Research

Wildfires and Smoke in North America

Severe fire seasons in recent years have underscored the increasing vulnerability of the western United States and Canada to wildfires, which deteriorate public health, disrupt livelihoods, and damage infrastructure. Against the backdrop of a warming climate and recent droughts, the buildup of fuels from the legacy of federal active fire suppression in the 1900s led to larger and more severe wildfires in recent decades.

In current work, we are tackling research on western U.S. wildfires on three fronts. First, we are using NOAA's Hazard Mapping System (HMS) smoke product as an indicator of fire-related smoke. The HMS smoke product contains polygons drawn by NOAA analysts who delineate smoke from non-smoke areas using primarily GOES satellite imagery. We found that wildfires in CA, OR, and WA in 2020 exacerbated short-term exposure to smoke particulate matter and the severity of COVID-19 pandemic (Zhou et al., 2021). While the HMS smoke product is increasingly used in public health and air pollution studies, the dataset is subject to caveats. For example, HMS may indicate aloft smoke that does not affect surface air quality, particularly within polygons categorized as light density. To investigate how accurately HMS reflects surface conditions, we are comparing HMS with airport observations, monitor measurements, and model estimates (Liu et al., in press). Second, we seek to quantify smoke impacts on public health, smoke risk, and smoke exposure mitigation from prescribed burning (Kelp et al., 2023). Third, we developed an algorithm using the GOES-East and GOES-West geostationary satellites to map the spatio-temporal progression of large wildfires on an hourly basis (Liu et al., 2024). With better estimates of the diurnal cycle of fire growth, we will investigate the meteorological and active fire suppression controls on fire spread.

Publications: Zhou et al. (2021, Sci. Adv.), Kelp et al. (2023, Earth's Future), Liu et al. (2024, ESSD), Liu et al. (in press, IJWF)

Agricultural Fires and Air Quality Degradation in Northwestern India

In northwestern India, the double crop system alternates between the monsoon crop (kharif, predominantly rice) and winter crop (rabi, predominantly wheat). The short turnaround between harvest and sowing pressures farmers to quickly remove crop residue from the field to prepare for the next planting. In addition, the use of combine harvesters, which saves time and decreases labor costs, leaves behind abundant root-bound residues that are difficult to remove manually. The prevailing, cost-effective method is to burn the crop residues. However, crop residue burning releases a suite of pernicious gases and aerosols into the atmosphere that degrades both rural and urban air quality. Although crop residue burning is just one of many pollution sources, it is highly episodic and seasonal. Every year, the burning of rice residue from October to November contributes to thick haze and poor air quality over the region (Lan et al., 2022). New Delhi, a mega-city already suffering from heavy local pollution, is located downwind of these seasonal agricultural fires. During the post-monsoon, meteorology (e.g. slow winds and temperature inversions) help trap pollutants near the surface in north India, increasing air pollution levels (Gautam et al., 2023).

We first used HYSPLIT atmospheric back trajectories to define Delhi's post-monsoon (October-November) and pre-monsoon (April-May) airsheds, or approximate regions where emissions can affect Delhi's local air pollution (Liu et al., 2018a). Focusing on the post-monsoon burning season, we then used the STILT model to estimate the PM2.5 enhancement in the Delhi National Capital Territory (Cusworth et al., 2018). We also developed a fusion method using MODIS (500 m x 500 m) and Landsat (30 m x 30 m) imagery to approximate burned area; however, we find that moderate-resolution sensors, such as MODIS, are unable to see many small fires, therefore underestimating overall fire activity (Liu et al., 2019). In a more recent approach, we used satellite and household survey data to build a new agricultural fire emissions inventory, SAGE-IGP, for north India (Liu et al., 2020b).

North India faces dual challenges of severe groundwater depletion and air pollution. To alleviate groundwater depletion in Punjab and Haryana, the Preservation of Sub-Soil Water Act of 2009 delayed rice sowing dates to align the timing closer to the monsoon onset. However, this delay in the monsoon rice growing season has led to a delay in the post-monsoon fire season of around two weeks from 2003-2016 (Liu et al., 2021a). Through atmospheric modeling of smoke PM2.5 driven by SAGE-IGP fire emissions, we find that the delay in the post-monsoon fire season has consistently exacerbated poor air quality in nearby downwind areas, such as in New Delhi (Liu et al., 2022). The unintended air pollution resulting from the cascading delays in the crop cycle and fire season suggests that future policies should consider groundwater depletion and smoke pollution as linked challenges.

Publications: Liu et al. (2018a, Atmos. Environ.), Cusworth et al. (2018, Environ. Res. Lett.), Liu et al. (2019, Environ. Res. Commun.), Liu et al. (2020b, Atmos. Environ. X), Liu et al. (2021a, Environ. Res. Lett.), Liu et al. (2022, J. Geophys. Res. Atmos.), Lan et al. (2022, Nat. Commun.), Gautam et al. (2023, Geophys. Res. Lett.)

Modeling the Public Health Impacts of Fires in Equatorial Asia

In strong El Niño years (typically very dry conditions), such as 2006 and 2015, fires associated with oil palm, timber, and logging plantations and deforestation in Indonesia lead to severe, persistent haze over Equatorial Asia from July to November. In part due to drought conditions associated with El Niño and the large area of fuel-rich peatlands, which contain partially decayed organic material, fires were able to persist and associated emissions exceeded yearly U.S. fossil fuel emissions on many days during the 2015 fire season (WRI, GFED). Peat sequesters carbon in meters-deep layers accumulated over thousands of years; peat fires, then, reduce the ability of Indonesia to be a carbon sink. In past work, we developed a framework to rapidly estimate the public health impacts of the 2015 severe haze episode in Equatorial Asia. Overall, smoke exposure, weighted by population, more than doubled from the 2006 to 2015 severe haze events. Based on chemical transport modeling, fire emissions in the provinces of Jambi and South Sumatra and West and Central Kalimantan contributed to the bulk of population-weighted smoke exposure in Indonesia, Singapore, and Malaysia (Koplitz et al., 2016). We then used Google Earth Engine to build the SMOKE Policy Tool, which estimates the public health benefits of reducing fires in various land use scenarios, including peatland restoration (Marlier et al., 2019). In more recent work, we find that using different global fire emissions inventories can significantly impact modeled estimates of smoke particulate matter, particularly in Equatorial Asia (Liu et al., 2020a). To help end-users quickly compare five widely-used inventories and diagnose their underlying assumptions, we developed the FIRECAM Tool (Liu et al., 2020a).

Publications: Koplitz et al. (2016, Environ. Res. Lett.), Marlier et al. (2019, GeoHealth), Liu et al. (2020a, Remote Sens. Environ.), Madrigano et al. (2024, GeoHealth)

Satellite Observations of Global Fire Activity: Trends, Implications, and Technical Challenges

Satellites are an invaluable tool for observing, monitoring, and mapping global fire activity (FIRMS). Sensors such as MODIS, aboard the NASA's Terra and Aqua satellites, and VIIRS, aboard the S-NPP and NOAA-20 satellites, detect active fires on a daily and global scale by tracking thermal anomalies. Burned area can be mapped at high spatial resolution (10-30 m) using Landsat and Sentinel-2 imagery by comparing pre- and post-fire infrared signatures. Several technical challenges complicate interpretation of satellite-derived fire datasets. For example, clouds and thick haze can obscure fires from satellite detection, leading to missed active fires. Dark soil and cloud shadows make it challenging to separate burned from unburned areas. Moreover, we find artificial discontinuities in MODIS burned area datasets due to the tile-dependent algorithms that process images in chunks for computational efficiency (Liu and Crowley, 2021; BA Tiling Explorer).

Humans increasingly control fire ignition and spread through deforestation, agriculture, and forest management. However, more people than ever are living in the wildland-urban interface and thus vulnerable to destructive fires. Not only does fire shape ecosystems and landscapes, but it also impacts air quality, public health, and climate through emission of greenhouse gases and aerosols. Although it is important to quantify trends in the magnitude of burned area, temporal shifts in the fire season add another dimension to the story. Following Liu et al. (2021a), we search globally for regions where the timing of the primary fire season has shifted, focusing on north Australia and southwest Russia (Liu et al., 2021b). In these two regions, we find that humans altered the temporal distribution of fire activity through increased prescribed burning and potentially modernization in agricultural practices.

Publications: Liu and Crowley (2021, IOP SciNotes), Liu et al. (2021b, Environ. Res. Lett.)