Tropical savannas have been increasingly targeted for carbon (C) sequestration from afforestation, assuming large gains in soil organic C (SOC) with increasing tree cover. Because savanna SOC is also derived from grasses, this assumption may not reflect real changes in SOC under afforestation, but grass contributions to SOC and changes in SOC with increasing tree cover remain poorly synthesized. Here, we combine a case study from Kruger National Park, South Africa, with data synthesized from tropical savannas globally to show that grass-derived C constitutes more than half of total SOC to a soil depth of 1-meter, even in soils directly under trees. The largest SOC concentrations were associated with the largest grass contributions (> 70% of total SOC). Regionally and across the tropics, SOC concentration was not explained by tree cover. Both SOC gain and loss were observed following increasing tree cover, and on average SOC storage within 1-meter profile only increased by a negligible and non-significant 6% (SE = 4%, n = 44). These results underscore the substantial contribution of grasses to SOC and the considerable uncertainty in SOC responses to increasing tree cover, challenging the widespread assumption that afforestation universally and substantially enhances SOC storage across tropical savannas.

SOC dynamics across Kruger: We extensively sampled soil cores across a semiarid savanna in Kruger National Park to explore factors that influence C₄-derived C (i.e., grass-derived C) and SOC concentration at the regional scale.  Kruger covers nearly 20,000 km² (lat. 22°20'-25°30'S, long. 31°10'-32°00'E) and consists of tropical and subtropical savannas. Mean annual rainfall increases from 350 mm in the north to 750 mm in the south with most precipitation falling from November to April. Elevation ranges from 260 to 839 m above sea level. Based on long-term fire records, the average fire return interval for Kruger is c. 3.5 years, but fire regimes vary spatially across the park, ranging from one fire per year to one every 34 years. Kruger is dominated by two underlying parent materials, granite and basalt, which strongly influence soil and vegetation properties. The flora of Kruger includes over 400 woody species and 200 herbaceous species. The herbaceous layer is dominated by the C4 graminoids, including Aristida congesta, Digitaria eriantha, and Panicum maximum. The most dominant woody encroachers are Dichrostachys cinerea and Combretum apiculatum.

To monitor grass biomass to inform fire management, Kruger established 533 Veld Condition Assessment (VCA) sites throughout the park in 1989.  At each VCA site, grass biomass was measured with a calibrated disc pasture meter every year in April from 1989 to 2008 within a plot of 50 m × 60 m. Plot-level grass biomass estimates were averaged from measurements taken every 2 m along four 50-m transects (i.e., 100 measurements in total), running at 0, 20, 40, and 60 m along the length of the plot. In 2008, a one-time woody plant survey was performed. Plot-level stem density and basal diameter were averaged from two measurements located along each transect (i.e., 8 measurements in total). Measurements for the woody plant survey were located at 30 and 50 m along the first and third transects, and 20 and 40 m along the second and fourth transects. Woody cover for each VCA site was extracted from a 10-m resolution remote sensing product across Kruger based on Sentinel-1 time series for the year 2016-2017 and LiDAR data (Fig. 1). 

During the rainy seasons of 2010 and 2011, soil samples were collected to a depth of 20 cm from 98 VCA sites that were easy to access (Fig. 1). At each site, soil samples were collected at each of four corners and in the middle of the 50 × 60 m plot, and those five samples were then homogenized and subsampled. All soil samples were dried at 60°C and sieved through a 2-mm sieve. A subset of each soil sample (~ 50 g) was used to measure soil texture using a hydrometer, with soil sand, silt, and clay content adding up to 100%. Another subset was ground to a fine powder, acid washed, and used to analyze SOC concentrations and stable carbon isotopic ratios using a Costech ECS 4010 Elemental Analyzer interfaced via a ConFlo III device with a Delta V Advantage isotope ratio mass spectrometer at the Yale Analytical and Stable Isotope Center. The stable carbon isotopic value was expressed as deviations of soil samples from an international standard (Vienna Pee Dee Belemnite) in parts per thousand (‰) using ⸹-notation (i.e., δ¹³C).

SOC dynamics across tropical and subtropical savannas: However, soil samples from Kruger only covered a narrow rainfall gradient (460-700 mm) and were constrained to surface soils. To get a fuller picture of grass-derived SOC across a full gradient of rainfall and to a soil depth of 1-m across global savannas, we used the keywords savanna AND soil AND 13C OR stable carbon isotope, to search relevant literature on Google Scholar on July 17, 2021. We used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol to screen and identify publications to be included in this analysis. Briefly, the title and abstract of each publication were initially screened for eligibility based on whether the publication included soil-stable carbon isotopic measurements within savanna soils. Eligible publications were further filtered based (1) on whether the herbaceous layer was dominated by C₄ grasses, and (2) on whether the publication included at least one soil profile where both soil δ¹³C and SOC values were measured to up to a depth of 1-m. Overall, a total of 148 soil profiles met these criteria. Among these soil profiles, 33 were sampled directly underneath individual tree canopies or woody patches within savannas. The geographic locations of soil profiles are shown in Fig. 1.

For each soil profile, we recorded locations (latitude and latitude), continents (Africa, Australia, North America, and South America), climatic variables (mean annual rainfall [MAP] and temperature [MAT]), elevation, soil sand content, tree cover, soil bulk density, SOC concentration (or storage), soil δ¹³C, and end members for calculating C₃- vs. C₄-derived carbon if reported. When results were presented graphically, we used WebPlotDigitizer 4.4 to digitize the data. If these data were not reported, we contacted the corresponding authors for additional information. Otherwise, MAT and MAP were extracted from the WorldClim database (https://www.worldclim.org/); soil sand content was extracted from SoilGrids database (https://soilgrids.org/); elevation was extracted from the Global Multi-resolution Terrain Elevation Data 2010 (https://earthexplorer.usgs.gov/); fire frequency (2000-2019) was extracted from MODIS Burned Area Product (https://lpdaac.usgs.gov/products/mcd64a1v006/); and tree cover was extracted from the Global Tree Cover 2010 (https://glad.umd.edu/dataset/global-2010-tree-cover-30-m).

SOC storage following woody encroachment: To quantify changes in SOC storage following woody (or forest) encroachment throughout the whole 1-m soil profile across tropical and subtropical grassy ecosystems, we used the key words woody encroachment OR woody thickening OR woody invasion AND soil profile OR soil column OR deep soil AND carbon, to search the relevant literature on Google Scholar on October 18, 2021. Similarly, we used the PRISMA protocol to screen and identify publications to be included in this analysis. Eligible publications were identified based (1) on whether SOC storage was measured from woody encroached and non-encroached sites, including the conversion of grasslands to woodlands/forests (hereafter encroached grasslands) and the thickening of savannas to woodlands/forests (hereafter encroached savannas), and (2) on whether SOC storage was measured throughout the whole 1-m soil profile. Overall, 44 paired-soil profiles with and without woody encroachment have been collected across tropical savannas and grasslands (Extended Data Table 3). Likewise, we extracted SOC storage from encroached and non-encroached sites, and all other information outlined in the previous section. The geographic locations of soil profiles related to woody encroachment are shown in Fig. 1.

The dataset comprises a CSV file with different tabs showing soil organic carbon measurements from Kruger National Park and more broadly across tropical and subtropical savannas.