Large Lake Statistical Water Balance Model (L2SWBM)

Before using this product, please view the laboratory's Disclaimer and Intellectual Property Notice.

Models and model generated data (Experimental)

In the table below you will find, in ZIP folders, the L2SWBM data and software for generating new historical records of the Laurentian Great Lakes' water balance.

Click here to view details on downloadable ZIP folders below.

The ZIP folders linked below contain:

  • Numerical summary data in comma-delimited (CSV) format of L2SWBM inferred values for the Great Lakes' water balance and components. These files should open easily in MS Excel or LibreOffice Calc for quick analysis. File names are formatted as follows:
    • Abbreviated lake name (superior, miHuron, clair, erie, ontario)
    • Component of interest, which include Precip, Evap, Runoff, Outflow, Diversion, NBS (for St. Clair), and StorageChange (change in water level)
    • '_L2SWBM_vX.csv', identifying the L2SWBM version (X) the model generated data are from
    Data are formatted with the following columns (labelled):
    • Year
    • Month
    • Median estimate
    • 2.5 Quantile estimate
    • 97.5 Quantile estimate
  • Plots in PDF format of the complete L2SWBM generated time-series along with model input data for comparison. These plots are done by decade, and thus there may be multiple plots per lake. File names are formatted as follows:
    • Abbreviated lake name (superior, miHuron, clair, erie, ontario)
    • '_YYYY_ZZZZ_' indicating the plot spans the year YYYY to ZZZZ
    • '_L2SWBM_vX.csv', identifying the L2SWBM version (X) the model generated data are from
    Plots are structured such that:
  • Model code, with documentation contained within the README.txt file inside the ZIP folder. Changes from previous versions are documented in the CHANGES_vA_vB.txt file, where A is the previous, and B is the downloaded version.

Version Historical range Time-series PDF plots Model code
v1 1950-2015 v1 time-series (CSV) v1 PDF plots v1 model code
Beta (v0) 1950-2015 v0 time-series (CSV) v0 PDF plots v0 model code


Water balance models are often employed to improve understanding of drivers of change in regional hydrologic cycles. Most of these models, however, are physically-based, and few employ state-of-the-art statistical methods to reconcile measurement uncertainty and bias.

Starting in 2015, NOAA-GLERL, along with its partners at CIGLR, began developing a water balance model under a Bayesian Markov chain Monte Carlo framework. Through this model, we generate new estimates of monthly runoff, over-lake evaporation, over-lake precipitation, and connecting channel flows for each of the Great Lakes. The new model reconciles discrepancies between model and measurement-based estimates of each component while closing the Laurentian Great Lakes water balance.

In 2017, funding from the International Joint Commission - through their International Watersheds initiative - was received to use the model in generating a new, balanced historical (1950 - 2015) record of the Laurentian Great Lakes water balance. The project will help in resolving the regional water budget across monthly and inter-annual time scales and represents an important stepping stone towards addressing a long-standing need in the Great Lakes for clear and defensible differentiation between hydrological, climatological, geological, and anthropogenic drivers behind seasonal and long-term changes in Laurentian Great Lakes water levels.

On this page, you will find publications, data, and select code to run versions of the Large Lake Statistical Water Balance Model (L2SWBM) developed here at GLERL. To run these models, you will need to do the following:

Publications and associated software


We thank Song Qian, Yves Atchade, Kerby Shedden, Edward Ionides, Vincent Fortin, Bryan Tolson, and Craig Stow for helpful discussions on Bayesian inference and alternative formulations of our water balance model. Jacob Bruxer, Frank Seglenieks, Tim Hunter, and Lauren Fry provided expert opinions and water balance data. Nicole Rice provided graphical and editorial support. Funding was provided by the International Joint Commission (IJC) International Watersheds Initiative (IWI) to NOAA and the Cooperative Institute for Great Lakes Research (CIGLR) through a NOAA Cooperative Agreement with the University of Michigan (NA12OAR4320071); many thanks to Wendy Leger and Mike Shantz. The use of product names, commercial and otherwise, in this paper does not imply endorsement by NOAA, NOAA-GLERL, CIGLR, or any other contributing agency or organization.