Dynamic savanna burning emission factors based on satellite data using a machine learning approach
<p><span id="page1040"/>Landscape fires, predominantly found in the frequently burning global savannas, are a substantial source of greenhouse gases and aerosols. The impact of these fires on atmospheric composition is partially determined by the chemical breakup of the constit...
| Main Authors: | , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2023-10-01
|
| Series: | Earth System Dynamics |
| Online Access: | https://esd.copernicus.org/articles/14/1039/2023/esd-14-1039-2023.pdf |
| _version_ | 1797662916775772160 |
|---|---|
| author | R. Vernooij T. Eames J. Russell-Smith J. Russell-Smith C. Yates C. Yates R. Beatty R. Beatty J. Evans J. Evans A. Edwards A. Edwards N. Ribeiro M. Wooster M. Wooster T. Strydom M. V. Giongo M. A. Borges M. Menezes Costa A. C. S. Barradas D. van Wees G. R. Van der Werf |
| author_facet | R. Vernooij T. Eames J. Russell-Smith J. Russell-Smith C. Yates C. Yates R. Beatty R. Beatty J. Evans J. Evans A. Edwards A. Edwards N. Ribeiro M. Wooster M. Wooster T. Strydom M. V. Giongo M. A. Borges M. Menezes Costa A. C. S. Barradas D. van Wees G. R. Van der Werf |
| author_sort | R. Vernooij |
| collection | DOAJ |
| description | <p><span id="page1040"/>Landscape fires, predominantly found in the frequently burning
global savannas, are a substantial source of greenhouse gases and aerosols.
The impact of these fires on atmospheric composition is partially
determined by the chemical breakup of the constituents of the fuel into
individual emitted chemical species, which is described by emission factors
(EFs). These EFs are known to be dependent on, amongst other things, the
type of fuel consumed, the moisture content of the fuel, and the
meteorological conditions during the fire, indicating that savanna EFs are
temporally and spatially dynamic. Global emission inventories, however, rely
on static biome-averaged EFs, which makes them ill-suited for the estimation
of regional biomass burning (BB) emissions and for capturing the effects of
shifts in fire regimes. In this study we explore the main drivers of
EF variability within the savanna biome and assess which geospatial proxies
can be used to estimate dynamic EFs for global emission inventories. We made
over 4500 bag measurements of CO<span class="inline-formula"><sub>2</sub></span>, CO, CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O EFs using
a UAS and also measured fuel parameters and fire-severity proxies during 129 individual fires. The measurements cover a
variety of savanna ecosystems under different seasonal conditions sampled
over the course of six fire seasons between 2017 and 2022. We complemented
our own data with EFs from 85 fires with locations and dates provided in the
literature. Based on the locations, dates, and times of the fires we retrieved
a variety of fuel, weather, and fire-severity proxies (i.e. possible
predictors) using globally available satellite and reanalysis data. We then
trained random forest (RF) regressors to estimate EFs for CO<span class="inline-formula"><sub>2</sub></span>, CO,
CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O at a spatial resolution of 0.25<span class="inline-formula"><sup>∘</sup></span> and a
monthly time step. Using these modelled EFs, we calculated their
spatiotemporal impact on BB emission estimates over the 2002–2016 period
using the Global Fire Emissions Database version 4 with small fires
(GFED4s). We found that the most important field indicators for the EFs of
CO<span class="inline-formula"><sub>2</sub></span>, CO, and CH<span class="inline-formula"><sub>4</sub></span> were tree cover density, fuel moisture content, and
the grass-to-litter ratio. The grass-to-litter ratio and the nitrogen-to-carbon ratio were important indicators for N<span class="inline-formula"><sub>2</sub></span>O EFs. RF models using
satellite observations performed well for the prediction of EF variability
in the measured fires with out-of-sample correlation coefficients between
0.80 and 0.99, reducing the error between measured and modelled EFs by
60 %–85 % compared to using the static biome average. Using dynamic EFs,
total global savanna emission estimates for 2002–2016 were 1.8 % higher
for CO, while CO<span class="inline-formula"><sub>2</sub></span>, CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O emissions were, respectively,
0.2 %, 5 %, and 18 % lower compared to GFED4s. On a regional scale we
found a spatial redistribution compared to GFED4s with higher CO, CH<span class="inline-formula"><sub>4</sub></span>,
and N<span class="inline-formula"><sub>2</sub></span>O EFs in mesic regions and lower ones in xeric regions. Over the
course of the fire season, drying resulted in gradually lower EFs of these
species. Relatively speaking, the trend was stronger in open savannas than
in woodlands, where towards the end of the fire season they increased again.
Contrary to the minor impact on annual average savanna fire emissions, the
model predicts localized deviations from static averages of the EFs of CO,
CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O exceeding 60 % under seasonal conditions.</p> |
| first_indexed | 2024-03-11T19:07:02Z |
| format | Article |
| id | doaj.art-2adbd1a809454feb837fe995cd511e91 |
| institution | Directory Open Access Journal |
| issn | 2190-4979 2190-4987 |
| language | English |
| last_indexed | 2024-03-11T19:07:02Z |
| publishDate | 2023-10-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | Earth System Dynamics |
| spelling | doaj.art-2adbd1a809454feb837fe995cd511e912023-10-10T05:53:07ZengCopernicus PublicationsEarth System Dynamics2190-49792190-49872023-10-01141039106410.5194/esd-14-1039-2023Dynamic savanna burning emission factors based on satellite data using a machine learning approachR. Vernooij0T. Eames1J. Russell-Smith2J. Russell-Smith3C. Yates4C. Yates5R. Beatty6R. Beatty7J. Evans8J. Evans9A. Edwards10A. Edwards11N. Ribeiro12M. Wooster13M. Wooster14T. Strydom15M. V. Giongo16M. A. Borges17M. Menezes Costa18A. C. S. Barradas19D. van Wees20G. R. Van der Werf21Department of Earth Sciences, Faculty of Science, Vrije Universiteit Amsterdam, Amsterdam, the NetherlandsDepartment of Earth Sciences, Faculty of Science, Vrije Universiteit Amsterdam, Amsterdam, the NetherlandsDarwin Centre for Bushfire Research, Charles Darwin University, Darwin, 0909 Northern Territory, AustraliaInternational Savanna Fire Management Initiative (ISFMI), Level 4, 346 Kent Street, Sydney, 2000 New South Wales, AustraliaDarwin Centre for Bushfire Research, Charles Darwin University, Darwin, 0909 Northern Territory, AustraliaInternational Savanna Fire Management Initiative (ISFMI), Level 4, 346 Kent Street, Sydney, 2000 New South Wales, AustraliaInternational Savanna Fire Management Initiative (ISFMI), Level 4, 346 Kent Street, Sydney, 2000 New South Wales, Australia321 Fire, Praia Do Tofo, Inhambane, 1300, MozambiqueDarwin Centre for Bushfire Research, Charles Darwin University, Darwin, 0909 Northern Territory, AustraliaInternational Savanna Fire Management Initiative (ISFMI), Level 4, 346 Kent Street, Sydney, 2000 New South Wales, AustraliaDarwin Centre for Bushfire Research, Charles Darwin University, Darwin, 0909 Northern Territory, AustraliaInternational Savanna Fire Management Initiative (ISFMI), Level 4, 346 Kent Street, Sydney, 2000 New South Wales, AustraliaFaculty of Agronomy and Forest Engineering, Eduardo Mondlane University, Maputo, MozambiqueEnvironmental Monitoring and Modelling Research Group, Department of Geography, King's College London, London, UKNational Centre for Earth Observation (NERC), Leicester, UKSouth African National Parks (SANParks), Scientific Services, Skukuza, South AfricaCenter for Environmental Monitoring and Fire Management (CEMAF), Federal University of Tocantins, Gurupi, BrazilChico Mendes institute for Conservation of Biodiversity (ICMBio), Rio da Conceição, BrazilChico Mendes institute for Conservation of Biodiversity (ICMBio), Rio da Conceição, BrazilChico Mendes institute for Conservation of Biodiversity (ICMBio), Rio da Conceição, BrazilDepartment of Earth Sciences, Faculty of Science, Vrije Universiteit Amsterdam, Amsterdam, the NetherlandsDepartment of Earth Sciences, Faculty of Science, Vrije Universiteit Amsterdam, Amsterdam, the Netherlands<p><span id="page1040"/>Landscape fires, predominantly found in the frequently burning global savannas, are a substantial source of greenhouse gases and aerosols. The impact of these fires on atmospheric composition is partially determined by the chemical breakup of the constituents of the fuel into individual emitted chemical species, which is described by emission factors (EFs). These EFs are known to be dependent on, amongst other things, the type of fuel consumed, the moisture content of the fuel, and the meteorological conditions during the fire, indicating that savanna EFs are temporally and spatially dynamic. Global emission inventories, however, rely on static biome-averaged EFs, which makes them ill-suited for the estimation of regional biomass burning (BB) emissions and for capturing the effects of shifts in fire regimes. In this study we explore the main drivers of EF variability within the savanna biome and assess which geospatial proxies can be used to estimate dynamic EFs for global emission inventories. We made over 4500 bag measurements of CO<span class="inline-formula"><sub>2</sub></span>, CO, CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O EFs using a UAS and also measured fuel parameters and fire-severity proxies during 129 individual fires. The measurements cover a variety of savanna ecosystems under different seasonal conditions sampled over the course of six fire seasons between 2017 and 2022. We complemented our own data with EFs from 85 fires with locations and dates provided in the literature. Based on the locations, dates, and times of the fires we retrieved a variety of fuel, weather, and fire-severity proxies (i.e. possible predictors) using globally available satellite and reanalysis data. We then trained random forest (RF) regressors to estimate EFs for CO<span class="inline-formula"><sub>2</sub></span>, CO, CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O at a spatial resolution of 0.25<span class="inline-formula"><sup>∘</sup></span> and a monthly time step. Using these modelled EFs, we calculated their spatiotemporal impact on BB emission estimates over the 2002–2016 period using the Global Fire Emissions Database version 4 with small fires (GFED4s). We found that the most important field indicators for the EFs of CO<span class="inline-formula"><sub>2</sub></span>, CO, and CH<span class="inline-formula"><sub>4</sub></span> were tree cover density, fuel moisture content, and the grass-to-litter ratio. The grass-to-litter ratio and the nitrogen-to-carbon ratio were important indicators for N<span class="inline-formula"><sub>2</sub></span>O EFs. RF models using satellite observations performed well for the prediction of EF variability in the measured fires with out-of-sample correlation coefficients between 0.80 and 0.99, reducing the error between measured and modelled EFs by 60 %–85 % compared to using the static biome average. Using dynamic EFs, total global savanna emission estimates for 2002–2016 were 1.8 % higher for CO, while CO<span class="inline-formula"><sub>2</sub></span>, CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O emissions were, respectively, 0.2 %, 5 %, and 18 % lower compared to GFED4s. On a regional scale we found a spatial redistribution compared to GFED4s with higher CO, CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O EFs in mesic regions and lower ones in xeric regions. Over the course of the fire season, drying resulted in gradually lower EFs of these species. Relatively speaking, the trend was stronger in open savannas than in woodlands, where towards the end of the fire season they increased again. Contrary to the minor impact on annual average savanna fire emissions, the model predicts localized deviations from static averages of the EFs of CO, CH<span class="inline-formula"><sub>4</sub></span>, and N<span class="inline-formula"><sub>2</sub></span>O exceeding 60 % under seasonal conditions.</p>https://esd.copernicus.org/articles/14/1039/2023/esd-14-1039-2023.pdf |
| spellingShingle | R. Vernooij T. Eames J. Russell-Smith J. Russell-Smith C. Yates C. Yates R. Beatty R. Beatty J. Evans J. Evans A. Edwards A. Edwards N. Ribeiro M. Wooster M. Wooster T. Strydom M. V. Giongo M. A. Borges M. Menezes Costa A. C. S. Barradas D. van Wees G. R. Van der Werf Dynamic savanna burning emission factors based on satellite data using a machine learning approach Earth System Dynamics |
| title | Dynamic savanna burning emission factors based on satellite data using a machine learning approach |
| title_full | Dynamic savanna burning emission factors based on satellite data using a machine learning approach |
| title_fullStr | Dynamic savanna burning emission factors based on satellite data using a machine learning approach |
| title_full_unstemmed | Dynamic savanna burning emission factors based on satellite data using a machine learning approach |
| title_short | Dynamic savanna burning emission factors based on satellite data using a machine learning approach |
| title_sort | dynamic savanna burning emission factors based on satellite data using a machine learning approach |
| url | https://esd.copernicus.org/articles/14/1039/2023/esd-14-1039-2023.pdf |
| work_keys_str_mv | AT rvernooij dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT teames dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT jrussellsmith dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT jrussellsmith dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT cyates dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT cyates dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT rbeatty dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT rbeatty dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT jevans dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT jevans dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT aedwards dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT aedwards dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT nribeiro dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT mwooster dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT mwooster dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT tstrydom dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT mvgiongo dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT maborges dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT mmenezescosta dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT acsbarradas dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT dvanwees dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach AT grvanderwerf dynamicsavannaburningemissionfactorsbasedonsatellitedatausingamachinelearningapproach |