Search

Summary Diverse anthropogenic activities are changing our natural environment, with important implications for human health. Successfully managing their impacts requires an understanding of the compounding hazards resulting from multi-faceted environmental changes. Here, we propose a human-environment systems lens comprising public health, climate, air quality, and agricultural land-use land management to characterize the combined health risks of anthropogenic environmental changes. Interactions within this system can amplify, diminish, or generate additional hazards associated with changes in any individual element. Using South Asia as an example—where rapid industrialization and the Green Revolution aided economic development and food production but inadvertently compromised multiple human health dimensions—we synthesize the influence of human-environment system interactions on environment-sensitive health outcomes. We further demonstrate the utility of this lens for evaluating the health outcomes of existing and planned regional policies and interventions to identify unintended negative consequences and solutions that realize co-benefits and minimize trade-offs.

India is largely devoid of high-quality and reliable on-the-ground measurements of fine particulate matter (PM2.5). Ground-level PM2.5 concentrations 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 first evaluate the information content of tropospheric trace gas columns for PM2.5 estimates 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 dataset. We then quantify the relative information content of tropospheric trace gas columns, AOD, meteorological fields, and emissions for estimating PM2.5 over four Indian sub-regions on daily and monthly time scales. Our findings suggest that, regardless of the specific model assumptions, incorporating trace gas modeled columns improves PM2.5 estimates. We use the ranking scores produced from the AutoML algorithm and Spearman's rank correlation to infer the relative dominance of primary versus secondary sources of PM2.5 as a first step towards estimating particle composition. Our comparison of AutoML-derived models to selected baseline machine learning models demonstrates that AutoML is at least as good as model selection and hyperparameter tuning prior to training. The idealized pseudo-observations 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.

Models of atmospheric composition rely on fire emissions inventories to reconstruct and project impacts of biomass burning on air quality, public health, climate, ecosystem dynamics, and land-atmosphere exchanges. Many such global inventories use satellite measurements of active fires and/or burned area from the Moderate Resolution Imaging Spectroradiometer (MODIS). However, differences across inventories in the interpretation of satellite imagery, the emissions factors assumed for different components of smoke, and the adjustments made for small and obscured fires can result in large regional differences in fire emissions estimates across inventories. Using Google Earth Engine, we leverage 15 years (2003–2017) of MODIS observations and 6 years (2012–2017) of observations from the higher spatial resolution Visible Imaging Infrared Radiometer Suite (VIIRS) sensor to develop metrics to quantify five major sources of spatial bias or uncertainty in the inventories: (1) primary reliance on active fires versus burned area, (2) cloud/haze burden on the ability of satellites to “see” fires, (3) fragmentation of burned area, (4) roughness in topography, and (5) small fires, which are challenging to detect. Based on all these uncertainties, we devise comprehensive “relative fire confidence scores,” mapped globally at 0.25° × 0.25° spatial resolution over 2003–2017. We then focus on fire activity in Indonesia as a case study to analyze how the choice of a fire emissions inventory affects model estimates of smoke-induced health impacts across Equatorial Asia. We use the adjoint of the GEOS-Chem chemical transport model and apply emissions of particulate organic carbon and black carbon (OC + BC smoke) from five global inventories: Global Fire Emissions Database (GFEDv4s), Fire Inventory from NCAR (FINNv1.5), Global Fire Assimilation System (GFASv1.2), Quick Fire Emissions Dataset (QFEDv2.5r1), and Fire Energetics and Emissions Research (FEERv1.0-G1.2). We find that modeled monthly smoke PM2.5 in Singapore from 2003 to 2016 correlates with observed smoke PM2.5, with r ranging from 0.64–0.84 depending on the inventory. However, during the burning season (July to October) of high fire intensity years (e.g., 2006 and 2015), the magnitude of mean Jul-Oct modeled smoke PM2.5 can differ across inventories by >20 μg m−3 (>500%). Using the relative fire confidence metrics, we deduce that uncertainties in this region arise primarily from the small, fragmented fire landscape and very poor satellite observing conditions due to clouds and thick haze at this time of year. Indeed, we find that modeled smoke PM2.5 using GFASv1.2, which adjusts for fires obscured by clouds and thick haze and accounts for peatland emissions, is most consistent with observations in Singapore, as well as in Malaysia and Indonesia. Finally, we develop an online app called FIRECAM for end-users of global fire emissions inventories. The app diagnoses differences in emissions among the five inventories and gauges the relative uncertainty associated with satellite-observed fires on a regional basis.

A rising source of outdoor emissions in northwestern India is crop residue burning, occurring after the monsoon (kharif) and winter (rabi) crop harvests. In particular, post-monsoon rice residue burning, which occurs annually from October to November and is linked to increasing mechanization, coincides with meteorological conditions that enhance short-term air quality degradation. Here we examine the Global Fire Emissions Database (GFED), whose bottom-up emissions are based on the 500-m burned area product, MCD64A1, derived from Moderate Resolution Imaging Spectroradiometer (MODIS) observations. Using a household survey from 2016, we find that MCD64A1 tends to underestimate burned area in many surveyed villages, leading to poor representation of small, scattered fires and consequent spatial biases in model results. To more accurately allocate such small fires and resolve sub-village heterogeneity, we use an experimental hybrid MODIS-Landsat method (ModL2T) to map burned area at 30-m spatial resolution, which results in 44 ± 21% higher burned area than MCD64A1 and up to 105 ± 52% increase in dry matter emissions over GFEDv4s. In our validation and assessments, we find that ModL2T performs better relative to MCD64A1 in terms of bias and omission error, but may introduce commission error due to conflation of burning with harvest and still underestimate burned area due to Landsat’s coarse temporal resolution (every 16 days). We conclude that while MODIS and Landsat provide more than two decades worth of observations, their spatio-temporal resolution is too coarse to overcome several region-specific challenges: small median landholding size (1-3 ha), quick harvest-to-sowing turnover period, prevalence of partial burning, and increasing haziness. To further constrain agricultural fire emissions in northwestern India and improve model estimates of associated public health impacts, integration of finer resolution imagery, as well as better understanding of the spatial patterns in burn rates, burn practices, and fuel loading, is requisite.

To evaluate uncertainty in the spatial distribution of air emissions over India, we compare satellite and surface observations with simulations from the U.S. Environmental Protection Agency (EPA) Community Multi-Scale Air Quality (CMAQ) model. Seasonally representative simulations were completed for January, April, July, and October 2010 at 36 km × 36 km using anthropogenic emissions from the Greenhouse Gas-Air Pollution Interaction and Synergies (GAINS) model following version 5a of the Evaluating the Climate and Air Quality Impacts of Short-Lived Pollutants project (ECLIPSE v5a). We use both tropospheric columns from the Ozone Monitoring Instrument (OMI) and surface observations from the Central Pollution Control Board (CPCB) to closely examine modeled nitrogen dioxide (NO2) biases in urban and rural regions across India. Spatial average evaluation with satellite retrievals indicate a low bias in the modeled tropospheric column (−63.3%), which reflects broad low-biases in majority non-urban regions (−70.1% in rural areas) across the sub-continent to slightly lesser low biases reflected in semi-urban areas (−44.7%), with the threshold between semi-urban and rural defined as 400 people per km2. In contrast, modeled surface NO2 concentrations exhibit a slight high bias of +15.6% when compared to surface CPCB observations predominantly located in urban areas. Conversely, in examining extremely population dense urban regions with more than 5000 people per km2 (dense-urban), we find model overestimates in both the column (+57.8) and at the surface (+131.2%) compared to observations. Based on these results, we find that existing emission fields for India may overestimate urban emissions in densely populated regions and underestimate rural emissions. However, if we rely on model evaluation with predominantly urban surface observations from the CPCB, comparisons reflect model high biases, contradictory to the knowledge gained using satellite observations. Satellites thus serve as an important emissions and model evaluation metric where surface observations are lacking, such as rural India, and support improved emissions inventory development.

Ambient air pollution in India contributes to negative health impacts and early death. Ground-based monitors often used to quantify health impacts are located in urban regions, yet approximately 70% of India's population lives in rural communities. We simulate high-resolution concentrations of fine particulate matter (PM) and ozone from the regional Community Multi-scale Air Quality model over northern India, including updated estimates of anthropogenic emissions for transportation, residential combustion and location-based industrial and electrical generating emissions in a new anthropogenic emissions inventory. These simulations inform seasonal air quality and health impacts due to anthropogenic emissions, contrasting urban versus rural regions. For our northern India domain, we estimate 463 200 (95% confidence interval: 444 600���482 600) adults die prematurely each year from PM2.5 and that 37 800 (28 500���48 100) adults die prematurely each year from O3. This translates to 5.8 deaths per 10 000 attributable to air pollution out of an annual rate of 72 deaths per 10 000 (8.1% of deaths) using 2010 estimates. We estimate that the majority of premature deaths resulting from PM2.5 and O3 are in rural (383 600) as opposed to urban (117 200) regions, where we define urban as cities and towns with populations of at least 100 000 people. These findings indicate the need for rural monitoring and appropriate health studies to understand and mitigate the effects of ambient air pollution on this population in addition to supporting model evaluation.

In a region heavily influenced by anthropogenic and biogenic atmospheric emissions, recent field measurements have attributed one-third of urban organic aerosol by mass to isoprene epoxydiols (IEPOX). These aerosols arise from the gas-phase oxidation of isoprene, the formation of IEPOX, the reactive uptake of IEPOX by particles, and finally the formation of new compounds in the aerosol phase. Using a continental-scale chemical transport model, we find a strong temporal correspondence between the simulated formation of IEPOX-derived organic aerosol and these measurements. However, because only a subset of isoprene-derived aerosol compounds have been specifically identified in laboratory studies, our simulation of known IEPOX-derived organic aerosol compounds predicts a mass 10-fold lower than the field measurements, despite abundant gas-phase IEPOX. Sensitivity studies suggest that increasing the effective IEPOX uptake coefficient and including aerosol-phase reactions that lead to the addition of functiona...

Satellite observations from the Ozone Monitoring Instrument are used in support of model evaluation of seasonal average results from the U.S. Environmental Protection Agency (EPA) Community Multi-scale Air Quality (CMAQ) model. Model evaluation was conducted with the purpose of identifying regional biases in model output compared to tropospheric columns. Comparison with tropospheric column NO 2 , an anthropogenic indicator, reveal that there are uncertainties regarding the emissions inventory input to CMAQ. Results have implications for developing accurate model inputs to produce accurate model output for relevant health impact assessments, which are increasingly important with increasing population, urbanization, and pollution in the region.

A rising source of outdoor emissions in northwestern India is crop residue burning, occurring after the monsoon (kharif) and winter (rabi) crop harvests. In particular, post-monsoon rice residue burning, which occurs annually from October to November and is linked to increasing mechanization, coincides with meteorological conditions that enhance short-term air quality degradation. Here we examine the Global Fire Emissions Database (GFED), whose bottom-up emissions are based on the 500-m burned area product, MCD64A1, derived from Moderate Resolution Imaging Spectroradiometer (MODIS) observations. Using a household survey from 2016, we find that MCD64A1 tends to underestimate burned area in many surveyed villages, leading to poor representation of small, scattered fires and consequent spatial biases in model results. To more accurately allocate such small fires and resolve sub-village heterogeneity, we use an experimental hybrid MODIS-Landsat method (ModL2T) to map burned area at 30-m spatial resolution, which results in 44 ± 21% higher burned area than MCD64A1 and up to 105 ± 52% increase in dry matter emissions over GFEDv4s. In our validation and assessments, we find that ModL2T performs better relative to MCD64A1 in terms of bias and omission error, but may introduce commission error due to conflation of burning with harvest and still underestimate burned area due to Landsat’s coarse temporal resolution (every 16 days). We conclude that while MODIS and Landsat provide more than two decades worth of observations, their spatio-temporal resolution is too coarse to overcome several region-specific challenges: small median landholding size (1-3 ha), quick harvest-to-sowing turnover period, prevalence of partial burning, and increasing haziness. To further constrain agricultural fire emissions in northwestern India and improve model estimates of associated public health impacts, integration of finer resolution imagery, as well as better understanding of the spatial patterns in burn rates, burn practices, and fuel loading, is requisite.

Research suggests non-experts associate different content with the terms “global warming” and “climate change.” We test this claim with Twitter content using supervised learning software to categorize tweets by topic and explore differences between content using “global warming” and “climate change” between 1 January 2012 and 31 March 2014. Twitter data were combined with temperature records to observe the extent to which temperature was associated with Twitter discussions. We then used two case studies to examine the relationship between extreme temperature events and Twitter content. Our findings underscore the importance of considering climate change communication on social media.

We present an examination of the potential emissions and air quality benefits of shifting freight from truck to rail in the upper Midwestern United States. Using a novel, freight-specific emissions inventory (the Wisconsin Inventory of Freight Emissions, WIFE) and a three-dimensional Eulerian photochemical transport model (the Community Multiscale Air Quality Model, CMAQ), we quantify how specific freight mode choices impact ambient air pollution concentrations. Using WIFE, we developed two modal shift scenarios: one focusing on intraregional freight movements within the Midwest and a second on through-freight movements through the region. Freight truck and rail emissions inventories for each scenario were gridded to a 12 km × 12 km horizontal resolution as input to CMAQ, along with emissions from all other major sectors, and three-dimensional time-varying meteorology from the Weather Research and Forecasting model (WRF). The through-freight scenario reduced monthly mean (January and July) localized concentrations of nitrogen dioxide (NO2) by 28% (-2.33 ppbV) in highway grid cells, and reduced elemental carbon (EC) by 16% (-0.05 μg/m(3)) in highway grid cells. There were corresponding localized increases in railway grid cells of 25% (+0.83 ppbV) for NO2, and 22% (+0.05 μg/m(3)) for EC. The through-freight scenario reduced CO2 emissions 31% compared to baseline trucking. The through-freight scenario yields a July mean change in ground-level ambient PM2.5 and O3 over the central and eastern part of the domain (up to -3%).