Investigating Smoke Aerosol Emission Coefficients Using MODIS Active Fire and Aerosol Products: A Case Study in the CONUS and Indonesia

Abstract

Smoke aerosols released from biomass burning greatly influence air quality, weather, and climate. The total particulate matter (TPM) of smoke aerosols has been demonstrated to be a linear function of fire radiative energy (FRE) during a period of biomass burning via a smoke aerosol emission coefficient (Ce). However, it remains challenging to quantify Ce appropriately through satellite observations. In this study, an innovative approach was put forward to explore Ce by establishing the relationships between FRE and TPM in two regions, the CONtiguous United States and Indonesia. Specifically, we identified 584 isolated fire clusters and smoke plumes in the CONtiguous United States and 248 in Indonesia using Moderate Resolution Imaging Spectroradiometer natural color images, and then calculated FRE from Moderate Resolution Imaging Spectroradiometer active fire product and TPM from Moderate Resolution Imaging Spectroradiometer aerosol optical depth product for each fire‐smoke matchup during Terra and Aqua overpasses. The relationships between TPM and FRE were constructed to determine Ce using an ordinary least squares regression. The results show that FRE and TPM are significantly correlated (r2 ≥ 0.63, p < 0.001) with the Ce varying across regions and fuel types. In the CONtiguous United States, forest Ce values are 21.3 and 34.1 g/MJ and savanna Ce values are 18.2 and 22.8 g/MJ for western and eastern regions, respectively; additionally, Ce is 20.9 g/MJ for grasslands and 5.0 g/MJ for shrublands. In Indonesia, Ce is 52.4 and 30.0 g/MJ for peatlands and forests, respectively. Overall, this study improves our understanding of Ce variations with fuel types and climate regions.

Top-Down Estimation of Particulate Matter Emissions from Extreme Tropical Peatland Fires Using Geostationary Satellite Fire Radiative Power Observations

“…Figure 3 shows the data from the thirteen matchup fires of Figure 2 , used to estimate C e as 16.8 ± 1.6 g·MJ −1 through OLS regression, with Table 1 reporting various metrics for each matchup. For comparison, in Table 2 are the existing TPM published for SE Asian peatland fires, either derived from the multiplication of an FRE to fuel consumption conversion factor with a TPM emission factor [ 27 ], or calculated directly via a “top-down” method relating FRE to TPM emissions [ 37 ]. Values range from 69.3 g·MJ −1 in the former case to 52.4 g·MJ −1 in the latter, and are far higher than the 16.8 ± 1.6 g·MJ −1 determined herein.…”

“…A recent update to the EF from [ 32 ] provides an explicit, though indirectly estimated, TPM EF value for peat of 27.5 g·kg −1 , resulting in an even larger C e estimate of 161 g·MJ −1 . The “top-down” C e of 52.4 g·MJ −1 was calculated from a set of 19 collocated smoke plumes and fires burning on SE Asian peatlands, with the TPM measurements derived from MODIS deep blue AOD products [ 55 ] and the FRE calculated from MODIS FRP data provided by successive overpasses of the Terra and Aqua satellites and an assumed FRE diurnal distribution [ 37 ].…”

“…Estimating appropriate AODs for missing pixels is perhaps the most challenging problem of the four. In [ 37 ], this issue was tackled using the maximum near-fire-observed AOD value, but since the cause of the missing data is often related to the fact that a pixel’s AOD exceeds the maximum possible with the retrieval approach, or that the smoke is optically thick and misidentified as cloud [ 51 ], assigning the maximum “unsaturated” AOD value found elsewhere in the plume will certainly lead to a low biased AOD estimate. The consequence would be a low biased in-plume TPM estimate and resulting C e coefficient.…”

“…Whilst significant advancements have been made during the last two decades in the active fire detection and fire characterisation algorithms that underlie FRP approaches (e.g., [ 29 , 30 ]), and landcover specific EF databases are becoming more detailed in their contents [ 31 , 32 , 33 ], there remain uncertainties in the FRP approach to fire emissions estimation. These can stem from (i) the relatively limited sampling frequency provided by the polar-orbiting satellites that currently dominate provision of FRP [ 34 ]; (ii) the fact that some of any surface-fire emitted FRP maybe intercepted by overlying tree canopies [ 35 ]; (iii) that some fires are too small or weakly burning to be detected in the most common spaceborne FRP data products (see comparisons by [ 36 ]); and (iv), in the case of tropical peatlands, from the rather limited (and until recently largely laboratory-derived) emission factors for peat burning [ 37 ]. Perhaps the most significant uncertainties are associated with the conversion between FRP and fuel consumption rate or totals.…”

“…The FREM approach uses species- and biome-specific to directly link FRP data to the emissions of any particular species within the smoke, thus removing the need for the interim fuel consumption estimation step. The method is “top-down” as it relies on atmospheric and terrestrial remote sensing observations only, as detailed in [ 37 , 38 , 44 , 45 ]. In contrast to the Moderate Resolution Image Radiometer (MODIS) based approach of [ 44 ], the FREM approach also attempts to minimise the number of assumptions required when deriving each biome-specific by exploiting the almost continuous, very high temporal resolution FRP data available from geostationary sensors.…”