Appendix D - NE Kansas Regional Model. Model Development Documentation. Table of Contents

Transcription

1 Appendix D - NE Kansas Regional Model Model Development Documentation Table of Contents NE Kansas Regional Model... D.1 Model Development Documentation... D.1 Table of Contents... D.1 Table Directory... D.2 Figure Directory... D.2 Network Setup... D.1 Existing Network... D.1 New Network Coding... D.2 SE Data... D.3 Trip Generation... D.3 Trip Generation Sub-Models:... D.3 Household Size Allocation Model... D.3 Household Income Allocation Model... D.4 Auto Ownership Sub-Model... D.5 Joint Household Distribution... D.8 Trip Production Rates... D.8 Trip Attraction Equations... D.9 Freight... D.9 Trip Distribution... D.9 Mode Choice... D.16 Time of Day... D.17 Assignment... D.19 Volume Delay Functions... D.19 External Trips... D.19 External to Internal Trip Distribution... D.2 External to External Trip Distribution... D.2 Special Generators... D.22 Application Program... D.22 Calibration Results... D.24 Synthetic Matrix Estimation (SME) Analysis... D.24 Percent Root Mean Squared Error (PRMSE)... D.25 5-County Regional Transportation Study D-1

2 Table Directory Table D.1: Link Type Definition... D.1 Table D.2: Volume Delay Function Definition... D.2 Table D.3: Node Numbering - Overall Layout... D.2 Table D.4: Socio-Economic Growth Rates... D.3 Table D.5: Auto Ownership Model Structure... D.6 Table D.6: Auto Ownership Model Constants and Coefficients... D.7 Table D.7: Trip Generation Rates (trips/household/day)... D.8 Table D.8: HBW Trip Attractions by Income... D.9 Table D.9: Distance to Time Factor for Distribution Cost... D.1 Table D.1: Trip Distribution Parameters... D.15 Table D.11: Mode Choice Model Coefficients... D.16 Table D.12: 26 Mode Choice Model Vehicle Trip Output... D.17 Table D.13: Time Of Day Factors by Trip Purpose... D.18 Table D.14: Capacity and Free Flow Speed Assumptions... D.19 Table D.15: External Station Trips... D.21 Table D.16: Special Generator List... D.22 Table D.17: Matrix Model Inputs... D.23 Table D.18: Johnson County Additional Trips... D.25 Table D.19: PRMSE by Average Daily Count... D.25 Table D.2: PRMSE by Facility Type*... D.26 Table D.21: Screenline 1 Volumes... D.29 Table D.22: Screenline 2 Volumes... D.29 Figure Directory Figure D.1: Household Size Sub-Model Size Distribution Curves... D.4 Figure D.2: Household Income Distribution... D.5 Figure D.3: Auto Ownership Structure... D.5 Figure D.4: HBW High Income TLFD, Observed and Estimated... D.1 Figure D.5: HBW Middle Income TLFD, Observed and Estimated... D.11 Figure D.6: HBW High Income TLFD, Observed and Estimated... D.12 Figure D.7: HBO TLFD, Observed and Estimated... D.12 Figure D.8: HBSocial/Recreational TLFD, Observed and Estimated... D.13 Figure D.9: HBShopping TLFD, Observed and Estimated... D.13 Figure D.1: HBSchool (K-12) TLFD, Observed and Estimated... D.14 Figure D.11: Non Home-Based Work TLFD, Observed and Estimated... D.14 Figure D.12: Non Home-Based Other TLFD, Observed and Estimated... D.15 Figure D.13: Diurnal Factors: Total Vehicles by Hour... D.18 Figure D.14: PRMSE by Average Daily Count... D.26 Figure D.15: PRMSE by Facility Type... D.27 Figure D.16: Screenlines... D.28 Figure D.17: Modeled vs. Daily Count Link Scattergram... D.31

3 Network Setup The starting point is the combination of two existing models from two software platforms: the 23 MARC regional model in EMME/2 (which includes Leavenworth, Wyandotte, Johnson, Platte, Clay, Jackson, and Cass counties), and the 23 Lawrence/Douglas Co. regional model in TransCad. It was agreed that the 5-County model would be developed in EMME, with the possibility of transferring the final databank into TransCAD. The Missouri network detail is not included in the study area, but most of the network detail will remain in the 5-County network. Existing Network The first step is to establish a relationship between coordinate systems of the two existing networks. Using linear regression, the coordinate conversion equations are derived as follows: (Longitude, Latitude) -> (xi,yi) Lon =.186*xi , R 2 =.9987 Lat =.147*yi , R 2 =.9976 Thus, the inverse functions provide a map from the Lawrence coordinate system into EMME xi = (Lon )/.186 yi = (Lat )/.147 Several steps are taken to transfer the Lawrence model into the MARC model: 1. Export links, centroids, and nodes into Excel format. 2. Apply conversion formula 3. Assign appropriate link attributes according to MARC system. 4. Re-number nodes and centroids to fit into MARC system. 5. Batch-in nodes and centroids into EMME environment, then batch in links. Link attributes defining the number of lanes and the area type are easily taken from the Lawrence model and input when the links are created in EMME. Definitions for type and volume delay function (vdf) attributes are described below. Table D.1: Link Type Definition Link Type = County*1 + Area*1 + Facility Classification County Area Facility Classification 1 Leavenworth 1 CBD 1 Interstate Freeway 2 Wyandotte 2 Fringe 2 Expressway 3 Johnson 3 Urban 3 Collector / Minor Arterial 4 Platte 4 Suburban 4 Collector / Minor Arterial One-Way 5 Clay 5 Rural 5 Centroid Connector 6 Jackson 6 Principal Arterial 7 Cass 7 Auto-Connect Only 8 Douglas 8 9 Miami 9 Walk Only 5-County Regional Transportation Study D-1

4 Table D.2: Volume Delay Function Definition VDF Code = Area*1 + Facility Classification Area Facility Classification 1 CBD 1 Interstate Freeway 2 Fringe 2 Expressway 3 Urban 3 Collector / Minor Arterial 4 Suburban 4 Collector / Minor Arterial One-Way 5 Rural 5 Centroid Connector 6 Principal Arterial New Network Coding There are currently parts of the study area that are not represented in any current travel demand model. Miami County, as well as areas in Leavenworth and Douglas Counties, will be created based on the current highway and arterial configuration, in addition to future roadway projects. Table D.3: Node Numbering - Overall Layout Centroids MARC Centroids (existing) MARC (new) Lawrence Centroids (existing) Lawrence (new) Leavenworth Miami Ottawa (existing) Added Zones (KDOT comments) External Zones MARC Nodes (existing) Lawrence Nodes (existing) Ottawa Nodes (existing) Douglas Nodes (added) Leavenworth Nodes (added) Miami Nodes (added) Johnson Nodes (added) The detail of the network will be one level down from the roadways that are of interest. MARC s model currently includes minor arterials and some local streets. Miami Co. should include state highways and some major and minor collectors. Douglas and Leavenworth Counties included highways and county maintained roads. Functional class maps are used as a reference to determine what roadways should be included in the model network. Additionally, some minor roadways may be included to appropriately distribute loading onto the highway network from the centroids. 5-County Regional Transportation Study D-2

5 A buffer area surrounding the 5-County region will also be developed to accommodate major trip generators and attractors. The city of Ottawa is included in the model as part of this buffer. Transit access walk links are added using the WLKLINKS.exe program. The program reads in EMME/2 output files and then generates the access links. For our model, we used different specifications for CBD areas and non-cbd areas. CBD regions have a walk links coded in, so the access links only need to provide connectivity from the centroids to the walk network. Outside areas don t have walk access, therefore connectivity from the centroids directly to the transit lines is necessary. CBD areas:.2 mile max distance, 1 link Non-CBD:.99 mile max distance, 2 links. SE Data The 26 data that is being used for this study was received from OA. New modeled areas need to have 23 data calculated. We will use the boundary from the MARC and LD-MPO models and aggregate to come up with the growth factors. Table D.4: Socio-Economic Growth Rates Population Households Retail Employment Non-Retail Employment Growth Factor Growth factors will be applied at a zonal level to the new zones. Trip Generation Our trip generation model was borrowed from the recent MARC model. The model uses a cross-classification structure for trip productions, and a set of linear models for trip attractions. In addition, submodels supply the distribution of trips by auto ownership, household size and (for home-based work trips) by income category. Trip Generation Sub-Models: Household Size Allocation Model The household size allocation model uses the average household size to compute the shares of households by size for each zone. Average household size is equal to the total population 5-County Regional Transportation Study D-3

6 Share of Households divided by the number of households for each zone. A set of household size curves are then used to calculate the number of households by size for each zone. Figure D.1 shows the household size distribution curves. Figure D.1: Household Size Sub-Model Size Distribution Curves Household Size Distribution Curves HHS=1 HHS=2 HHS=3 HHS=4 HHS= Average Household Size (at the zone level) Household Income Allocation Model The household income allocation model uses the average household income to compute the shares of households by three income groups for each zone. Average household income is a forecast variable, but is expressed in 24 constant dollars, so a forecast value that is unchanged represents an assumption that incomes will keep pace with the cost of living. Figure D.2 shows the household income distribution curves. Annual household income ranges, in 24 constant dollars are: Low income: <$25, Middle income: $25, - $75, High income: >$75, 5-County Regional Transportation Study D-4

7 Pct of households Figure D.2: Household Income Distribution 1% 9% 8% 7% 6% 5% 4% 3% 2% 1% % $18 $23 $28 $33 $38 $43 Auto Ownership Sub-Model Households by auto ownership groups (, 1, 2 and 3+ autos) is computed using a sequential logit formulation, where a series of binary logit choice models estimate probabilities of auto ownership that are dependent upon upper level choice groups. For example, Figure D.3 shows that the probability of owning 3 or more autos is dependent upon the probability of owning 2 or more autos, which in turn is dependent upon the probability of owning at least 1 auto. This formulation is consistent with the behavioral aspects of changes auto ownership, which is naturally incremental. Figure D.3: Auto Ownership Structure All Households HH Income Curves Low Income Mid Income High Income $48 $53 $58 $63 $68 $73 $78 $83 $88 $93 $98 $13 $18 $113 $118 $123 $128 Average Zone HH Income (Year 24 dollars, in thousands) Autos 1+ Auto 1 Auto 2+ Auto 2 Auto 3+ Auto 5-County Regional Transportation Study D-5

8 The utility equations used for the auto ownership model are shown in Table D.5 below: Table D.5: Auto Ownership Model Structure Utility Equations For /1+ nest: U( Auto) = r1*density + r2*taccess U(1+ Auto) = r111*hhs1 + r112*hhs2 + r113*hhs3 + r114*hhs4 + r211*inc1 + r212*inc2 + r11 For 1/2+ nest: U(1 Auto) = r121*hhs1 + r122*hhs2 + r123*hhs3 + r124*hhs4 + r221*inc1 + r212*inc2 + r21 U(2+ Auto) = r3*density + r3*taccess For 2/3+ nest: U(2 Auto) = r5*density + r6*taccess U(3+ Auto) = r131*hhs1 + r132*hhs2 + r133*hhs3 + r134*hhs4 + r231*inc1 + r232*inc2 + r31 where: Density = Population density in thousands of persons/square mile, based on zone in question and all adjacent (i.e., touching) zones Taccess = Ratio of walk transit employment accessibility and sum of walk transit and auto employment accessibility (am peak period) transit accessibility is the sum of the ratios of employment to destination zone divided by the composite transit time to each zone. Sum over all destination zones. Composite transit time is IVT + 1.5*(access, auxilliary and egress time) + 2*wait time + 5*transfers. Highway accessibility is the sum of the ratios of employment to destination zone divided by the highway travel time to each zone. Sum over all destination zones. HHSx = Dummy variable for household size market. 1 if hhsize=x, else Dummy variable for household income market. 1 if hhinc group=y, else (1=low, INCy <$25k; 2=mid, $25k-75k) Coeff & Coefficients and Constants (r***) as defined in Table 7- Auto Ownership Model Constants Constants and Coefficient Values Table D.6 shows the coefficients and constants, corresponding to the formulations in Table D.5. The coefficients and constants follow a logical relationship, as described below: The constants stratified by household size demonstrate that auto ownership is greater with greater household size The constants stratified by household income demonstrate that auto ownership is greater with greater income The density coefficient shows that auto ownership is less with increasing density The transit accessibility coefficient shows that auto ownership decreases with increasing transit accessibility. 5-County Regional Transportation Study D-6

9 Table D.6: Auto Ownership Model Constants and Coefficients Constants Description (Constant for ) Choice nest Calibrated Value r11 Share of 1+ Auto Households (all size and income markets) / r111 Share of 1+ Auto Households, HHSIZE=1 / r112 Share of 1+ Auto Households, HHSIZE=2 / r113 Share of 1+ Auto Households, HHSIZE=3 / r114 Share of 1+ Auto Households, HHSIZE=4 / r211 Share of 1+ Auto Households, Inc=1 (low) / r212 Share of 1+ Auto Households, Inc=2 (mid) / r21 Share of 1 Auto Households (all size and income markets) 1/ r121 Share of 1 Auto Households, HHSIZE=1 1/ r122 Share of 1 Auto Households, HHSIZE=2 1/ r123 Share of 1 Auto Households, HHSIZE=3 1/ r124 Share of 1 Auto Households, HHSIZE=4 1/ r221 Share of 1 Auto Households, Inc=1 (low) 1/ r222 Share of 1 Auto Households, Inc=2 (mid) 1/ r31 Share of 3+ Auto Households (all size and income markets) 2/ r131 Share of 3+ Auto Households, HHSIZE=1 2/ r132 Share of 3+ Auto Households, HHSIZE=2 2/ r133 Share of 3+ Auto Households, HHSIZE=3 2/ r134 Share of 3+ Auto Households, HHSIZE=4 2/ r231 Share of 3+ Auto Households, Inc=1 (low) 2/ r232 Share of 3+ Auto Households, Inc=2 (mid) 2/ Coefficients Description (Coefficient on ) Choice nest Estimated Value r1 Population Density; Share of Auto Households / r2 Transit Accessibility Index; Share of Auto Households / r3 Population Density; Share of 2+ Auto Households 1/ r4 Transit Accessibility Index; Share of 2+ Auto Households 1/ r5 Population Density; Share of 2 Auto Households 2/3+.5 r6 Transit Accessibility Index; Share of 2 Auto Households 2/ County Regional Transportation Study D-7

10 Joint Household Distribution The model uses a seed matrix (derived from the 2 census) and the household size and auto marginal distributions to compute a joint household distribution by size and income group for each zone. A matrix balancing procedure is used to ensure that the marginal distributions are maintained. A further stratification of households by income is performed prior to the balancing. This stratification is used only to generate home-based work trips. Trip Production Rates The trip production rates by income, size, and number of autos are multiplied by the stratified households and summed to generate trip productions by zone. Table D.7 shows the trip rates used in this model. Table D.7: Trip Generation Rates (trips/household/day) Size Autos HBW1 HBW2 HBW3 HBSR HBSCH HBO NHBW NHBO County Regional Transportation Study D-8

11 Trip Attraction Equations Trip attractions are calculated using a set of linear equations, as follows: HBWA =.832*TEMP HBSHOP = 1.59*RetEmp +.6*NRetEmp +.93*HHLDS HBSR =.71*RetEmp +.25*NRetEmp *HHLDS HBO =.11*RetEmp +.23*NRetEmp +.23*HHLDS NHBWA =.56*RetEmp +.4*NRetEmp +.3*HHLDS NHBOA =.95*RetEmp +.16*NRetEmp +.92*HHLDS Trips are normalized to production totals for trip distribution. The values in Table D.8 are used to split the home-based work attractions by income group. Table D.8: HBW Trip Attractions by Income Area Type Low Income Mid Income High Income Total CBD 13.9% 39.44% 47.47% 1.% Fr/Sub/Rural 9.3% 53.62% 37.35% 1.% Urban 6.98% 48.36% 44.66% 1.% Average Total 9.12% 5.32% 4.56% 1.% Freight Truck trip productions and attractions were generated based on the NCHRP Quick Response Freight Manual II. The document outlines a suggested rate for a metropolitan area based on employment and households. Resulting equation for truck trips: 1.387*(Retail Employment) *(Non-Retail Employment) +.446*(Households) Trip Distribution Trip distribution for the model is based on data collected from the Census Transportation Planning Package (CTPP) and MARC Home Interview Survey (HIS). We collected the most recent (year 2 census) home-to-work trip flow information for the modeled area from the CTPP. This data was used to generate a 26 trip table for internal to internal Home-Based Work (HBW) trips. This was used to compute an observed trip length frequency distribution for calibration of our HBW distribution model. Our measure for HBW trips was a composite time and distance impedance using the following equation: Cost = Travel Time + [(Value of Time)/(Auto Operating Cost per Mile)]* Distance. Value of time is specific to trip purpose while auto operating costs were assumed to be a fixed 8.76 cents per mile. Table D.9 shows the cost per mile factor used for each purpose. 5-County Regional Transportation Study D-9

12 Trip Share Table D.9: Distance to Time Factor for Distribution Cost Trip Purpose Value of Time ($/hr) Cost per Mile (min/mile) HBW Overall $ Income 1 $ Income 2 $ Income 3 $ Home-Base Non-Work $ Non-Home-Based $ For other trip purposes, we used the MARC HIS data to determine relative average trip lengths (in minutes) compared to HBW distribution. From this we could get an approximation of what the shape of the trip-length frequency distributions (TLFD) should look like for each purpose compared to our known HBW TLFD from the census bureau data. We used the relative distributions and the average trip length estimate to calibrate the non-hbw purposes. Figures D.4-D.12 show the resulting observed and estimated trip length frequency distributions by purpose. Figure D.4: HBW High Income TLFD, Observed and Estimated Observed and Estimated TLFD, HBW, Low Income Observed Estimated 12% 1% 8% 6% 4% 2% % Composite Impedance 5-County Regional Transportation Study D-1

13 Trip Share Figure D.5: HBW Middle Income TLFD, Observed and Estimated Observed and Estimated TLFD, HBW Middle Income Observed Estimated 12% 1% 8% 6% 4% 2% % Composite Impedance 5-County Regional Transportation Study D-11

14 Trip Share Trip Share Figure D.6: HBW High Income TLFD, Observed and Estimated Observed and Estimated TLFD, HBW, High Income Observed Estimated 12% 1% 8% 6% 4% 2% % Composite Impedance Figure D.7: HBO TLFD, Observed and Estimated Observed and Estimated TLFD, HBO Observed Estimated 6% 5% 4% 3% 2% 1% % Composite Impedance 5-County Regional Transportation Study D-12

15 Trip Share Trip Share Figure D.8: HBSocial/Recreational TLFD, Observed and Estimated Observed and Estimated TLFD, HBSocRec Observed Estimated 6% 5% 4% 3% 2% 1% % Composite Impedance Figure D.9: HBShopping TLFD, Observed and Estimated Observed and Estimated TLFD, HBSHOP Observed Estimated 7% 6% 5% 4% 3% 2% 1% % Composite Impedance 5-County Regional Transportation Study D-13

16 Trip Share Trip Share Figure D.1: HBSchool (K-12) TLFD, Observed and Estimated Observed and Estimated TLFD, HBSCHOOL Observed Estimated 7% 6% 5% 4% 3% 2% 1% % Composite Impedance Figure D.11: Non Home-Based Work TLFD, Observed and Estimated Observed and Estimated TLFD, NHBW Observed Estimated 7% 6% 5% 4% 3% 2% 1% % Composite Impedance 5-County Regional Transportation Study D-14

17 Trip Share Figure D.12: Non Home-Based Other TLFD, Observed and Estimated Observed and Estimated TLFD, NHBO Observed Estimated 7% 6% 5% 4% 3% 2% 1% % Composite Impedance Table D.1 shows the final trip distribution parameters used for the friction factor function, which is a gamma function of the form: Table D.1: Trip Distribution Parameters Trip Purpose Alpha Beta Gamma HBW HBW HBW HBW HBSc HBSh HBSR HBO NHBW NHBO FF = Alpha * Imp Beta * e (Gamma*Imp) The TAZ area was divided into districts by county and according to natural geographic boundaries. K-factor adjustments were tested, but ultimately not used. 5-County Regional Transportation Study D-15

18 Mode Choice The mode choice model was borrowed from and recently calibrated by MARC. As described in the MARC travel demand model documentation: The mode choice model uses a functional form known as a nested logit model. A nested logit model recognizes the potential for something other than equal competition among modes. This structure assumes that modes, submodes, and access modes are distinctly different types of alternatives that present distinct choices to travelers. In a nested model, the lower level choices (i.e., 2-person carpool versus 3+ person carpool; or pnr versus knr) are more elastic than they would be in a multinomial model. In synthesizing a nested logit model, both the variable coefficients, as well as the nesting coefficients must be specified. These were based on extensive national experience with originally estimated models. Table D.11: Mode Choice Model Coefficients Variable Description HBW HBSR HBSHop HBO NHB In-vehicle travel time Wait time (Initial) Wait time (Transfer) Walk time (To transit) Walk time (walk mode) Drive - In vehicle travel time Bike Out-of-pocket costs Parking costs The model contains three nesting coefficients: primary nesting coefficient,.55 submode nesting coefficient,.55 access path nesting coefficient,.75 There are six elemental alternatives in the mode choice model: da drive alone 2 pers shared ride, with two persons in the auto 3+ pers shared ride, with three or more people in the auto walk walking from home to a transit stop pnr driving to a park-and-ride lot and taking transit knr being dropped off somewhere and taking transit 5-County Regional Transportation Study D-16

19 Each of these elemental alternatives was represented by a path from home to work. Separate paths were developed for the highway alternatives (da, 2 pers, and 3+ pers), as well as the transit alternatives (pnr, knr, and walk). Table D.12: 26 Mode Choice Model Vehicle Trip Output Home Based Work HBW by Income Group Home Based Social/Rec Home Based Shop Home Based Other Non- Home Based Trip Type Drive Alone 961,857 55,47 494, , , , , ,469 2 Person Auto 137,148 18,358 8,756 38,35 488, , , , Person Auto 51,622 13,387 25,485 12,75 338,21 334, ,96 274,77 Walk 25,727 1,673 1,321 4,733 16,316 19,35 9,393 18,349 Bicycle 2, ,14 1, Local Bus (walk) 28,27 19,296 7,718 1,194 12,356 6,655 5,53 7,139 Local Bus (P&R) Local Bus (K&R) Express Bus (walk) Express Bus (P&R) Express Bus (K&R) TOTAL 1,28,59 118,243 62, ,571 1,484,531 1,176,5 478,467 86,594 Time of Day To generate hourly network demand, time of day factors were applied to daily vehicle demand as output by the mode choice model. Time-of-day factors were taken from the MARC Household Travel Survey. Factors are displayed in Table D.13 below. Daily vehicle distribution by hour is shown in Figure D.13 5-County Regional Transportation Study D-17

20 Table D.13: Time Of Day Factors by Trip Purpose Time hbw P->A hbw A->P hbshop P->A hbshop A->P hbsr P- >A hbsr A- >P hbother P->A hbother A->P nhb P- >A nhb A- >P hr 1 12am-1am hr 2 1am-2am hr 3 2am-3am hr 4 3am-4am hr 5 4am-5am hr 6 5am-6am hr 7 6am-7am hr 8 7am-8am hr 9 8am-9am hr 1 9am-1am hr 11 1am-11am hr 12 11am-12pm hr 13 12pm-1pm hr 14 1pm-2pm hr 15 2pm-3pm hr 16 3pm-4pm hr 17 4pm-5pm hr 18 5pm-6pm hr 19 6pm-7pm hr 2 7pm-8pm hr 21 8pm-9pm hr 22 9pm-1pm hr 23 1pm-11pm hr 24 11pm-12am Figure D.13: Diurnal Factors: Total Vehicles by Hour Total Trucks Autos County Regional Transportation Study D-18

21 Assignment The assignment procedure is a Multi-Class assignment of auto and trucks. Demand is split by purpose into daily demand using the time of day factors listed above. The 24 hourly assignments are summed to make up daily totals. Volume Delay Functions Volume Delay Functions (VDF) were defined using Bureau of Public Roads (BPR) formulation. The functions varied depending on area type as well as facility classification. Model capacity (vehicles per hour per lane) and free flow speed (miles per hour) assumptions were borrowed from the MARC and Lawrence models and are shown in Table D.14. Table D.14: Capacity and Free Flow Speed Assumptions Hourly Lane Capacity CBD Fringe Urban Suburban Rural Interstate Expressway Major Arterial Collector / Minor Arterial Collector / Minor Arterial One-Way Free Flow Speed (MPH) CBD Fringe Urban Suburban Rural Interstate Expressway Major Arterial Collector / Minor Arterial Collector / Minor Arterial One-Way Closure Criteria for equilibrium assignment. Max 3 iterations, gap=.5 External Trips Trips at external stations were generated based on 26 traffic counts at each station as well as using applicable external station data from the MARC, Lawrence-Douglas, and Ottawa regional models. 5-County Regional Transportation Study D-19

22 Two cases were considered: External to internal trips and external to external (through) trips. Because of the size of the modeled area, we made the assumption that through trips occurred only on interstate highways and expressways. We assigned 9% of trips at each interstate external station as through trips, 3% on the expressway external stations as through trips, and the rest as internal trips. External to Internal Trip Distribution External to internal trips were distributed based on households and total employment in each zone. We relied on MARC data to compute a relative share of trips with a household-based end versus trips with an employment-based end. We used the following MARC data: HBWP_s trips by persons living outside the area coming in for work an employment-based share HBNWP_s - trips by persons living outside the area coming in for other purposes mostly shopping and entertainment, -- also employment based HBWA_s trips by residents traveling outside the area for work a household-based share HBNWA_s trips by residents traveling outside the area for other purposes a household-based share NHB_s trips made by anyone with a non-home end internally and a non-home end externally an employment-based share. The employment-based share will be: EMPB = HBWP_s + HBNWP_s + NHB_s The household-based share will be: HHB = HBWA_s + HBNWA_s Relative attractions for external to internal trips: Households + (EMPB/HHB)*Total Employment External to External Trip Distribution Though trips were distributed in the following way: 1. A subset of only stations that corresponded with interstate highways and expressways were selected. 2. Trips were assumed to be more likely to stay on the same roadway if possible, so those trips were given higher preference. 3. All other through trips were considered equally likely, so they were weighted uniformly. Trips were then balanced against the total number of through trips at each respective station. The resulting through trip matrix was then added to the external to internal trip matrix, giving us a daily external trip matrix that we factor for hourly shares using Non-Home Based diurnal factors. 5-County Regional Transportation Study D-2

23 Table D.15: External Station Trips External Station Internal Trips Through Trips Internal Trips Through Trips County Regional Transportation Study D-21

24 Special Generators (tg procedures on case by case basis, as is distribution, auto-based) A set of special generators was added to the network to supplement the socio-economic data where the standard trip generation or trip distribution was unique from the rest of the zone. Trip generation and distribution methods were considered on a case by case basis for each generator. These trips were for only the auto mode. Diurnal factors were applied based on the implied trip purpose. (KCI and the Industrial development in Lawrence were given NHBO diurnal factors) A list of these generators is found below. Table D.16: Special Generator List TAZ 26 Daily Trips 23 Daily Trips Generator City County Distribution Attractor Gravity Model Formulation 156 2,49 na Mission Farms Leawood / Overland Park Johnson Households Use HBSHOP ,421 22,564 KCI Airport Platte Households No distance 1456 na 59,89* 876, 1452, 1453, 1454 na 91, na 9,412 BNSF Intermodal Gardner Johnson Employment Use HBW3 Sunflower Development Desoto Johnson Households Use HBSHOP Unnamed Industrial Development Lawrence Douglas Employment No Distance, just relative to employment 912 na 3,643 Lawrence Airport Development Lawrence Douglas Employment Use NHBO 131, 133, 134, 135, 164, 1126, 1247, 1152, 165, 17, 161, 1246, 162, ,568 55,21 KU Campus Lawrence Douglas *BNSF Intermodal trips assigned as truck trips Households (Lawrence) LDMPO formula Application Program Model application program was developed to run the model once all required inputs are in place and calibration has been completed. The application invokes EMME from the DOS prompt. A comprehensive list of required inputs is below: 5-County Regional Transportation Study D-22

25 Table D.17: Matrix Model Inputs mo2 pctwlk Percent Walk Origins mo19 4inc 24 Avg HHLD Income mo2 spcgen Special Generator Trips 26 mo27 totpop 26 Total Population mo28 tothh 26 Total Households mo29 Zarea Zone Area (sq. mi.) mo38 trnacc Transit Accessibility mo39 autacc Auto Accessibility mo4 accrat transit/auto accessibility ratio mo52 ex2ex External to External Vehicle Trips mo53 ex2int External to Internal Vehicle Trips mo615 extext Ext-Ext Origins mo763 jocop joco other trips pr md7 HBWaLi HBW Attractions Income 1 md8 HBWaMi HBW attractions Income 2 md9 HBWaHi HBW attractions Income 3 md11 attrhh Household "attractions" for Spec Gen md21 retemp 26 Retail Employment md24 totemp 26 Total Employment md29 nremp 26 Non-Retail Employment md5 extatt External Attr (HH+.16*TotEmp) md11 pcost3 zonal parking cost - 25 & 1997 md12 pctwlk Percent Walk Destinations md615 extext Ext-Ext Destinations md763 jocoa joco other trips attr mf1 amphvd Est.AM Peak hour vehicle demand mf2 CTPP6 26 CTPP-Based HBW Person-Trips, P/A mf1 comskm composite travel cost am pk hbw mf11 cmskm1 composite travel cost ampk I1 mf12 cmskm2 composite travel cost ampk I2 mf13 cmskm3 composite travel cost ampk I3 mf14 ccstnw composite travel cost non-work mf15 ccnhb composite travel cost NHB mf16 hwydst hwydist HBW mf2 area Area of Zones (sq. mi.) mf18 Adjmtx Adjacency Matrix: 1 if zones adjacent mf181 OBHWI1 Observed HBW Inc 1 Person-trips P/A mf182 OBHWI2 Observed HBW Inc 2 Person-trips P/A 5-County Regional Transportation Study D-23

26 mf183 OBHWI3 Observed HBW Inc 3 Person-trips P/A mf41 trfare PK local transit fare mf42 fare OP Local Transit Fare mf43 trfare Express Fare mf759 Indic Indicator for JoCo addl. Trips Other batch-in files: ZoneGroup.31 ProdRatesHBO.311 ProdRatesHBSCH.311 ProdRatesHBSHOP.311 ProdRatesHBSR.311 ProdRatesHBWI1.311 ProdRatesHBWI2.311 ProdRatesHBWI3.311 ProdRatesNHBO.311 ProdRatesNHBW.311 ProdRatesTruck.311 Calibration Results Synthetic Matrix Estimation (SME) Analysis We performed a synthetic matrix estimation analysis which generated a trip table to match the observed daily count values. This analysis provided a way to check if any of our count values were inconsistent. This showed places to double check the count value for accuracy. Also, it allowed us to observe where our model was over or under-estimating daily vehicle trips. In particular, we observed that a group of zones in Johnson County was being under estimated by our model by a significant amount more than surrounding areas. This Johnson County region had a particularly high concentration of work and retail centers, so it appeared that there may have been additional mid-day trips to lunch, meetings, or other discretionary travel that were not being well represented by the model. To accommodate these, we added new trips by taking the difference (by zone) in 26 between the regular 26 assignment and the SME estimation. The additional trips were then distributed according to total employment for each zone. To estimate additional 23 trips, the growth factor was taken by comparing total employment for 26 and 23 in the zone subset then applying the growth factor to the additional 26 trips. Trips were then distributed according to total employment by zone for County Regional Transportation Study D-24

27 Table D.18: Johnson County Additional Trips TAZ 26 Daily Trips 23 Daily Trips 158 6,385 7, ,792 11, ,925 2, ,121 5, ,379 4, ,78 4, ,4 5, ,689 9, ,182 18, ,476 3, ,728 7, ,569 2, ,227 5, ,397 6, ,344 2, ,151 25, ,987 1, ,85 11,968 Percent Root Mean Squared Error (PRMSE) One way to assess the performance of the model is to compute the Percent Root Mean Squared Error of the daily modeled volumes versus observed counts. In Figure D.14 and Figure D.15 below, the roadways are grouped by daily volume or by facility type. Table D.19: PRMSE by Average Daily Count Daily Count Avg. of Count Sq. Error RMSE PRMSE % % % % % % % % % % % % % % Grand Total % 5-County Regional Transportation Study D-25

28 Figure D.14: PRMSE by Average Daily Count 7.% 6.% 5.% 4.% 3.% 2.% 1.%.% Grand Total Table D.2: PRMSE by Facility Type* Facility Type Average of count Average of sq error RMSE PRMSE Minor Arterials % Major Arterials % Expressways % Interstates % Total % *Excludes centriod connectors 5-County Regional Transportation Study D-26

29 Figure D.15: PRMSE by Facility Type 9.% 8.% 7.% 6.% 5.% 4.% 3.% 2.% 1.%.% Minor Arterials Major Arterials Expressways Interstates Total Screen lines were defined as a way to look at regional East-West and North-South flows in the study area. Screen line volumes include all traffic from each roadway crossing the screen line. These volumes were compared against daily traffic counts from KDOT. Figure D16 shows the two screen lines defined for our study area. The East-to-West movements in the study area are captured by Screen Line 1, which cuts through Leavenworth, Douglas, and Miami Counties just to the west of the Johnson / Douglas county line. Major facilities that are included by this screen line include: US-24, I-7, K-1, US- 56, I-35, and US-169. North-to-South movements are observed crossing the Kansas River. Major roadways that cross this screen line include: I-7, K-7, I-635, and I County Regional Transportation Study D-27

30 Figure D.16: Screenlines Screen Line 1 Screen Line 2 5-County Regional Transportation Study D-28

31 Table D.21: Screenline 1 Volumes Roadway Name 26 ADT Count 26 Daily Volume 23 Daily Volume Growth K % Millwood Rd ,2 34% K-192 1,54 1,31 1,64 25% K-92 1, ,2 24% Dempsey Rd. 1,4 1,42 2,3 43% Parallel Rd % Toganoxie Rd. 3, 2,79 4,36 56% US 24/4 11,5 6,27 8,49 35% Evans Rd. 1,16 2,87 4,1 4% I-7 29,5 3,54 67,46 121% K-32 3,18 2,63 4,88 86% K-1 27,8 41,56 64,91 56% 143rd St. 66 4,39 11,68 166% US-56 4,4 7,1 1,89 53% I-35 2,6 25,3 32,9 27% 231st St % 247th St. 12 1,22 1,77 45% K-68 2,75 2,17 2,83 3% Jackson Rd % John Brown Rd ,2 33% 379th St % US-169 3,79 3,3 4,8 24% TOTAL 115,89 136,8 226,27 65% Table D.22: Screenline 2 Volumes Roadway Name 26 ADT Count 26 Model Output 23 Model Growth Lecompton Rd. 4,49 4,49 5,75 28% I-7 3, 42,35 61,35 45% US-59 23,74 25,2 29,2 16% E 22 Rd. 2,47 4,79 9,75 14% Wyandotte St. 2,36 4,44 1,25 131% K-7 19, 22,21 35,94 62% I ,1 57,3 85,88 51% Turner Diagonal Fwy. 16,3 8,2 13,74 68% I-635 8, 73,62 86,35 17% 18th Expy 36,9 38,34 42,27 1% 1th St. 8,78 2,45 2,64 8% 7th St. Trfy 21,8 15,89 17,46 1% I-67 5,9 82,34 94,3 15% James St. 5,86 2,99 3,89 3% I-7 53,9 35,49 39,42 11% TOTAL 421,6 419,83 538,19 28% Tables D21 and D22 shows all of the modeled roadways that are included as part of the screen line analysis. 5-County Regional Transportation Study D-29

32 Between 26 and 23, volume along Screen Line 1 is forecast to increase by 85, trips per day, an increase of 65%. Screen Line 2 is forecast to increase by 118, trips per day, which is a 28% increase. The growth in the north to south movement is affected primarily by large growth on I-435, as well as along I-635 and I-67. The east to west growth is mainly a function of large gains in daily volume between I-7 and K- 1. Overall observed daily counts when compared with estimated 26 daily volumes are displayed below in Figure D County Regional Transportation Study D-3

33 Figure D.17: Modeled vs. Daily Count Link Scattergram 5-County Regional Transportation Study D-31