1 Introduction
In Equatorial Africa, the edge of the evergreen forest massif consists of a mosaic of semi-deciduous forests and savannas. Savannas also occur within the forest massif in small patches but also, as in the Bateke plateaus in Congo, over wide areas. The origins of these open landscapes, which occur under present-day equatorial climate mostly suitable for forest presence, are still a matter of debate (Bayon et al., 2012). Isotope analyses on soil organic matter are widely used to discriminate between C3 (mostly forests) and C4 (mostly savannas) vegetation through δ13C measurements and thus are powerful tools to reconstruct the history of these former landscapes (Boutton, 1996). Although based on a few sites, the soil profiles from Congo (Schwartz, 1997; Schwartz et al., 1992) and Gabon (Delègue et al., 2001) recorded large-scale environmental changes in relation with the Late Quaternary climatic succession in specific areas of the Atlantic Coast and the Bateke Plateaus.
As soon as the late fifties, Bachelier et al. (1957) studied soils from the savanna–forest mosaic from the lowlands of central Cameroon. They concluded that the anthropogenic activities and mainly repeated burnings were responsible for the recent opening up of the forest. However, Youta Happi (1998) and Guillet et al. (2001) showed that the forest has been currently expanding since, at least, the last decades.
In this article, we present isotope analyses on six soil profiles recovered in central Cameroon along a south-north transect from Abong Mbang to the southeast (3°50′N–13°20′E) and Ngambe Tikar to the north-west (5°40′N–11°40′E), in an area where the lack of lacustrine or swamp deposits hampers our understanding of past environmental changes. The goal of our study is to discuss the changes in forest extents during the recent past, particularly at the age of the Late Holocene forest crisis evidenced throughout central Africa (Vincens et al., 1999). These changes are thought to have strongly influenced the local populations (Bayon et al., 2012; Verdu et al., 2009).
2 Materials and methods
2.1 Soil profile location
Two soil profiles were taken in semi-deciduous Sterculiaceae/Ulmaceae forests (at Abong Mbang and Ngambe Tikar) and four profiles were taken in savannas (at Ndjolé, Sangbé, Maboen, and Nditam) (Fig. 1). The savannas are of Guineo-Congolian phytogeographical type (White, 1983) and are characterized by tall grasses, mainly Hyparrhenia sp. with shrubs and trees. (Lophira lanceolata, Annona senegalensis, Bridelia ferruginea, Bauhinia thonningii, Terminalia sp.). All the studied soils were located in the highest part of the relief, and classified as oxisols, with texture varying from clayey to sandy-loamy.
Location map of the study sites. The simplified vegetation map is from Guillaumet et al. (2009).
Carte de localisation des sites étudiés. La carte de végétation simplifiée est reprise de Guillaumet et al. (2009).
2.2 Soil sampling
Soils were sampled using a 6-cm-diameter auger, to a depth of 390 cm, at 10-cm intervals between the surface and 50 cm, and at 20-cm intervals from 50 cm to the bottom. The soil samples were air-dried, sieved with a 2-mm mesh, homogenized and ground to < 200 μm. The forest and savanna core sites were chosen in low declivity zones, within undisturbed vegetation apart from the present savanna–forest boundary.
2.3 Stable isotope analyses
Since C3 plants (the dominant plants in forests) and C4 plants (the dominant plants in savannas) are isotopically distinct, it is possible to detect shifts from tropical forests to grasslands (or vice versa) from the δ13C signature of organic matter in soils (Desjardins et al., 1996; Mariotti, 1991; Runge, 2002; Sanaiotti et al., 2002; Schwartz et al., 1996).
The isotopic ratio (R = 13C/12C) is reported in standard delta notation (δ13C), defined as parts per thousand (‰) deviation from a standard (Vienna Pee Dee Belemnite) (Peterson and Fry, 1987):
The samples were air-dried and sieved at 2 mm. Aliquots of these samples were ground at 70 meshes and submitted to an acid attack by HCl 10%. The samples were combusted in an elemental analyzer (Costech) for carbon content analysis. δ13C was measured on a continuous-flow gas-ratio mass spectrometer (Finnigan Delta PlusXL) coupled with the elemental analyzer. Standardization is based on acetanilide for elemental concentration, NBS-22 and USGS-24 for δ13C. Precision is better than ± 0.10‰ for δ13C, based on repeated internal standards.
2.4 Radiocarbon dating
The radiocarbon measurements performed on the bulk soil organic matter represent a mixture of younger and older carbon pools whose residence time is increasing with depth and consequently do not correspond to the absolute age of soil organic matter (Guillet et al., 2001; Schwartz et al., 1996). As a consequence, the temporal resolution of past vegetation changes is low and must be interpreted with caution (Boutton et al., 1998; Trumbore et al., 1995; Wiedemeier et al., 2012). In addition, the calibration of 14C ages does not make sense for soil samples. For an easier comparison with earlier studies, 14C values are expressed in 14C yr BP (apparent age according to Guillet et al., 2001). AMS 14C measurements were performed on bulk soil organic matter samples with a 3 MV Pelletron (NEC, Middleton, Wisconsin, USA). The sample pre-treatment was the same as for the other analysis.
3 Results
3.1 Total organic carbon content
The six studied profiles showed a relatively similar distribution of organic carbon with depth (Fig. 2). In the surface (0–10 cm) layer, the organic carbon (C) content ranges from 18.8 to 27.9 mg·g−1. Then, the C content decreases abruptly until the 60–70-cm level, without evidence of any organic paleohorizon: at this depth, the C content ranges from 4.7 to 10.3 mg·g−1. Below this level, the C content decreases slightly to reach 1.6 to 4.1 mg·g−1 at 380–390-cm depth.
Variation with depth of the total carbon content for the six studied soil profiles (circles represent forest profiles and squares represent savanna profiles).
Variation de la teneur en carbone total en fonction de la profondeur pour les six profils de sol étudiés (les cercles représentent les profils de forêt et les carrés les profils de savane).
3.2 Depth variability in δ13C values of soil organic matter
Soil profiles from semi-deciduous forests (Fig. 3) show a typical δ13C pattern with values ranging from –28.4 to –26.9‰ at the surface, an increase of about 2‰ in the sub-surface levels between 10 and 30 cm, and then a slight increase until values ranging from −23 to −25.4‰ between 200 and 390 cm.
Variation of δ13C values with depth in the studied soil profiles (circles represent forest profiles and squares represent savanna profiles).
Variation du δ13C en fonction de la profondeur dans les profils étudiés (les cercles représentent les profils de forêt et les carrés les profils de savane).
δ13C values range from a minimum of −16.4 to a maximum of −13.9‰ in the surface layer from the savanna sites (Ndjolé, Sangbé, Maboen, Nditam). An increase of 1.4 to 1.6‰ is observed at the uppermost decimeters, with a maximum recorded at 20–30 or 30–40 cm (Fig. 3). Downward, a pronounced decrease is recorded, with δ13C values falling to −19.1 and −17.6‰ at 120–130 cm. Two of these profiles record a stabilization of the 13C values between 120–130 cm and 380–390 cm (at Sangbé, between −17.6 and 19.8‰, and at Maboen, between −17.2 and −18.2‰), while in the two other sites (Ndjolé and Nditam), the δ13C values decrease progressively until −21.9 and −22.3‰ at 360–370 cm.
3.3 14C dating of soil organic matter
As the incorporation of carbon fixed from the atmosphere since atomic weapons testing in the early 1960s, which nearly doubled the amount of 14C in the atmosphere, high values of radiocarbon (> 100% of the modern carbon) are generally found in the surface soil layers. No dating of surface layers was realized. However, even at the 20–30-cm level, three profiles show a modern 14C signature and the three other ages vary between 180 and 485 14C yr B.P (Table 1). Below this level, the radiocarbon ages increase downward at different rates: at 400-cm depth, Abong Mbang and Ndjolé profiles record radiocarbon ages of 3900–4050 14C yr BP, whereas ages from the other profiles are older and range from 5900 to 7000 14C yr BP. There is no differentiation between the forest and savanna profiles.
14C de la matière organique du sol dans les six profils, exprimé en pourcentage de carbone moderne (pMC) et en âge radiocarbone conventionnel en années (non calibré).
| Abong Mbang CAM-10-2-07 Forest 3°49′52,6′′N 13°19′53,9′′E |
Ngambe Tikar CAM-10-2-17 Forest 5°39′01,6′′N 11°28′58,0′′E |
Ndjolé CAM-10-2-09 Savanna 4°51′09,4′′N 11°58′35,9′′E |
Sangbé CAM-10-2-10 Savanna 5°59′16,3′′N 12°23′35,0′′E |
Maboen CAM-10-2-14 Savanna 5°35′04,8′′N 11°50′25,7′′E |
Nditam CAM-10-2-18 Savanna 5°27′06,1′′N 11°22′06,8′′E |
|||||||
| Depth (cm) | pMC | Radiocarbon age BP | pMC | Radiocarbon age BP | pMC | Radiocarbon age BP | pMC | Radiocarbon age BP | pMC | Radiocarbon age BP | pMC | Radiocarbon age BP |
| 20–30 | 94.16 | 485 | 97.80 | 180 | 100.30 | Modern | 101.17 | Modern | 10.14 | Modern | 94.88 | 420 |
| 80–90 | 84.12 | 1390 | 85.20 | 1285 | 83.70 | 1430 | 84.09 | 1390 | 73.06 | 2520 | 78.61 | 1935 |
| 160–170 | 75.52 | 2255 | 77.21 | 2080 | 75.54 | 2255 | 65.71 | 3375 | 56.80 | 4540 | 63.44 | 3655 |
| 260–270 | 64.47 | 3525 | 46.33* | 6180* | – | – | 57.66 | 4425 | 46.40 | 6165 | 48.63 | 5790 |
| 380–390 | 60.56 | 4030 | 45.44 | 6335 | 61.49** | 3905** | 47.78 | 5930 | 42.18 | 6935 | 44.16 | 6565 |
* 280–290 cm.
** 360–370 cm.
4 Discussion
As far as the total carbon content is concerned, no significant difference is observed between the forest and savanna soils. Our results show similar carbon content and distribution compared to other African oxisols with clayey or median texture (Brunet et al., 2007). Total carbon decreases in an exponential manner (Bennema, 1974).
Desjardins et al. (1991), Martinelli et al. (1996), Gouveia et al. (1997) and Pessenda et al. (1998) among others have already observed that the δ13C of soil organic matter increases with depth in forest profiles. It is accepted that a 3.5–4.0‰ isotopic enrichment is due to mineralization and humification decomposition processes (Desjardins et al., 1996; Mariotti and Petershmitt, 1994; Martinelli et al., 1996). The δ13C values observed in the forest soil profiles of Abong Mbang and Ngambe Tikar lie within the range of expected values for soils where C3 plants have remained the dominant vegetation cover over time. Similar values have been recorded by Schwartz et al. (1992) and Guillet et al. (2000, 2001) in stable forests from Central Africa.
The δ13C values measured in the upper layers of the savanna profiles are typical of C4-dominated vegetation. They closely match values from savannas already encountered in Gabon, Congo and Cameroon by Delègue et al. (2001), Schwartz et al. (1992), Schwartz (1997) and Guillet et al. (2001) (−12 to −17‰). At these later sites, the δ13C values changed from −22 to −24‰ between 2 m and 4 m depth, reflecting, according to the authors, the presence of an organic matter inherited from former C3-dominated vegetation. In our savanna profiles of Ndjolé and Nditam, quite similar δ13C trends are observed, with values reaching from −21 to −22‰ at the bottom of the profiles. Although these values are slightly less negative than those observed by Delègue et al. (2001), Schwartz et al. (1992), and Schwartz (1997), they likely reflect dominant C3 vegetation. The savanna sites of Sangbé and Maboen show a strong decrease in δ13C values between 30 and 130 cm. However, the δ13C values stabilize downward, around −18 to −20‰ and −17 to −18‰, respectively, which testify the former presence of mixed vegetation with C3 plants and C4 plants. These intermediate δ13C values are not reported in the literature and, to our knowledge, no corresponding vegetation landscape exists in central Africa today, except at a single site, located at the forest–savanna boundary at Kandara (Guillet et al., 2001). Schwartz (1991) and Guillet et al. (2001), also suggested, based on almost similar δ13C values, the presence of a savanna more densely forested than today. Similar 13C values from −19.7 to −20.5‰ were also recorded between 0 and 2 m depth in the southeastern region of Brazil, on a native Cerrado vegetation site (Wilcke and Lilienfein, 2004). This Cerrado is grassland with up to 40% cover of 3 to 5 m tall trees.
Another interpretation of such δ13C values has been proposed by Schwartz (1991): they could represent an average δ13C from different vegetation types, which have succeeded over time in relation with paleoclimatic fluctuations.
The age distribution of the analyzed organic matter is not significantly different from those reported in central Atlantic Africa. In Gabon, Delègue et al. (2001) mentioned a mean age of 4200 14C yr BP at 180 cm. In Cameroon, Guillet et al. (2001) mentioned a mean age of 2640 and 4855 14C yr BP at 200 cm and Schwartz et al. (1992) mentioned ages slightly older: 6960 14C yr BP at 200 cm and 8300 14C yr BP cm at 300 cm in Congo.
Because of the heterochrony of the soil organic matter, these mean ages do not indicate directly the age of vegetation changes. However, they give an indication of the timing of such changes (Sanaiotti et al., 2002). Based on our 14C data, we can deduce that the extension of savannas in central Cameroon did not begin before the Late Holocene (ca 4000–3500 14C yr BP). This is in agreement with pollen studies, which record the collapse of the forest massif from 3300 to 2400 cal yr BP (Lézine et al., 2013 this volume and reference therein). Further north in the Adamaoua region, Vincens et al. (2010) showed that the semi-deciduous forest begun to decline earlier at 6000 cal BP.
At Ndjolé and Nditam, the δ13C profiles indicate that the former vegetation was a forest of semi-deciduous type, which was present at the Early to Middle Holocene. Differently, at this time, the isotopic values at Sangbé and Maboen show a more complex landscape formed by a mixture of C3 and C4 plants.
5 Conclusion
We obtained for the first time Holocene paleoenvironmental data in the savanna–forest mosaic zone of central Cameroon. Our results indicate that this modern landscape is mostly inherited from the last environmental crisis, which led to a widespread forest collapse in the Atlantic Equatorial Africa at the end of the African Humid Period 4000–3500 years ago. A stable environment is recorded in the present-day forests of Ngambe Tikar and Abong Mbang. The savannas soil profiles display a more complicated picture with δ13C values pointing to specific environmental conditions which are not known today: either densely forested grasslands or an alternation of successive savanna–forests phases at the sample sites. More profiles are required to understand better this history.
Acknowledgments
This study has been carried out in the frame of C3A project (ANR 09-PEXT-0). We thank the IRD representation at Yaoundé (Cameroon), René Tsiquia and Valentin Angoni for their assistance in the field. We are grateful to the Laboratoire de mesure du carbone 14 (LMC14)–UMS 2572 for AMS dating and to NSF Arizona for carbon and δ13C data. TD, BT, MMY, FC are funded by IRD, AML by CNRS, JPN by Ngaoundéré University and GA by the National Herbarium of Yaoundé.