skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.
Attention:The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 7:00 AM ET to 7:30 AM ET on Friday, April 24 due to maintenance. We apologize for the inconvenience.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


  • NSF Public Access
  • Search Results
  • Automated Machine Learning to Evaluate the Information Content of Tropospheric Trace Gas Columns for Fine Particle Estimates Over India: A Modeling Testbed
Title: Automated Machine Learning to Evaluate the Information Content of Tropospheric Trace Gas Columns for Fine Particle Estimates Over India: A Modeling Testbed
Abstract India is largely devoid of high‐quality and reliable on‐the‐ground measurements of fine particulate matter (PM2.5). Ground‐level PM2.5concentrations are estimated from publicly available satellite Aerosol Optical Depth (AOD) products combined with other information. Prior research has largely overlooked the possibility of gaining additional accuracy and insights into the sources of PM using satellite retrievals of tropospheric trace gas columns. We evaluate the information content of tropospheric trace gas columns for PM2.5estimates over India within a modeling testbed using an Automated Machine Learning (AutoML) approach, which selects from a menu of different machine learning tools based on the data set. We then quantify the relative information content of tropospheric trace gas columns, AOD, meteorological fields, and emissions for estimating PM2.5over four Indian sub‐regions on daily and monthly time scales. Our findings suggest that, regardless of the specific machine learning model assumptions, incorporating trace gas modeled columns improves PM2.5estimates. We use the ranking scores produced from the AutoML algorithm and Spearman’s rank correlation to infer or link the possible relative importance of primary versus secondary sources of PM2.5as a first step toward estimating particle composition. Our comparison of AutoML‐derived models to selected baseline machine learning models demonstrates that AutoML is at least as good as user‐chosen models. The idealized pseudo‐observations (chemical‐transport model simulations) used in this work lay the groundwork for applying satellite retrievals of tropospheric trace gases to estimate fine particle concentrations in India and serve to illustrate the promise of AutoML applications in atmospheric and environmental research.  more » « less
Award ID(s):
2020677
PAR ID:
10484001
Author(s) / Creator(s):
; ; ; ; ; ; ; ; ; ; ;
Publisher / Repository:
JAMES
Date Published:
Journal Name:
Journal of Advances in Modeling Earth Systems
Volume:
15
Issue:
3
ISSN:
1942-2466
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ; ; (, Frontiers in Environmental Science)
    Introduction:Traditional methods to estimate exposure to PM2.5(particulate matter with less than 2.5 µm in diameter) have typically relied on limited regulatory monitors and do not consider human mobility and travel. However, the limited spatial coverage of regulatory monitors and the lack of consideration of mobility limit the ability to capture actual air pollution exposure. Methods:This study aims to improve traditional exposure assessment methods for PM2.5by incorporating the measurements from a low-cost sensor network (PurpleAir) and regulatory monitors, an automated machine learning modeling framework, and big human mobility data. We develop a monthly-aggregated hourly land use regression (LUR) model based on automated machine learning (AutoML) and assess the model performance across eight metropolitan areas within the US. Results:Our results show that integrating low-cost sensor with regulatory monitor measurements generally improves the AutoML-LUR model accuracy and produces higher spatial variation in PM2.5concentration maps compared to using regulatory monitor measurements alone. Feature importance analysis shows factors highly correlated with PM2.5concentrations, including satellite aerosol optical depth, meteorological variables, vegetation, and land use. In addition, we incorporate human mobility data on exposure estimates regarding where people visit to identify spatiotemporal hotspots of places with higher risks of exposure, emphasizing the need to consider both visitor numbers and PM2.5concentrations when developing exposure reduction strategies. Discussion:This research provides important insights for further public health studies on air pollution by comprehensively assessing the performance of AutoML-LUR models and incorporating human mobility into considering human exposure to air pollution. 
    more » « less
  2. ; ; ; ; ; ; ; ; ; ; et al (, Atmospheric Chemistry and Physics)
    Abstract. The climatic and health effects of aerosols are strongly dependent on the intra-annual variations in their loading and properties. While the seasonal variations of regional aerosol optical depth (AOD) have been extensively studied, understanding the temporal variations in aerosol vertical distribution and particle types is also important for an accurate estimate of aerosol climatic effects. In this paper, we combine the observations from four satellite-borne sensors and several ground-based networks to investigate the seasonal variations of aerosol column loading, vertical distribution, and particle types over three populous regions: the Eastern United States (EUS), Western Europe (WEU), and Eastern and Central China (ECC). In all three regions, column AOD, as well as AOD at heights above 800m, peaks in summer/spring, probably due to accelerated formation of secondary aerosols and hygroscopic growth. In contrast, AOD below 800m peaks in winter over WEU and ECC regions because more aerosols are confined to lower heights due to the weaker vertical mixing. In the EUS region, AOD below 800m shows two maximums, one in summer and the other in winter. The temporal trends in low-level AOD are consistent with those in surface fine particle (PM2.5) concentrations. AOD due to fine particles ( < 0.7µm diameter) is much larger in spring/summer than in winter over all three regions. However, the coarse mode AOD ( > 1.4µm diameter), generally shows small variability, except that a peak occurs in spring in the ECC region due to the prevalence of airborne dust during this season. When aerosols are classified according to sources, the dominant type is associated with anthropogenic air pollution, which has a similar seasonal pattern as total AOD. Dust and sea-spray aerosols in the WEU region peak in summer and winter, respectively, but do not show an obvious seasonal pattern in the EUS region. Smoke aerosols, as well as absorbing aerosols, present an obvious unimodal distribution with a maximum occurring in summer over the EUS and WEU regions, whereas they follow a bimodal distribution with peaks in August and March (due to crop residue burning) over the ECC region. 
    more » « less
  3. ; ; ; ; (, GeoHealth)
    Abstract Globally, solid biofuels (SB) have been widely used for household cooking and energy production for decades due to electricity shortages and socio‐economic barriers to adopting renewable energy alternatives. This has detrimental effects on air quality, human health, and climate through trace gas and aerosol emissions. Despite numerous studies, the long‐term consequences of SB emissions remain poorly understood. Here, we use the Community Earth System Model and the Community Emissions Data System emission inventory to investigate the SB emission impacts on air quality and human health for 2000–2019. Global SB emission increased the ambient PM2.5(particulate matter with aerodynamic diameters 2.5 μm) and ozone (O3) concentrations up to 23.61 /m3and 13.69 ppbv, with significant effects found in India, China, and the Rest of Asia (ROA). Our study estimates total annual premature deaths (APDs) associated with global SB‐attributable PM2.5and O3exposure as 1.11 million [95% confidence interval (95% CI): 1.00–1.22 million] in 2000 up to 1.43 million (95% CI: 1.30–1.56 million) in 2019. China's SB emissions and associated APDs have reduced substantially, whereas India and ROA had a major leap in both estimates in 2019 compared to 2000. China's progress in cutting residential SB emissions accounts for its improvements. Our study urges the reduction of SB usage and emissions to potentially improve overall air quality and human health conditions, especially in highly populated, low‐ and middle‐income countries, where the poor air quality and associated health burden attributable to SB emissions are estimated to be higher. 
    more » « less
  4. ; ; ; (, Remote Sensing)
    The 2020 wildfire season in the Western U.S. was historic in its intensity and impact on the land and atmosphere. This study aims to characterize satellite retrievals of carbon monoxide (CO), a tracer of combustion and signature of those fires, from two key satellite instruments: the Cross-track Infrared Sounder (CrIS) and the Tropospheric Monitoring Instrument (TROPOMI). We evaluate them during this event and assess their synergies. These two retrievals are matched temporally, as the host satellites are in tandem orbit and spatially by aggregating TROPOMI to the CrIS resolution. Both instruments show that the Western U.S. displayed significantly higher daily average CO columns compared to the Central and Eastern U.S. during the wildfires. TROPOMI showed up to a factor of two larger daily averages than CrIS during the most intense fire period, likely due to differences in the vertical sensitivity of the two instruments and representative of near-surface CO abundance near the fires. On the other hand, there was excellent agreement between the instruments in downwind free tropospheric plumes (scatter plot slopes of 0.96–0.99), consistent with their vertical sensitivities and indicative of mostly lofted smoke. Temporally, TROPOMI CO column peaks were delayed relative to the Fire Radiative Power (FRP), and CrIS peaks were delayed with respect to TROPOMI, particularly during the intense initial weeks of September, suggesting boundary layer buildup and ventilation. Satellite retrievals were evaluated using ground-based CO column estimates from the Network for the Detection of Atmospheric Composition Change (NDACC) and the Total Carbon Column Observing Network (TCCON), showing Normalized Mean Errors (NMEs) for CrIS and TROPOMI below 32% and 24%, respectively, when compared to all stations studied. While Normalized Mean Bias (NMB) was typically low (absolute value below 15%), there were larger negative biases at Pasadena, likely associated with sharp spatial gradients due to topography and proximity to a large city, which is consistent with previous research. In situ CO profiles from AirCore showed an elevated smoke plume for 15 September 2020, highlighted consistency between TROPOMI and CrIS CO columns for lofted plumes. This study demonstrates that both CrIS and TROPOMI provide complementary information on CO distribution. CrIS’s sensitivity in the middle and lower free troposphere, coupled with TROPOMI’s effectiveness at capturing total columns, offers a more comprehensive view of CO distribution during the wildfires than either retrieval alone. By combining data from both satellites as a ratio, more detailed information about the vertical location of the plumes can potentially be extracted. This approach can enhance air quality models, improve vertical estimation accuracy, and establish a new method for assessing lower tropospheric CO concentrations during significant wildfire events. 
    more » « less
  5. ; ; ; ; ; ; ; ; ; ; et al (, Nature Communications)
    Abstract Ambient fine particulate matter (PM2.5) is the world’s leading environmental health risk factor. Quantification is needed of regional contributions to changes in global PM2.5exposure. Here we interpret satellite-derived PM2.5estimates over 1998-2019 and find a reversal of previous growth in global PM2.5air pollution, which is quantitatively attributed to contributions from 13 regions. Global population-weighted (PW) PM2.5exposure, related to both pollution levels and population size, increased from 1998 (28.3 μg/m3) to a peak in 2011 (38.9 μg/m3) and decreased steadily afterwards (34.7 μg/m3in 2019). Post-2011 change was related to exposure reduction in China and slowed exposure growth in other regions (especially South Asia, the Middle East and Africa). The post-2011 exposure reduction contributes to stagnation of growth in global PM2.5-attributable mortality and increasing health benefits per µg/m3marginal reduction in exposure, implying increasing urgency and benefits of PM2.5mitigation with aging population and cleaner air. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Text Encoded Conversion
  • XML
  • Save / Share this Record
  • Send to Email