Two Decades of Fire‐Induced Albedo Change and Associated Short‐Wave Radiative Effect Over Sub‐Saharan Africa

Abstract

We present an analysis of 20 years of fire and albedo data in Africa. We show that, in the mean, the sub‐Saharan Africa post‐fire surface albedo anomaly can be parameterized using an exponential recovery function, recovering from a decrease of 0.019±0.001 0.019±0.0010.019\pm 0.001 immediately after a fire with a time constant of 34.0±0.4 34.0±0.434.0\pm 0.4 days. Although the magnitude of albedo changes shows large spatial and temporal variations and a strong land cover type (LCT) dependency, exponential recovery is observed in the majority of LCTs. We show that fires cause long‐term surface brightening, with an Africa‐wide albedo increase of (9.5±0.2)×10−4 (9.5±0.2)×10−4(9.5\pm 0.2)\times 1{0}^{-4} 10 months after a fire, but we find this is driven almost exclusively by slow vegetation recovery in the Kalahari region, confirming previous findings. Using downward surface shortwave flux (DSSF) estimates, we calculate the fire‐induced surface radiative forcing (RF), peaking at 5±2 5±25\pm 2 Wm⁻² in the burn areas, albeit with a significantly smaller effect when averaged temporally and spatially. We find that the long‐term RF in months 5–10 after a burn averaged over the continent is negative because of the brightening observed. Despite a well‐documented reduction in burning in Africa in the recent decades, our temporal analysis does not indicate a decrease in the overall fire‐induced RF likely due to large interannual variability in albedo anomaly and DSSF data. However, we observe a decline in the short‐term RF in southern hemisphere Africa, driven by both a reduction in fires and changes in LCT distributions.

Full text

Two Decades of Fire‐Induced Albedo Change and Associated
Short‐Wave Radiative Effect Over Sub‐Saharan Africa
Michaela Flegrová
1,2
and Helen Brindley
2
1
Department of Physics, Leverhulme Centre for Wildfires, Environment & Society, Imperial College London, London, UK,
2
Department of Physics, NERC National Centre for Earth Observation, Imperial College London, London, UK
Abstract We present an analysis of 20 years of fire and albedo data in Africa. We show that, in the mean,
the sub‐Saharan Africa post‐fire surface albedo anomaly can be parameterized using an exponential recovery
function, recovering from a decrease of 0.019 ±0.001 immediately after a fire with a time constant of
34.0±0.4 days. Although the magnitude of albedo changes shows large spatial and temporal variations and a
strong land cover type (LCT) dependency, exponential recovery is observed in the majority of LCTs. We show
that fires cause long‐term surface brightening, with an Africa‐wide albedo increase of (9.5±0.2) × 104
10 months after a fire, but we find this is driven almost exclusively by slow vegetation recovery in the Kalahari
region, confirming previous findings. Using downward surface shortwave flux (DSSF) estimates, we calculate
the fire‐induced surface radiative forcing (RF), peaking at 5 ±2 Wm
2
in the burn areas, albeit with a
significantly smaller effect when averaged temporally and spatially. We find that the long‐term RF in months 5–
10 after a burn averaged over the continent is negative because of the brightening observed. Despite a well‐
documented reduction in burning in Africa in the recent decades, our temporal analysis does not indicate a
decrease in the overall fire‐induced RF likely due to large interannual variability in albedo anomaly and DSSF
data. However, we observe a decline in the short‐term RF in southern hemisphere Africa, driven by both a
reduction in fires and changes in LCT distributions.
Plain Language Summary This study examines 20 years of data on fires and albedo (surface
reflectivity) in Africa. We find that fires in the region cause, on average, an immediate surface darkening,
reducing albedo by 0.019 ±0.001. After a fire the albedo initially recovers rapidly before the recovery slows;
this temporal behavior is seen over most land cover types but the magnitude depends on land cover.
Interestingly, the research reveals that fires lead to some long‐term surface brightening, with an African average
rise of (9.5±0.2) × 104as measured 10 months after a fire. This is primarily attributed to slow vegetation
recovery in the Kalahari region. The study also calculates the surface shortwave radiative forcing (RF) induced
by fires—a measure of how much more of the incoming solar radiation is absorbed at the surface due to a fire—
which peaks at 5 ±2 Wm
2
in the burned areas. However, when averaged over time and the whole continent,
the overall effect is smaller. Despite a documented decrease in the number of fires in Africa over the last
20 years, a corresponding trend in the continent‐wide fire‐induced RF is not observed likely because of large
interannual variability in albedo and incoming radiation.
1. Introduction
Fire is an important wide‐spread Earth‐system process, affecting local ecosystems and climate around the globe.
An estimated 350 Mha or 2% of all global land is burned annually, and wildfire activity is observed on all
continents (Giglio et al., 2013). Wildfires release CO2and other gases and are the biggest contributor to black
carbon and organic carbon emissions, accounting for over half of these aerosols (Crutzen & Andreae, 1990;
Mahowald et al., 2011). Fire also affects the underlying surface; burning results in the removal of vegetation and
the deposition of dark ash and char (Smith et al., 2005; Stronach & McNaughton, 1989). Some plant types and
ecosystems directly rely on regular landscape burning (Pausas & Keeley, 2019). Although it is observed that fires
cause significant radiative forcing (RF), the net effect and its magnitude remain uncertain (Voulgarakis &
Field, 2015).
Human activity is one of the most important drivers of fires in Africa with most fires started either deliberately or
accidently by people (Andela et al., 2017; Archibald, 2016). Fire practices have been changing in recent years
partly due to the expansion and intensification of agriculture, affecting the frequency of fires (Earl &
RESEARCH ARTICLE
10.1029/2024JD041491
Key Points:
•Changes to surface albedo following
landscape fires can be accurately
modeled using an exponential recovery
function
•African landscape fires cause
significant surface radiative forcing
(RF) for several months after a burn
with notable spatial variations
•The short‐term surface RF due to fires
in southern‐hemisphere Africa has
declined over the past two decades
Correspondence to:
M. Flegrová,
michaela.flegrova16@imperial.ac.uk
Citation:
Flegrová, M., & Brindley, H. (2025). Two
decades of fire‐induced albedo change and
associated short‐wave radiative effect over
sub‐Saharan Africa. Journal of
Geophysical Research: Atmospheres,130,
e2024JD041491. https://doi.org/10.1029/
2024JD041491
Received 30 APR 2024
Accepted 25 DEC 2024
© 2025. The Author(s).
This is an open access article under the
terms of the Creative Commons
Attribution License, which permits use,
distribution and reproduction in any
medium, provided the original work is
properly cited.
FLEGROVÁ AND BRINDLEY 1 of 14

Simmonds, 2018). Investigations of African fire occurrence show that burned area (BA) has been declining by
about 1.2%–1.5% per year in the past two decades, with 80% of the decline occurring in Northern Hemisphere
Africa (Andela et al., 2017; Zubkova et al., 2019).
Surface albedo is defined as the fraction of incoming solar radiation at the surface that is reflected and is
wavelength‐ and angle‐of‐incidence‐dependent (Ångström, 1925). Albedo changes cause RF at the surface, and
their effect on the energy balance has been shown to affect climate and ecosystem processes, including atmo-
spheric circulation and rainfall (Ceppi & Shepherd, 2019; Saha et al., 2017). Previous studies have investigated
the effect of African landscape fires on surface albedo using satellite data. However, there is currently
disagreement when it comes to the direction of the change, with some studies reporting darkening of the surface
post‐fire, whereas others have observed significant brightening. Most notably, Dintwe et al. (2017) used satellite
observations made by the Moderate Resolution Imaging Spectroradiometer (MODIS) to evaluate albedo change
after fire and its associated RF, finding the short‐term effect to be up to 5.41 Wm2in burned areas. Their findings
suggested fire‐induced albedo change was entirely negative, corroborating similar findings by Gatebe
et al. (2014). More recent studies by Saha et al. (2017, 2019) found evidence of long‐term brightening, implying a
negative RF in the longer term. As surface albedo naturally varies throughout the year and interannually (Palle
et al., 2016), the effect of wildfires can be difficult to isolate.
Although fire is an important process affecting the Earth's climate, its representation in Earth System Models
(ESMs) has been limited to date. The current generation of ESMs is not developed enough to account for fire
feedback, including fire influence on atmospheric composition, climate, and vegetation (Teixeira et al., 2021). For
example, some models currently prescribe or simulate vegetation distributions but fail to account for further fire‐
vegetation feedback where the interaction is two‐way (Hantson et al., 2020; Rabin et al., 2017). Other factors,
such as demographic changes, affect fire significantly, especially in Africa, where population size, distribution,
and economic activity are rapidly changing; those indicators are, however, not well‐represented in ESMs at the
moment (Teixeira et al., 2021).
This work expands on previous studies by looking at albedo changes following fires in all sub‐Saharan Africa. We
set out to evaluate what the instantaneous effects of fire on albedo are, and what the recovery looks like, focusing
on determining whether the change is unidirectional, or if brightening is observed long‐term. We also aim to
determine what the associated impact on surface shortwave fluxes of these changes is and what would be the
implications of potential brightening for the surface radiative balance. Further, we aim to parametrize the albedo
recovery period, simplifying how the time‐dependent surface albedo behavior is represented for potential
implementation in climate models. We ask how these effects vary spatially, temporally, and with land‐cover type.
2. Data Used
We use 20 years of data from NASA's MODIS instrument on board the Terra and Aqua Sun‐synchronous sat-
ellites, namely the BA (MCD64A1), Albedo (MCD43A3), and land cover type (LCT) (MCD12Q1) products,
Collection 6.1, covering the period 2001–2020. Additionally, we use the downward surface shortwave radiation
flux (DSSF) produced by EUMETSAT's Land Satellite Application Facility and derived from measurements by
the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) on board the geostationary Meteosat Second
Generation satellite series.
2.1. Burned Area
The MCD64A1 product uses the daily MODIS land surface reflectance data at 500 ×500 m pixel resolution to
estimate fire location and day of burn (Giglio et al., 2018). The algorithm focuses on identifying spectral,
structural, and temporal changes following a fire, which result from ash deposition, charring, and vegetation
removal or alteration. Compared to an active fire product, which works by detecting actively burning fires during
a satellite overpass (Giglio et al., 2003) in the absence of cloud a BA product is able to detect a burned pixel even
after burning concludes. The majority of fires in Africa are short‐lived, lasting less than 15 min (Roberts
et al., 2009); an active fire product, unlike a BA product, may therefore not detect these if they do not burn during
the satellite overpass.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 2 of 14

2.2. Albedo
The MCD43A3 product calculates the bidirectional reflectance distribution function (BRDF) in three spectral
bands: 0.30.7 μm (visible), 0.75.0 μm (near‐infrared), and 0.35.0 μm (shortwave) (Schaaf et al., 2002).
Black Sky and White Sky albedo values are available to represent the two extreme cases of a point illumination
and a homogeneous hemispherical illumination, respectively. Although daily values are available for the
MCD43A3 product, each value is calculated using 16 days of data to account for different angles of observation
(Schaaf et al., 2002); each daily albedo value therefore effectively has an associated ±8 days uncertainty. Our
analysis used the shortwave Black‐Sky albedo values, as the arid and semi‐arid regions of Africa are usually
characterized by low cloud cover (Wylie et al., 2005). However, additional tests showed that the corresponding
White Sky albedo product produced very similar albedo anomalies when processed using the method we outline
in Section 3.2.
2.3. Land Cover Type
To investigate post‐fire albedo development in different land cover types (LCTs), we used the MCD12Q1 annual
product at 500 m resolution, developed originally by Friedl et al. (2002) and later improved by Friedl et al. (2010).
Their algorithm uses information from all 7 MODIS land channels (6202155 nm), focusing mainly on the
surface reflectance and vegetation index data. Several LCT classifications are available; here, we use the In-
ternational Geosphere‐Biosphere Programme (IGBP) classification, which recognizes 17 different LCTs based on
the vegetation composition in the ecosystem (Loveland & Belward, 1997).
2.4. Downward Shortwave Radiation Flux
We use the daily downward surface shortwave flux (DSSF) product version 0.6.1 derived from SEVIRI (Geiger
et al., 2008). This product combines the extraterrestrial solar radiation with the solar zenith angle and an estimate
of the effective atmospheric transmittance under either clear or cloudy conditions to derive a broadband surface
flux covering the range 0.3–4 microns. The product is output at the SEVIRI pixel scale which is 3 ×3 km at the
subsatellite point, and data are available at 30 min intervals. Here, we use the temporally integrated daily product
spatially regridded to match the MODIS products using a nearest neighbor algorithm. The DSSF data are
available from part way through 2004 onwards; hence, our analysis covers the period 2005–2020.
3. Method
3.1. Study Area and Period
Our analysis covers sub‐Saharan Africa, defined as all African land areas south of 20°N but excluding
Madagascar. We divide each of the 1,200 km MODIS tiles into 12 ×12 segments to perform the analysis on a
grid of 100 km resolution. The grid cell size was chosen to provide a decent spatial resolution akin to the order of
magnitude resolution of current global climate models, while also ensuring a sufficient number of fires were
present in the majority of cells, giving statistical robustness to the results obtained. The data from 40,000 MODIS
pixels (500 m) are averaged in each 100 km grid cell; on average, 11% of these pixels burn every year, but this
number varies significantly spatially.
In Figure 1, we show the fire frequency and dominant LCT according to the IGBP classification, as derived from
MODIS data, in each 100 km grid cell. Here, fire locations identified agree with the analysis by Giglio
et al. (2013), with significant burning observed in the northern as well as the southern hemisphere. As most fires
occur in the savannas and grasslands, little burning is observed in the forested areas directly around the equator.
In most regions in Africa, burning happens predominantly in the dry season with fire season typically lasting only
a few months (Giglio et al., 2006). For this reason, we analyze albedo changes 2 months prior to and 10 months
following a fire, covering a period of 1 year but aiming to minimize the interference of the following fire season.
The top 5 LCTs—Savannas (46% of all fires), Grasslands (34%), Forests (7%), Woody Savannas (6%), and
Croplands (4%)—are also analyzed separately to investigate how the albedo effects vary depending on the
vegetation type. Our analysis combines the deciduous broadleaf and mixed forests LCTs, accounting for 4.2% and
2.7% of fires, respectively, to create the Forests LCT map. For both albedo and DSSF, each calendar year is
analyzed separately, enabling us to track interannual variability of albedo response to fire and identify potential
drivers of recovery.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 3 of 14

Recent decline in BA in Africa is well‐documented, driven by a notable decline in fire activity in NHA (e.g.,
Zubkova et al. (2019), Jiang et al. (2020)) as illustrated in Figure 2. We split our analysis by northern and southern
hemisphere Africa (NHA and SHA, respectively), as different trends are observed in the two hemispheres. In
SHA, fire activity increases until 2011 and decreases thereafter. Savage and Strydom (2016) hypothesize that the
unusually active fire seasons in 2010 and 2011 amplified by a strong La Niña event (a mechanism also
demonstrated by Amiri‐Farahani et al. (2020)) may have led to an improvement in fire mitigation and man-
agement practices, resulting in an overall reduction in BA thereafter. We analyze fire‐induced albedo changes and
their associated surface RF over two decades (2001–2020) to evaluate whether the observed change in fire activity
affects the local impact of individual fires as well as the continent‐wide effects.
3.2. Albedo Anomaly Definition
To determine the effect of fire and isolate it from the natural annual albedo variability, post‐fire albedo needs to be
compared to albedo under similar conditions but unaffected by burning. Here, we follow a temporal anomaly
comparison approach similar to Dintwe et al. (2017), where the albedo of a burned 500 m MODIS pixel is
compared to its own albedo in previous and future years when burning did not take place.
It has been shown (e.g., Andela et al. (2016)) that fire return periods are influenced by fuel (vegetation) build up
over several years; surface conditions are therefore expected to be different the year before or the year after a fire
Figure 1. Study area subdivided into 100 km grid cells, showing the average annual burned area percentage in each cell,
averaged over 2001–2020 (a) and the corresponding dominant international geosphere‐biosphere program land cover type
classification for the same period (b), both based on Moderate Imaging Spectroradiometer (MODIS) data. MODIS tiles
shown for reference.
Figure 2. Total burned area during the fire season by year, corresponding to the total area of pixels identified as burnedin the
MCD64A1 data set. In northern hemisphere Africa (NHA, blue), a decline of 2.1×10
4km2per year is observed
p=1.2×106). In southern hemisphere Africa (SHA, red), opposing trends are observed before and after 2011, with an
increase of 2.1×104km2/year (p=0.018)and a decrease of 2.7×104km2/year (p=0.003)respectively. Trends are
determined using an ordinary least squares algorithm.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 4 of 14

due to different levels of vegetation present. To minimize this effect when calculating a pixel's baseline clima-
tology, we use the data from all years unaffected by fire, that is, where the most recent fire was longer than
10 months prior to and the next fire no sooner than 2 months following the albedo measurement, to generate a “no‐
fire” albedo climatology for each 500 m MODIS pixel. This albedo climatology αcis calculated per pixel pper
day of year (DOY) as
αc(DOY,p) = 1
Y∑
Y
y
α(DOY,p,y) (1)
where αis the albedo value in a year yunaffected by fire, as defined above, and Yis the number of fire‐free years
contributing to the calculation.
We anticipate that the albedo climatology might change over the full study period, as Africa's landscape develops
and changes. Climatologies were calculated separately for 2001–2010 and 2011–2020 to account for any un-
derlying drift in albedo over the two decades. With a typical pixel burning every 3–5 years (Archibald
et al., 2010), we expect Yto be about 7–8, providing a sufficient number of datapoints to smooth out interannual
variability effects in each of the climatologies. To avoid basing a climatology on too few samples, we only create
climatologies for and analyze pixels with Y≥3.
3.3. Fire‐Induced Albedo Change Calculation
After calculating the non‐fire albedo climatology for each 500 m MODIS pixel for 2001–2010 and 2011–2020
following the approach outlined above, we considered each year of the record, identifying all fires per pixel.
Where more than one burning occurred within the same pixel in a given year, the last of the burns was analyzed
with the previous burns discarded to avoid the interference of subsequent fires with the albedo recovery.
For each fire event, the albedo anomaly on day tafter the fire, αa(t,p), was calculated using the corresponding
albedo and climatology values as
αa(t,p) = α(t,p)  αc(t,p),(2)
with 60 ≤t≤300. The results of this calculation for one pixel are illustrated in Figure 3. Individual pixel data are
dominated by noise. To obtain clearer spatial data, we average albedo anomalies for all fires within each 100 km
grid cell across the study period. We also calculate averages for each of the top 5 land‐cover types, combining the
albedo anomalies from all burns within the LCT and the study period. The goal is to obtain clear signals and
Figure 3. Calculation of albedo anomaly for a given 500 mModerate Imaging Spectroradiometer pixel. Top panel shows the
climatology derived from non‐fire years (blue), with the same pattern repeating every 12 months, and the actual pixel albedo
values (red). Bottom panel shows the difference between the values ‐ the albedo anomaly. Orange vertical lines indicate a fire
event.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 5 of 14

potentially develop empirical relationships that could easily be implemented in global climate models, while
noting the spread in the findings as well as temporal and spatial differences and possible LCT dependency.
3.4. Estimating the Downward Surface Shortwave Flux
Radiative impact of fires is directly proportional to the DSSF levels, which vary with DOY and atmospheric
conditions and will therefore depend on fire location and date. We determine the relevant levels of DSSF in a
similar way to our albedo analysis. To calculate the average DSSF value over the fire‐affected MODIS 500 m
pixels, we first identify all pixels that have burned in a given time period. We then determine the DSSF at these
pixels from 60 days before to 300 days after each burn by finding the closest available SEVIRI datapoint cor-
responding to a given MODIS pixel, again averaging as we go along. Using this method, we calculate separate
averages for each 100 km grid cell, LCT, and year.
3.5. Evaluating the Radiative Forcing
We define the surface RF due to a fire‐induced change in surface albedo αaon day tafter a fire as
RF(t) =  αa(t)DSSF(t).(3)
The negative sign is to preserve the convention that an enhanced albedo leads to a reduction in absorbed solar
radiation at the surface or a negative RF.
We evaluate RF at three temporal scales: short‐term RF, RFS, defined as the RF at 8 days after a burn; medium‐
term RF, RFM, defined as the average RF between the fire event and 150 days after a fire; and longer‐term RF,
RFL, averaged between days 150 and 300. We define RFSbased on the ±8 days uncertainty of the albedo data,
arising from the way albedo values are calculated, which means RF peaks, on average, on day 8. The 150‐day
cutoff for medium‐range RF falls in the middle of the post‐fire period investigated but also marks the point
where post‐fire albedo anomaly becomes positive on the continental scale as will be illustrated in Figure 4.
We define the short‐term RF as
RFS=RF(t=8)fb(4)
and the medium‐ and long‐term RF as
RFM,L=1
150∫150,300
0,150
RF(t)fbdt (5)
where fbis the burn fraction, that is, the percentage of MODIS 500 m pixels that experienced burning in the area
considered in a given year.
4. Results and Discussion
4.1. Post‐Fire Albedo Anomaly
Our analysis covered a total of 163 million fires between the years 2001 and 2020. Although large spatial and
temporal variations are present, all LCTs, areas, and years analyzed show a similar trend, where the average
albedo drops sharply immediately after a fire event and recovery back to baseline can be approximated as an
exponential function. After the fire, the average albedo anomaly αaat time tcan be described as
αa(t) =  Δα et
τ(6)
where Δαis the maximum albedo drop or albedo anomaly at t=8 days, and τrefers to the albedo recovery time
constant, where αa(t=τ)≈0.37 Δα.
Equation 6assumes complete albedo recovery, that is, limt→∞αa(t) = 0, with no positive values of αa. However,
our analysis shows that some regions exhibit strong evidence of albedo brightening, where αa“over‐recovers” and
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 6 of 14

its value is positive and relatively constant by the end of the period we considered. To correct for this, we
introduce an albedo brightening constant αb, modifying Equation 6as
αa(t) =  (Δα+αb)et
τ+αb.(7)
Here, the first instance of αbensures Δαis defined with respect to 0 not αb.
Although we assume that, by definition, albedo values must eventually recover back to the climatology baseline,
Equation 7was found to be a good fit for the time period considered, showing that in some regions the observed
brightening persisted beyond 300 days following a fire. Possible reasons for this brightening are discussed in
Section 4.3. Data averaged over the study area and period are shown in Figure 4.
Equation 7is a good representation of the average albedo recovery following fires, suitable for the study period
and area, when a large number of fires are considered. However, individual 500 m MODIS pixel recovery is rarely
smooth instead typically following a noise‐dominated trajectory. Although the proposed relationship is suitable
for the evaluation of albedo impacts on a larger scale, it should not be used to predict surface recovery following
individual fires.
4.2. Fire‐Induced Surface Radiative Forcing
In Figure 5, we show the average RF in locations affected by fire as a function of time since burn, averaged over
the continent and calculated using Equation 3. Although some variation in DSSF is present in the 300 days
following a fire, the shape of the RF(t)curve is dominated by αa(t)values, following a shape similar to Figure 4
but inverted. Similarly to αa(t), RF peaks 8 days after the burn, with RF =4.5 Wm2and a spatial variation
characterized by a standard deviation σRF =1.7 Wm2. These results are comparable to the findings of Dintwe
et al. (2017), who found RF to peak at 5.98 ±6.12 Wm2. The annual burn fraction in the study area averaged
over 2004–2020 is fb=0.11; using this value, we calculate the area‐average RFS=0.50 ±0.19 Wm2,
RFM=0.13 ±0.02 Wm2and RFL=  0.02 ±0.02 Wm2.
4.3. Spatial Variation
Fitting Equation 7to albedo anomaly values using the SciPy Trust Region Reflective algorithm (Virtanen
et al., 2020) allows us to extract the fit parameters and compare them across the different locations analyzed. In
Figure 6, we show the fitted Δα,τand αbvalues across Africa. The data show the need for spatial discrimination,
as the results differ significantly across the study area.
Albedo recovery is slowest in the Sahel, with τvalues of over 100 days in many cells, as illustrated in Figure 6b.
The magnitude of the instantaneous change in albedo following a fire, represented by Δα, also shows large spatial
Figure 4. Average albedo anomaly following a fire over the whole study period and region. Day of burn uncertainty (yellow
shading) refers to the ±8 day uncertainty on albedo values. Albedo uncertainty (gray shading) represents the spatial variation.
Exponential fit, found using the SciPy Trust Region Reflective algorithm (Virtanen et al., 2020), follows Equation 7, with
Δα=0.019 ±0.001, α
b= (9.5±0.2) × 104and τ=34.0±0.4 days.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 7 of 14

variations, with values peaking in the Sahel and around the Kalahari Desert in Botswana, as shown in Figure 6a.
Comparison with Figure 1a indicates that grid cells with more fires are found to have both faster recovery times
and smaller overall albedo change per fire. Figure 6c shows that the need for the brightening correction in
Equation 7is driven mostly by two regions; as previously identified by Saha et al. (2017, 2019), brightening is
observed in the Kalahari region in southern Africa with αbvalues of 0.02 or more in several grid cells. Similar αb
values are observed in some parts of the Sahel; however, due to high values of τin the region, positive values of αa
are often not observed there in the first 300 days following a fire.
Saha et al. (2019) hypothesize that soil color plays an important role in long‐term surface brightening following
fires. The Sahel and Kalahari regions are characterized by brighter sands (Dewitte et al., 2013); when fires remove
vegetation and underlying soil is exposed, brightening can be observed once charring subsides. Additionally, both
regions are characterized by low annual precipitation (Lian et al., 2022), leading to slower vegetation recovery. It
is therefore likely that the brightening observed is due to a combination of the two factors; vegetation does not
recover fast enough after a burn, exposing especially bright soil.
Figure 7shows the spatial variation of RF. At all timescales investigated, values of RF are dependent on fband
Δα, increasing linearly with both. However, RFM,Lalso have a linear, inverse relationship with αb, and increase
Figure 5. Average radiative forcing of albedo changes following fires at t=0 across all burn locations, calculated according
to Equation 3. Gray shading represents the spatial standard deviation. Day of burn uncertainty (yellow shading) refers to the
±8 day uncertainty on albedo values.
Figure 6. Map of Δα(a), τ(b) and α
b(c) values extracted from albedo anomaly fits (Equation 7), averaged over 2001–2020.
Only grid cells with a good fit χ2≤2)are shown. Large spatial variability is evident in all three parameters, with extreme
values present in areas with fewer fires, as indicated in Figure 1a.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 8 of 14

with longer recovery times. We observe significant negative RF, both in the medium‐ and long‐term, in the
Kalahari region; indeed the negative continental RFLvalue is strongly driven by results from this region. Although
αbvalues are also high in the Sahel, the magnitude of the RF there is not comparable to the results seen in the
Kalahari region; this is due to both a smaller number of fires and relatively high τand Δαvalues.
4.4. Land Cover Type Dependency
We expect different ecosystems to react and recover from fire differently. Here, we break down post‐fire albedo
recovery by the five LCTs responsible for 97% of all fires in sub‐Saharan Africa. In Figure 8, the effect of fire on
albedo is shown for Savannas (a), Grasslands (b), Woody Savannas (c), Croplands, (d) and Forests (e) averaged
over the whole study region.
The data show strong LCT dependency for all fit parameters as seen in Table 1. The most significant fire effect on
albedo is observed in Grasslands, followed by Croplands, Savannas, Woody Savannas, and Forests. Apart from
Croplands, there is an apparent inverse relationship between percentage tree cover and all three fit parameters Δα,
τand αb. Land cover types with a higher tree cover (Woody Savannas, Forests) demonstrate a smaller immediate
fire effect on albedo, faster recovery and less brightening, compared with less woody LCTs (Grasslands, Sa-
vannas). This implies that the fires largely only affect the undergrowth with the trees themselves not significantly
affected. Tree survival of fires is a well‐documented phenomenon (e.g., Balfour and Midgley (2006);
Bond (2008); Gignoux et al. (1997)), with savanna fires affecting mostly grass and small trees only. As the
percentage of tree cover increases, the effect of fires on the surface as seen from above diminishes, as the sur-
viving and less fire‐affected trees effectively cover a larger proportion of the surface.
Albedo recovery following Cropland fires needs to be interpreted with caution. From Figure 8, it is clear that
although albedo recovery in other LCTs follows the exponential fit described by Equation 7almost exactly, the fit
is less suitable for Croplands, where recovery seems to be accelerated in the first few weeks, slowing down
thereafter. This may be explained by human interference. Although we expect the majority of fires in all LCTs to
be human‐induced (Andela et al., 2017), we assume that croplands will be managed more intensely, as agri-
cultural fires are commonly used for field clearing and preparation (Clerici et al., 2004) and the remains of a fire
are likely to be cleared soon after the burn, speeding up the natural recovery of the surface. It should also be noted
that Cropland burning is known to be detected unreliably by the MODIS BA product due to problems with
agricultural burning detections (Hall et al., 2016)
Table 2shows the RF due to post‐fire albedo changes for different LCTs. We observe that the highest RFS,Mis
observed in the savannas due to the high burn fraction of fb=0.30 in this LCT. This is consistent with the
findings of Dintwe et al. (2017). RFLis most significant in grasslands, where the values of αbare an order of
magnitude higher than in other LCTs. Grasslands are also the only LCT with RFM<0, highlighting the impact of
surface brightening on RF in both the medium‐ and long‐term. Grasslands are one of the most dominant LCTs in
the Sahel and Kalahari regions, where the highest αbvalues are observed.
Figure 7. Map of short‐ (a), medium‐ (b), and long‐term (c) radiative forcing values across Africa.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 9 of 14

4.5. Temporal Variations
The temporal extent of our analysis allows us to investigate possible trends in
albedo recovery and its RF over the study period. We average albedo
anomaly data over the study area for each year between 2001 and 2020 and
find the corresponding albedo recovery fit parameters from Equation 7shown
in Figure 9. The data is dominated by interannual variability, leading to low
confidence for most trends apart from Δαin SHA, which is shown to
decrease by 2.2×104/year (p=0.04)using an ordinary least squares
(OLS) fitting algorithm, indicating that the immediate darkening following
individual fires in the southern hemisphere is decreasing. There is no sig-
nificant change to this parameter in the northern hemisphere, and no sig-
nificant change to the recovery time constant or level of brightening observed
in either hemisphere.
In common with other studies, we found that fires are declining in all LCTs
apart from forests. However, we also find that the relative contribution of
fires over different LCTs to the overall total is changing. For example,
grasslands account for 34% of landscape fires on average, but this proportion
is declining by 0.13%per year (p=0.02), whereas the relative contribu-
tion of savanna fires increases by 0.16%per year p=2×105). These
trends are mostly driven by changes to the fire distribution in SHA. As more
woody LCTs are becoming responsible for a higher proportion of fires, we
expect the albedo impact of a typical fire to decrease in line with our findings
in Section 4.4.
We expect changes to the fire‐induced albedo effect to have an impact on the
RF at surface. Although RFSis directly proportional to Δα, the relationship
between RFM,Land αband τis more complex. In Figure 10, we show yearly
RFS,M,Lbetween 2005 and 2020 split by hemispheres. A notable reduction in
fb(see Figure 2) affects all RF values in both hemispheres. However, as the
RF value is also proportional to the albedo anomaly and DSSF levels, which
both exhibit high variability over the study period, the confidence in RF
trends is generally low despite the high‐confidence fbdecrease.
In NHA, the observed fbdecrease is partially offset by some increase in
individual‐fire albedo effects at all temporal scales; this results in only a small
RFSreduction (8 mWm2/year, p=0.31) and a small observed increase in
RFM(+3 mWm2/year, p=0.60). In NHA, the levels of brightening
observed are low over the whole study period, resulting in near‐zero values of
RFLas also shown in Figure 7c. Over the study period, we see an overarching
trend indicating a transition from positive to negative NHA albedo anomaly
values 5–10 months post‐fire (not shown) as also indicated by the decrease in
αb(see Figure 9c), resulting in the transition from negative to positive RFLin
NHA (+5 mWm2/year, p=0.15). However, these trends in NHA RF are
likely mostly driven by interannual variability and not indicative of a long‐
term shift.
In SHA, we see a significant decrease in RFS(36 mWm2/year, p=0.01),
which is both due to a decrease in fbas well as a decline in Δα.RFMtrends
follow changes to fire activity in SHA (Figure 2), with an increase before
2011 (49 mWm2/year, p=0.01) and a decrease thereafter (31
mWm2/year, p=0.01). The future evolution of RFMwill likely depend on
whether the decrease in fire activity continues. In the long‐term, we see an
insignificant increase in RFL(+2 mWm2/year, p=0.84), essentially
amounting to a reduction of the RFLcooling effect. Although we would
Figure 8. Average albedo anomaly before and after a fire in the top 5 land
cover types. Data in black fit (Equation 7) in orange. Day of burn uncertainty
(yellow shading) refers to the ±8 days uncertainty on albedo values. Albedo
uncertainty (gray shading) represents the spatial variation between different
Moderate Imaging Spectroradiometer tiles. See Table 1for the fit parameters.
Table 1
Albedo Recovery Fit Parameters From Equation 7, as Shown in Figure 8, for
Different Land Cover Types (LCTs)
Δα×102)τ(days) αb×104)Tree cover (%)
Grasslands 2.49 ±0.15 41.0±0.4 16.1±0.3<10
Savannas 1.48 ±0.08 28.4±0.3 3.3±0.5 10–30
Woody Savannas 1.03 ±0.04 18.2±0.1 3.9±0.5 30–60
Forests 0.86 ±0.03 19.7±0.3 1.8±0.1>60
Croplands 2.20 ±0.14 79.6±1.3 3.2±0.7 0
Note. The percentage tree cover associated with each LCT is also provided.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 10 of 14

expect a reduction in the effect due to a decrease in fb, this is partially negated by increase in surface brightening
by individual fires.
5. Conclusions
We have presented a comprehensive analysis of surface albedo changes following landscape fires in sub‐Saharan
Africa, assessing the associated surface shortwave RF at different temporal scales and for different LCTs. Over
the two decades studied, we find that instantaneously after a fire, surface albedo always reduces prior to a longer
timescale recovery (of order 3–5 months). For four out of the five LCTs primarily responsible for burning in the
region (grasslands, savannas, woody savannas, and forests), the recovery follows an exponential form and can be
successfully modeled using three parameters: the initial reduction in albedo, the e‐folding time of albedo recovery
and a brightening constant, introduced to capture cases where the recovery overshoots the original albedo prior to
the burn. In contrast, cropland albedo recovery does not follow the simple mathematical representation introduced
here. This may be related to increased uncertainty in the burnt area product in these locations (Hall et al., 2016) but
may also reflect more intense human intervention in the initial period after a fire.
Table 2
Short‐, Medium‐ and Long‐Term Radiative Forcing and Burn Fraction in Each of the 5 Land Cover Types Analyzed
RFSmWm2)RFMmWm2)RFLmWm2)fb
Grasslands 600 ±220 41 ±11 187 ±8 0.12
Savannas 900 ±300 77 ±11 118 ±6 0.30
Woody Savannas 310 ±90 10 ±346 ±1 0.14
Forests 120 ±30 7 ±117 ±1 0.06
Croplands 260 ±90 11 ±442 ±3 0.09
Figure 9. Fit parameters Δα(a), τ(b) and α
b(c) describing average albedo recovery over the study period split by
hemispheres. The data is dominated by interannual variability and only Δαin southern hemisphere Africa shows a
statistically significant trend, with a decline of 2.2×104per year (p=0.04). Trends are determined using an ordinary least
squares algorithm.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 11 of 14

Excepting the cropland land cover class, both the magnitude of the instantaneous albedo reduction and the albedo
recovery time depend on percentage tree cover, where a larger proportion of trees results in both a smaller albedo
drop and faster recovery. This is consistent with previous findings showing that it is the undergrowth rather than
trees themselves that tends to burn in managed dryland fires (Bond, 2008). This also has the knock‐on effect of
surviving tree cover partially obscuring the surface burnt area as seen from space. In addition, as noted above, in
some locations the albedo recovery can lead to an overshoot of the original albedo prior to the burn, effectively
resulting in surface brightening. Consistent with previous studies, this longer‐term brightening is predominantly
located in the Kalahari and on the southern margins of the Sahara Desert (Saha et al., 2019).
Our evaluation of the RF of fires due to these surface albedo changes indicates that the effect of a reduced surface
albedo immediately after a fire is amplified by seasonally higher values of DSSF in the burn season. We find that
the short‐term (instantaneous) RF averaged across sub‐Saharan African regions affected by fire is +5±2 Wm
2
consistent with previous findings by Dintwe et al. (2017). This reduces to +0.5±0.2 Wm
2
when averaged
across all of sub‐Saharan Africa. The sub‐Saharan Africa average RF in the 5 months following a fire reduces
further to +0.13 ±0.02 Wm
2
before changing sign due to longer‐term brightening, with the corresponding
average RF in months 5–10 after a fire being 0.02 ±0.02 Wm
2
.
Exploiting the length of the data records to assess longer term patterns of behavior, we replicate the known
decrease in number of African landscape fires over recent years most notably in the northern hemisphere (Jiang
et al., 2020; Zubkova et al., 2019) and in the southern hemisphere from 2011 onwards (Savage & Stry-
dom, 2016). In SHA, we observe a statistically significant reduction in the short‐term RF associated with fires,
driven both by the reduction in fband a decrease in albedo drop immediately following a burn. The latter can be
partially attributed to an observed change in proportion of fires in different LCTs with woody LCTs responsible
for an increasing fraction of fires. A similar trend is observed in the medium‐term RF in SHA post 2011, here
likely driven primarily by the fbreduction. Although the number of fires is decreasing, when compounded with
Figure 10. Short‐ (a), medium‐ (b), and long‐term (c) radiative forcing between 2005 and 2020 split by the hemispheres. Large
interannual variability is observed. The only statistically significant trend over the full time period is RF
Sin southern
hemisphere Africa (SHA), which shows a decline of 36 mWm2per year (p=0.01). Following fire trends in Figure 2, SHA
RFMis found to increase before 2011 (49 mWm2/year, p=0.01) and decrease after 2011 (31 mWm2/year, p=0.01).
Similar analysis does not produce more statistically significant trends for RFSor RFL. Trends are found using an ordinary least
squares algorithm.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 12 of 14

highly variable albedo changes and DSSF levels their associated continent‐wide albedo‐induced RF is not
showing a statistically significant decline in the long‐term in either hemisphere or in the short‐ and medium‐
term in NHA.
As albedo recovery after fires is generally not fully represented in the current generation of ESMs, we suggest
further work should investigate the effect of including these post‐fire effects and evaluating their further impact
on the surface radiation budget as well as additional climate variables, such as atmospheric circulation, rainfall,
and soil moisture. For example, it has already been previously shown that surface brightening following fires may
lead to rainfall suppression (Saha et al., 2017). Given that we've shown that surface RF levels due to fires are
changing, it is important we understand their associated impacts to evaluate how these might also change in the
future.
Data Availability Statement
All MODIS data used in this project are available via the Level‐1 and Atmosphere Archive & Distribution System
Distributed Active Archive Center: https://ladsweb.modaps.eosdis.nasa.gov (Friedl & Sulla‐Menashe, 2019;
Giglio et al., 2015; Schaaf & Wang, 2015). Downward Surface Shortwave Flux data were provided by the
EUMETSAT Satellite Application Facility on Land Surface Analysis (LSA SAF; Trigo et al., 2011)https://lsa‐
saf.eumetsat.int.
References
Amiri‐Farahani, A., Allen, R. J., Li, K., Nabat, P., & Westervelt, D. M. (2020). A La Niña‐Like climate response to South African biomass burning
aerosol in CESM simulations. Journal of Geophysical Research: Atmospheres,125(6), e2019JD031832. https://doi.org/10.1029/
2019JD031832
Andela, N., Morton, D. C., Giglio, L., Chen, Y., van der Werf, G. R., Kasibhatla, P. S., et al. (2017). A human‐driven decline in global burned area.
Science (American Association for the Advancement of Science),356(6345), 1356–1361. https://doi.org/10.1126/science.aal4108
Andela, N., Van Der Werf, G. R., Kaiser, J. W., Van Leeuwen, T. T., Wooster, M. J., & Lehmann, C. E. R. (2016). Biomass burning fuel
consumption dynamics in the tropics and subtropics assessed from satellite. Biogeosciences,13(12), 3717–3734. https://doi.org/10.5194/bg‐13‐
3717‐2016
Ångström, A. (1925). The albedo of various surfaces of ground. Geografiska Annaler,7(4), 323–342. https://doi.org/10.2307/519495
Archibald, S. (2016). Managing the human component of fire regimes: Lessons from Africa. Philosophical Transactions of the Royal Society B:
Biological Sciences,371(1696), 20150346. https://doi.org/10.1098/rstb.2015.0346
Archibald, S., Scholes, R., Roy, D., Roberts, G., & Boschetti, L. (2010). Southern African fire regimes as revealed by remote sensing. Inter-
national Journal of Wildland Fire,19(7), 861–878. https://doi.org/10.1071/WF10008
Balfour, D. A., & Midgley, J. J. (2006). Fire induced stem death in an African acacia is not caused by canopy scorching. Austral Ecology,31(7),
892–896. https://doi.org/10.1111/j.1442‐9993.2006.01656.x
Bond, W. J. (2008). What limits trees in C‐4 grasslands and savannas? Annual Review of Ecology Evolution and Systematics,39(1), 641–659.
https://doi.org/10.1146/annurev.ecolsys.39.110707.173411
Ceppi, P., & Shepherd, T. (2019). Contributions of climate feedbacks to changes in atmospheric circulation. Journal of Climate,30(22), 9097–
9118. https://doi.org/10.1175/JCLI‐D‐17‐0189.1
Clerici, N., Boschetti, L., Eva, H., & Gregoire, J.‐M. (2004). Assessing vegetation fires activity and its drivers in West‐Central Africa using
MODIS and TRMM data. IGARSS 2004. 2004 IEEE International Geoscience and Remote Sensing Symposium,3, 2087–2090. https://doi.org/
10.1109/IGARSS.2004.1370768
Crutzen, P., & Andreae, M. (1990). Biomass burning in the tropics: Impact on atmospheric chemistry and biogeochemical cycles. Science
(American Association for the Advancement of Science),250(4988), 1669–1678. https://doi.org/10.1126/science.250.4988.1669
Dewitte, O., Jones, A., Spaargaren, O., Breuning‐Madsen, H., Brossard, M., Dampha, A., et al. (2013). Harmonisation of the soil map of Africa at
the continental scale. Geoderma,211–212, 138–153. https://doi.org/10.1016/j.geoderma.2013.07.007
Dintwe, K., Okin, G. S., & Xue, Y. (2017). Fire‐induced albedo change and surface radiative forcing in sub‐saharan Africa savanna ecosystems:
Implications for the energy balance. Journal of Geophysical Research: Atmospheres,122(12), 6186–6201. https://doi.org/10.1002/
2016JD026318
Earl, N., & Simmonds, I. (2018). Spatial and temporal variability and trends in 2001–2016 global fire activity. Journal of Geophysical Research:
Atmospheres,123(5), 2524–2536. https://doi.org/10.1002/2017JD027749
Friedl, M., & Sulla‐Menashe, D. (2019). MCD12Q1 MODIS/Terra+Aqua Land Cover Type Yearly L3 Global 500m SIN Grid V006 [Dataset].
NASA EOSDIS Land Processes Distributed Active Archive Center.https://doi.org/10.5067/MODIS/MCD12Q1.006
Friedl, M. A., McIver, D., Hodges, J., Zhang, X. Y., Muchoney, D., Strahler, A. H., et al. (2002). Global land cover mapping from MODIS:
Algorithms and early results. Remote Sensing of Environment,83(1), 287–302. https://doi.org/10.1016/S0034‐4257(02)00078‐0
Friedl, M. A., Sulla‐Menashe, D., Tan, B., Schneider, A., Ramankutty, N., Sibley, A., & Huang, X. (2010). MODIS collection 5 global land cover:
Algorithm refinements and characterization of new datasets. Remote Sensing of Environment,114(1), 168–182. https://doi.org/10.1016/j.rse.
2009.08.016
Gatebe, C. K., Ichoku, C. M., Poudyal, R., Román, M. O., & Wilcox, E. (2014). Surface albedo darkening from wildfires in northern sub‐saharan
Africa. Environmental Research Letters,9(6), 065003. https://doi.org/10.1088/1748‐9326/9/6/065003
Geiger, B., Meurey, C., Lajas, D., Franchistéguy, L., Carrer, D., & Roujean, J.‐L. (2008). Near real‐time provision of downwelling shortwave
radiation estimates derived from satellite observations. Meteorological Applications,15(3), 411–420. https://doi.org/10.1002/met.84
Giglio, L., Boschetti, L., Roy, D. P., Humber, M. L., & Justice, C. O. (2018). The Collection 6 MODIS burned area mapping algorithm and
product. Remote Sensing of Environment,217, 72–85. https://doi.org/10.1016/j.rse.2018.08.005
Acknowledgments
M. Flegrová and H. Brindley were funded
as part of NERC's support of the National
Centre for Earth Observation under Grant
NE/R016518/1. M. Flegrová was also
funded by the Leverhulme Centre for
Wildfires, Environment, and Society
through the Leverhulme Trust under Grant
RC‐2018‐023.
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 13 of 14

Giglio, L., Csiszar, I., & Justice, C. O. (2006). Global distribution and seasonality of active fires as observed with the Terra and Aqua Moderate
Resolution Imaging Spectroradiometer (MODIS) sensors: Global fire distribution and seasonality. Journal of Geophysical Research,111(G2),
G02016. https://doi.org/10.1029/2005JG000142
Giglio, L., Descloitres, J., Justice, C. O., & Kaufman, Y. J. (2003). An enhanced contextual fire detection algorithm for MODIS. Remote Sensing of
Environment,87(2), 273–282. https://doi.org/10.1016/S0034‐4257(03)00184‐6
Giglio, L., Justice, C., Boschetti, L., & Roy, D. (2015). MCD64A1 MODIS/Terra+Aqua Burned Area Monthly L3 Global 500m SIN Grid V006
[Dataset]. NASA EOSDIS Land Processes Distributed Active Archive Center.https://doi.org/10.5067/MODIS/MCD64A1.006
Giglio, L., Randerson, J. T., & van der Werf, G. R. (2013). Analysis of daily, monthly, and annual burned area using the fourth‐generation global
fire emissions database (GFED4). Journal of Geophysical Research: Biogeosciences,118(1), 317–328. https://doi.org/10.1002/jgrg.20042
Gignoux, J., Clobert, J., & Menaut, J. (1997). Alternative fire resistance strategies in savanna trees. Oecologia,110(4), 576–583. https://doi.org/
10.1007/s004420050198
Hall, J. V., Loboda, T. V., Giglio, L., & McCarty, G. W. (2016). A MODIS‐based burned area assessment for Russian croplands: Mapping
requirements and challenges. Remote Sensing of Environment,184, 506–521. https://doi.org/10.1016/j.rse.2016.07.022
Hantson, S., Kelley, D. I., Arneth, A., Harrison, S. P., Archibald, S., Bachelet, D., et al. (2020). Quantitative assessment of fire and vegetation
properties in simulations with fire‐enabled vegetation models from the fire model intercomparison project. Geoscientific Model Development,
13(7), 3299–3318. https://doi.org/10.5194/gmd‐13‐3299‐2020
Jiang, Y., Zhou, L., & Raghavendra, A. (2020). Observed changes in fire patterns and possible drivers over central Africa. Environmental
Research Letters,15(9), 0940b8. https://doi.org/10.1088/1748‐9326/ab9db2
Lian, X., Zhao, W., & Gentine, P. (2022). Recent global decline in rainfall interception loss due to altered rainfall regimes. Nature Communi-
cations,13(1), 7642. https://doi.org/10.1038/s41467‐022‐35414‐y
Loveland, T. R., & Belward, A. S. (1997). The IGBP‐DIS global 1 km land cover data set, DISCover: First results. International Journal of
Remote Sensing,18(15), 3289–3295. https://doi.org/10.1080/014311697217099
Mahowald, N., Ward, D. S., Kloster, S., Flanner, M. G., Heald, C. L., Heavens, N. G., et al. (2011). Aerosol impacts on climate and biogeo-
chemistry. Annual Review of Environment and Resources,36(1), 45–74. https://doi.org/10.1146/annurev‐environ‐042009‐094507
Palle, E., Goode, P. R., Montañés‐Rodríguez, P., Shumko, A., Gonzalez‐Merino, B., Lombilla, C. M., et al. (2016). Earth’s albedo variations
1998–2014 as measured from ground‐based earthshine observations. Geophysical Research Letters,43(9), 4531–4538. https://doi.org/10.1002/
2016GL068025
Pausas, J. G., & Keeley, J. E. (2019). Wildfires as an ecosystem service. Frontiers in Ecology and the Environment,17(5), 289–295. https://doi.
org/10.1002/fee.2044
Rabin, S., Melton, J., Lasslop, G., Bachelet, D., Forrest, M., Hantson, S., et al. (2017). The fire modeling intercomparison project (FireMIP), phase
1: Experimental and analytical protocols with detailed model descriptions. Geoscientific Model Development,10(3), 1175–1197. https://doi.
org/10.5194/gmd‐10‐1175‐2017
Roberts, G., Wooster, M. J., & Lagoudakis, E. (2009). Annual and diurnal African biomass burning temporal dynamics. Biogeosciences,6(5),
849–866. https://doi.org/10.5194/bg‐6‐849‐2009
Saha, M. V., D’Odorico, P., & Scanlon, T. M. (2017). Albedo changes after fire as an explanation of fire‐induced rainfall suppression.
Geophysical Research Letters,44(8), 3916–3923. https://doi.org/10.1002/2017GL073623
Saha, M. V., D’Odorico, P., & Scanlon, T. M. (2019). Kalahari wildfires drive continental post‐fire brightening in sub‐saharan Africa. Remote
Sensing,11(9), 1090. https://doi.org/10.3390/rs11091090
Savage, M. J., & Strydom, S. (2016). A spatio‐temporal analysis of fires in South Africa. South African Journal of Science,112(11), 83–90. https://
doi.org/10.17159/sajs.2016/20150489
Schaaf, C., & Wang, Z. (2015). MCD43A3 MODIS/Terra+Aqua BRDF/Albedo Daily L3 Global ‐ 500m V006 [Dataset]. NASA EOSDIS Land
Processes Distributed Active Archive Center.https://doi.org/10.5067/MODIS/MCD43A3.006
Schaaf, C. B., Gao, F., Strahler, A. H., Lucht, W., Li, X., Tsang, T., et al. (2002). First operational BRDF, albedo nadir reflectance products from
MODIS. Remote Sensing of Environment,83(1), 135–148. https://doi.org/10.1016/S0034‐4257(02)00091‐3
Smith, A. M., Wooster, M. J., Drake, N. A., Dipotso, F. M., Falkowski, M. J., & Hudak, A. T. (2005). Testing the potential of multi‐spectral remote
sensing for retrospectively estimating fire severity in African savannahs. Remote Sensing of Environment,97(1), 92–115. https://doi.org/10.
1016/j.rse.2005.04.014
Stronach, N. R. H., & McNaughton, S. J. (1989). Grassland fire dynamics in the Serengeti ecosystem, and a potential method of retrospectively
estimating fire energy. Journal of Applied Ecology,26(3), 1025–1033. https://doi.org/10.2307/2403709
Teixeira, J., Folberth, G., O’Connor, F., Unger, N., & Voulgarakis, A. (2021). Coupling interactive fire with atmospheric composition and climate
in the UK Earth System Model. Geoscientific Model Development,14(10), 6515–6539. https://doi.org/10.5194/gmd‐14‐6515‐2021
Trigo, I. F., Dacamara, C. C., Viterbo, P., Roujean, J.‐L., Olesen, F., Barroso, C., et al. (2011). The satellite application facility for land surface
analysis. International Journal of Remote Sensing,32(10), 2725–2744. https://doi.org/10.1080/01431161003743199
Virtanen, P., Gommers, R., Oliphant, T. E., Haberland, M., Reddy, T., Cournapeau, D., et al., SciPy 1.0 Contributors. (2020). SciPy 1.0:
Fundamental algorithms for scientific computing in Python. Nature Methods,17(3), 261–272. https://doi.org/10.1038/s41592‐019‐0686‐2
Voulgarakis, A., & Field, R. D. (2015). Fire influences on atmospheric composition, air quality and climate. Current pollution reports,1(2), 70–
81. https://doi.org/10.1007/s40726‐015‐0007‐z
Wylie, D., Jackson N, D. L., Menzel, W. P., & Bates, J. J. (2005). Trends in global cloud cover in two decades of HIRS observations. Journal of
Climate,18(15), 3021–3031. https://doi.org/10.1175/JCLI3461.1
Zubkova, M., Boschetti, L., Abatzoglou, J. T., & Giglio, L. (2019). Changes in fire activity in Africa from 2002 to 2016 and their potential drivers.
Geophysical Research Letters,46(13), 7643–7653. https://doi.org/10.1029/2019GL083469
Journal of Geophysical Research: Atmospheres
10.1029/2024JD041491
FLEGROVÁ AND BRINDLEY 14 of 14