Citation:

Nojavan, F., B. Kreakie, Jeff Hollister, AND S. Qian. Rethinking the lake trophic state index. PeerJ. PeerJ Inc., Corte Madera, CA, , e7936, (2019). https://doi.org/10.7717/peerj.7936

Impact/Purpose:

This research revisits classical methods of lake trophic state classification. The classical schemes have been criticized for developing indices that are single variable, discrete, and/or deterministic. Herein, we present an updated lake trophic classification model using a Bayesian multilevel ordered categorical regression. Our presented work addresses the critics of past classification methods.

Description:

Lake trophic state classifications provide information about the condition of lentic ecosystems and are indicative of both ecosystem services (e.g., clean water, recreational opportunities, and aesthetics) and disservices (e.g., cyanobacteria blooms). The current classification schemes have been criticized for developing indices that are single variable, discrete, and/or deterministic. Herein, we present an updated lake trophic classification model using a Bayesian multilevel ordered categorical regression. The model consists of (1) a proportional odds logistic regression (POLR) that models ordered, categorical, lake trophic state using Secchi disk depth, elevation, nitrogen concentration, and phosphorus concentration and (2) models linking universally available GIS variables to nitrogen (N) and phosphorus (P). The linkage makes it simple and cost e?ective to predict lake trophic state, especially for lakes without costly monitoring data. We used the National Lake Assessment 2007 data set, a probabilistic sample of lakes across the continental US, to develop and validate the model. The overall accuracy for POLR model was 0.7 and the balanced accuracy ranged between 0.74 and 0.84 among classes. For the POLR model combined with the models of N and P the overall accuracy was 0.6 and balanced accuracy was between 0.68 and 0.78 for each of the classes. This work delivers an index that is multivariate, continuous, and classifies lakes in probabilistic terms (i.e., a measure of uncertainty). While our model addresses all the limitations of the current approach to lake trophic classification, the addition of uncertainty quantification is exceedingly important, because the trophic state response to predictors varies among lakes. Our model successfully addresses concerns with the current approach, and additionally performs well across trophic states in a large spatial extent.

URLs/Downloads:

Record Details: