study area is between mountain and lowland forests (Letouzey and Aubréville 1968;
White 1983). Within the lowland forest category, swamp forests and mangroves are
distinguished from terra firme forests (Letouzey and Aubréville 1968), and can be
remotely sensed (Verhegghen et al. 2012). Even though such forests occupy vast areas
in the Congo basin, and hold huge amount of carbon (Dargie et al. 2017), we focused
here only on terra firme vegetation, excluding mangrove and swamps from our
analysis.
We further studied how tree cover and tree cover changes are distributed across
forests and savannas (grouping classes 5–7 outlined above). This is because tree
cover is a key variable for both forest and savanna functioning, enabling the
differentiation of forests and savannas at a broader scale (Staver et al. 2011a; Aleman
et al. 2016; Aleman and Staver 2018). Tree cover within savannas has been shown to
be highly variable (Sankaran et al. 2005) and largely determined by annual rainfall
(Bucini and Hanan 2007). However, even though annual rainfall indeed constraints
maximum tree cover, disturbances (especially fire and herbivory), complex interactions between disturbances (Hanan et al. 2008; Staver and Bond 2014) and soil
composition (Sankaran et al. 2008) can also reduce tree cover from its climatic
potential in a less predictable way (Sankaran et al. 2005, 2008; Staver 2018).
We used remotely sensed information on tree cover, initially available at 30-m
resolution, and aggregated at the resolution of the vegetation types (900 m) (Hansen
et al. 2013). Tree cover losses between 2000 and 2015 are also available from the
same dataset (Hansen et al. 2013) and were used to identify the forest and savanna
areas that have been heavily modified during that period.
9.2.2.3 Identification of Natural and Anthropogenic Drivers
of Vegetation Change
A wide array of drivers is currently influencing vegetation structure (Aleman et al.
2017) and composition in forests (Fayolle et al. 2012, 2014b) and savannas (Fayolle
et al. 2019). We focused here on annual rainfall and fire as drivers of vegetation
change (Sect. 9.3.3), as they are important in determining forest and savanna as
alternative stable states in the climate zone where both forests and savannas co-occur
(Staver et al. 2011a).
We used for annual rainfall the 3B43 Monthly gridded rainfall product
from TRMM, available at 0.25
resolution (Nicholson et al. 2003). For fires, we
used the burned area monthly product from the MODerate-resolution Imaging
Spectroradiometer (MODIS) sensor available at 500-m resolution (Giglio et al.
2010). Both datasets are resampled at the resolution of the vegetation types
(900 m) (Sect. 9.2.2.2), using a nearest neighbor procedure and the package “raster”
in the R environment (R Core Team 2015). We computed annual rainfall between
2002 and 2015, and the number of times an area represented by a pixel was burned
between 2002 and 2015 as an index of fire frequency.
As mentioned in Sect. 9.2.1, ecosystems in central Africa are facing an increasingly high pressure from human activity, but are also threatened by ongoing climate
290
J. C. Aleman and A. Fayolle
White 1983). Within the lowland forest category, swamp forests and mangroves are
distinguished from terra firme forests (Letouzey and Aubréville 1968), and can be
remotely sensed (Verhegghen et al. 2012). Even though such forests occupy vast areas
in the Congo basin, and hold huge amount of carbon (Dargie et al. 2017), we focused
here only on terra firme vegetation, excluding mangrove and swamps from our
analysis.
We further studied how tree cover and tree cover changes are distributed across
forests and savannas (grouping classes 5–7 outlined above). This is because tree
cover is a key variable for both forest and savanna functioning, enabling the
differentiation of forests and savannas at a broader scale (Staver et al. 2011a; Aleman
et al. 2016; Aleman and Staver 2018). Tree cover within savannas has been shown to
be highly variable (Sankaran et al. 2005) and largely determined by annual rainfall
(Bucini and Hanan 2007). However, even though annual rainfall indeed constraints
maximum tree cover, disturbances (especially fire and herbivory), complex interactions between disturbances (Hanan et al. 2008; Staver and Bond 2014) and soil
composition (Sankaran et al. 2008) can also reduce tree cover from its climatic
potential in a less predictable way (Sankaran et al. 2005, 2008; Staver 2018).
We used remotely sensed information on tree cover, initially available at 30-m
resolution, and aggregated at the resolution of the vegetation types (900 m) (Hansen
et al. 2013). Tree cover losses between 2000 and 2015 are also available from the
same dataset (Hansen et al. 2013) and were used to identify the forest and savanna
areas that have been heavily modified during that period.
9.2.2.3 Identification of Natural and Anthropogenic Drivers
of Vegetation Change
A wide array of drivers is currently influencing vegetation structure (Aleman et al.
2017) and composition in forests (Fayolle et al. 2012, 2014b) and savannas (Fayolle
et al. 2019). We focused here on annual rainfall and fire as drivers of vegetation
change (Sect. 9.3.3), as they are important in determining forest and savanna as
alternative stable states in the climate zone where both forests and savannas co-occur
(Staver et al. 2011a).
We used for annual rainfall the 3B43 Monthly gridded rainfall product
from TRMM, available at 0.25
resolution (Nicholson et al. 2003). For fires, we
used the burned area monthly product from the MODerate-resolution Imaging
Spectroradiometer (MODIS) sensor available at 500-m resolution (Giglio et al.
2010). Both datasets are resampled at the resolution of the vegetation types
(900 m) (Sect. 9.2.2.2), using a nearest neighbor procedure and the package “raster”
in the R environment (R Core Team 2015). We computed annual rainfall between
2002 and 2015, and the number of times an area represented by a pixel was burned
between 2002 and 2015 as an index of fire frequency.
As mentioned in Sect. 9.2.1, ecosystems in central Africa are facing an increasingly high pressure from human activity, but are also threatened by ongoing climate
290
J. C. Aleman and A. Fayolle