A Synthetic Dataset for Decarbonization Policy: Integrating Monte Carlo Simulations and Rebound Effects
Description
This research hypothesizes that traditional, deterministic Marginal Abatement Cost Curves (MACC) consistently overestimate the mitigation potential of urban decarbonization strategies by ignoring parametric uncertainty and behavioral feedback, particularly the rebound effect. We argue that a data-driven, probabilistic framework that combines Monte Carlo (MC) simulations and Bayesian Networks (BN) is crucial for assessing the vulnerability of "win-win" measures and for establishing a more robust basis for climate policy in Smart Cities. This dataset includes a comprehensive set of technical, economic, and environmental variables for nine urban mitigation strategies across the residential, commercial, and transportation sectors. The data were collected using a bottom-up approach that combined documented scientific literature with real-world urban case studies. These inputs were analyzed with a probabilistic engine to produce empirical distributions of the Marginal Abatement Cost (MAC) and other performance indicators, replacing static point estimates with a multivariate probabilistic knowledge structure.
Files
Steps to reproduce
Summary of the Methodological Workflow The dataset was created using a probabilistic framework that combines traditional cost engineering with advanced inference models to evaluate urban climate resilience. The process is outlined in four main stages: 1. Baseline Modeling: Data on technical and economic aspects were gathered for nine mitigation strategies across key sectors (residential, commercial, and transportation), comparing a conventional technology setup with an efficient technology setup. Using market parameters (energy prices, discount rates, and emission factors), the deterministic indicators of Total Annualized Cost (TAC), Emissions (Em), and initial Marginal Abatement Cost (MAC) were calculated. 2. Stochastic Simulation (Monte Carlo): To manage parametric uncertainty, a Monte Carlo simulation was run with 135,000 independent runs for each strategy. Investment, lifespan, and energy consumption variables were modeled using probability distributions (triangular and truncated normal) from a matrix featuring 81 market scenarios and behavioral response levels (η). 3. Integration of the Rebound Effect and Risk: The model adjusts efficient consumption through the parameter η, thereby recalculating the MAC to identify weaknesses in climate policy. At this stage, three key risk metrics were quantified: the Financial Backfire Probability (FBP), the Environmental Backfire Probability (EBP), and the Abatement Erosion Probability (AEP). 4. Structuring Bayesian Networks: Continuous simulation results were converted into ordinal categories to train two Bayesian networks (Financial and Environmental). This step transforms synthetic data into a knowledge structure capable of making conditional inferences and conducting "what-if" sensitivity analyses in response to changes in user behavior or economic conditions. 5. Software and Reproducibility: The entire workflow was implemented in Python, using fixed seeds to ensure that the resulting stochastic distributions and conditional probability tables (CPTs) are fully reproducible by other researchers.
Institutions
- Autonomous University of TamaulipasTamaulipas, Ciudad Victoria