A Retrospective Numerical Weather Simulation of the 1913 Storm
Click on the image above to view the simulation
Extreme weather events, like the Great Lakes Storm of November 7-11, 1913, provide unique challenges to meteorologists. They are usually rare events and often behave in ways that are not easily predictable. Sometimes meteorologists are able perform “forensic” studies of significant weather episodes in order to better understand how they occur and to gain greater context into the extreme conditions they produce. One of the most prominent tools in these studies is the computer model. Atmospheric and oceanic motions of air and water are governed by highly complex physical processes that can be described through mathematical relationships. The emergence of advanced computer processing over the last half century has revolutionized weather forecasting by allowing real-time calculations of weather patterns and sea states through these mathematical relationships. This has allowed modern-day meteorologists to predict weather patterns with a remarkable degree of accuracy – provided the observed weather, or “initial conditions”, can be accurately measured and injected into the start of the computer simulation.
One hundred years ago weather forecasters did not have the luxury of computer models, nor the detailed surface and upper-air observations needed to make the most accurate predictions. The lack of observations, especially upper air and satellite-derived data, are also major obstacles to accurately simulating historic weather episodes like the Great Storm of 1913. Fortunately, within the last decade, a group of meteorologists developed an ingenious method of estimating upper atmospheric conditions for historical periods prior to the advent of weather balloon and satellite observations. The 20th Century Reanalysis Project now provides us with an adequate representation of the state of the atmosphere in early November 1913 to prescribe “initial conditions” from which to generate a computer simulation.
The main purpose of the numerical model retrospective of the Great Storm of 1913 was to gain insights into the timing and severity of the conditions experienced by Great Lakes mariners. Of particular interest were wave conditions, as several large boats were caught unprepared for such extreme conditions. In any sort of numerical model simulation, there can be several sources of error and a perfect simulation is usually unattainable. This is especially true of a one hundred year retrospective. Nonetheless, even a less-than-perfect simulation affords important context into what happened and when.
This study leveraged the capabilities of the Weather Research and Forecast (WRF) modeling system to produce a detailed reconstruction of atmospheric conditions; and the NOAA Great Lakes Environmental Research Laboratory – Donelan Wave Model (GDM) to reconstruct the resultant sea state.
The WRF model, for the Great Storm of 1913 presentation, produces hourly forecasts of: 1) Mean Sea Level Pressure in Millibars (MSLP), 2) sustained wind speeds at the surface in Knots, 3) wind gusts at the surface in Knots, and 4) 3-dimensional equivalent potential temperature. The GDM wave model was configured using a 5km rectangular grid on all of the Great Lakes to simulate the wave conditions during the November 1913 storm. Surface wind and temperature output from the WRF simulation was used to drive the wave model simulation. The GDM provides approximations for significant wave height (average of the highest ⅓), dominant wave period, and wind wave direction. As a companion calculation, an estimate of the average highest 5th percentile wave height from wave energy distribution is produced to characterize reasonably observed “peak” or “maximum” wave conditions. The return frequency of the peak wave is also calculated based upon the dominant wave period and the statistical occurrence of the the average highest 5th percentile wave. The return frequency gives an estimate of how frequently ships experienced larger waves during the storm.