For enhanced public safety and water resource managerial and scientific purposes, with the development of computational ‘data transfer’ languages Python and data processing tools MATLAB, we created a transformative research cluster on predictive Computational Fluid Dynamic (CFD) models,  carrying on the high resolution (500m horizontal grid) real-time and forecasting simulations.

Lake Erie was chosen as the pilot lake as the CFD model have been successfully applied to hindcast the hydrodynamics and water quality of the lake. Currently, the model enables predictive simulation of physical processes in Lake Erie, like current circulation, surface seiches, and spring turnover. In the future, by adding the water quality module within AEM3D, the model will predict the development of harmful algae blooms in western and central basin of Lake Erie, and distributions of fish habitat, as well as the key water quality indicators required for management. 


AEM3D is a 3-Dimensional coupled Hydrodynamic-Aquatic Ecosystem Model, based on the internationally recognized model ELCOM-CAEDYM developed by the Center for Water Research. Distributed by HydroNumetrics (http://www.hydronumerics.com.au/), AEM3D is used to simulate the velocity, temperature, salinity, nutrients and biogeochemistry in surface waters that are subjected to environmental and anthropogenic forcing such as wind, tides, surface heating and cooling, inflows, withdrawals, bubblers and mixers.

Model runs with 500-m horizontal grid, and 45 vertical layers, with thickness varying between 0.5-m and 5-m.  The layers near the lake surface and bottom, as well as thermocline depth were set to 0.5-m. Model ran from 2019 Sep. 1st with 5-min time steps. 

We use a Python script to retrieve online daily historical climate data (-24h – 0h) and predictive meteorological model outputs (0h – 240h) , and prepare the input files for hind-cast, real-time and fore-cast CFD models via a MATLAB script.



Models are driven by meteorological forcing variables including: wind speed, wind direction, air temperature, solar radiation, relative humidity, air pressure, and cloud cover.

In order to address the spatial variability of meteorological conditions across the lake, the computational domain was divided into several sections. In each section, the forcing variables was set to be uniform. We applied 25-km and 15-km resolution meteorological outputs from Global Deterministic Forecast System (GDPS) as forcing variables, so there are 23 and 31 meteorological sections, respectively.


Water level and temperature output from forecast models are validated with observations from Fisheries and Oceans Canada (https://marees.gc.ca/eng/find/zone/44), and Great Lakes Observing System (GLOS: https://www.glos.us/).


We gratefully acknowledge the  Dean’s Research Fund from Faculty of Engineering and Applied Science, Queen’s University.


Lin, S. 2019. Turbulence and sediment resuspension modeling in Lake Erie. PhD thesis, Dep. Of Civ. Eng., Queen’s Univ. at Kingston, Canada.

Gaudard, A. et al., 2019. Toward an open-access of high-frequency lake modelling and statistics data for scientists and practitioners. The case of Swiss Lakes using Simstrat v2.1. Geosci. Model Dev., Volume 12, p. 3955–3974.