Multiscale Carbon Burden of Infrastructure in the United States
WCTR, Toulouse, France
2026-07-08
Overview
- Study land use & climate mitigation relationship by combining Vulcan 1-km gridded fossil fuel CO2 (FFCO2) & EPA Smart Location Database land use features
- Partition emissions into transportation, residential energy, & scope 2 residential electricity
Motivation
Newman & Kenworthy (1989)
Literature
- Urban economics: Glaeser & Kahn (2008) used an ad-hoc CO2 estimation strategy to compare 66 metropolitan areas
- Lowest CO2 rates in California
- Highest CO2 rates in Texas & Oklahoma
- Development restrictions push new development into high CO2 regions
- Urban planning: Kockelman (1997) + Cervero & Ewing (2010) examine 3/5Ds of land use
- Density, Diversity, Design, Destination Access, & Distance to Transit
- Urban science: Fragkias et al. (2013) used earlier version of Vulcan CO2 dataset (1999-2008) to study urban scaling in total CO2
- Found proportional scaling of 0.95% increased CO2 per 1% increase in population – i.e., larger cities not significantly more efficient
Model Development Process
flowchart TD
A[Input FFCO2 by sector] –> B[Add climate & 5D land use data]
B –> C[Correlation analysis
Identify redundant variables
for GPS construction]
C –> D[Confounder analysis using DAGs
Assess strata-sum importance scores]
D –> E[Test GPS functions
GLM, BART, etc.]
E –> F[Estimate doubly robust
outcome model with BART]
F –> G[Estimate outcome model
for all 5D treatment variables
with linear regression]
Confounder Identification
Causal Identification Strategy
Doubly Robust Outcome Model
\[\begin{split} \ln(CO_2 \text{ per capita}) =& \text{ln(total CBSA population)} + \text{ln(CBSA population density)} \\&+ \text{ln(average CBG treatment level in CBSA)} + \text{propensity score} \\&+ \text{CBSA fixed effect} \end{split}\]
\[\hat{\mu}_{i}^{DR}(t) = m(t,X_i) + w_i(t)\left\{Y_i - m(T_i,X_i)\right\}\]
Transportation Results
Residential Electricity Results
Residential Energy Results
Discussion
Next Steps
Thank you
Questions? Reach out any time.
Presentation & paper available:
https://hawkins-tech-lab.github.io/
