth, Hanover, NH

Background/Question/Methods
Clear freshwater lakes and reservoirs are important ecological systems that contribute to local and regional economies but are threatened by anthropogenic change. For example, land conversion for agriculture and development is a key driver of the transition from a high-value, clear-water condition to a eutrophic state plagued by cyanobacterial blooms. Because standard classifications of land use/land cover based on Landsat satellite imagery do not incorporate potentially important features that are undetectable from space, such as the density of houses located underneath primarily forested cover, we hypothesized that incorporating housing density into land-cover classifications would improve our ability to predict a key metric of water clarity, Secchi depth. We tested this hypothesis using general linear mixed models and average summer Secchi depths (2005-2016) recorded for 56 lakes across Maine, New Hampshire, Vermont, and New York by state agencies, researchers, lake associations, and volunteer community scientists. The models compared two suites of land cover classifications – metrics derived directly from the 2011 National Land Cover Database (NLCD) and “demographically enhanced” metrics that incorporate housing density based on the 2010 US Census into NLCD categories – within 100 m, 1000 m, and 5 km buffers around each lake.
Results/Conclusions
Our analyses provide some support for the hypothesis that demographically enhanced land cover categorizations improve our ability to predict average summer Secchi depth in the focal northeastern US lakes. We found that clearer water (i.e., deeper Secchi depth) was associated with lower percent cover in wetlands and agricultural land uses (pasture and crops), and higher percent cover in forest, within the 100 m riparian buffer. The best-fitting model with NLCD data included the percent cover in forest and wetland (F2,53=13.8, P<0.0001, R2=0.34); inclusion of information about high-density housing (>80 homes km-2) under the forested land cover improved R2 slightly (to 0.36), though the AICc and BIC changed little. The modest improvement in explanatory power gained with the demographically enhanced land cover categorizations may be good news, as obtaining these data is time-consuming and can only be done on a decadal time scale following the US Census, whereas NLCD-type data can be derived far more frequently if desired. We are now turning our attention to modeling interannual variability and temporal trends in Secchi depth to determine whether demographically enhanced land cover data improve upon models that also include weather variables such as cumulative spring precipitation, average summer air temperatures, extreme precipitation events, and heat waves.