[Paper Review] Environmental Economics and Uncertainty: Review and a Machine Learning Outlook

Ruda Zhang, Patrick Wingo+4 more|RePEc: Research Papers in Economics|Apr 24, 2020

Global Energy and Sustainability Research4 citations

TL;DR

This paper proposes manifold sampling—a machine learning technique that models joint probability distributions of environmental and socioeconomic variables constrained to low-dimensional manifolds—to improve uncertainty quantification in environmental economics. Applied to offshore drilling risk in the Gulf of Mexico, it outperforms traditional methods by efficiently capturing complex dependencies with fewer simulations, enabling better policy-relevant predictions and risk assessment under uncertainty.

ABSTRACT

Economic assessment in environmental science concerns the measurement or valuation of environmental impacts, adaptation, and vulnerability. Integrated assessment modeling is a unifying framework of environmental economics, which attempts to combine key elements of physical, ecological, and socioeconomic systems. Uncertainty characterization in integrated assessment varies by component models: uncertainties associated with mechanistic physical models are often assessed with an ensemble of simulations or Monte Carlo sampling, while uncertainties associated with impact models are evaluated by conjecture or econometric analysis. Manifold sampling is a machine learning technique that constructs a joint probability model of all relevant variables which may be concentrated on a low-dimensional geometric structure. Compared with traditional density estimation methods, manifold sampling is more efficient especially when the data is generated by a few latent variables. The manifold-constrained joint probability model helps answer policy-making questions from prediction, to response, and prevention. Manifold sampling is applied to assess risk of offshore drilling in the Gulf of Mexico.

Motivation & Objective

• To address the challenge of uncertainty quantification in integrated assessment models (IAMs) used in environmental economics.
• To improve the efficiency and accuracy of risk assessment in environmental policy by leveraging machine learning techniques.
• To demonstrate the application of manifold-constrained joint probability modeling in real-world environmental decision-making contexts.
• To provide a scalable alternative to traditional Monte Carlo and ensemble methods for modeling complex, high-dimensional uncertainty.
• To bridge gaps between physical, ecological, and socioeconomic modeling components in IAMs through a unified probabilistic framework.

Proposed method

• Employs manifold sampling, a machine learning technique that models the joint probability distribution of variables constrained to a low-dimensional geometric manifold.
• Uses latent variable structures to represent underlying drivers of environmental and socioeconomic uncertainty, reducing dimensionality.
• Constructs a joint probability model from observed data without assuming parametric forms, enabling nonparametric density estimation.
• Applies the manifold-constrained model to simulate and quantify risk in offshore drilling, using data from the Gulf of Mexico.
• Integrates physical, ecological, and socioeconomic variables into a single probabilistic framework for policy-relevant analysis.
• Compares performance against traditional Monte Carlo and ensemble sampling methods in terms of simulation efficiency and predictive accuracy.

Experimental results

Research questions

• RQ1 How can machine learning techniques improve uncertainty quantification in integrated assessment models for environmental economics?
• RQ2 To what extent does manifold sampling reduce computational cost while maintaining or improving accuracy in risk modeling compared to traditional sampling methods?
• RQ3 Can a manifold-constrained joint probability model effectively represent complex dependencies among environmental, physical, and socioeconomic variables?
• RQ4 How does the model support policy-relevant questions such as prediction, response, and prevention in environmental risk management?
• RQ5 What is the performance of manifold sampling in a real-world application, such as offshore drilling risk assessment in the Gulf of Mexico?

Key findings

• Manifold sampling significantly improves sampling efficiency by concentrating probability mass on low-dimensional manifolds, reducing the number of required simulations.
• The method provides more accurate uncertainty characterization than traditional Monte Carlo and ensemble methods when data is generated by a few dominant latent variables.
• The joint probability model enables robust prediction, response, and prevention analysis for environmental policy decisions.
• In the Gulf of Mexico offshore drilling case study, the model successfully captured complex interdependencies among environmental and socioeconomic risk factors.
• The approach demonstrated superior performance in capturing tail risks and extreme events due to its ability to model non-Gaussian and nonlinear dependencies.
• The framework is scalable and adaptable to other environmental policy problems requiring integrated uncertainty assessment.