A tree-ring based comparison of Terminalia superba climate–growth relationships in West and Central Africa
Tropical lowland forests are characterized by humid climate conditions with interannual variations in amount of precipitation, length of dry season, and relative humidity. The African tree species, Terminalia superba Engl. & Diels has a large distribution area and potentially incorporates these variations in its tree rings. Tree ring analysis was performed on 60 plantation trees (increment cores) and 41 natural trees (stem disks) from Ivory Coast and the Congolese Mayombe Forest. Natural forests and old plantations (50–55 years) showed similar growth patterns. Regional chronologies were developed for the two sample regions and showed a long-distance relationship for the period 1959–2008. Growth in the Mayombe was associated with early rainy season precipitation, but no relation was found between tree growth and precipitation in Ivory Coast. Congolese trees possibly show a higher climate-sensitivity than Ivorian trees, because precipitation in the Mayombe is more limiting, and Congolese T. superba trees are found closer to the margins of their distribution. Likewise, tree growth in the Mayombe was also influenced by the SSTs of the Gulf of Guinea and the South Atlantic Ocean during the early rainy season. However, tree growth was influenced by ENSO in both regions. In the Mayombe, La Niña years were associated with stronger tree growth whereas in Ivory Coast, El Niño years corresponded with stronger tree growth. The presented relation between ENSO, precipitation and tree growth is original for equatorial African forests, suggesting an influence of global climate variability on tree growth.
KeywordsAfrica Dendroclimatology ENSO Sea surface temperature Tree rings Tropical forest
Tropical climates are characterized by high temperatures, humidity and precipitation, light frosts and an average temperature of at least 18 °C during the coolest month (McKnight and Hess 2000). However, tropical climates are far from homogeneous and large differences in precipitation periodicity, amount of precipitation, air humidity and annual temperature occur, related to the distance from the equator, ocean and altitude (McKnight and Hess 2000; van Oldenborgh and Burgers 2005). Within the Köppen–Geiger climate classification, these differences result in three tropical climate subtypes: equatorial, monsoon and savannah climates (Köppen and Geiger 1930). Notwithstanding these climatic differences, several tree species have a broad distribution area within the tropics and are able to grow in diverse climate types. Furthermore, they can grow in a large range of local site conditions, including dense mixed forest stands and widely spaced man-made plantations (CTFT 1983). Light-demanding species such as Terminalia superba Engl. & Diels (Groulez and Wood 1985) are expected to be sensitive to such variability in climate and site conditions. The link between light-demanding trees and significant climate correlations was already presumed in the 1920s (Douglass 1920).
Dendrochronology in tropical regions has a history of more than 100 years (Worbes 2002), but a lot more tree-ring research was conducted in tropical forests of Latin America and Asia than in Africa. Several studies on different species resulted in tree-ring chronologies that were verified with radiocarbon dating (Worbes et al. 2003) or with relations to climate (Bräuning et al. 2009; Brienen and Zuidema 2005; Dünisch et al. 2003; Pumijumnong et al. 1995; Worbes 1999). For the African continent, tree-ring chronologies were mostly developed for semi-arid savannas (Fichtler et al. 2004; Gebrekistos et al. 2008; Gourlay 1995), the miombo woodlands in southern Africa (Stahle et al. 1999; Trouet et al. 2001, 2006, 2010) and Ethiopia (Couralet et al. 2005; Sass-Klaassen et al. 2008; Wils et al. 2011). However, the number of exactly dated tree-ring chronologies from West and Central African species is limited (Couralet et al. 2010; Schöngart et al. 2006; Worbes et al. 2003), despite clear potential as pointed out by the exploratory research by Hummel (1946) and Détienne and Mariaux (1970, 1975, 1976, 1977). For the tropics, especially the work of Détienne and Mariaux was important because they showed the existence of annual tree rings in different climates and countries by applying cambial wounding, phenological observations, and monthly diameter measurements. More recently, Tarhule and Hughes (2002) published a list of West African species with dendrochronological potential. Only 7 out of more than 70 species were labelled ‘potentially useful’, indicating that potential exists but the choice of study species is limited.
The majority of dendroclimatological studies in tropical regions focus on one climate type and compare tree growth of different species within this climate, e.g. Devall et al. (1995) and Brienen and Zuidema (2005). Still, when tree ring analysis is applied and results in long time-series, it offers a powerful tool to analyze the responses of trees, species and stands to different and changing climatic conditions. Detailed knowledge on the growth patterns of tropical tree species can thus be obtained, that is necessary for reforestation programs and REDD + files (reduced emissions from deforestation and degradation). In addition, tree ring time-series enables the development of dated proxies for climate reconstruction (Schöngart et al. 2006; Therrell et al. 2006) and can extend instrumental climate time-series that are often sparse in tropical regions. The comparison of those climatic reconstructions based on tree rings with atmospheric circulation patterns, sea surface temperatures (SSTs) and the El Niño—Southern Oscillation (ENSO) provides a useful bridge between past and future trends in global climate change and its implications for human welfare and socio-economic development (Boninsegna et al. 2009).
The interannual precipitation variability in West Central Africa is extremely complex, contrary to the rest of tropical Africa, where the variability is coherent over very large sectors (Balas et al. 2007). Research on the causes of this variability has been focused on teleconnections with tropical SSTs rather than on land–atmosphere interactions (Joly et al. 2007). The SSTs of the equatorial Pacific (Balas et al. 2007; Camberlin et al. 2001; Joly et al. 2007; Paeth and Friederichs 2004), the tropical Atlantic in general (Camberlin et al. 2001; Paeth and Friederichs 2004), and the Benguela Coast including the Gulf of Guinea in particular (Balas et al. 2007; Paeth and Friederichs 2004; Joly et al. 2007) are among the most important factors governing the interannual variability in western and equatorial Africa. The influence of these oceans is seasonally dependent (Balas et al. 2007) and resulting relations also vary between studies based on the season, the time span, the study region, and the methods selected for analysis.
Moreover, the abovementioned anomalies of SSTs of the equatorial Pacific Ocean are generally used to calculate ENSO indices (Trenberth 1997). ENSO is the result of the coupling between oceans and the atmosphere (Nicholson and Entekhabi 1987), but most ENSO indices, including the Niño3.4 index, are based on anomalies of SSTs in the region between 5°N–5°S and 120°–170°W (Trenberth 1997). Camberlin et al. (2001) studied the gridded response of African precipitation on Niño3 indices and found no consistent large-scale relation in West Africa. Surprisingly, they observed a clear negative relation in March–June and August–November in western equatorial Africa [Cameroon, Gabon, Republic of Congo and the coastal region of the Democratic Republic of Congo (DRC)], a region which had never before been shown to exhibit significant correlations with ENSO.
Planted and natural T. superba trees reveal different growth patterns due to different site conditions and/or different responses to climate.
Trees of T. superba show cambial growth in one hemisphere, but not in the other hemisphere and vice versa due to the shift of 6 months in the precipitation seasonality.
This means that trees react sensitively to climate at one site, but do not incorporate these climate conditions at the other site and vice versa. Climate conditions can include regional climate variability, and can be influenced by SST anomalies in the surrounding tropical ocean sectors as well as by ENSO events.
Materials and methods
All study sites belong to the Guineo-Congolian regional centre of endemism (White 1983). The Ivorian study sites are classified as evergreen moist rainforest, but the Congolese study sites are situated within a drier semi-evergreen rainforest.
Characteristics of climate stations around Ivorian study sites and the Mayombe (station, precipitation, dry season (<60 mm), distance to the closest study site)
Annual precipitation (mm)
Distance to closest study site (km)
25 (Bin Houye)
5 (natural forest of Luki)
Summary of the details of the sampling campaigns in the Mayombe and Ivory Coast
Number of sampled trees
Tree ring analysis
Before measurement, all tree ring boundaries were marked with pencil under a stereomicroscope or magnifying glass. On stem discs, three radii were chosen and every tenth ring was followed along the circumference to check for ring anomalies. In case of anomalies (false rings and particularly wedging rings), every tree ring in this zone was followed separately. When unclear, the particular ring number(s) were archived to facilitate corrections during subsequent tree-ring analysis. Ring widths were measured to the nearest 0.01 mm using a stereo-microscope and a Lintab measuring device with TSAP-Win software (Rinn 2003).
The age of plantation trees was compared with the planting period if two conditions were fulfilled: at least one of the two wood cores included the pith and the trees were sampled within the fixed planting scheme of 8 × 12 m. Outside of this scheme, T. superba trees could be regenerated naturally or planted later. If the age corresponded with the planting period, tree rings were considered annual (Stahle 1999).
Temporal correspondence between the various tree ring series was checked by cross-dating (Douglass 1941). Cross-dating is a technique that ensures that each individual tree ring is assigned to the exact year of its formation. This is accomplished by matching patterns of wide and narrow rings between cores/radii from the same tree. Subsequently, the tree ring series of each tree are merged into an average tree ring series and then cross-dated with the other trees of the same site. Successful cross-dating of several trees indicates that there is a common external factor influencing the growth of these trees (Cook and Kairiukstis 1990; Worbes 1995). Those trees can then be averaged to a site chronology or even to a regional tree-ring chronology.
Two statistical parameters are taken into account to evaluate the success of cross-dating: the Pearson’s correlation (r) and the t value of Baillie–Pilcher (Baillie and Pilcher 1973). Another non-statistical parameter is Gleichläufigkeit (Eckstein and Bauch 1969) or percentage of parallel run (ppr), which reflects the percentage of oscillations in the same direction within the overlapping interval. Still, especially for tropical trees, visual control and expert knowledge of Terminalia wood anatomy and tree rings are indispensable. The work of Trouet et al. (2010) is referred to for a detailed explanation on the lower thresholds for ppr (≥60 %) and t values (≥2) in tropical trees compared to temperate regions. Autocorrelation (AC) was checked to filter possible age trends and standardize the tree ring series to enable dendroclimatological analysis (Fritts 1976). For site and regional chronologies, series were standardized using a one-sided moving average over 5 years (Baillie and Pilcher 1973). The mean sensitivity (MS) of these standardized chronologies reflects the sensitivity of tree growth to a common external factor (Fritts 1976). For each site chronology, the expressed population signal (EPS) was also calculated (Haneca et al. 2005; Wigley et al. 1984) to quantify uncertainty due to decreasing sample replication. As such, the EPS indicates how well the finite-sample chronology matches the theoretical population chronology. A value of 0.85 was considered as a reasonable limit for a reliable chronology (Wigley et al. 1984).
Analysis of climate–growth relationships
A climatological year was considered as a sequence of a dry and a rainy season. As such, climatological years for the Mayombe Forest were delimited from Mayn−1 to Mayn where n is the calendar year to which the tree ring is assigned. For western Ivory Coast, years were delimited from Decembern−1 to Decembern. First, annual, seasonal (dry/rainy season) and monthly precipitation data were correlated with regional tree-ring chronologies. Depending on the resulting Pearson’s correlations, months were grouped to test the influence of the early rainy season, transition from dry to rainy season, etc. Precipitation data were derived from the aforementioned climate stations (Table 1; Fig. 1). In DRC, the regional chronology (plantation and natural forest chronologies combined) was correlated with precipitation data of the Luki station. The regional tree-ring chronology for western Ivory Coast was correlated with the mean precipitation data of six surrounding climate stations. Both growth–precipitation relations were calculated over the common time span limited by the length of the regional chronologies and precipitation time series (1959–1996).
We conducted spatial correlation analyses to study the influence of SSTs on precipitation and tree growth in the two regions. Spatial correlation maps were generated using the KNMI explorer (van Oldenborgh and Burgers 2005, Trouet and van Oldenborgh (2013) http://climexp.knmi.nl) and were based on gridded 5 × 5 monthly and seasonal SST fields (Hadley Centre HadSST3; Kennedy et al. 2011a, b). Correlations were calculated over the period from 1959 onwards. These correlation maps give a clear view on which oceanic regions influence precipitation patterns and tree growth.
Furthermore, we used the SST indices for the Niño3.4 region (Trenberth and Stepaniak 2001) to study ENSO influence on regional precipitation patterns and tree growth. For this purpose, we calculated spatial correlation maps between the annual (July–June) Niño 3.4 index and gridded 1 × 1° monthly and seasonal precipitation fields (CRU T.S3.0; Mitchell and Jones 2005). ENSO signals in the regional tree-ring chronologies were analyzed following the method described in Schöngart et al. (2004, 2006). Two sample tests between tree growth in ENSO years and other years were performed (Schöngart et al. 2004, 2006). In this analysis, ENSO years were defined by 5-month running means of SST anomalies in the region of 5°N–5°S and 120°–170°W exceeding +0.4 °C for 6 or more consecutive months (El Niño years) or −0.4 °C for 6 or more consecutive months (La Niña years) (Trenberth 1997).
Historical phenological data were available for the Congolese Mayombe study site (Couralet 2010). From 1948 to 1957, phenological data on 3,750 woody plants were collected in Biosphere Reserve of Luki. Patterns of defoliation, flowering, fruiting and dissemination of fruits were studied, also for T. superba (n = 184 trees). These data were compared to precipitation and tree-ring chronologies within the overlapping time span. No phenological data were collected or available for western Ivory Coast.
Tree-ring related wood anatomy, ring anomalies and the use of wood cores
Marking tree-ring boundaries was rather straightforward in young trees (<30 years), from both regions. Occasionally, the first rings showed intra-annual density variations but they were not considered problematic on stem discs because they were not present on the whole circumference. After a period of fast growth with wide tree rings, ring width drops significantly in the most recent rings. This decrease usually occurs gradually and could be linked to the age trend. False rings and double rings were easy to detect but wedging rings formed the main problem in identifying annual tree rings in T. superba. Wood cores from the planted trees in the Mayombe offered the opportunity to quantify these ring anomalies. A tree-ring chronology of stem discs from the natural Mayombe forest was developed (see below) and subsequently used to test through cross-dating if tree-ring series from wood cores were accurate enough to construct a tree-ring chronology. More than 28 % of the measured wood cores showed no false or wedging tree rings upon comparison with the natural forest chronology. Those wood cores formed the foundation of a plantation chronology that spans from 1959 to 2007. On average, per tree, 1 false ring (3 %) occurred and 0.5 ring (1 %) was at least partially missing. In the latter case, the tree ring was visible on one wood core of a tree but not on the other one. Analysis of the exact location of tree-ring anomalies shows that 45 % of the anomalies were found between the age of 21 and 30 years (0.15 anomalies/ring). In sections over 30 years, the number of anomalies decreased to 21 % or 0.11 anomalies/ring (31–40 years), thus the most recent formed rings were not necessarily more sensitive to anomalous rings. However, plantation trees were not older than 55 years and wedging rings also caused many problems in 9 stem discs >100 years. Unfortunately, quantifying these wedging rings was less straightforward due to the low growth rates and therefore, low visibility.
Growth patterns in natural and planted African forests
Site and regional chronology characteristics of natural forests in Ivory Coast (IC) and the Mayombe (n samples, diameter, number of tree rings, mean growth, MS (mean sensitivity), AC (auto-correlation), time span, Pearson correlation, EPS (expressed population signal)
Natural forest IC
Natural forest Mayombe
Total n sampled trees
No. of samples in chronology
22 (81 %)
28 (47 %)
11 (92 %)
Mean diameter (cm)
56 ± 11
41 ± 12a
57 ± 15
Mean no. of tree rings
46 ± 34
40 ± 6
47 ± 38
Time span chronology
Mean growth of chronology (mm)
4.53 ± 1.64
5.45 ± 2.60
7.19 ± 2.67
MS after standardisation
Ppr (percentage of parallel run), t value Baillie–Pilcher and Pearson’s correlation coefficient for cross-dating of local chronologies in the Mayombe and Ivory Coast
t value Baillie–Pilcher
Tshela vs. Luki
6 vs. 5
Tshela vs. Monzi
6 vs. 28
Luki vs. Monzi
5 vs. 28
Bin Houye vs. Danane
4 vs. 6
Bin Houye vs. Goya
4 vs. 6
Bin Houye vs. Scio
4 vs. 7
Danane vs. Goya
6 vs. 6
Danane vs. Scio
6 vs. 7
Scio vs. Goya
7 vs. 6
Table 3 shows that the percentage of trees included in the stem disc-based chronologies is higher than in the wood core-based chronology. In the Mayombe, the mean growth of natural T. superba trees is remarkably higher than for plantation trees (p < 0.01). For comparable diameters, T. superba trees from the Mayombe grow faster (p < 0.001) than Ivorian trees. Plantation and natural forest chronologies from the Mayombe region showed high autocorrelations, indicating a clear age trend. The Ivorian chronology has a lower AC and therefore a higher MS. After standardization, AC was close to zero for all chronologies. All EPS values approach the threshold of 0.85.
Response of T. superba growth to interannual climate variability
Visualization of two sample tests evaluating the influence of ENSO events on tree growth
Tree-ring related wood anatomy, ring anomalies and the use of wood cores
Mariaux (1969) was the first to describe annual tree rings in T. superba. His description was based on young trees with a lot of juvenile wood. Mature tree rings appear to differ substantially. In trees older than 30 years, bands of parenchyma were less abundant and the marginal parenchyma (if present) was not always continuous. However, the large variations between years that Mariaux (1969) reported were also found in our study. In some years, there is a vessel-less area at the start of a growing season. Also, the normal anatomical structure within a ring can be inversed, as was the case in some Ivorian tree rings. Clearly, the anatomical tree-ring structure can differ in T. superba from different regions.
Tree rings were easier to detect in the Mayombe compared to Ivory Coast. This is potentially related to a difference in leaf shedding behaviour in the two regions and the generally drier conditions of the Mayombe. In the Mayombe, defoliation is simultaneous in the late dry season (up to 50 % of T. superba trees lose their leaves in September (Couralet 2010)) and complete while in Guinea (adjacent to Ivory Coast), the leafless period occurs not simultaneously in every tree, is shorter in time and less complete (CTFT 1959). Personal observations confirm that defoliation in the DRC is complete in June–July for all trees of one whole plateau or valley while in Ivory Coast, it differs strongly between individual trees: trees with new leaves, leafless trees and trees that are still shedding leaves are found at the same site at the same time.
Problems to define tree rings were caused by the first unclear juvenile tree rings, false rings, double rings and especially ring wedging (Mariaux 1969). Worbes (2002) also classified Terminalia as a taxon that tends to form wedging rings. Ring wedging has been noted in many tropical tree species (e.g., Trouet et al. 2010; Tarhule and Hughes 2002), but the exact cause for this phenomenon is unknown. Our study confirms that wedging mostly occurs in the outer parts of the tree circumference, when the tree starts to grow slower due to its inherent growth trend and forms buttresses (Worbes 2002). Buttresses become larger when tree diameters increase (De Ridder et al. 2010), and could also cause wedging. Worbes (2002) related ring wedging to trees that grew under poor light conditions and under competition. T. superba, abundantly present in fallows and forest gaps, mostly receives full light during its juvenile phase. Afterwards, when canopy closure occurs or trees of the climax forest reach their full height, competition for light could cause wedging, especially in the most recent tree rings. To answer Worbes’ hypothesis on competition, more ecological data on T. superba and the accompanying tree species need to be collected within the sampled forests.
Because of the frequent occurrence of wedging rings and its interference with cross-dating, Worbes (2002), Trouet et al. (2006), and Brienen and Zuidema (2005) recommend the use of stem discs rather than wood cores for tropical tree ring analysis. Our study shows that the percentage of cross-dated samples in our wood-core based plantation chronology is significantly lower than in the stem disc-based chronology of the natural forest. However, there were three main reasons why plantation trees were excluded from the chronology: (1) trees had an eccentric centre (the orientation of wood cores is difficult in this case), (2) wood cores were broken, resulting in invisible tree rings due to cracks, or (3) tree ring series from opposing cores from the same tree did not cross-date. Tree ring anomalies were the main cause for omitting tree ring series in the latter case. However, when a first chronology can be built from stem discs with accurate mapping and documenting of ring anomalies (especially wedging and false rings), tree ring series of wood cores can be checked and implemented in or omitted from this existing chronology. In the tropics, Schöngart et al. (2006), Stahle et al. (1999) and Worbes (1999) also successfully analyzed tree growth using only wood cores and one or a few stem discs. Moreover, sampling wood cores are less destructive and substantially facilitate sample transport.
Growth patterns in natural and planted African forests
In the Luki Reserve (DRC), Couralet et al. (2010) found between-trees correlations for understory trees that were similar in magnitude to our results. The correlation strength in our study was also in accordance with the findings of Bräuning et al. (2009) and Brienen and Zuidema (2005), lower than those of Stahle et al. (1999) and Therrell et al. (2006), and higher than the correlations of Trouet et al. (2006, 2010). Auto-correlation is high compared to other tropical tree species (Bräuning et al. 2009; Couralet et al. 2010; Trouet et al. 2006, 2010) but not unusual because T. superba has a strong and known age trend, colonizing fallows with fast growth during the first years and strongly decreasing in growth after this juvenile phase.
Fifty-year-old plantations and 40- to 50-year-old natural forest of the same tree species in the Mayombe enable an original comparison of growth patterns in natural versus planted trees. Eshete and Stahl (1999) as well as Worbes (1999) used plantation trees but their time-series were too short for chronology development (<25 years). In our study, the overlap between planted and natural trees is 36 years, and planted and natural trees showed similar growth patterns. The studied plantation is a production forest that is not intensively managed, but nevertheless, the natural forests were more diverse with a more explicit competition for light and nutrients. The average growth as well as the initial growth (first 20 years) was lower in the plantation compared to the natural forest. Two extremely dry years (1956 and 1958) and the use of inferior planting material have been shown to explain the loss of plantation trees immediately after the installation (De Ridder et al. 2010) and have likely also influenced the growth of surviving trees.
In temperate regions, reference chronologies are available from a dense network of dendrochronological records and can be used for provenancing and dating wooden objects (Haneca et al. 2009). These chronologies can span more than 10,000 years [e.g., oak (Friedrich et al. 2004)]. Such long time spans are difficult to accomplish in the tropics, but the development of a regional network of species-specific chronologies that include older trees is feasible. T. superba appears to be a suitable tree species for this goal, because it shows distinct tree rings, has a wide distribution range, and displays cross-dating potential throughout a large section of this range. The longest regional chronology here developed only spans 113 years, but the oldest sampled tree was 165 years in the Mayombe and 182 years in Goya (Ivory Coast) and nine trees in total in our dataset covered >100 years. Unfortunately, the sample replication for such long tree-ring chronologies was too low to incorporate the oldest trees in the regional chronologies. Our study shows that wood cores give an accurate indication of the number of tree rings and thus the age of a tree and we suggest to first core trees before cutting for chronology development. With additional samples, it should be possible to reconstruct a regional chronology for T. superba with a length of up to 200 years, which would greatly increase the time span covered by instrumental climate records and thus allow for paleoclimatic analysis.
Our regional chronologies for the Mayombe and Ivory Coast were located more than 2,600 km apart, overlapped in time for 50 years (1959–2008), and showed a weak synchronization. Fichtler et al. (2004) did not find a match between two sites in southern Africa that were 900 km apart and used spectral analysis to reveal similar long-term oscillation patterns. Trouet et al. (2010) cross-dated chronologies from five Miombo woodland sites in Zambia over a distance of more than 1,000 km and found long-distance growth correlations between some sites. To our knowledge, our study is the first to indicate long-distance growth correspondence in rainforests. One hypothesis for these synchronized growth variations could be the presence of a common growing season as Mariaux (1969) confirmed with dendrometer studies. On a limited number of trees, he found that tree growth in both the Mayombe and central Ivory Coast mainly occurred during October–December. Tree growth and November–April precipitation are not directly related in either of the regions, but both tree growth and November–April precipitation showed similar variability. Other relations between tree growth and climate are discussed in detail in the next paragraph.
Response of T. superba growth to interannual climate variability
In the Mayombe, early rainy season precipitation is an important driver of tree growth. This matches with the abovementioned dendrometer measurements in the Mayombe Forest that show the strongest growth during the onset of the rainy season (Mariaux 1969). A positive response to the first rains was also found in one understory species [Aidia ochroleuca (K. Schum.)] in the Mayombe (Couralet et al. 2010) and tropical forests in Bolivia (Brienen and Zuidema 2005) and Thailand (Pumijumnong et al. 1995). In Ivory Coast, there is no important influence of precipitation on tree growth.
Thus, no comparable relation appears to exist between tree growth and precipitation in the two regions, despite long-distance connections for growth on one hand and November–April precipitation on the other hand. Differences in response to climate could be related to climate sensitivity of the trees or the use of short precipitation time series. Regarding climate sensitivity, western Ivory Coast is located in a wetter region than the Mayombe. This could cause a stronger climate signal in the Mayombe following the principle of limiting factors (Fritts and Swetnam 1989). Also, trees near the margins of their climatic distribution have a higher climate-sensitivity (Cook and Kairiukstis 1990). The microclimate in the Congolese Mayombe still enables the growth of tropical trees while in fact, annual precipitation is too limited for their growth (optimal annual precipitation for T. superba is 1,500 mm; Groulez and Wood 1985). In addition, our climate–growth relations were only tested over a rather short but common time span to enable comparison between the two regions (1959–1996). We correlated growth and precipitation from 1959 to 2006 with data from the Luki climate station (Mayombe) to illustrate the influence of longer climate series. In addition to a stronger relation with November precipitation (r = 0.42; p < 0.01), summed October–December precipitation also showed a strong positive relation with tree growth (r = 0.37; p < 0.01). So, more and stronger climate correlations in the Mayombe were found if the most recent precipitation data were included. This difference could not be tested in Ivory Coast (shorter precipitation time series, more stations with lacking data), but we hypothesize that stronger relations with precipitation in Ivory Coast could come out if longer precipitation time series, including the most recent years, were available. This is indirectly confirmed by the ENSO study (see below), which was run for the two regions from 1959 to 2008, resulting in a significant relation between tree growth and ENSO climate indices in both regions.
Based on the local relation between climate and growth in the Mayombe, the regional influence of SSTs on tree growth and precipitation was analyzed with correlation maps, in order to provide a clear view on the influencing oceanic regions. The South Atlantic Ocean, including parts of the Gulf of Guinea, appears the most important driver for the Mayombe precipitation as well as tree growth. Balas et al. (2007) and Paeth and Friederichs (2004) also confirmed the influence of the Gulf of Guinea and the tropical Atlantic Ocean on precipitation in western Central Africa and indicated the ENSO mode as another important factor. In addition to precipitation, the SSTs of the Gulf of Guinea and the South Atlantic Ocean in October–December influenced tree growth in the Mayombe. This was also the case in Benin with significant correlations between tree growth and May–August SSTs from the Gulf of Guinea for four tree species (p < 0.05): Afzelia, Pterocarpus, Daniellia, and Isoberlinia (Schöngart et al. 2006).
Finally, anomalous ENSO conditions were associated with growth effects in both Ivory Coast and the Mayombe, even though local precipitation patterns did not influence tree growth in both regions. So far, the effect of ENSO on African tree growth in two different hemispheres has not been encountered in literature. Because the influence of ENSO on precipitation occurred during the growing season, tree growth could be influenced by ENSO. Schöngart et al. (2006) only found an influence of ENSO on precipitation during the dry season and subsequently, no relationship with tree growth. In general, the Mayombe region appears to be more sensitive to climate variability than Ivory Coast, with higher October–November precipitation and higher growth clearly observed during La Niña years. In Ivory Coast, higher precipitation during El Niño years slightly northward of the sample sites was found during the growing season, probably resulting in higher growth during El Niño years. In literature, most studies in West and Central Africa are focused on the relation between precipitation and ENSO. Camberlin et al. (2001) found a negative correlation between ENSO and precipitation in August–November for Cameroon, Gabon, the Republic of Congo and coastal DRC, confirming the higher precipitation during La Niña years in the Mayombe, as did Paeth and Friederichs (2004). For Ivory Coast, Paeth and Friederichs (2004) did not observe a consistent relationship for West Africa between ENSO and precipitation. However, we compared the actual zones of higher precipitation in October–November during El Niño years in Fig. 7c with the precipitation response map of Camberlin et al. (2001), Fig. 4, resulting in a response type with some dry months during El Niño years but with wet conditions in October–November. Still, this type of response in October–November is not as strong as the response in the Mayombe, both for precipitation as well as growth. Again, the principle of limiting factors (Fritts and Swetnam 1989) and the presence of T. superba trees near the margins of their climatic distribution (Cook and Kairiukstis 1990), could explain the higher climate-sensitivity in the Mayombe.
From the abovementioned studies on precipitation and SSTs, it is evident that results are often linked to the methods used and the area and season studied. The diversity in results underlines the complexity of our study region in its response to seasonal and interannual SST changes and ENSO events. Furthermore, Camberlin et al. (2001) did not observe consistent relationships with any of the SST indices in some parts of Africa, suggesting that precipitation variability responds to other oceanic patterns, or only to local/regional atmospheric dynamics.
Conclusion and perspectives
Although planted and natural forests with T. superba have distinct differences in site conditions such as structure and light competition, both forest types showed significantly similar growth patterns and responses to climate. Cambial growth probably occurs at the same time in planted and natural forests and was influenced by the onset of the rainy season in the Mayombe. Unfortunately, precipitation did not correlate with the Ivorian regional chronology. This is possibly the result of limited precipitation data, rendering the hypothesis on cambial growth shifts with precipitation inconclusive. However, regional climate variability was influenced by SST anomalies in the Mayombe whereas ENSO indices controlled climate variability over both hemispheres in October–November with, on the one hand, higher growth at the northern hemisphere sites during El Niño years and, on the other hand, higher growth at the southern hemisphere sites during La Niña years.
Importantly, if length or quality of precipitation data hampers proper correlation with tree ring data, comparison with large-scale climate factors as SSTs is valuable and can pinpoint at important influences on precipitation patterns and tree growth. Therefore, in tropical regions, where climate data are often scarce, comparison with these SSTs and derived ENSO indices offers interesting possibilities for future dendroclimatological studies in the tropics, even in equatorial Africa.
This research project is funded by a PhD grant (M. De Ridder) of the Flemish Interuniversity Council (VLIR). The fieldwork in Ivory Coast was supported by a grant from the King Leopold III Fund for Nature Exploration and Conservation and the Congolese fieldwork was possible with the help of a grant from VLIR. We are indebted to the Special Research Fund of Ghent University for financing the PhD project of W. Hubau. We would also like to thank the teams of WWF DRC, WWF Belgium, Soforma, the ERAIFT (École Régionale post-universitaire d’Aménagement et de gestion Intégrés des Forêts et Territoires tropicaux), Thanry and Bomaco for their financial support and their guidance throughout the fieldwork. Special thanks goes out to Guy Bayens for all possible help on organizing the Ivory Coast fieldwork and to Laurent Nsenga, Geert Lejeune, Bruno Pérodeau and Prof. Shango Mutambwe, whose efforts were indispensable for the success of the field campaigns in the DRC. Also a warm thank you to the local crew who guided us through the forests of West and Central Africa.
- Baillie MGL, Pilcher JR (1973) A simple program for tree-ring research. Tree Ring Bull 33:7–14Google Scholar
- Boninsegna JA, Argollo J, Aravena JC, Barichivich J, Christie D, Ferrero ME, Lara A, Le Quesne C, Luckman BH, Masiokas M, Morales M, Oliveira JM, Roig F, Srur A, Villalba R (2009) Dendroclimatological reconstructions in South America: a review. Paleogeogr Paleoclimatol Paleoecol 281:210–228CrossRefGoogle Scholar
- Couralet C (2010) Community dynamics, phenology and growth of tropical trees in the rain forest Reserve of Luki, Democratic Republic of Congo. PhD thesis, Ghent UniversityGoogle Scholar
- CTFT (1959) Monographie du limba (Terminalia superba Engler et Diels). CTFT, ParisGoogle Scholar
- CTFT (1983) Bois tropicaux (5ième edition). CTFT, Nogent-sur-MarneGoogle Scholar
- Détienne P, Mariaux A (1970) La périodicité de formation des cernes dans le bois d’okoumé. Bois For Trop 131:37–50Google Scholar
- Détienne P, Mariaux A (1975) Nature et periodicité des cernes dans le bois de Niangon. Bois For Trop 159:29–37Google Scholar
- Détienne P, Mariaux A (1976) Nature et périodicité des cernes dans le bois de samba. Bois For Trop 169:29–35Google Scholar
- Détienne P, Mariaux A (1977) Nature et périodicité des cernes dans les bois rouges de méliacées africaines. Bois For Trop 175:52–61Google Scholar
- Devall MS, Parresol BR, Wright JS (1995) Dendroecological analysis of Cordia alliodora, Pseudobombax septenatum and Annona spraguei in Central Panama. IAWA J 16:411–424Google Scholar
- Douglass AE (1941) Crossdating in dendrochronology. J For 39:825–831Google Scholar
- Dünisch O, Montoia VR, Bauch J (2003) Dendroecological investigations on Swietenia macrophylla King and Cedrela odorata L. (Meliaceae) in the central Amazon. Trees 17:244–250Google Scholar
- FAO (1986) Atlas of African agriculture. FAO, RomeGoogle Scholar
- FAO (2008) Digital soil map of the world. FAO, RomeGoogle Scholar
- Friedrich M, Remmele S, Kromer B, Hofmann J, Spurk M, Kaiser KF, Orcel C, Küppers M (2004) The 12,460-year Hohenheim oak and pine tree-ring chronology from Central Europe—a unique annual record for radiocarbon calibration and paleoenvironment reconstructions. Radiocarbon 46:1111–1122Google Scholar
- Fritts HC (1976) Tree rings and climate. Academic Press, LondonGoogle Scholar
- Gourlay ID (1995) The definition of seasonal growth zones in some African Acacia species—a review. IAWA J 16:353–359Google Scholar
- Groulez J, Wood PJ (1985) A monograph on Terminalia superba. Centre Technique Forestier Tropical & Commonwealth Forestry Institute, Nogent-sur-MarneGoogle Scholar
- Hawthorne WD (1995) Ecological profiles of Ghanaian forest trees. Tropical forestry papers 29, University of OxfordGoogle Scholar
- Humblet P (1946) Aménagement des forêts climatiques tropicales au Mayumbe. Bull Agric Congo Belge 37:15–87Google Scholar
- Hummel FC (1946) The formation of growth rings in Entandrophragma macrophyllum A. Chev. and Khaya grandifolia C. DC. Int For Rev 25:103–107Google Scholar
- Köppen WP, Geiger R (1930) Handbuch der Klimatologie. Verlag von Gebrüder Borntraeger, BerlinGoogle Scholar
- Mariaux A (1969) La périodicité des cernes d’accroissement dans le bois de Terminalia superba. Bois For Trop 128:39–54Google Scholar
- McKnight TL, Hess D (2000) Climate zones and types: the Köppen system. Physical geography: a landscape appreciation. Prentice Hall, Upper Saddle RiverGoogle Scholar
- Pumijumnong N, Eckstein D, Sass U (1995) Tree-ring research on Tectona grandis in northern Thailand. IAWA J 16:385–392Google Scholar
- Rinn F (2003) TSAP-WinTM user reference. Rinntech, HeidelbergGoogle Scholar
- Stahle DW (1999) Useful strategies for the development of tropical tree-ring chronologies. IAWA J 20:249–253Google Scholar
- Tarhule A, Hughes MK (2002) Tree-ring research in semi-arid West Africa: need and potential. Tree Ring Res 58:31–46Google Scholar
- White F (1983) The vegetation of Africa. UNESCO, SwitzerlandGoogle Scholar
- Worbes M (1995) How to measure growth dynamics in tropical trees. A review. IAWA J 16:337–351Google Scholar
- Trouet V, Van Oldenborgh GJ (2013) KNMI Climate Explorer: a web-based research tool for high-resolution paleoclimatology. Tree Ring Res 69:3–13Google Scholar