Quantifying the role of geographic regionalizations for national lake classification and assessment

Background/Question/Methods

Geographic regionalization frameworks (i.e. ecoregions) cluster geographical data to create contiguous regions of similar climate, geology and hydrology by delineating land into discrete regions that are often hierarchically nested. Even though most regionalization frameworks were not originally designed for aquatic ecosystem classification or management, they are often used for such purposes, with surprisingly few explicit tests of them. The implicit assumption in using regionalization frameworks for understanding lake variability is that lakes are spatially organized across the landscape such that lakes in similar ‘regions’ share similar hydrogeomorphic controls on lake chemistry and quality. However, the spatial scale with which these regions should be delineated, or the hydrogeomorphic features that these regions should be based on have not been identified for lakes. In this study, we tested the relative ability of eleven different regionalizations to group 2500 northern U.S. lakes for water quality monitoring and assessment. We examined which of eleven different regionalization schemes at two spatial scales best captured the maximum amount of regional variation in water quality for total nutrients, water clarity, chlorophyll, overall trophic state, and alkalinity. We conducted our analyses twice, once using raw response variables and once using residuals from regressions of each response variable with human land use/cover (percent of urban and agricultural land within each lake’s 500m buffer). The latter set of response variables is one way to define lake reference conditions and is useful for examining how much variation among lakes is due to human disturbance as compared to natural hydrogeomorphology. For both analyses, we used multi-level models that partition the total variance into within-region and among-region components.

Results/Conclusions

We found 1) a significant amount of variation among regions for multiple regionalization frameworks, both spatial scales and all response variables (15-80%), 2) that the “best” regionalization scheme depended upon the response variable of interest (e.g., highest for alkalinity and lowest for total phosphorus), and 3) that the results were dependent upon whether or not human land use was accounted for (among region variation was lower for land use residuals than for raw data and the “best” regionalization was different for the two datasets for many response variables). These results suggest that regionalization can provide a useful framework for asking ecological questions about lakes and for water quality assessment and monitoring, but that we must identify the appropriate spatial scale for the questions being asked, the management application, and the metrics being assessed.