1. Introduction
The Congo basin rain forest is the second largest tropical forest in the world. This highly diverse ecosystem provides resources to more than 30 million people (Riddell 2013; Díaz et al. 2019). It also contains the Cuvette Centrale peatland that stores a significant amount of the world’s forest carbon (Dargie et al. 2017), playing an important role in the global carbon budget (Harris et al. 2021).
Several papers indicate that the Congo rain forest is becoming increasingly fragmented (Hansen et al. 2014; Shapiro et al. 2021). Human deforestation is certainly playing a role in the decline (Tyukavina et al. 2018; Molinario et al. 2020), and there may also be contributions from climate change due to increasing greenhouse gases (Asefi-Najafabady and Saatchi 2013; Zhou et al. 2014). However, caution is recommended in examining precipitation trends over the Congo basin. Relatively low numbers of ground-based observations limit opportunities to calibrate and evaluate satellite-based observations (Washington et al. 2013; Nicholson et al. 2019). In addition, Maidment et al. (2015) examined precipitation trends over central Africa in eight datasets from 1983 to 2010 and found that they disagree. Further, the number of reporting stations in the Congo basin has been decreasing in recent decades, and this can give rise to false negative trends in blended satellite–rain gauge datasets (Maidment et al. 2015; Dinku et al. 2018). While Cook et al. (2020) identified a potential mechanism for Congo basin drying in association with increasing greenhouse gases, more complete attribution studies are hampered by the limitations of the available precipitation datasets.
Here a holistic approach is used to identify a potential climate change component for the observed decline in the Congolese rain forest; climate stresses to ecosystems are associated not only with precipitation amounts but also with their seasonality and other factors such as temperature. We drive a potential vegetation model (PVM) of the rain forest (Oyama and Nobre 2004) with the ERA5 (Hersbach et al. 2020) reanalysis to translate decadal-mean climates (1980–2019) to natural vegetation distributions in equilibrium with the climate state. Then we identify what features of the climate state are related to any changes in the tropical rain forest produced by the PVM. Confidence in the results is strengthened by cross-referencing our findings with independent observational datasets.
The background is provided in section 2. Section 3 describes the methodology including a description of the potential vegetation model (PVM) and datasets utilized, while the results are presented in section 4. Finally, in section 5, the results are summarized and conclusions are stated.
2. Background
The Congo rain forest spans nearly 3 million km2 over equatorial Africa (FAO 2011). Given its size, the forest is highly diverse, ranging from evergreen trees in the west and swamp forests in the central areas to semi-evergreen trees in the east (Shapiro et al. 2021). It is relatively drier than other tropical rain forests such as the Amazon, so it lies closer to a climatic threshold for composition type (Guan et al. 2015; Philippon et al. 2019). This indicates that the Congo rain forest is not only sensitive to changes in the climate, but it may also lack resilience and could eventually reach a tipping point, especially in its interior (Staal et al. 2020).
Tree density also varies over the Congo basin. Interior forest areas are most dense, while the canopy cover thins with more frequent and larger gaps closer to the forest peripheries and adjacent to major rivers (Shapiro et al. 2021). It is believed that rain forests are most stable when the tree cover is most dense as trees in these areas are better suited to adapt to climate variations by accessing soil moisture while the dense canopies slow atmospheric moisture loss (Singh et al. 2021; Shapiro et al. 2021). In contrast, areas with sparser tree density have more difficulty adapting and hence show a greater propensity for composition changes. This suggests that areas more prone to change would likely lie along the forest margins or in areas where the trees have been thinned and the forest fragmented (Haddad et al. 2015; Betts et al. 2019).
Given the global importance of the Congo rain forest, there has been a concerted effort to quantify forest changes. Earlier studies (Hansen et al. 2014; Potapov et al. 2017) have observed declines in forest size and tree cover over the past 20 years, as estimated forest loss rates are among the highest in the world. More recently, Shapiro et al. (2021) estimates that currently less than 70% of the Congo rain forest remains fully intact, which is a decrease from 78% in 2000. This rate of forest change is not spatially uniform as Global Forest Watch (2022) indicates that between 2000 and 2021 the Democratic Republic of Congo (DRC) experienced a 5.6% decrease in total primary forest area. Cameroon and Equatorial Guinea have the next highest loss rates at 4.2% and 2.9%.
Potential anthropogenic factors influencing rain forest health include population growth, migration, and economic activity including logging and mining (Wilkie et al. 1999; Sonwa et al. 2012, 2020; Kranz et al. 2018). It is estimated that 84% of the forest disturbance between 2000 and 2014 was due to an increase in small-scale deforestation for agriculture (Turubanova et al. 2018; Tyukavina et al. 2018). Impacts include more fragmented forests, especially near transportation routes (Kleinschroth et al. 2019; Shapiro et al. 2021), and changes in the local water cycle (Baker and Spracklen 2022) as soil moisture depletes faster in deforested areas with remaining trees struggling to adapt (Singh et al. 2021).
Most of the moisture necessary for Congo basin rainfall comes from local evapotranspiration or advection from the adjacent oceans (Dyer et al. 2017; Burnett et al. 2020; Tuinenburg et al. 2020; Worden et al. 2021). The former is thought to be the primary source during the dry seasons. Recycling contributions are estimated at 80% during boreal winter, 60% in summer, and 30%–45% during the spring and fall wet seasons (Nicholson et al. 1997; Pokam et al. 2012; Risi et al. 2013; Sorí et al. 2017; Worden et al. 2021). Moisture advection is greatest during the wet seasons, with the Indian Ocean contributing around 21% of the moisture (Dyer et al. 2017) and lesser amounts from the Atlantic (Pokam et al. 2012, 2014; Dezfuli and Nicholson 2013; Dyer et al. 2017). This suggests that a relationship between SST anomalies and rainfall variability over the basin may exist (Hua et al. 2016, 2018; Wang et al. 2021; Nicholson et al. 2022).
Warm season rainfall in each hemisphere is supported when circulation about the continental thermal lows, namely the boreal summer Saharan thermal low (Parker et al. 2005; Cook and Vizy 2015; Vizy and Cook 2017) and the austral summer Angola thermal low (Zunckel et al. 1996; Vizy and Cook 2016; Munday and Washington 2017; Howard and Washington 2018), converges with cross-equatorial flow from the winter hemisphere. Cook et al. (2020) identify a poleward shift in the thermal lows in each hemisphere in association with greenhouse gas forcing and suggest that such a shift can be associated with Congo basin drying. Thus, heterogeneous warming of the land surface (i.e., some areas warm faster than other areas) is important regardless of season, and its role in influencing Congo basin climate variability/change needs to be understood.
There is a lack of agreement in future climate projections over the Congo basin (Christensen et al. 2013; Haensler et al. 2013; Creese et al. 2019). Developing our understanding of the relationship between the climate and the rain forest ecosystem is crucial for improving our understanding of this regional climate and advancing our ability to predict how it may change.
3. Methodology
a. Datasets
The datasets utilized are described below. They include the ECMWF ERA5 reanalysis (Hersbach et al. 2020). This global atmospheric reanalysis provides 0.25° resolution monthly mean output of atmospheric and land surface fields from 1979 to the present. It is used to force the PVM (see section 3b) as well as to physically understand how climate variations impact PVM simulated vegetation changes. A second reanalysis, the NASA Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA-2; Gelaro et al. 2017), is also inspected to confirm that the physical changes in the climate identified in ERA5 and understand the uncertainty.
To optimize and validate the PVM, three land-use datasets are evaluated. They are as follows:
-
ERA5 vegetation types (Hersbach et al. 2020): ERA5 provides a 0.25° resolution low and high static vegetation type based on Global Land Cover Characteristics V1.2 (GLCC1.2) data derived from Advanced Very High Resolution Radiometer data from 1992 to 1993 (Loveland et al. 2000). These values represent the dominant type for each grid cell and remain static over time. The 20 vegetation categories are reassigned to match the 12 PVM vegetation types.
-
Moderate Resolution Imaging Spectroradiometer (MODIS) vegetation types (Broxton et al. 2014): This uses Collection 5.1 MODIS output to determine dominant vegetation type via a supervised decision-tree algorithm (Quinlan 1993; Broxton et al. 2014) beginning in 2001. The data are reassigned to the PVM’s 12 vegetation types for 2001 for our comparison.
-
European Space Agency Climate Change Initiative (ESA-CCI) land cover dataset (ESA 2017): This 300-m resolution 22 vegetation class dataset provides annual land cover maps from 1992 to 2015 using version 2.0.7 and Version 2.1.1 after 2016. The data are reassigned to the 12 PVM vegetation classes.
Multiple precipitation datasets are also evaluated. They include the following:
-
Climate Hazards group Infrared Precipitation with Stations version 2 (CHIRPS2; Funk et al. 2015): 0.1° resolution monthly rainfall values from 1981 to 2019.
-
NOAA Climate Prediction Center African Rainfall Climatology version 2 (ARC2; Novella and Thiaw 2013): 0.1° resolution monthly rainfall over Africa from 1983 to 2019.
-
Tropical Applications of Meteorology using Satellite Data dataset (TAMSAT; Maidment et al. 2017): Daily 0.0375° resolution rainfall estimates are used to generate monthly averages for 1983–2019.
-
Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks–Climate Data Record (PERSIANN-CDR; Sorooshian et al. 2000): 0.25° resolution monthly rainfall from 1983 to 2019.
b. The potential vegetation model
To translate climate states into natural vegetation for the Congo basin, we adapt Oyama and Nobre’s (2004) Centro de Previsao de Tempo e Estudios Climaticos PVM (CPTEC PVM). This model is similar to the biome model of Prentice et al. (1992), except it does not account for plant ecological competition. The model is not as complex as other PVMs, making it easier to track climate-forced vegetation changes back to the climate factor(s). Past studies (Oyama and Nobre 2004; Cardoso et al. 2008; Cook and Vizy 2008) show that this PVM realistically produces the vegetation cover over the tropics, meaning that it is suitable for use over the Congo basin.
The five environmental variables are calculated from ERA5 data. ERA5 provides high-resolution gridded estimates of atmospheric and land surface fields by blending available satellite and in situ observations with model output in its assimilation system. Studies (e.g., Gleixner et al. 2020; Wright et al. 2020; Steinkopf and Engelbrecht 2022) indicate that improvements to ERA5 have reduced known tropical atmospheric biases. For this reason, and because ERA5 has a relatively high spatial resolution (0.25°), at least by atmospheric reanalysis standards, we select to use it. Here climatological (1980–2019) and decadal (1980–89, 1990–99, 2000–09, and 2010–19) monthly mean values are utilized. PVM output from the climatologically forced simulation is used to optimize the PVM algorithm, while the decadallyforced simulations are used to assess the changes in the rain forest over the past 40 years.
Various ERA5 fields are used to estimate the five PVM environmental parameters. Skin temperature is used to calculate G0, G5, and Tc, while wi in Eq. (2) is estimated from the volumetric soil water in the soil layers and is dependent on the assigned soil and vegetation types for each grid point. Note that E and Emax in Eq. (1) are estimated by using the surface heat fluxes, 2-m temperature, skin temperature, and surface pressure via the Penman–Monteith equation [see, e.g., Oyama and Nobre’s (2004) Eq. (6)]. The temperature fields are generally more reliable as they are dependent on direct observations, while the surface heat fluxes and volumetric soil water are model derived with the ERA5 algorithm framework and are more loosely constrained by direct observations.
Figure 1a shows the PVM algorithm used to determine the vegetation type. The algorithm classifies the land use based on 12 different vegetation types. In the case of the tropical evergreen broadleaf rain forest, if the near-surface climate is not wet enough and H is below the threshold, then there will not be enough water in the environment to support the rain forest.
(a) The PVM algorithm utilized. Environmental variables include mean temperature of the coldest month (Tc), growing degree days (G0 and G5), wetness index (H), and a seasonality index (D). Red denotes algorithm adjustment made to the PVM. (b) D and (c) H index values (unitless) determined using the ERA5 1980–2019 climatological forcing. White line denotes the PVM algorithm threshold value for each index as stated in (a), while blue line shows the Congo rain forest margin as indicated by the ERA5 vegetation cover.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
c. PVM algorithm adjustments
Since the CPTEC PVM was originally optimized for water balance model output, adjustments are needed to produce a realistic Congo basin vegetation cover when forced with ERA5 since each dataset will have its own inherent biases. Based on Fig. 1a, H, D, and Tc determine whether there is evergreen broadleaf forest, so these are the thresholds that are most crucial. Our approach is to use the 1980–2019 ERA5 climatological values to determine climatological estimates of H, D, and Tc, and then compare these estimates to the PVM thresholds to determine whether adjustments are warranted.
First, we consider Tc. Oyama and Nobre (2004) use a 284.15-K threshold, although Cook and Vizy (2008) find that using Prentice et al.’s (1992) 288.65-K threshold improves the PVM accuracy over the Amazon. It turns out that Tc does not fall below either of these thresholds for the Congo basin (not shown), indicating that Tc is not an important determinant for this rain forest. Here the 288.65-K threshold value is used.
Figure 1b shows D estimated from the ERA5 climatology. Over much of the Congo basin D is greater than 0.95, indicating minimal seasonality, while there is little spatial correspondence between D and the actual areal rain forest extent (blue line). Based on this evaluation the D threshold is left at 0.81.
Figure 1c shows the estimated H. There is a strong spatial correspondence between H and the rain forest areal extent. If the original threshold of 0.80 is used (white line), the forest is too large. To determine an appropriate value, we iteratively run the PVM using different H threshold values and difference the forest extent produced by the PVM with that from the static ERA5 vegetation type (blue line) over the Congo basin area (8°S–8°N, 8°–29°E) to find the optimal value for H that minimizes this difference. Doing so, we obtain an H threshold of 0.88 that is applied here.
d. Analysis and confidence
After first validating the PVM vegetation cover results associated with climatological forcing, our analysis approach in section 4 will involve using decadally forced PVM output to identify areas where changes in the climate result in Congo rain forest changes. Then for these identified areas, we examine the individual environmental parameters that comprise the PVM to identify which are responsible for the vegetation cover change, and pinpoint month(s) the climate forcing changes are most relevant. These month(s) are finally analyzed to physically understand the change in the climate.
The approach above is designed to build confidence in the results. Rather than simply report on how a single PVM forced by one reanalysis predicts vegetation cover changes, we focus on developing a physical understanding of the changes in the climate that are responsible for the PVM predicted rain forest changes. Then aspects of the physical processes identified are evaluated in independent data sources to assess the consistency of the response identified from ERA5.
4. Results
a. ERA5 climatological forcing
Figure 2a shows the primary rivers in the Congo basin while Figs. 2b–d show the observed vegetation from ERA5, MODIS, and ESA-CCI. Dark green indicates the equatorial African rain forest which is primarily surrounded by savanna. ESA-CCI, with its finer spatial resolution, is more detailed with a large area of wetland rain forest along the DRC/Republic of Congo border. All three datasets also indicate varying amounts of croplands.
(a) Primary rivers in equatorial Africa, and estimated vegetation cover type from (b) ERA5 (1992–93) at ∼27.8-km resolution, (c) NASA MODIS (2001) at 27-km resolution, (d) ESA-CCI (1992) at 0.30-km resolution, and (e) original PVM and the (f) updated PVM forced by ERA5 1980–2019 climatological data at ∼27.8-km resolution.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Figure 2e shows the PVM results forced by the ERA5 climatological forcing without the H adjustment. This confirms that using 0.80 for the H threshold results in an overestimate of the rain forest. Table 1 quantifies the forest areal extent, with the PVM forest 27%–36% larger than the observations. Since ESA-CCI is more detailed, rain forest estimates from this dataset also need to include the wetland rain forest and cropland area to be consistent with the other observations. Both estimates are provided in Table 1.
Tropical boreal rain forest areal extent over equatorial Africa (8°S–8°N, 8°–29°E).
Figure 2f shows the PVM results with the H threshold adjusted to 0.88. Forest margins are now in closer agreement with ERA5 and MODIS as the estimated forest size is now within 1% of these datasets (Table 1). The difference is larger (∼7%) compared with ESA-CCI but is still an improvement.
Overall, the adjusted PVM realistically represent the Congo rain forest extent when forced with climatological ERA5 conditions. This gives us confidence that the PVM will be useful for helping us understand how decadally forced changes in the climate affect the Congolese rain forest.
b. ERA5 decadal forcing
Figures 3a–d show the rain forest when the PVM is forced with ERA5 1980–89, 1990–99, 2000–09, and 2010–19 decadal averages, respectively. While subtle, there are decadal differences found primarily along the forest peripheries. To highlight these changes Figs. 3e–g show decadal differences in the rain forest extent between the 1980s and 1990s, between the 1990s and 2000s, and between the 2000s and 2010s. From the 1980s to the 1990s (Fig. 3e) there is a concentrated area in south-central Cameroon, and narrow, broken zones along the northern and southern margins along 4°N and near 7°S where the rain forest disappears by the 1990s. Additionally, there is a coherent area over coastal DRC and northwestern Angola where the rain forest re-emerges. From the 1990s to the 2000s (Fig. 3f) forest declines along the northern and southern margins become more coherent, while changes over Cameroon and DRC/Angola remain robust. Between the 2000s and 2010s (Fig. 3g) the decline in the southern margin rain forest shifts equatorward, while the northern margin decline along 4°N is primarily east of 20°E. The decline over Cameroon is now more localized, intermixed with areas where rain forest has reemerged. Rainforest also continues to reappear over coastal DRC/northwestern Angola by the 2010s. In the interior a coherent decline in the forest between 20° and 25°E appears.
PVM estimated vegetation cover for the decade of the (a) 1980s (1980–89), (b) 1990s (1990–99), (c) 2000s (2000–09), and (d) 2010s (2010–19). Additionally decadal changes in rain forest size (e) between the 1980s and 1990s, (f) between the 1990s and 2000s, (g) between the 2000s and 2010s, and (h) between the 1980s and 2010s with red representing the rain forest disappears in the latter decade and blue representing the rain forest reappearing in the latter decade. Boxes denote locations of averaging regions used in Table 2.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Changes in the rain forest extent between the 1980s and 2010s are shown in Fig. 3h. Seven coherent regions are identifiable, indicated by the boxes in Fig. 3h, and are listed in Table 2 along with decadal areal extents of the rain forest for each region as well as regional area decadal percent changes. Along the northern margin, the largest change over R1 occurs between the 1990s and 2000s, with a 61.1% decrease in the rain forest size between the 1980s and 2010s. The decrease over R2 has been getting larger each decade, resulting in a 57.7% decrease. Southern margin regions R3 and R4 also indicate that forest loss has been increasing each decade with larger change in R4 (−66.4%) compared to R3 (−32%). There is an increase in the rain forest size (+113.8%) in R5 with the largest change occurring between the 1980s and 1990s. R6 (−27.4%) and R7 (−18.2%) both indicate a decline in the interior rain forest size with most of the change occurring in the 2010s.
PVM decadal analysis of evergreen broadleaf vegetation type areal extent (km2) for various regions.
Next, we address how changes in the climate over the past four decades are associated with these rain forest changes produced by the PVM. The following discussion is organized regionally into the northern rain forest margin (R1–R2), southern rain forest margin (R3–R5), and interior rain forest (R6–R7).
1) Changes in the northern Congo rainforest margin
Figures 4a–d show the areas in each decade where H exceeds the 0.88 threshold, and Figs. 4e–h show areas where D exceeds the 0.81 threshold. A similar plot for Tc is not shown because the Tc threshold is never exceeded in the Congo basin for all four decades. For R1 and R2, much of the change over the four decades is associated with changes in H. The exception is around 5°N and 22°E where D falls below 0.81 by the 2010s.
PVM threshold for H ≥ 0.88 for (a) D1 (1980–89), (b) D2 (1990–99), (c) D3 (2000–09), and (d) D4 (2010–19). (e)–(h) As in (a)–(d), but for D ≥ 0.81. Boxes show locations of regions identified in Fig. 3.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Table 3 quantifies these decadal H and D values for the regions. For R1, H falls below the threshold in the 1990s while there is little change in D across all four decades. For R2, H does not fall below the threshold until the 2000s. Note that D remains close to 1 throughout the four decades but decreases to 0.973 by the 2010s, reflecting the localized area mentioned earlier.
Regional decadal analysis of the PVM’s H and D parameters. Bold values denote when H and D fall below their critical threshold values of 0.88 and 0.81, respectively, for the entire area average. The asterisks (*) denote when H and D fall below their critical threshold values for the 2010s − 1980s changepoints only shaded in Fig. 3h for the given region.
Figure 5a shows the area-averaged monthly E and Emax for R1 for each decade to better understand this change in H. The annual profiles indicate two peaks in E coinciding with the spring and fall wet seasons. It is at these times E is closest to Emax; Emax peaks during the late winter when conditions are warmer then remains relatively flat during the summer and fall.
(a) Area-averaged monthly actual evapotranspiration (E; solid lines; W m−2) and maximum evapotranspiration (Emax; dashed line; W m−2) for 1980s (black), 1990s (red), 2000s (light blue), and 2010s (dark blue) decadal means area averaged over R1: North Cameroon. (b) The corresponding area-averaged monthly 2010s − 1980s difference in E (solid line; W m−2) and Emax (dashed line; W m−2) for R1. (c),(d) As in (a) and (b), but for R2: the North Interior region.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Figure 5a also indicates that E and Emax have changed more in some months over the four decades. To understand these changes Fig. 5b shows the 2010s − 1980s monthly decadal differences for E (solid line) and Emax (dashed line). Based on Eq. (1), the largest change in H will occur when the monthly E differences and Emax differences have the opposite sign. So when the E difference is negative and the Emax difference is positive for a given month, or vice versa, it will contribute to a decrease in H. Thus, the changes during December–March are associated with the decrease in H. For the remaining months E and Emax differences are both positive and approximately equal in magnitude, contributing little to changing H in R1.
Figures 5c and 5d show the R2 area-averaged monthly E and Emax for each decade, and the 2010s − 1980s monthly decadal differences for E and Emax. The E and Emax profiles for R2 (Fig. 5c) resemble those in R1. The largest E and Emax differences for R2 where they have the opposite signs also occur during boreal winter/early boreal spring (Fig. 5d).
Based on Fig. 5 the changes in H associated with the decline in the northern rain forest are primarily occurring during December–March, and it is these months that we need to focus on to understand the physical mechanism behind this change.
Figure 6a shows the ERA5 December 1980s decadal skin temperature, while Fig. 6b shows the 2010s − 1980s skin temperature difference. Skin temperatures over much of the Congo basin between 7°S and 4°N range between 296 and 298 K (Fig. 6a), while temperatures are 1–2 K cooler (1–4 K warmer) than this between 7° and 11°S (north of 4°N). Surface temperatures have been significantly warmed, up to 1 K, by the 2010s over most of the basin (Fig. 6b). North of 3°N the warming is even greater, up to 4 K near 8°N and 32°E where the atmosphere is even drier than at 3°N. This area of enhanced warming lies coincident with the northern rain forest margin and indicates an increase in the meridional temperature gradient over this region.
(a) 1980–89 (D1) average December surface skin temperature (K) and (b) 2010–19 minus 1980–89 (D4 − D1) December surface skin temperature difference. (c) D1 December 900-hPa geopotential heights (m) and winds (m s−1) and (d) D4 − D1 December 900-hPa geopotential heights and winds differences. (e) D1 average December 900-hPa specific humidity (g kg−1) and (f) D4 − D1 December 900-hPa specific humidity difference. Black stippling/red vectors denote differences that are significant at the 95% level of confidence. Unshaded areas indicate data in rock for the 900-hPa level.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Figure 6c shows the ERA5 December 900-hPa geopotential heights and winds for the 1980s, while Fig. 6d shows their 2010s − 1980s decadal differences. The 900-hPa heights are less than 1006 m over the Congo basin and the low-level flow is relatively weak (Fig. 6c). Along 4°N a trough extends from 32° to 15°E with strong northeasterly flow from northeastern Africa converging with weak southerly flow. By the 2010s the trough deepens (Fig. 6d) and is associated with a significant increase in anomalous meridional flow directed into the trough along 4°N.
This circulation change affects the low-level atmospheric moisture content in the region. Figure 6e shows the 1980s decadal December 900-hPa specific humidity, while Fig. 6f shows the 2010s − 1980s decadal difference in specific humidity. There is a strong meridional moisture gradient located over this forest margin with drier (wetter) air poleward (equatorward) of 4°N (Fig. 6e). Specific humidity values are 1–2.5 g kg−1 lower over the region by the 2010s (Fig. 6f) as enhanced low-level northeasterly flow increases the transport of drier air from the north.
Over the past four decades surface temperatures over the Congo basin have been warming due to global warming. The magnitude of this warming is larger north of 4°N on the dry side of the forest margin. This differential surface warming is associated with an intensification of the low-level trough along 4°N at the start of the boreal winter, which enhances northeasterly flow of drier air over the northern rain forest margin, resulting in lower values of H over this region.
This anomalous pattern persists into January (Fig. 7). Surface temperatures over the Congo basin in the 1980s (Fig. 7a) still exhibit the same general structure as December but are now warmer along the DRC/Republic of Congo border and the meridional temperature gradient along 4°N is stronger. The area of significant warming greater than 1 K expands westward to the coast and southward to the equator (Fig. 7b). Likewise, the low-level trough along 4°N in the 1980s (Fig. 7c) also deepens by the 2010s (Fig. 7d), but, unlike in December, significant anomalous flow is now primarily from the south and the west while the 900-hPa meridional specific humidity gradient shifts equatorward by January (Fig. 7e). By the 2010s, conditions are still anomalously dry along 4°N, but the anomalies are smaller than in December (Fig. 7f) due in part to the drier climatological conditions over the northern rain forest margin.
(a) 1980–89 (D1) average January surface skin temperature (K) and (b) 2010–19 minus 1980–89 (D4 − D1) January surface skin temperature difference. (c) D1 January 900-hPa geopotential heights (m) and winds (m s−1) and (d) D4 − D1 January 900-hPa geopotential heights and winds differences. (e) D1 average January 900-hPa specific humidity (g kg−1) and (f) D4 − D1 January 900-hPa specific humidity difference. Black stippling/red vectors denote differences that are significant at the 95% level of confidence.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
As the winter progresses and the scale of the amplified warming and the low-level trough along 4°N expands by the 2010s, anomalous northerly inflow into the trough weakens, while the anomalous southerly flow south of 4°N remains steady in January compared to December. As this transition occurs, the drier conditions begin to subside in January, briefly becoming wet in February, before reverting to dry by March as the low-level trough shifts poleward to 8°N (not shown). Evidence of this abrupt month-to-month change is identifiable in the R1 and R2 2010s − 1980s Emax differences (Fig. 5) and coincides with a gradual reduction in the negative E differences in Figs. 5b and 5d. This brief wet February period is consistent with results from Taylor et al. (2018) and Ward et al. (2022), who observe an increase in intense MCS frequency in February over the northern Congo basin since 1999 while the rest of the boreal winter behaves differently.
2) Changes in the southern Congo rainforest margins
The PVM-produced vegetation changes are less uniform in the southern rain forest with declines in the eastern and central forest margin areas (R3 and R4) but expansion for the coastal region R5 (Fig. 3h). Based on Fig. 4 and Table 3 these changes are primarily associated with changes in H. For R3, H does not fall below the threshold for any decade if the entire averaging region is considered. However, if a refined averaging region consisting of only the red shaded grid points in the R3 box in Fig. 3h is used, then H exceeds the threshold by the 2010s. For R4, H falls below the threshold in the 2000s, while D continues to decrease and approach but not cross the threshold. For R5 H eventually exceeds the threshold by the 2010s.
Figure 8a shows the R3 area-averaged monthly E and Emax for each decade, while Fig. 8b shows the E and Emax 2010s − 1980s monthly differences. From October to May E is nearly equal to Emax as the atmosphere is almost saturated (Fig. 8a). From June to September E decreases more than Emax, indicating a drier atmosphere with this drying increasing over the decades. Figure 8b shows a consistent increase in Emax in every month by the 2010s, but E decreases during the austral winter. The profiles of E and Emax in R4 (Fig. 8c) resemble those from R3, while the E and Emax 2010s − 1980s monthly differences for R4 (Fig. 8d) also indicate a decrease in E during the boreal winter by the 2010s.
(a) Area-averaged monthly actual evapotranspiration (E; solid lines; W m−2) and maximum evapotranspiration (Emax; dashed line; W m−2) for the 1980s (black), 1990s (red), 2000s (light blue), and 2010s (dark blue) decadal means area averaged over R3: South Interior East region. (b) The corresponding area-averaged monthly 2010s–1980s difference in E (solid line; W m−2) and Emax (dashed line; W m−2) for R3. (c),(d) As in (a) and (b), but for the R4 (South Interior West) region. (e),(f) As in (a) and (b), but for R5 (South Coastal) region.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Decadal changes in R5 (Figs. 8e,f) differ from those in R3 and R4. While the annual cycle is generally the same as R3 and R4, over the decades there is an increase in E in most months, and minimal change in Emax except in June and August when there is a decrease by the 2010s (Fig. 8f).
Overall, Fig. 8 indicates that the changes in H occur from June to September for the southern rain forest region, so we need to focus on these months.
Figure 9a shows the ERA5 June 1980s decadal surface temperatures, while Fig. 9b shows their 2010s − 1980s differences. Temperatures are warmest over the equatorial Congo basin and decrease from the equator to 16°S, setting up a positive meridional temperature gradient during June (Fig. 9a). Elevation is a factor as the terrain rises over 1000 m from the equator to 12°S. By the 2010s significant surface warming occurs over the entire Congo basin (Fig. 9b) with warming up to +2 K located over the central Congo basin. Elsewhere the warming is <1 K.
(a) 1980–89 (D1) average June surface skin temperature (K) and (b) 2010–19 minus 1980–89 (D4 − D1) June surface skin temperature difference. (c) D1 June 900-hPa geopotential heights (m) and winds (m s−1) and (d) D4 − D1 June 900-hPa geopotential heights and winds differences. (e) D1 average June 900-hPa specific humidity (g kg−1) and (f) D4 − D1 June 900-hPa specific humidity difference. Black stippling/red vectors denote differences that are significant at the 95% level of confidence.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Figures 9c and 9d show the ERA5 June 1980s decadal 900-hPa geopotential heights and winds, and their 2010s−1980s differences. The lowest heights over the Congo basin are east of 18°E, setting up a zonal height gradient with weak westerly flow directed down this gradient (Fig. 9c). South of 8°S the topography acts as a barrier with southeasterly downslope flow over southeasternmost DRC. Heights rise over both the Congo basin and the equatorial Atlantic by the 2010s (Fig. 9d), but the increase is greater over the eastern Congo basin weakening the zonal height gradient. These height changes are associated with significant changes in the low-level flow south of 4°S. Here westerly flow over the southwestern Congo basin and southeasterly flow along coastal Angola weaken by the 2010s.
In terms of low-level moisture during June, specific humidity is highest over the equatorial Congo basin, and there is a sharp meridional moisture gradient between 5° and 10°S (Fig. 9e). The moisture content is also lower along the eastern Atlantic coast, associated with strong southeasterly flow promoting coastal upwelling (Fig. 9a). The weaker onshore westerly flow between 4° and 8°S by the 2010s is associated with a significant decrease in the low-level moisture inland, as specific humidity values are 2–2.5 g kg−1 lower than the 1980s (Fig. 9f). Along the coast, there is a significant increase in the specific humidity by the 2010s as the relaxing of the low-level southeasterly flow weakens the coastal upwelling and warms the coastal SSTs.
Figure 10 is the same as Fig. 9, but for July. By July a robust area of significant warming develops over the western DRC by the 2010s, indicating that the warmest areas of the Congo basin are getting warmer (Figs. 10a,b). This affects the zonal height gradient and westerly flow into the interior Congo basin (Fig. 10c) as the anomalous low-level flow remains easterly but expands equatorward and eastward to 26°E (Fig. 10d). Additionally, the southeasterly low-level flow remains weaker along the Angola coast. Likewise, anomalous low-level specific humidity patterns (Figs. 10e,f) resemble the June patterns, but magnitudes are larger over the southern DRC and eastern South Atlantic Ocean.
(a) 1980–89 (D1) average July surface skin temperature (K) and (b) 2010–19 minus 1980–89 (D4 − D1) July surface skin temperature difference. (c) D1 July 900-hPa geopotential heights (m) and winds (m s−1) and (d) D4 − D1 July 900-hPa geopotential heights and winds differences. (e) D1 average July 900-hPa specific humidity (g kg−1) and (f) D4 − D1 July 900-hPa specific humidity difference. Black stippling/red vectors denote differences that are significant at the 95% level of confidence.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
These anomalies persist into August (Fig. 11). During August the low-level height field exhibits more meridional structure with lower heights at the base of the highlands along 8°S (Fig. 11c). Thus, enhanced surface warming over the southern Congo basin by the 2010s (Fig. 11b) now is also associated with the strengthening of northeasterly cross-equatorial flow down the anomalous height gradient (Fig. 11d). The enhanced northeasterly flow is directed from a relatively moist region climatologically (Fig. 11e), helping to mitigate the drying over the southern DRC (Fig. 11f) by the end of austral winter.
(a) 1980–89 (D1) average August surface skin temperature (K) and (b) 2010–19 minus 1980–89 (D4 − D1) August surface skin temperature difference. (c) D1 August 900-hPa geopotential heights (m) and winds (m s−1) and (d) D4 − D1 August 900-hPa geopotential heights and winds differences. (e) D1 average August 900-hPa specific humidity (g kg−1) and (f) D4 − D1 August 900-hPa specific humidity difference. Black stippling/red vectors denote differences that are significant at the 95% level of confidence.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Figures 9–11 indicate that changes in the circulation associated with surface warming affect the low-level flow, which reduces moisture transport over the southern Congolese rain forest interior during the early austral winter. This dries the region, making it more conducive for continued warming that helps maintain the anomalous easterly flow and drier conditions over the southern rain forest margin east of 15°E during winter.
Another factor is the decadal change in the South Atlantic subtropical anticyclone (SASH). The SASH plays a significant role in regional climate variability over Africa (Vigaud et al. 2009; Hermes and Reason 2009; Sun et al. 2017) and coastal upwelling variability along the southeastern Atlantic coast (Richter et al. 2010; Lübbecke et al. 2010, 2014; Vizy et al. 2018). Figure 12 shows the July 900-hPa 2010s − 1980s geopotential height differences. By the 2010s the SASH shifts west-southwest to 26°S and 23.5°W with significant height increases south of 20°S, consistent with other studies (Vizy and Cook 2016; Vizy et al. 2018; Sun et al. 2017; Reboita et al. 2019). Between 15° and 25°S height increases are larger on the western side of the basin, around 7 m west of 10°W compared to 3.5 m east of 0°E. The height gradient, and hence the southeasterly flow along the eastern flank of the SASH, weakens associated with this shift. It is this relaxation of coastal winds that is associated with the warming of coastal SSTs and an increase in low-level atmospheric moisture over the coastal rain forest margin by the 2010s. Furthermore, we hypothesize that it is this coastal warming of SSTs that is associated with the weakening of the low-level onshore flow at the beginning of austral winter (Fig. 9b) that reduces the low-level moisture transport into the southern Congo basin (Fig. 9f), leading to the development of amplified warming centered around 4°S and 20°E.
D4 − D1 July 900-hPa geopotential heights differences. White and black stippling denote differences that are significant at the 90% and 95% level of confidence, respectively. Open and closed circles denote the position of the D1 and D4 subtropical high based on the 900-hPa height fields, respectively. Solid contours are drawn every 5 m, while dashed contours are every 2 m between 0 and 10 m to highlight differences close to the tropics.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
3) Changes in the interior Congo rainforest
PVM-produced changes in the rain forest interior (R6 and R7) are localized until the 2010s when the rain forest decline becomes more widespread (Figs. 3e–h and Table 2). These declines are primarily associated with changes in D (Fig. 4 and Table 3). Figures 13a and 13b show the area-averaged monthly value for F(0.5 − wi) from Eq. (2) for each decade, as well asd the 2010s − 1980s monthly differences, for R6. Figures 13c and 13d show the same fields, but for R7. All of the change in D occurs during the 2010s as the soil saturation is now less than 50% for all months for both regions.
(a) Area-averaged monthly F(0.5 − wi) (unitless) for the 1980s (black), 1990s (red), 2000s (light blue), and 2010s (dark blue) decadal means area averaged over R6: the North Equatorial Central region. (b) The corresponding area-averaged monthly 2010s − 1980s difference in F(0.5 − wi) (unitless) for R6. (c),(d) As in (a) and (b), but for R7: the South Equatorial Central region.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
These changes in the Congo basin interior are an indirect consequence of the land use changes between the 1980s and the 2010s. While ERA5 implements static vegetation cover, the remotely sensed measurements of atmospheric fields used as input for the reanalysis algorithm are not constrained by static vegetation. This means that ERA5 atmospheric conditions will change over time in response to dynamic and human-induced changes in the underlying land surface even though ERA5 sets vegetation cover constant in time. The PVM is capturing the consequences of the conversion of the rain forest to cropland along the major transportation routes (Fig. 2a). For example, ESA-CCI landcover data indicate that the cropland in R6 and R7 has increased by approximately 19% and 9% respectively, between 1992 and 2019. Others have noted the accelerating disappearance of the rain forest area in this basin (Ernst et al. 2013; Tyukavina et al. 2018; Kleinschroth et al. 2019; Molinario et al. 2020). This indicates that as more of the forest is cleared, the associated changes in seasonality are likely to become more relevant for determining whether the climate can support the interior rain forest ecosystem.
c. Consistency with independent datasets
Our findings are compared with independent rainfall, MERRA-2 reanalysis, and landcover data to evaluate the extent to which the forest dieback signal is evident. Figures 14a–d show the 2010s − 1980s December decadal rainfall differences for CHIRPS, ARC2, TAMSAT, and PERSIANN-CDR, respectively. While the rainfall anomaly magnitudes vary among the datasets, all four indicate that reduced rainfall along the northern margin (3°–6°N). ARC2 has the largest negative anomaly magnitudes, while CHIRPS has the largest significant area. There is good agreement of significant drying over northern DRC between 22° and 28°E, the area where ERA5 indicates there is reduced low-level atmospheric moisture (Fig. 6f).
2010s minus 1980s December precipitation difference (mm day−1) from (a) CHIRPS2, (b) ARC2, (c) TARCAT, and (d) PERSIANN. (e)–(h) As in (a)–(d), but for August. Black stippling denotes differences that are significant at the 90% level of confidence. Relevant averaging region boxes for the northern and southern margins are drawn in red.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
Figures 14e–h show August 2010s − 1980s decadal rainfall differences for the same datasets. While CHIRPS, ARC2, and PERIANN have reduced August rainfall from 18° to 29°E over the southern margin by the 2010s, only CHIRPS indicates a coherent significant region. TAMSAT (Fig. 14g) hints at a decrease in August rainfall across the margin, but it is narrower, centered near 4°S, and only extends to 25°E. Rainfall is reduced over R3 in each precipitation dataset, but there are disagreements regarding magnitude, spatial scale, and significance. Over R4, agreement is better, but only CHIRPS and ARC2 indicate significant differences. For R5, climatological rainfall is minimal in August, so rainfall differences are small. Overall, these rainfall anomalies are consistent with the ERA5 atmospheric moisture changes.
Skin temperature and low-level specific humidity in the MERRA-2 reanalysis are compared with ERA5. Figure 15a shows the December 1980s decadal surface skin temperature, while Fig. 15b shows their 2010s − 1980s decadal difference. While the 1980s December surface temperatures are cooler than ERA5 between 3° and 8°N (Fig. 15a), there is still evidence of stronger surface warming along 4°N by the 2010s (Fig. 15b) that become more robust in size during January and February (not shown). Consistent with ERA5 (Fig. 6f), MERRA-2 also indicates a significantly robust dry signal along 4°N (Figs. 15c,d) that persists through January before transitioning to a wet anomaly (not shown).
MERRA2 (a) 1980–89 (D1) average December surface skin temperature (K), and (b) 2010–19 minus 1980–89 (D4 − D1) December surface skin temperature difference. (c) D1 average December 900-hPa specific humidity (g kg−1), and (d) D4 − D1 December 900-hPa specific humidity difference. (e)–(h) As in (a)–(d), respectively, but for August. Black stippling denotes differences that are significant at the 95% level of confidence.
Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-23-0275.1
For August (Figs. 15e–h) the 1980s MERRA-2 surface temperatures in the Congo basin (Fig. 15e) are close to that in ERA5 (Fig. 11a), while temperatures on the elevated land south and east of the basin are cooler. The warming signal is stronger in MERRA-2 (Fig. 15f), reaching above 3 K near 8°S and 20°E, and extending farther south into northeastern Angola and southernmost DRC. MERRA-2 also indicates an anomalous pattern of lower (higher) specific humidity over southeastern DRC (along the coastal areas) consistent with ERA5, although the lower specific humidity anomaly is smaller in spatial extent and weaker in MERRA2 compared to ERA5.
Finally, ESA-CCI landcover data are evaluated over the seven regions. There are limitations to comparing this observed landcover dataset with the PVM results since both climate and deforestation changes are present. Also, the difference in spatial resolution between ESA-CCI and the PVM forced by ERA5 is large, and ESA-CCI only provides annual vegetation estimates back to 1992. The point of the evaluation is to see if changes in the rain forest area occur in the regions identified using the PVM.
Table 4 provides the decadal average rain forest estimates for the seven regions from ESA-CCI along with an estimate of the rain forest area percent change between the 2010s and 1990s. The PVM area percent change sign (see last column in Table 2) agrees with the sign change in ESA-CCI for every region except R4. Magnitudes of the changes are larger in the PVM, but the distribution of the change is represented. In R4, ESA-CCI indicates the rain forest increased by 7.7%, opposite to the PVM results. It is unclear exactly why there is this difference.
ESA-CCI decadal analysis of evergreen broadleaf vegetation type areal extent (km2).
Overall, decadal anomalies in precipitation observations, the vegetation cover dataset, and MERRA-2 are largely consistent with the findings from ERA5 that identify physical processes responsible for decadal climate changes in the northern and southern fringes of the Congolese rain forest. Agreement is stronger for the boreal winter changes along 4°N than for the austral winter changes over the southern Congo basin.
5. Summary and conclusions
A potential vegetation model is used to translate decadal mean climate states derived from ERA5 into natural vegetation distributions to identify regions where climate change could have played a role in changing vegetation and to understand how and why these climate changes could affect the Congolese rain forest coverage.
We optimize the CPTEC PVM and force it with ERA5 data to translate decadal-mean climate states into natural vegetation distributions for the 1980s–2010s. The PVM vegetation distributions are analyzed to identify areas of the rain forest that are susceptible to climate forcing, and to understand how climate conditions are changing in a way that may alter climate support for the rain forest.
PVM results indicate that the most robust Congo rain forest changes occur on the northern and southern forest margins (Fig. 3). For the former, two prominent areas of PVM produced forest decline emerge, one over Cameroon (R1) and the other along the northern DRC/southern Central African Republic border (R2). For both areas the forest decline in the PVM is associated with a decrease in the wetness index H [Eq. (1)] due to a decrease in E combined with an increase in Emax during December–March (Fig. 5). Boreal winter surface temperatures have increased over the entire Congo basin since the 1980s (Fig. 6b), consistent with the observed global warming trend (Eyring et al. 2021; IPCC 2021). However, the rate of warming is larger north of 4°N on the dry side of the forest margin (Fig. 6e). This warming is associated with an intensification of the low-level trough along 4°N, helping to increase northeasterly flow of drier air over the northern margin that reinforces the drier conditions (Fig. 6f), and supports the continued surface warming along the forest margin during December. As the trough continues to deepen, the anomalous northeasterly flow weakens north of 4°N allowing the drier conditions to begin subsiding in January (Fig. 7f) and become wetter in February before becoming dry once again in March as the trough seasonally shifts poleward.
For the southern margin (5.5°S), the PVM indicates a decrease in rain forest size east of 16°E (R3 and R4) by the 2010s (Fig. 3h). This decrease is associated with changes in H during the austral winter (Fig. 8). Enhanced surface warming near 4°S (Fig. 9b) weakens the low-level zonal height gradient over the Congo basin, reducing the eastward low-level moisture transport into the interior (Fig. 9f). As the winter progresses surface warming over the southwestern DRC becomes more robust and the associated circulation response reinforces the anomalous easterly flow and drier conditions before breaking down in September (Fig. 8). Agreement with independent datasets is weaker for the austral winter changes over the southern Congo basin than the boreal winter changes along 4°N, indicating that more work is still needed to completely understand the austral winter mechanism.
Over coastal DRC/Angola (R5) the PVM indicates a rain forest expansion by the 2010s (Fig. 3h). Changes in H and D (Table 3) indicate the environment is becoming wetter with less seasonal influence since the 1980s. This forest expansion is associated with a west-southwestward shift in the SASH, particularly during the austral winter, that weakens the flow along the coast leading to reduced coastal upwelling, warmer SSTs, and a wetter environment.
Localized declines in the interior rain forest (R6 and R7) become more robust by the 2010s (Fig. 3h) and are found to be an indirect effect of the land use changes over the past four decades. The PVM captures the impact of the rain forest to cropland conversion along transportation routes in the Congo basin as actual observational and satellite-derived measurements are used to constrain ERA5 despite their being static vegetation cover in ERA5 itself. By the 2010s the changes are on a large enough scale to affect D, resulting in the loss of forest in the PVM. Thus, human-induced deforestation in the rain forest interior is leading to an increase in the seasonality of soil moisture availability that is likely to threaten the remaining rain forest if the scale of deforestation becomes large enough.
The following conclusions are drawn:
-
The climate appears to be becoming less able to support the tropical rain forest ecosystem along the northern and southern forest margins. In these areas, environmental conditions are closest to thresholds in surface wetness (H) and, at least locally, the seasonality of soil saturation (D). Furthermore, sharp contrasts in temperature exist across these margins, particularly during the winter/dry seasons (Figs. 6e and 9e). This differential heating across the boundaries influences the low-level circulation and moisture transport over the margins. In the northern margin, stronger warming north of the rain forest in early boreal winter enhances drier northeasterly flow into the region. For the southern margin, enhanced warming in the southern Congo basin weakens the low-level westerly flow of moisture, reinforcing the dry austral winter conditions.
-
This surface warming mechanism suggests attribution to greenhouse gas–induced climate change. Surface temperatures will continue to increase through the century (IPCC 2021), modifying meridional temperature gradients as some areas over Africa are warming faster than others (Cook and Vizy 2015; Vizy and Cook 2017; Cook et al. 2020).
-
Our analysis indicates that climate change over the coast of Angola may provide strengthening support for the rain forest ecosystem. Southeasterly trade winds on the eastern flank of the SASH weaken over time resulting in reduced coastal upwelling and warmer SSTs that enhance the atmospheric moisture content over the adjacent land, making it a more suitable environment for the rain forest ecosystem.
Along the transportation routes in the rain forest interior (Figs. 3g and 4h), the PVM-generated rain forest declines after 2010. This change is forced by the remotely sensed observations that constrain ERA5, and it is not related to a large-scale change in the circulation as is identified over the northern and southern rain forest boundaries. This implies that human intervention (deforestation) has caused changes in the seasonality of the soil moisture to the point where the climate is no longer able to sustain the rain forest, at least locally along the interior transportation routes. As more rain forest is converted to cropland, the repercussions for the health of the rain forest may increase. In fact, we may already be witnessing this effect in R4 (Fig. 3h), which is near the Brazzaville/Kinshasa metropolitan area and has seen a population increase from 2.05 million in 1980 to 13.74 million in 2019. More work is needed to better understand this interaction and its inherent thresholds as deforestation and greenhouse gas–induced climate change continue. Likewise, this study is focused on analyzing the changes in the climate on monthly to seasonal time scales. The next step is to evaluate the mechanisms identified here, including the development of anomalous southerly flow over the central and eastern Congo basin, at shorter, synoptic time scales to better understand how they operate to affect the regional climate variability of the Congo basin.
Acknowledgments.
This work was funded by NSF Award 1939880. We acknowledge the Texas Advanced Computing Center at The University of Texas at Austin for providing HPC resources that have contributed to the results reported within this paper (http://www.tacc.utexas.edu). The authors declare no conflict of interests.
Data availability statement.
All datasets analyzed in this study are freely available from their original sources.
REFERENCES
Asefi-Najafabady, S., and S. Saatchi, 2013: Response of African humid tropical forest to recent rainfall anomalies. Philos. Trans. Roy. Soc., B368, 20120306, https://doi.org/10.1098/rstb.2012.0306.
Baker, J. C. A., and D. V. Spracklen, 2022: Divergent representation of precipitation recycling in the Amazon and Congo in CMIP6 models. Geophys. Res. Lett., 49, e2021GL095136, https://doi.org/10.1029/2021GL095136.
Betts, M. G., and Coauthors, 2019: Extinction filters mediate the global effects of habitat fragmentation on animals. Science, 366, 1236–1239, https://doi.org/10.1126/science.aax9387.
Broxton, P. D., X. Zeng, D. Sulla-Menashe, and P. A. Troch, 2014: A global land cover climatology using MODIS data. J. Appl. Meteor. Climatol., 53, 1593–1605, https://doi.org/10.1175/JAMC-D-13-0270.1.
Burnett, M. W., G. R. Quetin, and A. G. Konings, 2020: Data-driven estimates of evapotranspiration and its controls in the Congo basin. Hydrol. Earth Syst. Sci., 24, 4189–4211, https://doi.org/10.5194/hess-24-4189-2020.
Cardoso, M. F., C. A. Nobre, D. M. Lapola, M. D. Oyama, and G. Sampaio, 2008: Long-term potential for fires in estimates of the occurrence of savannas in the tropics. Global Ecol. Biogeogr., 17, 222–235, https://doi.org/10.1111/j.1466-8238.2007.00356.x.
Christensen, J. H., and Coauthors, 2013: Climate phenomena and their relevance for future regional climate change. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 1217–1308.
Cook, K. H., and E. K. Vizy, 2008: Effects of 21st century climate change on the Amazon rainforest. J. Climate, 21, 542–560, https://doi.org/10.1175/2007JCLI1838.1.
Cook, K. H., and E. K. Vizy, 2015: Detection and analysis of an amplified warming of the Sahara Desert. J. Climate, 28, 6560–6580, https://doi.org/10.1175/JCLI-D-14-00230.1.
Cook, K. H., Y. Liu, and E. K. Vizy, 2020: Congo basin drying associated with poleward shifts of the African thermal lows. Climate Dyn., 54, 863–883, https://doi.org/10.1007/s00382-019-05033-3.
Creese, A., R. Washington, and R. Jones, 2019: Climate change in the Congo basin: Processes related to wetting in the December–February dry season. Climate Dyn., 53, 3583–3602, https://doi.org/10.1007/s00382-019-04728-x.
Dargie, G. C., S. L. Lewis, I. T. Lawson, E. T. A. Mitchard, S. E. Page, Y. E. Bocko, and S. A. Ifo, 2017: Age, extent and carbon storage of the central Congo basin peatland complex. Nature, 542, 86–90, https://doi.org/10.1038/nature21048.
Dezfuli, A. K., and S. E. Nicholson, 2013: The relationship of rainfall variability in western equatorial Africa to the tropical oceans and atmospheric circulation. Part II: The boreal autumn. J. Climate, 26, 66–84, https://doi.org/10.1175/JCLI-D-11-00686.1.
Díaz, S., and Coauthors, 2019: Summary for policymakers of the global assessment report on biodiversity and ecosystem services of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services. 39 pp., https://www.ipbes.net/sites/default/files/downloads/spm_unedited_advance_for_posting_htn.pdf.
Dinku, T., C. Funk, P. Peterson, R. Maidment, T. Tadesse, H. Gadain, and P. Ceccato, 2018: Validation of the CHIRPS satellite rainfall estimates over eastern Africa. Quart. J. Roy. Meteor. Soc., 144, 292–312, https://doi.org/10.1002/qj.3244.
Dyer, E. L. E., D. B. A. Jones, J. Nusbaumer, H. Li, O. Collins, G. Vettoretti, and D. Noone, 2017: Congo basin precipitation: Assessing seasonality, regional interactions, and sources of moisture. J. Geophys. Res. Atmos., 122, 6882–6898, https://doi.org/10.1002/2016JD026240.
Ernst, C., P. Mayaus, A. Verhegghen, C. Bodart, M. Christophe, and P. Defourny, 2013: National forest cover change in Congo basin: Deforestation, reforestation, degradation and regeneration for the years 1990, 2000 and 2005. Global Change Biol., 19, 1173–1187, https://doi.org/10.1111/gcb.12092.
ESA, 2017: Land Cover CCI Product User Guide version 2. Tech. Rep., 87 pp., https://www.esa-landcover-cci.org/?q=webfm_send/84.
Eyring, V., and Coauthors, 2021: Human influence on the climate system. Climate Change 2021: The Physical Science Basis, V. Masson-Delmotte, et al. eds., Cambridge University Press, 423–552, https://doi.org/10.1017/9781009157896.005.
FAO, 2011: The state of the Amazon Basin, Congo Basin and Southeast Asia. Food and Agriculture Organization of the United Nations, 80 pp., https://www.fao.org/3/i2247e/i2247e00.pdf.
Funk, C., and Coauthors, 2015: The Climate Hazards Infrared Precipitation with Stations—A new environmental record for monitoring extremes. Sci. Data, 2, 150066, https://doi.org/10.1038/sdata.2015.66.
Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 5149–5454, https://doi.org/10.1175/JCLI-D-16-0758.1.
Gleixner, S., T. Demissie, and G. T. Diro, 2020: Did ERA5 improve temperature and precipitation reanalysis over East Africa? Atmosphere, 11, 996, https://doi.org/10.3390/atmos11090996.
Global Forest Watch, 2022: Global Forest Watch: Forest monitoring designed for action. World Resources Institute, accessed 24 October 2022, https://www.globalforestwatch.org.
Guan, K., and Coauthors, 2015: Photosynthetic seasonality of global tropical forests constrained by hydroclimate. Nat. Geosci., 8, 284–289, https://doi.org/10.1038/ngeo2382.
Haddad, N., and Coauthors, 2015: Habitat fragmentation and its lasting impact on Earth’s ecosystems. Sci. Adv., 1, e1500052, https://doi.org/10.1126/sciadv.1500052.
Haensler, A., F. Saeed, and D. Jacob, 2013: Assessing the robustness of projected precipitation changes over central Africa on the basis of a multitude of global and regional climate projections. Climatic Change, 121, 349–363, https://doi.org/10.1007/s10584-013-0863-8.
Hansen, M. C., and Coauthors, 2014: High resolution global maps of 21st-century forest cover change. Science, 342, 850–853, https://doi.org/10.1126/science.1244693.
Harris, N. L., and Coauthors, 2021: Global maps of twenty-first century forest carbon fluxes. Nat. Climate Change, 11, 234–240, https://doi.org/10.1038/s41558-020-00976-6.
Hermes, J. C., and C. J. C. Reason, 2009: Variability in sea-surface temperature and winds in the tropical south-east Atlantic Ocean and regional rainfall relationships. Int. J. Climatol., 29, 11–21, https://doi.org/10.1002/joc.1711.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Howard, E., and R. Washington, 2018: Characterizing the synoptic expression of the Angola low. J. Climate, 31, 7147–7165, https://doi.org/10.1175/JCLI-D-18-0017.1.
Hua, W., L. Zhou, H. Chen, S. E. Nicholson, A. Raghavendra, and Y. Jiang, 2016: Possible causes of the central equatorial African long-term drought. Environ. Res. Lett., 11, 124002, https://doi.org/10.1088/1748-9326/11/12/124002.
Hua, W., L. Zhou, H. Chen, S. E. Nicholson, Y. Jiang, and A. Raghavendra, 2018: Understanding the central equatorial Africa long-term drought using AMIP-type simulations. Climate Dyn., 50, 1115–1128, https://doi.org/10.1007/s00382-017-3665-2.
IPCC, 2021: Climate Change 2021: The Physical Science Basis. V. Masson-Delmotte et al., Eds., Cambridge University Press, 2391 pp., https://doi.org/10.1017/9781009157896.
Kleinschroth, F., N. Laporte, W. F. Laurance, S. J. Goetz, and J. Ghazoul, 2019: Road expansion and persistence in forests of the Congo basin. Nat. Sustainability, 2, 628–634, https://doi.org/10.1038/s41893-019-0310-6.
Kranz, O., E. Schoepfer, R. Tegtmeyer, and S. Lang, 2018: Earth observation based multi-scale assessment of logging activities in the Democratic Republic of the Congo. ISPRS J. Photogramm. Remote Sens., 144, 254–267, https://doi.org/10.1016/j.isprsjprs.2018.07.012.
Loveland, T. R., B. C. Reed, J. F. Brown, D. O. Ohlen, Z. Zhu, L. Youing, and J. W. Merchant, 2000: Development of a global land cover characteristic database and IGB6 DISCover from the 1 km AVHRR data. Int. J. Remote Sens., 21, 1303–1330, https://doi.org/10.1080/014311600210191.
Lübbecke, J. F., C. W. Böning, N. S. Keenlyside, and S.-P. Xie, 2010: On the connection between Benguela and equatorial Atlantic Niños and the role of the South Atlantic anticyclone. J. Geophys. Res., 115, C09015, https://doi.org/10.1029/2009JC005964.
Lübbecke, J. F., N. J. Burls, C. J. C. Reason, and M. J. McPhaden, 2014: Variability in the South Atlantic anticyclone and the Atlantic Niño mode. J. Climate, 27, 8135–8150, https://doi.org/10.1175/JCLI-D-14-00202.1.
Maidment, R. I., R. P. Allan, and E. Black, 2015: Recent observed and simulated changes in precipitation over Africa. Geophys. Res. Lett., 42, 8155–8164, https://doi.org/10.1002/2015GL065765.
Maidment, R. I., and Coauthors, 2017: A new, long-term daily satellite-based rainfall dataset for operational monitoring in Africa. Sci. Data, 4, 170063, https://doi.org/10.1038/sdata.2017.63.
Molinario, G., M. Hansen, P. Potapov, A. Tyukavina, and S. Stehman, 2020: Contextualizing landscape-scale forest cover loss in the Democratic Republic of Congo (DRC) between 2000 and 2015. Land, 9, 23, https://doi.org/10.3390/land9010023.
Munday, C., and R. Washington, 2017: Circulation controls on southern African precipitation in coupled models: The role of the Angola low. J. Geophys. Res. Atmos., 122, 861–877, https://doi.org/10.1002/2016JD025736.
Nicholson, S. E., J. Kim, M. B. Ba, and A. R. Lare, 1997: The mean surface water balance over Africa and its interannual variability. J. Climate, 10, 2981–3002, https://doi.org/10.1175/1520-0442(1997)010<2981:TMSWBO>2.0.CO;2.
Nicholson, S. E., D. Klotter, L. Zhou, and W. Hua, 2019: Validation of satellite precipitation estimates over the Congo basin. J. Hydrometeor., 20, 631–656, https://doi.org/10.1175/JHM-D-18-0118.1.
Nicholson, S. E., D. Klotter, L. Zhou, and W. Hua, 2022: Recent rainfall conditions in the Congo basin. Environ. Res. Lett., 17, 054052, https://doi.org/10.1088/1748-9326/ac61c4.
Novella, N. S., and W. M. Thiaw, 2013: African Rainfall Climatology version 2 for famine early warning systems. J. Appl. Meteor. Climatol., 52, 588–606, https://doi.org/10.1175/JAMC-D-11-0238.1.
Oyama, M. D., and C. A. Nobre, 2004: A simple potential vegetation model for coupling with Simple Biosphere Model (SiB). Braz. J. Meteor., 19, 203–216.
Parker, D., and Coauthors, 2005: The diurnal cycle of the West African monsoon circulation. Quart. J. Roy. Meteor. Soc., 131, 2839–2860, https://doi.org/10.1256/qj.04.52.
Philippon, N., and Coauthors, 2019: The light-deficient climates of western Central African evergreen forests. Environ. Res. Lett., 14, 034007, https://doi.org/10.1088/1748-9326/aaf5d8.
Pokam, W. M., L. A. T. Djiotang, and F. K. Mkankam, 2012: Atmospheric water vapor transport and recycling in equatorial Central Africa through NCEP/NCAR reanalysis data. Climate Dyn., 38, 1715–1729, https://doi.org/10.1007/s00382-011-1242-7.
Pokam, W. M., C. L. Bain, R. S. Chadwick, R. Grahm, D. J. Sonwa, and F. M. Kamga, 2014: Identification of processes driving low-level westerlies in west equatorial Africa. J. Climate, 27, 4245–4262, https://doi.org/10.1175/JCLI-D-13-00490.1.
Potapov, P., and Coauthors, 2017: The last frontiers of wilderness: Tracking loss of intact forest landscapes from 2000 to 2013. Sci. Adv., 3, e1600821, https://doi.org/10.1126/sciadv.1600821.
Prentice, I. C., W. Cramer, S. P. Harrison, R. Leemans, R. A. Monserud, and A. M. Solomon, 1992: A global biome model based on plant physiology and dominance, soil properties, and climate. J. Biogeogr., 19, 117–134, https://doi.org/10.2307/2845499.
Quinlan, J. R., 1993: C4.5: Program for Machine Learning. Morgan Kaufmann, 302 pp.
Reboita, M. S., T. Ambrizzi, B. A. Silva, R. F. Pinheiro, and R. P. da Rocha, 2019: The South Atlantic subtropical anticyclone: Present and future climate. Front. Earth Sci., 7, 8, https://doi.org/10.3389/feart.2019.00008.
Richter, I., S. K. Behera, Y. Masumoto, B. Taguchi, N. Komori, and T. Yamagata, 2010: On the triggering of Benguela Niños: Remote equatorial versus local influences. Geophys. Res. Lett., 37, L20604, https://doi.org/10.1029/2010GL044461.
Riddell, M., 2013: Assessing the impacts of conservation and commercial forestry on livelihoods in northern Republic of Congo. Conserv. Soc., 11, 199, https://doi.org/10.4103/0972-4923.121002.
Risi, C., D. Noone, C. Frankenberg, and J. Worden, 2013: Role of continental recycling in intraseasonal variations of continental moisture as deduced from model simulations and water vapor isotopic measurements. Water Resour. Res., 49, 4136–4156, https://doi.org/10.1002/wrcr.20312.
Shapiro, A. C., H. S. Grantham, N. Aguilar-Amuchastegui, N. J. Murray, V. Gond, D. Bonfils, and O. Rickenbach, 2021: Forest condition in the Congo basin for the assessment of ecosystem conservation status. Ecol. Indic., 122, 107268, https://doi.org/10.1016/j.ecolind.2020.107268.
Singh, C., R. van der Ent, L. Wang-Erlandsson, and I. Fetzer, 2021: Hydroclimatic adaptation critical to the resilience of tropical forests. Global Change Biol., 28, 2930–2939, https://doi.org/10.1111/gcb.16115.
Sonwa, D. J., J. N. Nkem, M. E. Idinoba, M. Y. Bele, and C. Jum, 2012: Building regional priorities in forests for development and adaptation to climate change in the Congo basin. Mitigation Adapt. Strategies Global Change, 17, 441–450, https://doi.org/10.1007/s11027-011-9335-5.
Sonwa, D. J., M. O. Farikou, G. Martial, and F. L. Félix, 2020: Living under a fluctuating climate and a drying Congo basin. Sustainability, 12, 2936, https://doi.org/10.3390/su12072936.
Sorí, R., R. Nieto, S. M. Vicente-Serrano, A. Drumond, and L. Gimeno, 2017: A Lagrangian perspective of the hydrological cycle in the Congo River basin. Earth Syst. Dyn., 8, 653–675, https://doi.org/10.5194/esd-8-653-2017.
Sorooshian, S., K.-L. Hsu, X. Gao, H. V. Gupta, B. Imam, and D. Braithwaite, 2000: Evaluation of PERSIANN system satellite-based estimates of tropical rainfall. Bull. Amer. Meteor. Soc., 81, 2035–2046, https://doi.org/10.1175/1520-0477(2000)081<2035:EOPSSE>2.3.CO;2.
Staal, A., I. Fetzer, L. Wang-Erlandsson, J. H. C. Bosmans, S. C. Dekker, E. H. van Nes, J. Rockström, and O. A. Tuinenburg, 2020: Hysteresis of tropical forests in the 21st century. Nat. Commun., 11, 4978, https://doi.org/10.1038/s41467-020-18728-7.
Steinkopf, J., and F. Engelbrecht, 2022: Verification of ERA5 and ERA-Interim precipitation over Africa at intra-annual and interannual timescales. Atmos. Res., 280, 106427, https://doi.org/10.1016/j.atmosres.2022.106427.
Sternberg, L. S. L., 2001: Savanna–forest hysteresis in the tropics. Global Ecol. Biogeogr., 10, 369–378, https://doi.org/10.1046/j.1466-822X.2001.00243.x.
Sun, X., K. H. Cook, and E. K. Vizy, 2017: The South Atlantic subtropical high: Climatology and interannual variability. J. Climate, 30, 3279–3296, https://doi.org/10.1175/JCLI-D-16-0705.1.
Taylor, C. M., A. H. Fink, C. Klein, D. J. Parker, F. Guichard, P. P. Harris, and K. R. Knapp, 2018: Earlier seasonal onset of intense mesoscale convective systems in the Congo basin since 1999. Geophys. Res. Lett., 45, 13 458–13 467, https://doi.org/10.1029/2018GL080516.
Tuinenburg, O. A., J. J. E. Theeuwen, and A. Staal, 2020: High-resolution global atmospheric moisture connections from evaporation to precipitation. Earth Syst. Sci. Data, 12, 3177–3188, https://doi.org/10.5194/essd-12-3177-2020.
Turubanova, S., P. V. Potapov, A. Tyukavina, and M. C. Hansen, 2018: Ongoing primary forest loss in Brazil, Democratic Republic of Congo, and Indonesia. Environ. Res. Lett., 13, 074028, https://doi.org/10.1088/1748-9326/aacd1c.
Tyukavina, A., M. C. Hansen, P. Potapov, D. Parker, C. Okpa, S. V. Stehman, I. Kommareddy, and S. Turubanova, 2018: Congo basin forest loss dominated by increasing smallholder clearing. Sci. Adv., 4, eaat2993, https://doi.org/10.1126/sciadv.aat2993.
Vigaud, N., Y. Richard, M. Rouault, and N. Fauchereau, 2009: Moisture transport between the South Atlantic Ocean and southern Africa: Relationships with summer rainfall and associated dynamics. Climate Dyn., 32, 113–123, https://doi.org/10.1007/s00382-008-0377-7.
Vizy, E. K., and K. H. Cook, 2016: Understanding long-term (1982–2013) multi-decadal change in the equatorial and subtropical South Atlantic climate. Climate Dyn., 46, 2087–2113, https://doi.org/10.1007/s00382-015-2691-1.
Vizy, E. K., and K. H. Cook, 2017: Seasonality of the observed amplified Sahara warming trend and implications for Sahel rainfall. J. Climate, 30, 3073–3094, https://doi.org/10.1175/JCLI-D-16-0687.1.
Vizy, E. K., K. H. Cook, and X. Sun, 2018: Decadal change of the South Atlantic Ocean Angola-Benguela frontal zone since 1980. Climate Dyn., 51, 3251–3273, https://doi.org/10.1007/s00382-018-4077-7.
Wang, X.-Y., J. Zhu, M. Xin, C. Song, Y. Li, Y. Zhou, and X. Li, 2021: Weakened seasonality of the African rainforest precipitation in boreal winter and spring driven by tropical SST variabilities. Geosci. Lett., 8, 22, https://doi.org/10.1186/s40562-021-00192-w.
Ward, N., A. H. Fink, R. J. Kene, and D. J. Parker, 2022: Upper-level midlatitude troughs in boreal winter have an amplified low-latitude linkage over Africa. Atmos. Sci. Lett., 24, e1129, https://doi.org/10.1002/asl.1129.
Washington, R., R. James, H. Pearce, W. M. Pokam, and W. Moufouma-Okia, 2013: Congo basin rainfall climatology: Can we believe the climate models? Philos. Trans. Roy. Soc., B368, 20120296, https://doi.org/10.1098/rstb.2012.0296.
Wilkie, D., G. Morelli, F. Rotberg, and E. Shaw, 1999: Wetter isn’t better: Global warming and food security in the Congo basin. Global Environ. Change, 9, 323–328, https://doi.org/10.1016/S0959-3780(99)00021-7.
Worden, S., R. Fu, S. Chakraborty, J. Liu, and J. Worden, 2021: Where does moisture come from over the Congo basin? J. Geophys. Res. Biogeosci., 126, e2020JG006024, https://doi.org/10.1029/2020JG006024.
Wright, J. S., and Coauthors, 2020: Differences in tropical high clouds among reanalyses: Origins and radiative impacts. Atmos. Chem. Phys., 20, 8989–9030, https://doi.org/10.5194/acp-20-8989-2020.
Zhou, L. M., and Coauthors, 2014: Widespread decline of Congo rainforest greenness in the past decade. Nature, 509, 86–90, https://doi.org/10.1038/nature13265.
Zunckel, M., Y. Hong, K. Brassel, and S. O’Beirne, 1996: Characteristics of the nocturnal boundary layer: Okaukuejo, Namibia, during SAFARI-92. J. Geophys. Res., 101, 23 757–23 766, https://doi.org/10.1029/95JD00624.
