TY - JOUR
T1 - Statistical analysis of primary and secondary atmospheric formaldehyde
AU - Friedfeld, Stephen
AU - Fraser, Matthew
AU - Ensor, Kathy
AU - Tribble, Seth
AU - Rehle, Dirk
AU - Leleux, Darrin
AU - Tittel, Frank
N1 - Funding Information:
This work was funded by the Gulf Coast Hazardous Substance Research Center, the Texas Advanced Technology Program, NASA, the Welch Foundation and the National Science Foundation. The authors would like to acknowledge Jairam Nadkarny of the TNRCC for his assistance with this project. The authors are especially grateful to the statistical analyses and results from the Rice University Statistical Consulting Laboratory.
PY - 2002/10
Y1 - 2002/10
N2 - Regression models coupled with time series data were used to analyze the contribution of primary and secondary sources to formaldehyde (HCHO) concentrations, as determined by statistical analogy to primary (carbon monoxide, CO) and secondary (ozone, O3) compounds measured simultaneously in Houston, TX. Time series analyses substantiated the need for statistical methods of analysis, given the complexity of the data and the rapid fluctuations that occur in atmospheric concentrations. A positive relationship was found for both the auto-correlation function (ACF) and partial auto-correlation function (PACF) of HCHO with either CO or O3. Regression models used to distinguish primary and secondary contributions included a simple linear regression of the three compounds (one lag unit of time, 5min) on current HCHO concentrations, resulting in a ratio of secondary formation to primary emission of 1.7. A second, more robust model utilized auto-correlated error processes to approximate the true nature of the linear regression; this model also indicates the ratio of secondary to primary contribution at 1.7 as the mean of ten model simulations. From the error processes model, one lag unit of time was most significant for CO predicting HCHO, while simultaneous measurements (lag 0) were most significant for O3 predicting HCHO. Outlying O3 and HCHO concentrations were shown not to affect the results.
AB - Regression models coupled with time series data were used to analyze the contribution of primary and secondary sources to formaldehyde (HCHO) concentrations, as determined by statistical analogy to primary (carbon monoxide, CO) and secondary (ozone, O3) compounds measured simultaneously in Houston, TX. Time series analyses substantiated the need for statistical methods of analysis, given the complexity of the data and the rapid fluctuations that occur in atmospheric concentrations. A positive relationship was found for both the auto-correlation function (ACF) and partial auto-correlation function (PACF) of HCHO with either CO or O3. Regression models used to distinguish primary and secondary contributions included a simple linear regression of the three compounds (one lag unit of time, 5min) on current HCHO concentrations, resulting in a ratio of secondary formation to primary emission of 1.7. A second, more robust model utilized auto-correlated error processes to approximate the true nature of the linear regression; this model also indicates the ratio of secondary to primary contribution at 1.7 as the mean of ten model simulations. From the error processes model, one lag unit of time was most significant for CO predicting HCHO, while simultaneous measurements (lag 0) were most significant for O3 predicting HCHO. Outlying O3 and HCHO concentrations were shown not to affect the results.
KW - Formaldehyde
KW - Houston
KW - Secondary formation
KW - Statistical modeling
KW - TX
UR - http://www.scopus.com/inward/record.url?scp=0036795066&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0036795066&partnerID=8YFLogxK
U2 - 10.1016/S1352-2310(02)00558-7
DO - 10.1016/S1352-2310(02)00558-7
M3 - Article
AN - SCOPUS:0036795066
SN - 1352-2310
VL - 36
SP - 4767
EP - 4775
JO - Atmospheric Environment
JF - Atmospheric Environment
IS - 30
ER -