Urban Mobility Carbon Stress-Tester

Concept

A B2B scenario engine for urban planners to run 'what-if' stress tests on road transport emissions. Users vary socio-economic drivers (population growth, new residential zones, employment shifts) and policy levers (congestion pricing, transit service levels, zoning), and the tool predicts the resulting travel demand and GHG (CO2, CH4) impact — always as a range with a confidence interval, never a single point estimate.

The modeling backbone follows the DRIVE v1.0 architecture [S1]: a multi-modal macroscopic traffic model calibrated with vehicle-class-specific counting data and standard emission factors (HBEFA), producing hourly, road-link-level estimates with quantified activity uncertainty. On top of this, an ensemble/Monte Carlo uncertainty-propagation layer [S2] carries traffic-model calibration and validation errors through to the emission outputs, so every scenario reports how confident it is — and runs fast enough for interactive scenario iteration. Transit policy levers include a service-quality feedback module: because overcrowded buses push riders into private cars [S3], mode-shift estimates incorporate bus occupancy/comfort predictions (e.g., Markov-chain demand forecasting), so a service-expansion scenario is only credited with emission reductions if the added capacity actually absorbs demand comfortably.

Peer-city comparative data is used to calibrate scenario assumptions (e.g., observed demand responses to surges or pricing in similar metropolises) rather than to generate a standalone benchmark score.

Why now

• DRIVE v1.0 [S1] demonstrates that a macroscopic traffic model plus counting-station data and HBEFA factors can deliver hourly, link-level GHG estimates with quantified uncertainty — and that it captures irregular events such as COVID lockdowns, exactly the surge behavior a stress-tester must reproduce.
• Uncertainty-propagation research [S2] provides a working method (ensemble optimization plus Monte Carlo sampling) to turn traffic-data uncertainty into honest emission confidence intervals — the credibility feature that separates a decision-support tool from a black box.
• Markov-chain occupancy/comfort prediction frameworks [S3] make it feasible to model the transit-quality-to-private-car feedback inside policy scenarios, addressing a known blind spot in purely fare/frequency-based mode-shift models.

Honest limits and how the tool adapts

Two recent studies bound what this tool can credibly claim:

• Accuracy degrades precisely where stress tests operate. A coupled macroscopic traffic-emission system for Barcelona [C2] found that discrepancies versus a microscopic model grow with congestion level (up to +65% for NOx), and that outputs are highly sensitive to spatially resolved fleet composition and meteorology (+19% NOx for diesel effects). Demand-surge scenarios sit in this high-congestion regime. Adaptations: (1) city-specific fleet composition is a mandatory input, not an assumption; (2) uncertainty bands widen automatically for surge scenarios; (3) hotspot links flagged as high-congestion are routed to microscopic validation before results inform zoning decisions. Headline claims are scoped to CO2/CH4, whose speed sensitivity is smoother than that of NOx/PM.
• Air-pollutant co-benefits cannot be promised. Non-exhaust sources (tire, brake, road wear, resuspension) carry very high uncertainties — 80% of total PM in Barcelona [C2] — and street-scale models systematically underestimate local concentrations, as shown for black carbon in Paris [C1]. The tool therefore reports GHG outcomes as its primary output; NOx/PM figures are labeled indicative, and street-canyon exposure analysis is explicitly out of scope.

Open questions

• How large is the macro-vs-microscopic discrepancy for CO2 specifically (versus the documented NOx/PM gap), and can a hybrid correction be calibrated cheaply per city?
• Can open counting data capture behavioral responses to pricing, or do pricing elasticities always require local stated-preference surveys?
• Does induced demand (new capacity creating more traffic) systematically bias scenario forecasts, and can peer-city before/after data correct for it?
• Will planners act on interval forecasts for high-stakes zoning decisions, and does uncertainty framing increase or reduce trust versus legacy point-estimate models?
• What is municipal willingness to pay versus free academic frameworks (DRIVE itself is publishable research) or internal legacy systems?
• Are there regulatory 'certified forecast' requirements that would exclude inferred origin-destination demand?