Traffic Forecasting Tool

Transcription

1 Traffic Forecasting Tool Applications for Indiana, Kentucky and Tennessee Statewide Forecasting presented to Southeast Florida Model Users Group presented by Ken Kaltenbach, The Corradino Group November 16, 2018

2 Overview INDOT maintains a TFT to facilitate traffic forecasting activities. Earlier TFT developed in In use for 12 years. TFT v3.0 Compatible with the latest Indiana Statewide Travel Demand Model in TransCAD 6 The most current traffic count data Enhanced user interface Improved historical-count-based forecasting models Multiple links (up to 15) for analysis at a time (previously limited to 1 at a time) Reports more link attributes More effective & manageable outputs (.BIN reports & graphs) Minimizes staff requirements, reduces potential method & calculation errors Similar tools under development for Kentucky and Tennessee

3 ISTDM 7 4-step model (TransCAD 6) Coverage entire IN + partial IL, KY, MI, OH 4879 zones & links Modes - auto, freight, LD transit - tolling Model years with 5-year increment Currently being updated. TFT v3.0 uses ISTDM & 2035 model run results.

4 Traffic Data INDOT s Traffic Count Database System (TCDS) Use recent data (2001 current) Tag count stations GIS layer & data to ISTDM 7 network 11,558 links has data

5 Linear Forecasting Models in TFT v3.0 Linear Regression Exponential Adjusted Exponential - Bounded by the Indiana Traffic Growth Profile NCHRP 255 Calibration Procedure: average of difference and growth factor methods.

6 Indiana Traffic Growth Profile A set of reasonable ranges of traffic growth rates improve the Exponential model in TFT v3.0 In current TCDS, 69% of count stations and 76% of AADT records only have data of 2010 or later. Short-term historical counts may introduce abnormal growth rates (outliers) that are associated with transient events. For all states, most of the consultant effort is in linking the count database to the tool, and creating the growth profile.

7 Indiana Traffic Growth Profile Use Q-Q plot to filter outliers (normally distributed data should follow a straight line with correlation coef. = 1.0). Find & delete the outliers from the analysis (Excel-based). Example: Rural Interstate (raw data) Example: Rural Interstate (cleaned data) correlation coef. = 0.89 correlation coef. = 0.99

8 Cleaned Data Indiana Traffic Growth Profile (cont.)

9 Indiana Traffic Growth Profile (cont.) Cleaned Data vs. Raw Data

10 Indiana Traffic Growth Profile (cont.) Cleaned Data CI 99%,l, GF adj = GF, CI 99%,u, if GF < CI 99%,l if CI 99%,l GF CI 99%,u if GF > CI 99%,u

11 NCHRP 255 Adjustment Procedure Ratio adjustment Difference adjustment A ratio = COUNT A b A f A difference = COUNT A b + A f Final adjustment RA f = A ratio + A difference 2 where: COUNT = base year traffic count A b = base year model volume A f = future year model volume A ratio = future year volume based on ratio adjustment A difference = future year volume based on difference adjustment RA f = final adjusted future year volume

12 TFT v3.0 About Tab

13 TFT v3.0 Preprocess Tab All-in-one button o Joins: ISTDM loaded network ISTDM traffic loads - Historical traffic counts o Green-highlights links with historical traffic data. o Displays centroid connectors as grey dashed lines

14 TFT v3.0 Preprocess Tab (cont.) A message box confirms the completion of preprocessing data.

15 TFT v3.0 Estimation Tab Select link(s) Remove link(s) Methods used for estimation Generate traffic estimation report Generate historical counts report adjusted by IN growth profile Generate traffic estimation graph

16 Select & Clear Links Select Link button - select links. - TFT3 automatically redhighlights selected links and label link IDs. Clear Link button - remove all selected links

17 TFT v3.0 Historical Counts Report Link #1 Link #2 Historical Counts button generate traffic count report - Info organized by link - Key link attributes AADT Report must be closed before having another one. Save As as needed

18 TFT v3.0 - Estimation Report Report button generate traffic estimation report Report must be closed before having another one. Save As as needed. Link #1 Historical data ISTDM data Linear model Regression model Link #2 Exponential model Adjusted AGR based on IN traffic growth profile

19 TFT v3.0 - Estimation Graph Graph(s) show: Historical traffic counts Fitted traffic volumes by selected method

20 TFT v3.0 Forecast Tab Select link(s) Remove link(s) Methods used for forecasting Generate traffic forecast report Future year (>= 2030, default 2040) adjusted by IN growth profile Generate traffic forecast graph

21 based on historical data based on ISTDM data NCHRP 255 adjustment based on historical data based on ISTDM data NCHRP 255 adjustment based on historical data based on ISTDM data NCHRP 255 adjustment based on historical data TFT v3.0 - Forecast Report Report button generate traffic forecast report Report must be closed before opening another one. Save As as needed. Link #1 Linear Regression Exponential Adjusted Exponential Link #2

22 TFT v3.0 - Forecast Graph Graph(s) show: Traffic forecast by historical data and selected method ISTDM 7 traffic forecast ISTDM 7 traffic forecast adjusted by NCHRP 255 Capacities of LOS A - E

23 Improvements to TFT v3.0 under development Cover all segments of state owned roads by creating and using a new linear reference system. Compatible with the newer version of ISTDM (being updated) Incorporate more information in the tool to meet INDOT s traffic forecasting needs Customize reports for each state Revised interface to pick segments by first and last nodes Separate profiles and procedures for autos, SU trucks and combination trucks in Kentucky as requested by Sunil Saha. Tennessee allow the user to specify alternative forecast networks and year (not a static forecast target year).

24 Acknowledgements Johnny Han Corradino Indianapolis. Roy Nunnally Indianapolis DOT Scott Thomson Kentucky Transportation Cabinet David Lee Tennessee Department of Transportation

25 Questions?

26 Q-Q PLOT Use Excel 1. Order the original observations (growth factors) to get x(1), x(2),, x(n) and their corresponding probability values (1-0.5)/n, (2-0.5)/n,, (n-0.5)/n; 2. Calculate the standard normal quantiles q(1), q(2),, q(n); use excel function NORMSINV(prob). NORMSINV is the inverse of the cumulative standardized normal distribution. 3. Plot the pairs of observations (q(1), x(1)), (q(2), x(2)),, (q(n), x(n)); and 4. Examine the linearity of the plot