How to predict water temperatures

Posted on March 27, 2017


Because temperature is a key variable affecting a myriad of ecological processes, there’s a need for reliable ways of predicting the temperature of water bodies. Such temperatures can be influenced by both natural factors (e.g., air temperature, local topography, stream discharge, groundwater interactions) and anthropogenic factors (e.g., deforestation, thermal pollution, flow alteration, and runoff from impervious surfaces). A recent study set out to compare the performance of three commonly-used methods of water temperature prediction, these being (1) simple linear regressions of water temperature against air temperature; (2) non-linear (S-shaped) regressions, which include minimum and maximum stream temperatures as well as air temperature; and (3) a process-based model that accounts for snowmelt, groundwater flow, soil water lateral flow, discharge, and surface water runoff. The accuracy of each model was assessed by reference to hydroclimate data from the Hinkson Creek watershed in Missouri, USA. Over most of the range of data, all three models estimated daily water temperatures well, but above 20 °C water temperatures were significantly underestimated by the linear model, and to a lesser extent the non-linear model.  If applied in an ecological context (e.g., by land managers), these biases would produce underestimations of algal growth, nutrient cycling, biochemical oxygen demand, and the rates of other aquatic ecosystem processes. At the other end of the thermal range (temperatures below10 °C), both linear and non-linear regressions greatly overestimated water temperatures. By contrast, the process model was relatively unbiased in the high temperature range but below 10 °C it tended to yield underestimates.  In summary, all types of predictive models should be used with caution, especially near the temperature extremes. While the process-based model is the most transferable, it does depend on a number of climatic and flow-related inputs.

Reference: Zeiger, S. et al. 2016. Quantifying and modelling urban stream temperature: a central US watershed study
Hydrological Processes 30, 503–514. https://www.researchgate.net/publication/280916417_Quantifying_and_modelling_urban_stream_temperature_A_central_US_watershed_study

Advertisements