2018-165 Transportation Plan - Laserfiche WebLink

Home Browse Search

Calibration Approach Calibration is an iterative process and includes upgrading or adjusting entered data, program coefficients, or parameters and assumptions on successive simulation runs until the volumes and traffic patterns produced by the model approximate known volumes within an "acceptable level of error." The acceptable level of error for calibrated model data has been recommended in the National Cooperative Highway Research Program Report No. 255 entitled Highway Traffic Data for Urbanized Area Project Planning and Design. The primary premise behind these guidelines is that simulated model data should not significantly differ from actual count data thereby causing inappropriate under-or over-design of roadway facilities. Differences between modeled volumes and actual counts may look significant; however, in everyday practice, these differences should not cause unsuitable roadway facility planning. There are three significant points to consider. The first is "acceptable level of error" and "How good are the counts?" Given that this is the basis for calibration, are these counts good enough for the process? If some count data is questionable, can the model be asked to simulate a condition better than the condition is known? Considering these questions, it has been found through experience in modeling that an "acceptable level of error" is directly related to the existing traffic volumes on a certain link. Through the course of calibration, higher volume streets can be expected to have better results. Acceptable limits may be that a 20% error can be expected on heavily used arterials, 40% on primary collectors, and perhaps as high as 200% on little-used rural collectors. Although the latter level of error may seem high, a variation of 200% on little-used roadways may mean a difference of 25 to 100 vehicles, insufficient to cause inappropriate facility planning when the model results are used. The second point to consider is the adjustment of entered data, program parameters, and model assumptions. After entering all the data and making the initial model assumptions, the simulation distribution and assignment run is made. The desired outcome is that the results will perfectly match all the counts and the model will be calibrated. Usually, though, some data or assumptions (the "rules" of the model) are incorrect. On locating the errors from the distribution and assignment, causes are identified. The rules are reconsidered and adjusted. Kittitas County, Washington Transportation Model-May 2009 Page 36