|
A
<br /> ., , '1 ..;. t. . : I 1f• ;li , LA At 1L -ti<ti;;i Some example dynamic validation tests and expected `
<br /> Fehr&Peers will perform static and dynamic validation outcomes include:
<br /> on the updated base year model. o Adding housing at increments between 100 and
<br /> Task 4.1—Static Model Validation 1,000 to a TAZ with a corresponding increase In 1
<br /> The model will be calibrated and validated to a level trip generation based on the trip rates
<br /> meeting or exceeding the suggested targets published 0 Removing a roadway link with a corresponding a
<br /> by Caltrans.The validation criteria will include volume- shift in traffic to parallel routes
<br /> to-count ratios,maximum percentage deviation by 0 Adding a roadway link with a corresponding 12
<br /> functional class,correlation coefficient,and percentage shift in traffic from parallel routes
<br /> root mean square error.The validation tests will be O Decreasing capacity on a link operating at
<br /> applied both to screenlines and to each roadway link capacity with a corresponding shift in traffic to
<br /> where count data is available.These typical validation parallel routes
<br /> techniques are referred to as"static"validation,as they
<br /> reflect a model's ability to replicate existing conditions. The results of the dynamic validation task will inform the
<br /> The validation tests are summarized below: team about the ability of the model to evaluate -
<br /> • The maximum desirable error allowed between the constructing(or not constructing)Gateway Intersection
<br /> improvements. II
<br /> model volume and the actual count at individual
<br /> locations varies based on the count volume,with &
<br /> larger deviations allowed for lower-volume counts. We will work with City staff to review model information
<br /> At least 75%of the links should deviate by less than throughout the model update process.Additionally,we
<br /> the maximum desirable error. have developed our own quality assurance system to
<br /> • The sum of the model volumes on all links for which ensure reasonableness of assumptions.This quality
<br /> counts are available should be within 10%of the assurance system consists of reviewing model inputs,
<br /> sum of the counts. comparing area-wide land use control totals to typical `1
<br /> • The correlation coefficient,which estimates the planning control totals,reviewing model performance
<br /> correlation between the counts and the model to ensure accurate forecast projections,and ,--1
<br /> volumes,should be at least 88%. documenting assumptions and methodologies. r.,
<br /> • The%Root Mean Squared Error(%RMSE),a measure
<br /> of overall model performance,should be less than We will prepare a Model Development Report with
<br /> 40%. appropriate tables and graphics.The model
<br /> development report will describe the methodology,
<br /> In addition to the validation tests outlined above,we data sources,and assumptions used to develop the
<br /> will also review non-assignment related model statistics, model.Flow chart diagrams will be included that will
<br /> such as production-attraction balancing,trips per illustrate the relationship of data inputs,model steps,
<br /> household and trip lengths.If the productions and and output.The model code or script files will be
<br /> attractions do not balance in the model,the result could included in the appendix of the report.The model
<br /> bean under assignment of vehicle trips to the roadway development report will also act as a Model User's Guide ..]
<br /> network. intended to assist technical staff in modifying data
<br /> inputs,running the model,and performing selected
<br /> Task 4.2—Dynamic Model Validation basic model applications.A complete set of model files
<br /> Fehr&Peers also recommends conducting"dynamic" will be provided. ]
<br /> validation methodologies to ensure that travel demand . _ _ _. . _ .. _ __
<br /> models can be used to appropriately forecast future year ;,, . : , •, ,,•, . , . , t;, r i-,i., ,. ,,
<br /> travel demand.We developed these methodologies
<br /> because we have found that travel demand models can . .
<br /> generate counterintuitive results when making , ; ,', ; ' ; 1 i • i , ,
<br /> individual land use or roadway capacity changes.Some :.9
<br /> of the dynamic test applied to the Pleasanton model will
<br /> be systematic changes in residential and non-residential ,r ; • . ; , ; , ,
<br /> land uses in Individual TAZs,as well as changes in the
<br /> capacity of major roadway links,to ensure that the • '- . , ; ,. -' 1 ' ' .
<br /> •
<br /> model responds appropriately to these typical input
<br /> changes. v
<br /> -a
<br /> 8
<br /> 3
<br />
|