TY - JOUR
T1 - Day-of-week and seasonal patterns of PM2.5 concentrations over the United States
T2 - Time-series analyses using the Prophet procedure
AU - Zhao, Naizhuo
AU - Liu, Ying
AU - Vanos, Jennifer K.
AU - Cao, Guofeng
N1 - Publisher Copyright:
© 2018 Elsevier Ltd
PY - 2018/11
Y1 - 2018/11
N2 - Fluctuations of ambient fine particulate matter (PM2.5) concentrations show clear yearly and weekly patterns, which has been revealed by previous studies. However, reliability of those studies may be affected by their small research areas, short observation periods, and/or the lack of using specialized statistical approaches for time series. The current study applies a recently developed time-series analysis procedure, Prophet, to investigate seasonality of daily PM2.5 concentrations over nine years (2007–2015) measured at 220 monitoring stations across the United States. Prophet is a new tool for producing high quality forecasts from time series data that have characteristics of multiple temporal patterns with either linear or non-linear growth/decline. Through decomposing each PM2.5 time series into three major components (i.e., trend, seasonality, and holidays), we observed periodically changing patterns of PM2.5 concentrations weekly and yearly consistent with previous findings. Specifically, relatively high PM2.5 concentrations tend to appear in the month of January and on Fridays, and PM2.5 concentrations on Sunday are generally lower than those on most other days of the week. However, we discovered that high PM2.5 concentrations are also likely to appear in July. Additionally, compared to Fridays in most studies, the highest PM2.5 concentrations are found to more likely occur on Saturdays, while the lowest concentrations are found on Monday as universally as on Sunday. Beyond understanding the seasonality of PM2.5 concentrations, this study revealed the potential use of Prophet, originally designed for business time series, for detecting periodicities of environmental phenomena.
AB - Fluctuations of ambient fine particulate matter (PM2.5) concentrations show clear yearly and weekly patterns, which has been revealed by previous studies. However, reliability of those studies may be affected by their small research areas, short observation periods, and/or the lack of using specialized statistical approaches for time series. The current study applies a recently developed time-series analysis procedure, Prophet, to investigate seasonality of daily PM2.5 concentrations over nine years (2007–2015) measured at 220 monitoring stations across the United States. Prophet is a new tool for producing high quality forecasts from time series data that have characteristics of multiple temporal patterns with either linear or non-linear growth/decline. Through decomposing each PM2.5 time series into three major components (i.e., trend, seasonality, and holidays), we observed periodically changing patterns of PM2.5 concentrations weekly and yearly consistent with previous findings. Specifically, relatively high PM2.5 concentrations tend to appear in the month of January and on Fridays, and PM2.5 concentrations on Sunday are generally lower than those on most other days of the week. However, we discovered that high PM2.5 concentrations are also likely to appear in July. Additionally, compared to Fridays in most studies, the highest PM2.5 concentrations are found to more likely occur on Saturdays, while the lowest concentrations are found on Monday as universally as on Sunday. Beyond understanding the seasonality of PM2.5 concentrations, this study revealed the potential use of Prophet, originally designed for business time series, for detecting periodicities of environmental phenomena.
KW - PM concentrations
KW - Prophet
KW - Seasonality
KW - Time series
UR - http://www.scopus.com/inward/record.url?scp=85052760194&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85052760194&partnerID=8YFLogxK
U2 - 10.1016/j.atmosenv.2018.08.050
DO - 10.1016/j.atmosenv.2018.08.050
M3 - Article
AN - SCOPUS:85052760194
SN - 1352-2310
VL - 192
SP - 116
EP - 127
JO - Atmospheric Environment
JF - Atmospheric Environment
ER -