2026-09-19T05:03:19Zhttps://ecommons.cornell.edu/server/oai/requestoai:ecommons.cornell.edu:1813/1108092026-05-15T19:49:33Zcom_1813_35col_1813_47
Montazeri, Amir
Albertson, John D.
Bewley, Gregory Paul
Cowen, III, Edwin (Todd)
2022-01-24T18:07:43Z
2022-01-24T18:07:43Z
2021-12
ProQuest Submission ID: 12814
ProQuest Publication ID: 28772030
https://hdl.handle.net/1813/110809
https://doi.org/10.7298/9jkc-4m18
15312667
161 pages
Releases of pollutants into the atmosphere pose risks to human health, the environment, and the economy. Fine-scale monitoring of air pollution and efforts to characterize the impact of the local environment on pollutant concentrationshave led to great developments in sensor technology and the generation of a staggering amount of new data. While the advent of large datasets has led to more emphasis on statistical modeling, the disconnectedness of these models from underlying physical laws degrades their generalizability. Consequently, the objective of the present research is to apply a combination of statistical and deterministic models to quantify the effects of the environment at varying scales on: 1) our ability to observe and measure pollutant concentration profiles, 2) our ability to make inferences about the state of the pollutant sources, and 3) The evolution of pollutant concentration profiles. This dissertation is divided into three parts. The first part focuses on a theoretical analysis of the uncertainties involved in leak quantification via gas imaging techniques. These uncertainties are quantified through statistical analysisof Large Eddy Simulation (LES) data. Our results show that uncertainties that are due to inferring the 3D plume structure from 2D projections become smaller as measurements are made at larger downwind distances from the emission source. Further, acquisition times on the order of tens of seconds are sufficient to significantly reduce these uncertainties. The second part employs a recursive Bayesian scheme to infer the varying states of a gas emission source observed through downwind mobile measurements. Our findings suggest that the statistics of the measurements, such as the coefficient of variation and range are good predictors of the performance of the Bayesian algorithm. In addition, the algorithm shows a high success rate in detecting the state change when the emission rate is tripled. The third part introduces a spatial clustering framework developed for studying the role of land-use in mediating the effect of meteorology on urban air quality. Our study is based on long-term mobile measurements of Nitrogen Dioxide (NO2) concentrations in Oakland, California. We find strong correlations between wind speed and NO2 concentrations in the absence of other causes of strong vertical mixing such as highway traffic and higher surface heatfluxes in summer. In addition, an analysis of the exceedance probabilities shows that wind speed is effective in lowering the highest concentrations even in regions where mean concentrations are not responsive to wind speed. These findings coupled with projections of climate (e.g., wind speed forecasts) and urban development can be used to make predictions regarding future air quality in urban areas.
en
Attribution 4.0 International
https://creativecommons.org/licenses/by/4.0/
Applied Statistics
Environmental Sensing
Gas emission
Physics-guided inference from gas emission data for source characterization and air pollution mapping
dissertation or thesis
https://newcatalog.library.cornell.edu/catalog/15312667
application/pdf
Mechanical Engineering
Cornell University
Doctor of Philosophy
Ph. D., Mechanical Engineering
https://hdl.handle.net/1813/59810.2
Publication
Cornell Theses and Dissertations
Montazeri, Amir
5893a6ea-7af3-41d7-abc6-04bcd26ab5df
open.access
open.access
<Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="00a2feb3-4748-4925-b8d7-cb16bb545016">
<Type xmlns="https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types">http://purl.org/coar/resource_type/c_1843</Type>
<Language>en</Language>
<Title>Physics-guided inference from gas emission data for source characterization and air pollution mapping</Title>
<PublishedIn>
<Publication>
</Publication>
</PublishedIn>
<PublicationDate>2021-12</PublicationDate>
<DOI>https://doi.org/10.7298/9jkc-4m18</DOI>
<Authors>
<Author>
<DisplayName>Montazeri, Amir</DisplayName>
<Affiliation>
<OrgUnit>
</OrgUnit>
</Affiliation>
</Author>
</Authors>
<Editors>
</Editors>
<Publishers>
<Publisher>
<OrgUnit />
</Publisher>
</Publishers>
<License>https://creativecommons.org/licenses/by/4.0/</License>
<Keyword>Applied Statistics</Keyword>
<Keyword>Environmental Sensing</Keyword>
<Keyword>Gas emission</Keyword>
<Abstract>Releases of pollutants into the atmosphere pose risks to human health, the environment, and the economy. Fine-scale monitoring of air pollution and efforts to characterize the impact of the local environment on pollutant concentrationshave led to great developments in sensor technology and the generation of a staggering amount of new data. While the advent of large datasets has led to more emphasis on statistical modeling, the disconnectedness of these models from underlying physical laws degrades their generalizability. Consequently, the objective of the present research is to apply a combination of statistical and deterministic models to quantify the effects of the environment at varying scales on: 1) our ability to observe and measure pollutant concentration profiles, 2) our ability to make inferences about the state of the pollutant sources, and 3) The evolution of pollutant concentration profiles. This dissertation is divided into three parts. The first part focuses on a theoretical analysis of the uncertainties involved in leak quantification via gas imaging techniques. These uncertainties are quantified through statistical analysisof Large Eddy Simulation (LES) data. Our results show that uncertainties that are due to inferring the 3D plume structure from 2D projections become smaller as measurements are made at larger downwind distances from the emission source. Further, acquisition times on the order of tens of seconds are sufficient to significantly reduce these uncertainties. The second part employs a recursive Bayesian scheme to infer the varying states of a gas emission source observed through downwind mobile measurements. Our findings suggest that the statistics of the measurements, such as the coefficient of variation and range are good predictors of the performance of the Bayesian algorithm. In addition, the algorithm shows a high success rate in detecting the state change when the emission rate is tripled. The third part introduces a spatial clustering framework developed for studying the role of land-use in mediating the effect of meteorology on urban air quality. Our study is based on long-term mobile measurements of Nitrogen Dioxide (NO2) concentrations in Oakland, California. We find strong correlations between wind speed and NO2 concentrations in the absence of other causes of strong vertical mixing such as highway traffic and higher surface heatfluxes in summer. In addition, an analysis of the exceedance probabilities shows that wind speed is effective in lowering the highest concentrations even in regions where mean concentrations are not responsive to wind speed. These findings coupled with projections of climate (e.g., wind speed forecasts) and urban development can be used to make predictions regarding future air quality in urban areas.</Abstract>
<Access xmlns="http://purl.org/coar/access_right"
>
</Access>
</Publication>