Plant City Travel Demand Model

Sketch-Level Subarea TDM: Gravity Distribution, BPR Assignment, Congestion Feedback & External Trips

Author

Planning Team

Published

March 5, 2026

NoteModel Scope

This model includes internal-internal (I-I) trips distributed via doubly-constrained gravity, plus external (EI/IE/EE) trips loaded directly from the cordon survey OD table. Internal trip ends are factored by zone-level internal shares before gravity so productions and attractions represent only trips staying within the study area. External gate zones (IDs ≥ 150) are connected to the network via snap connectors built from gate centroids.

ImportantPrerequisites

Run 00_prep_network.R before rendering. It builds the directed network, reuses curated centroid connectors from the standardized shapefile (with spatial crosswalk to our 149-zone system), removes phantom/future links (to_remove=1, Selmon Expressway toll plaza links, and I-4 express/managed lane links), reclassifies retained toll links as Freeway, aligns freeway per-lane capacity to TBRPM v10 values, builds gate connectors for external zones, and routes the free-flow OD skim. Everything else — including TBRPM benchmark matching — runs inline during this render.

1 Introduction

This document implements a sketch-level subarea travel demand model for Plant City (149 internal zones), informed by NCHRP 716/365 guidance and calibrated against TBRPM and Replica benchmarks.

Step Method
Trip Generation TBRPM rates, 7 purposes — trip ends factored to internal share only
Trip Distribution Doubly-constrained gravity, gamma friction (NCHRP functional form)
Demand Scaling Person-trips scaled to TBRPM vehicle-trip control total (aggregate scalar)
External Trips EI/IE from Replica cordon × 2.0 (select-link correction); EE × 0.80 (FDOT-informed)
Traffic Assignment MSA with BPR volume-delay (\(\alpha/\beta\) by facility type — FDOT-calibrated)
Feedback 3-iteration congestion feedback: re-skim → re-distribute (I-I only)
NoteKey Assumptions

Demand scaling: Internal person-trips converted to vehicle-trips using a single aggregate scalar from the TBRPM control total (199,206 interzonal vehicle-trips), implicitly accounting for mode share and vehicle occupancy.

Friction factors: Gamma \(a\) and \(b\) transferred from NCHRP 365 Table 5.5 (small/medium urban areas); \(c\) calibrated locally to match TBRPM average trip lengths by purpose.

The travel time skim uses routed shortest paths on the actual network. A distribution feedback loop re-skims on congested BPR times so gravity trip lengths respond to real network conditions.

2 Step 0: Inputs

2.1 Zone System

Figure 1: 149-zone system (internal zones only)
Table 1: Zone system summary
Zones Min Area (sq mi) Median Area Max Area Total Area
149 0.03 0.93 35.27 297

2.2 Network

The study-area network is derived from the TBRPM regional model. Link attributes include facility class, free-flow speed, lane count, and hourly capacity. No intersection control, turn penalties, or signal timing.

Table 2: Network summary by facility class (excluding centroid connectors, Selmon + I-4 express toll links, and to_remove links; remaining toll links reclassified as Freeway)
Facility Dir. Edges Total Miles Avg Speed (mph) Avg Hourly Cap
Collector 862 450.7 29.6 637
Divided Arterial 343 164.6 39.0 1,979
Undivided Arterial 250 141.9 36.3 894
Freeway 134 56.6 62.6 5,630
Ramp 53 20.8 29.2 1,855
One-Way Arterial 22 3.7 32.0 1,390
Table 3: Reverse-record completeness by facility class
Facility Edges Has Reverse Missing Rev % Complete
Freeway 134 0 134 0.0
Ramp 53 0 53 0.0
One-Way Arterial 22 0 22 0.0
Divided Arterial 343 330 13 96.2
Collector 862 862 0 100.0
Undivided Arterial 250 250 0 100.0
WarningReverse-Record Gap: 222 of 1664 edges missing reverse

Edges without a reverse record are effectively one-way in assignment. Bidirectional volume aggregation reports only one direction, likely underestimating two-way volume by ~50%% on affected links.

Figure 2: Road network colored by facility class

2.3 External Gate Connectors

Figure 3: External gate connectors — highest class within snap radius, then expand
Table 4: External gate connector summary — functional class of snap target

2.4 Routed Skim

Table 5: Routed skim summary (free-flow, internal zones only)
Type Pairs Min (min) Median Mean Max
Interzonal 22052 1.5 18.3 19 60.5
Intrazonal 149 0.4 2.5 3 15.4
Figure 4: Distribution of interzonal routed travel times (internal zones)

3 Step 1: Trip Generation

Internal and external demand are handled separately. Internal demand uses TBRPM production/attraction rates with gravity distribution; external demand uses Replica gateway movements. Each component is anchored to its most reliable control total.

3.1 External Share Data

Table 6: Internal share summary across 149 zones
Zones Avg Prod % Internal Min Prod % Internal Max Prod % Internal Avg Attr % Internal Min Attr % Internal Max Attr % Internal
149 69 30.6 91.6 68.1 33.1 88.4
Figure 5: Internal production share by zone — zones with lower % send more trips outside the study area

3.2 Raw Trip Ends

Table 7: Raw trip generation by purpose (before external factoring)
Purpose Type Productions Attractions P/A Ratio
HBW HB 90,725 76,274 1.189
HBSH HB 37,447 36,345 1.030
HBSR HB 104,833 116,663 0.899
HBSC HB 58,665 49,845 1.177
HBO HB 78,988 54,571 1.447
NHBW NHB 18,698 18,973 0.986
NHBO NHB 25,332 26,387 0.960
**TOTAL** — 414,688 379,058 1.094

3.3 Factor Trip Ends to Internal Only

Table 8: Factored trip ends — internal trips only
Purpose Raw Prod Internal Prod Reduction Raw Attr Internal Attr Attr Reduction
HBW 90,725 64,276 29.2% 76,274 53,947 29.3%
HBSH 37,447 26,668 28.8% 36,345 25,701 29.3%
HBSR 104,833 74,342 29.1% 116,663 82,654 29.2%
HBSC 58,665 41,691 28.9% 49,845 36,807 26.2%
HBO 78,988 56,123 28.9% 54,571 38,566 29.3%
NHBW 18,698 13,034 30.3% 18,973 13,366 29.6%
NHBO 25,332 17,380 31.4% 26,387 18,740 29%
Figure 6: Total trip productions by zone (factored — internal only)

4 Step 2: Trip Distribution

The gravity model uses a gamma friction function calibrated on the free-flow skim. Only the factored (internal-only) trip ends feed into gravity.

\[F(t) = a \cdot t^b \cdot e^{ct}\]

4.1 Core Gravity Functions

4.2 Feedback Loop Helper Functions

4.3 TBRPM Benchmark

NoteTBRPM Demand Calibration
  • I-I control total: 199,206 interzonal vehicle-trips (from TBRPM trip tables: AM+MD+PM+EV, SOV+HOV+TRK across 149 TAZs)
  • TBRPM OD RDS: 124,140 interzonal trips across 114 zones (partial — used for distribution benchmark only)
  • External trips: EI/IE × 2 (select-link correction); EE × 0.8 (FDOT count-informed)

4.4 Gamma Parameters

4.5 One-Time Calibration

Table 9: Locked gamma friction parameters (calibrated on free-flow skim)
Purpose Type a b c (calibrated)
HBW HB 28,507 -0.02 -0.2293
HBSH HB 29,013 -0.40 -0.5000
HBSR HB 63,000 -0.70 -0.5000
HBSC HB 29,013 -0.40 -0.5000
HBO HB 29,013 -0.35 -0.5000
NHBW NHB 139,173 -0.80 -0.3144
NHBO NHB 219,113 -0.90 -0.5000

5 Step 3: External Trips

5.1 Load & Summarise

Table 10: External demand — Replica raw vs TBRPM-scaled
Flow Type Replica Raw Scaled Factor
EI (enter) 53,957 107,914 2.00
IE (leave) 50,877 101,754 2.00
EE (through) 124,205 99,364 0.80
**TOTAL** 229,039 309,032 1.35
ImportantExternal Demand Scaling — Rationale and Evidence

EI/IE scaled 2× to correct for Replica select-link methodology effects (highway class filter, block-group area weighting, regex prefilter).

EE scaled 0.8× to align freeway through-traffic with FDOT ground-truth counts. Prior to this adjustment, I-4 volumes exceeded 2020-adjusted FDOT AADT by ~25-30%. EE trips (dominated by I-4 Z150↔︎Z155 through-movements) were the primary contributor.

Sensitivity note: External trips constitute a majority of total assigned demand. Users should test sensitivity by varying these factors (EI/IE: 1.5–2.5 range; EE: 0.75–1.0 range).

Table 11: External trip summary by flow type (after TBRPM scaling)
Flow Type OD Pairs Total Trips Avg Trips/Pair Unique Internal Zones
ENTER_STUDY 1621 107,914 66.6 149
LEAVE_STUDY 1540 101,754 66.1 149
THROUGH_STUDY 137 99,364 725.3 0
Table 12: EI/IE correction factor — audit trail
Metric Value Notes
Replica raw EI trips 53,957 Select-link: enter study area
Replica raw IE trips 50,877 Select-link: leave study area
Replica raw EE (through) trips 124,205 Select-link: pass through
Replica raw total (all flow types) 229,039
EI/IE correction factor ×2 Corrects for select-link undercount
EE scale factor ×0.8 FDOT count-informed reduction
Scaled EI trips 107,914 53,957 × 2
Scaled IE trips 101,754 50,877 × 2
Scaled EE trips 99,364 124,205 × 0.8
**Scaled external total** 309,032 Scaled external OD total (before same-gate cleanup)

5.2 Check Network Coverage

NoteExternal Trip Network Coverage
  • Total external trips in OD table: 309,032
  • Assignable (both endpoints have network nodes, distinct O/D): 307,738 (99.6%)
  • Same-gate self-loops dropped: 1,294 EE trips where origin gate = destination gate (unroutable by definition — cordon survey artifact)
  • Unassignable (missing gate connector): 0

6 Step 4: Demand Scaling

TipDemand Scaling to TBRPM Control Totals

All demand calibrated to TBRPM v10 2020 trip-table magnitudes:

  • I-I: gravity person-trips scaled to 199,206 interzonal vehicle-trips
  • EI/IE: Replica cordon trips × 2 to correct for select-link methodology undercount
  • EE: Replica through-trips × 0.8 (FDOT count-informed reduction)

Replica provides the spatial distribution (gate × zone); TBRPM provides magnitude.

7 Step 5: Traffic Assignment & Feedback

Trips are loaded using shortest-path routing with BPR volume-delay functions and MSA averaging across 25 iterations per feedback pass. Each feedback iteration assigns I-I gravity trips and external OD trips together; only the I-I distribution updates as the congested skim evolves.

7.1 BPR Volume-Delay

\[t(V) = t_0 \left(1 + \alpha \left(\frac{V}{C}\right)^\beta\right)\]

where \(\alpha\) and \(\beta\) vary by facility type (see table below).

MSA: \(V^{(n)} = V^{(n-1)} + \frac{1}{n}\left(V^{AON} - V^{(n-1)}\right)\)

Table 13: BPR volume-delay parameters by facility type (FDOT BDK83-977-14)
Facility α β
Freeway 0.263 6.869
Ramp 0.263 6.869
Divided Arterial 0.240 7.895
Undivided Arterial 0.240 7.895
One-Way Arterial 0.240 7.895
Collector 0.215 8.135

Parameters \(\alpha\) and \(\beta\) are from FDOT Research Project BDK83-977-14 (Moses & Mtoi, Dec 2013), Table 3 — “Estimated VDF Parameters” for the modified BPR function. Values reflect Florida-specific calibration against TTMS speed-volume data. Urban area type used for freeways; signalized arterial classes used for interrupted-flow facilities. The higher \(\beta\) values (5–9 range vs. the standard 4.0) produce steeper speed drops near capacity, which is consistent with observed Florida traffic behavior and better matches the TBRPM regional model’s congestion response.

NotePre-Assignment QC
  • Road edges: 1,664
  • Internal zone connectors: 730
  • Gate connectors: 76 (38 gates × 2 directions)
  • K-factor: 0.095 → daily capacity ≈ 10.5× hourly
  • Centroid passthrough prevention: 30-min penalty on centroid/gate connector edges — makes centroid passthrough unattractive in normal cases (Snap Connectors are not penalized; they are network repair links)
  • Centroid passthrough QC: PASS — 50 paths sampled, no intermediate Z* nodes found
  • Internal path test (Z1 → Z10): 12/12 edges matched
  • Gate path test (Z150 → Z5): 27 edges — gate connector working

7.2 Feedback Loop

7.3 Feedback Convergence

Table 14: Feedback loop — demand composition and system performance by iteration
FB Iter Skim I-I Trips Ext Trips Scale Factor Avg TT (min) VMT Max V/C MSA Iters
1 Free-flow 197,842 307,738 0.8339 19.00 4,474,454 1.609 21
2 Congested 197,897 307,738 0.8391 20.54 4,456,956 1.620 16
3 Congested 197,897 307,738 0.8386 20.53 4,457,297 1.621 16

7.4 Assignment Diagnostics

ImportantAssignment Loading Check (Final Iteration)
  • I-I demand: 197,897 vehicle-trips
  • External demand (assigned): 307,738 vehicle-trips (99.6% of cordon total)
  • Total assigned: 505,634
  • Unroutable demand: 0 vehicle-trips (0 OD pairs with no network path)
  • Road links with volume > 1: 1,509
  • Feedback iterations: 3 | Final MSA iterations: 16
Table 15: MSA convergence log (final feedback iteration)
Iter Rel Gap Max V/C
1 0.000e+00 4.901
2 8.441e-01 7.960
3 9.926e-01 5.307
4 9.125e-01 3.980
5 5.777e-01 3.184
6 2.093e-01 2.653
7 6.155e-02 2.274
8 1.697e-02 1.990
9 8.034e-03 1.769
10 4.134e-03 1.592
11 2.652e-03 1.545
12 2.090e-03 1.504
13 1.764e-03 1.494
14 1.517e-03 1.529
15 1.198e-03 1.524
16 2.734e-04 1.621

The relative gap is computed as |TSTT_AON − TSTT_current| / TSTT_current, where both total system travel times are evaluated on the post-update cost vector. In a strict Beckmann equilibrium formulation (e.g., Frank-Wolfe), the AON assignment and gap would use the pre-update costs. For MSA the step size is fixed at 1/n regardless of the gap, so this distinction does not affect the assignment result — the gap is a monitoring metric, not a control variable. The monotonically decreasing gap confirms convergence.

Relative gap also includes penalized centroid/gate connectors. Because connector capacity is unconstrained (BPR cost ≈ t0), the penalty adds a nearly constant term to both AON and current total costs; the difference (numerator) is driven by road link cost changes. The inflated denominator may make the gap appear slightly smaller than a road-only calculation, but convergence behavior is unaffected. Reported travel times, VHT, and speeds in the results section exclude connectors entirely and are not affected by the penalty.

Figure 7: MSA relative gap by iteration (final feedback pass)

7.5 Distribution Benchmark Comparison (I-I Only)

Gravity model OD flows are compared against the TBRPM v10 and Replica OD tables as consistency checks. These are model-to-model comparisons, not validation against observed data. Correlation indicates whether the gravity model produces spatial patterns consistent with the regional models.

Table 16: Benchmark consistency — I-I OD flows (final feedback iteration)
Comparison r (OD) r (Zone)
Gravity vs TBRPM 0.789 0.801
Gravity vs Replica 0.716 0.843
TBRPM vs Replica 0.779 0.858
Figure 8: Trip length distribution — gravity model vs benchmark models

7.6 Distribution QC Checks

Table 17: Distribution quality checks
Check Value Status
Intrazonal trip share 19.1% (target: 10–20%) OK
I-I interzonal vehicle-trips 199,206 —
External vehicle-trips (assigned) 307,738 —
External share of total demand 60.7% —
Gravity ATL (free-flow skim) 9.2 min —
TBRPM benchmark ATL (free-flow skim) 9.5 min —
Figure 9: Demand composition — I-I vs external trip types

7.7 Flowmap — Internal Trips Only

Figure 10: I-I OD flow map — internal zones only (external gate trips excluded)

8 Assignment Results

Table 18: Volume summary by facility class (I-I + external combined)
Facility Links Total Vol Avg Vol Max Vol Avg V/C Max V/C
Collector 399 1,481,399 3713 19,212 0.288 1.335
Divided Arterial 161 2,327,630 14457 34,820 0.371 1.041
Undivided Arterial 114 1,038,353 9108 27,742 0.544 1.312
Freeway 104 4,175,505 40149 82,473 0.634 1.274
Ramp 53 281,349 5308 17,430 0.273 0.852
One-Way Arterial 22 200,356 9107 14,744 0.624 0.985
Table 19: V/C distribution
V/C Range n %
<0.50 591 69.3
0.50–0.75 109 12.8
0.75–0.90 59 6.9
0.90–1.00 39 4.6
1.00–1.25 45 5.3
>1.25 10 1.2
Table 20: System performance
Measure Value
Daily VMT 4,457,296 mi
Daily VHT 122,534 hrs
Avg Network Speed 36.4 mph
Loaded Road Links 853
I-I Vehicle-Trips 197,897
External Vehicle-Trips (assigned) 307,738
Total Assigned 505,634

8.1 Maps

Figure 11: Assigned daily volumes (I-I + external)
Figure 12: Volume-to-capacity ratio
Figure 13: Congested speed

8.2 Top Loaded Corridors

9 TBRPM Link-Level Benchmark Comparison

Model volumes are compared against the TBRPM v10 2020 loaded network as a benchmark consistency check (model-to-model, not observed data). Links are matched using line-to-line nearest-feature matching (robust to node renumbering and link splitting during network standardization). Only high-confidence matches are retained: <200 ft snap distance, exact facility class match, lane count within ±1. Volume comparison uses two facility-appropriate modes: bidirectional sum for collectors and arterials (both networks have paired AB/BA records), and directional for freeways and one-way arterials (both networks have single-direction records; TBRPM reverse records are dummies with FACL_TYPE=0). Ramps and toll facilities excluded; TBRPM dummy links (FACL_TYPE=0) excluded.

Interpreting fit. This is a reasonableness check on spatial loading patterns and corridor magnitudes, not a link-level validation. A 149-zone model will inherently produce less spatial differentiation than a ~4,000-zone regional model — this manifests as volume compression (less spread between high- and low-volume links of the same class).

Table 23: Facility class cross-tabulation — our model vs TBRPM (benchmark)
Our Facility TBRPM Facility n
Collector Collector 240
Divided Arterial Divided Arterial 74
Undivided Arterial Undivided Arterial 64
Freeway Freeway 17
One-Way Arterial One-Way Arterial 15
Table 24: Volume correlation and %RMSE by facility class — our model vs TBRPM (benchmark)
Facility Links r RMSE %RMSE Our Avg TBRPM Avg
Collector 240 0.328 4,052 83.2 3,221 4,872
Divided Arterial 74 0.184 14,388 53.8 19,995 26,734
Undivided Arterial 64 0.059 6,767 67.1 10,441 10,088
Freeway 17 0.831 4,259 5.7 72,359 74,396
One-Way Arterial 15 0.924 2,242 29.6 9,372 7,563
ImportantInterpreting Correlation Results

Pooled across all 410 matched links, overall r = 0.901. This headline number is inflated by cross-facility scale differences (freeways at 80k vs collectors at 3k). Within-class correlations: One-Way Arterial r = 0.924 (15 links), Collector r = 0.328, Divided Arterial r = 0.184, Undivided Arterial r = 0.059. Lower-class correlations reflect volume compression from the coarser 149-zone structure compared to TBRPM’s ~4,000 zones.

Figure 14: Model volume vs TBRPM volume (bidir for arterials/collectors, directional for freeways)

9.0.1 TBRPM Loaded Volume Map

Figure 15: TBRPM v10 loaded volumes in study area (clipped, all non-connector facilities)

9.0.2 Volume Difference Map — Our Model vs TBRPM

Matched links are mapped by absolute difference (model − TBRPM). Blue indicates our model is lower than TBRPM; red indicates higher. Colour breaks use Jenks natural breaks classification to highlight the natural clustering in the difference distribution. This map uses the same high-confidence matched link set as the correlation tables above (< 200 ft snap, exact facility class, lanes ± 1).

Figure 16: Absolute volume difference (Model − TBRPM) by matched link — Jenks natural breaks
Table 25: Screenline comparison — directional model volume vs TBRPM by screenline
Screenline Links Our Vol TBRPM Vol Ratio
4 6 99,958 87,645 1.14
7 1 10,510 6,195 1.70
15 6 9,592 18,410 0.52
18 1 1,342 3,320 0.40
22 8 183,573 185,965 0.99

10 FDOT Count Station Comparison

Unlike the TBRPM benchmark in the preceding section (which is a model-to-model consistency check), this section compares model volumes against observed FDOT traffic counts — the closest available ground truth for volume magnitude. Sixteen FDOT count stations within the study area are matched to specific network links. FDOT AADT is bidirectional for two-way facilities; for one-way facilities (D Factor ≈ 100%), the AADT effectively represents directional volume. All FDOT counts are 2024 AADT; the model represents a 2020 base year, so some growth-related discrepancy is expected. Polk Parkway links are classified as Toll Facility in the source network but reclassified as Freeway in our model.

Table 27: Model volume vs FDOT AADT at count stations
Site Road FDOT Site Type Facility FDOT AADT Model Vol Ratio % Diff
1 US-92 (W of Turkey Creek) 100080 Telemetered Undivided Arterial 14,286 14,362 1.01 +0.5%
2 I-4 (under Bethlehem Rd) 109962 Telemetered Freeway 141,500 160,264 1.13 +13.3%
3 US-92/E Baker St (W of Maryland) 105008 Portable Divided Arterial 13,800 13,085 0.95 -5.2%
4 SR-39/S Alexander St 108621 Portable Divided Arterial 30,500 27,570 0.90 -9.6%
5 Coronet Rd (E of Park Rd) 109191 Portable Collector 7,300 4,448 0.61 -39.1%
6 Charlie Taylor Rd 109618 Portable Collector 3,300 1,023 0.31 -69%
7 SR-39/S James L Redman Pky 106009 Portable Divided Arterial 25,500 20,753 0.81 -18.6%
8 Sydney Rd (W of S Forbes Rd) 109160 Portable Collector 6,900 2,315 0.34 -66.5%
9 E Baker St (E of Collins St) 105290 Portable One-Way Arterial 9,700 7,622 0.79 -21.4%
10 N Wheeler St (N of Baker St) 105006 Portable Undivided Arterial 6,300 7,677 1.22 +21.9%
11 E Reynolds St (EB, W of Gordon) 105042 Portable One-Way Arterial 8,100 6,043 0.75 -25.4%
12 SR-39/Buchman Hwy (N of I-4) 106017 Portable Undivided Arterial 10,200 11,077 1.09 +8.6%
13 I-4 (SW of SR 546/Memorial) 161005 Portable Freeway 120,500 120,155 1.00 -0.3%
14 I-4 (W of Countyline Rd) 100084 Portable Freeway 149,000 142,833 0.96 -4.1%
15 Polk Pkwy (E of CR-542 ramp) 974015 Portable Freeway 40,100 43,991 1.10 +9.7%
16 E Park Rd (E of Redman Hwy) 105720 Portable Collector 6,500 13,022 2.00 +100.3%
Figure 17: Model bidir volume vs FDOT AADT (16 count stations)
NoteFDOT Count Comparison Summary

Across 16 count stations, the mean model/AADT ratio is 0.94 with %RMSE of 15.4%. FDOT counts are 2024 AADT while the model represents 2020 conditions; approximately 4 years of background growth (typically 1–2% per year for this area) would account for some of the difference. The I-4 telemetered site (Site 2) provides the highest-confidence comparison given its permanent installation and high traffic volume.

Figure 18: FDOT count station locations
NoteCount Station Data Source

FDOT count data from the Traffic Data Application (TDA): Portable sites and Telemetered sites. Station-to-link matching was performed manually by spatial proximity and road name. All counts are 2024 AADT (bidirectional); model volumes represent 2020 base-year daily demand.

11 Limitations

Network simplifications. No intersection control, turn penalties, or signal timing. BPR congestion with MSA averaging; not a converged user-equilibrium solution. Results are corridor-level magnitudes, not link-specific forecasts.

Aggregate scaling. Mode split, time-of-day, and vehicle occupancy handled via aggregate scalars from the regional model — not explicit choice models.

Transferred parameters. Gamma friction from NCHRP 365 (\(c\) calibrated locally). BPR α/β from FDOT BDK83-977-14.

External demand. External trips are a majority of assigned demand, controlled by a ×2.0 EI/IE factor and ×0.80 EE factor. Users should test sensitivity (EI/IE: 1.5–2.5; EE: 0.75–1.0).

Intended use. Corridor screening, comparative scenario testing, and order-of-magnitude reasonableness checks. Not suitable for intersection operations, parcel-level impact assessment, or mode choice applications.

12 Summary

Table 28: Model steps and results
Step Method Result
Trip Generation TBRPM rates, 7 purposes — trip ends factored to internal share per zone 237,534 interzonal I-I person-trips
Trip Distribution Doubly-constrained gravity, gamma friction (NCHRP 365 Table 5.5), 3-iteration congestion feedback r = 0.789 vs TBRPM
Demand Scaling Aggregate person→vehicle scalar to TBRPM control total (factor = 0.8386) 199,206 I-I vehicle-trips
External Trips 38 gate zones; EI/IE × 2; EE × 0.8 307,738 vehicle-trips (99.6% of cordon total assigned)
Traffic Assignment MSA × 16 iters (final pass), BPR (FDOT-calibrated α/β by facility) VMT = 4,457,296 | Avg speed = 36.4 mph
TBRPM Benchmark 410 TBRPM-matched links across 5 facility classes TBRPM benchmark: mixed within-class fit (FW r=0.831 core, Coll r=0.447 core, DA r=0.347 core, UA r=0.69 core); overall r dominated by cross-facility scale differences
FDOT Counts 16 FDOT count stations (2024 AADT vs 2020 model) Mean ratio = 0.94 | %RMSE = 15.4%

Output files in out_quick_gravity/:

File Description
od_internal.csv I-I OD matrix (final feedback iteration, scaled to TBRPM control total)
od_external.csv EI/IE/EE OD table (from cordon survey, vehicle-trips)
assigned_links.csv Link volumes, V/C, loaded speeds
gate_connectors.csv Gate snap connector lookup (gate ID → network node)
routed_skim.csv Network-routed free-flow travel times (internal zones)
travel_time_matrix.csv Final congested zone-to-zone travel times (after feedback)
validation_matched_links.csv TBRPM benchmark matched links (if TBRPM shapefile available)
prepped_edges.rds Directed edge table