You are here:
Lake Photic Zone Temperature Across the Conterminous United States
Citation:
Shivers, S., B. Kreakie, W. Milstead, AND J. Hollister. Lake Photic Zone Temperature Across the Conterminous United States. 2019 North American Lake Management Society Meeting, Burlington, VT, November 11 - 15, 2019.
Impact/Purpose:
As lake water temperature increases we often see enhanced phytoplankton productivity (algal growth) and the enhanced risk for the development of harmful algal blooms. Although water temperature can be easily measured in the field we often wish to know the history of lake temperatures or the temperatures of lakes that have not been sampled. To help with this we have developed a model to predict current and historical lake temperatures from data that are readily available. We used measured lake temperatures from the 2007 and 2012 National Lake Assessment, estimated air temperatures, and lake morphometry measurements (e.g., depth, area, volume, etc) to develop a model to predict lake temperatures for all lakes in the conterminous United States for summer (June 1 to September 30) for 1981 to 2017 with a high degree of accuracy. This model can be used by lake managers, state departments of environmental management, and researchers to obtain current and historical estimates of lake photic zone temperatures from most of the lakes in the United States.
Description:
As global air temperatures increase, lake surface water temperatures also increase. These increases will cause ecosystem changes that will impact all aspects of lake management, particularly related to harmful algal blooms (HABs). Increasing surface water temperatures can cause an increase in the frequency, duration, and severity (i.e. toxicity) of HABs, which affect both human and environmental health. Our ability to predict and manage HABs is dependent on our ability to accurately predict temperature; thus the goal of this project was to develop a simple, robust photic zone temperature model for all lakes in the conterminous United States. Data from the 2007 and 2012 US EPA National Lake Assessment was used for model development. Random forest, a machine learning approach based on an aggregation of 10,000 regression trees, was used to develop our final predictive model. The final model had a mean square error of 2.17 °C and adjusted R^2 of 0.89. Sample date, average daily ambient air temperature, longitude, and the 30-day average ambient air temperature were the most important variables to overall model accuracy. We are currently backcasting temperature over the past thirty years for all lakes in the NHDPlus database to look for trends at regional and continental scales.
