Quantifying Uncertainty in CO2 Emissions with a Restricted Number of Remote Sensors: A Comparison of Model Calibration and Kalman Filtering Techniques

• Activity Number: 14
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Matthew Pratola*+ and Jon Reisner and M.K. Dubey and Dave Higdon
• Companies: and Los Alamos National Laboratories and Los Alamos National Laboratories and Los Alamos National Laboratories
• Keywords: Ensemble Kalman Filter ; Computer Model Calibration ; Markov Chain Monte Carlo ; CO2 Emissions ; Uncertainty Quantification ; Climate Treaty Verification

Abstract

Estimating the rate at which CO2 is emitted by human industrial activities, such as factories and power plants, is important from an impacts and regulatory standpoint. However, as direct measurement is usually not possible, the ability to use remote and a limited number of in-situ distributed sensors to infer this information is key. In addition, physical models of the emitted CO2 plume can be parameterized to account for effects such as wind forcing and local geography. The goal is to combine such models with limited observational data in order to solve the relevant inverse problem, such as estimating the rate of CO2 emission at the source. We compare two statistical approaches to remotely estimating emissions that leverage such physical models while quantifying the uncertainty in the parameter estimates: Kalman filtering and a novel statistical calibration approach. Our motivating example considers estimating the CO2 emission rate of a power plant whose exhaust plume is modeled by HIGRAD, while observations consist of column integrated CO2 measured by a single FTS sensor located downstream of the plant.