The study was conducted in the Yangambi Biosphere Reserve (YBR), located centrally in the Congo Basin, within the territories of Isangi and Banalia, Tshopo province in the Democratic Republic of Congo (Ebuy et al. 2016; Kearsley et al. 2017). The YBR lies between 00°38’ and 01°10’N and 24°16’ and 25°08’E, with altitudes varying between 350 and 500 m (Amani 2011). The climate of Yangambi belongs to the Af type according to the Köppen classification (Mohymont and Demarée 2006). The monthly average temperature ranges between 22.4 and 29.3°C, with an annual average of around 25°C (Alongo et al. 2013). In Yangambi, annual rainfall varies from 1500 to 2400 mm, with an average rainfall of 1800 mm. The Yangambi region is dominated by ferralsols, with a clay content varying between 20 and 45%. The soils of Yangambi are reputed to be poor in assimilable mineral matter, due to their high acidity and low retention capacity (Ebuy et al. 2016). They are acidic with a pH between 3 and 4. The vegetation of Yangambi is characterised by a mixture of old-growth and regrowth forests (Tarelkin et al. 2016).

In this study, we used leaves from mature plants, collected fresh and subsequently dried as outlined below, as well as dried leaves from historic herbarium specimens. Fresh leaves were collected from five tropical understory woody species in 2019–2022 (Table 1). These species were selected because they are among the most abundant in the understory of the YBR and belong to different plant families.

Ten individuals per species were selected in old-growth forests in the YBR for leaf sampling. The height of the sampled shrubs varied between 3 and 12 m, as to avoid ontogenetic bias by sampling of young shrubs. From each shrub, three leaves were collected at the top of the crown, three in the middle, and three at the base. Leaves that were diseased or had been browsed by animals were not sampled. All leaves were pressed in a herbarium press and dried in an oven at a temperature of 60°C for 3 days and subsequently air dried for at least four more weeks, in order to make them comparable to the historic herbarium specimens.

For each species, 20 herbarium specimens originally collected in the same section of the Yangambi Biosphere Reserve before 1960 (hereafter named pre-1960) were selected from the Yangambi (YBI) and the Meise Botanic Garden (BR) herbaria. From each herbarium specimen, a leaf was sampled while taking care to not destroy the voucher. All leaves used in this study were originally collected in the Yangambi region according to their labels.

Leaf trait analyses were performed on 459 newly collected leaves (9 leaves per individual, 10 or 11 shrubs per species, and 5 species) and on 135 old herbarium leaves (1 leaf per specimen, 5 species and 20 leaves for C. canephora and S. thonneri, 27 for H. gabonii, 42 for T. penduliflora and 26 for U. solheidii). The dry mass of stalkless leaves was determined using a precision balance (precision: 0.0001 g). The upper surface of the leaves was scanned at a resolution of 300 dpi using an EPSON 10000 XL scanner. The surface area of the leaves was calculated using ImageJ v.1.52a (US National Institutes of Health; https://imagej.nih.gov/ij/). The specific leaf area (SLA), which corresponds here to the surface area of one side of the dried leaf divided by its dry mass, was calculated.

On the abaxial surface, a thin layer of colourless nail varnish was applied on both sides of the main vein and dried overnight. The nail varnish was then meticulously detached using transparent tape and glued on a microscopy slide. Two impressions were made for each extant leaf. On each herbarium leaf, two prints at the top, two prints in the middle, and two prints at the base of the leaf were made. Three photos per impression were taken per leaf print at a ×1,000 lens magnification using a digital microscope (VH-5000, Ver 1.5.1.1, Keyence Corporation). The stomata were counted on the photos using ImageJ v.1.51n in a grid of 40,000 µm² surface area. The stomatal density (SD) of each species was calculated and expressed per mm². On each of the photos taken, the length and width of one single representative and clear stoma was measured. At the same time, the length and width of the same stoma pore was measured using the ObjectJ plugin in ImageJ (Pérez-Harguindeguy et al. 2013). The stomatal size was calculated by multiplying stomata length by stomata width (Franks and Beerling 2009). These data have been deposited in Zenodo and are available at https://doi.org/10.5281/zenodo.8130615.

Climatological data consisting of average monthly rainfall, minimum and maximum temperatures, and monthly averages, covering the period 1961–2021, obtained directly from the INERA Yangambi climatology station (KP5), located at 00°49’12” N, 24°27’18” E (see Yakusu et al. 2022 and supplementary material 2 for details).

Changes in mean annual temperature and rainfall in Yangambi over the past 60 years were highlighted by analysing temperature and rainfall data. The identification of any trend in the time series was done using the Mann-Kendall trend test, performed using the Mann-Kendall function of the package Kendall v.2.2.1 in R v.4.2.1. It was applied separately to the precipitation and temperature series.

In Yangambi, the average annual temperature recorded for the period 1961–2021 was 24.98°C. The annual minimum temperature of 18.93°C was recorded in 1985, while the annual maximum temperature of 30.8°C was recorded in 2016. There was an upward trend in the temperature time series for this period (Kendall’s Tau = 0.63, p value < 0.001; Fig. 1A). An increase of 1°C was recorded at Yangambi for the entire period: the mean annual temperature was 25.75°C in 2021, compared to 24.48°C in 1961.

The minimum annual rainfall recorded at Yangambi was about 1418 mm in 2017, while the maximum annual rainfall was about 2432 mm in 1966, with an annual mean of about 1817 mm for the entire period. No trend was found in the rainfall distribution series at Yangambi during the period 1961–2021 (Kendall’s Tau = 0.047, p value = 0.59; Fig. 1B).

All data were analysed using R v.4.2.1. The normality of the data and the distribution of the residuals was checked using a Shapiro-Wilk test for all traits before other statistical test were performed A logarithmic transformation was performed on all trait data to meet normality requirements. In order to quantify the sources of variability in the traits depending on the position of the leaf in the canopy, we fitted linear mixed-effects models using the lmer function of the package lme4 v.1.1-31 (Bates et al. 2015), with leaf position and species considered as fixed effect and individuals as random effect. Afterwards, the results of all models were evaluated using the Anova function of the package car v.3.1-2 (Fox and Weisberg 2018). To compare the historic herbarium samples to the extant leaves, and to understand if there was variability in the traits depending on the two periods considered in this research, we fitted the linear models. As we do not know the position within the shrub where the material was sampled for the historic herbarium samples, we first tested the influence of the position within the crown on the traits. As there was no effect of leaf position, we compared the historic herbarium samples with recent samples and excluded a possible sampling bias.

There were no significant differences in leaf traits located at different levels in the crown, for any of the traits measured, nor a significant interaction between position and species (Table 2). We did, however, find significant differences between species for all leaf traits.

The largest leaves in terms of area and dry mass were observed in Coffea canephora (Fig. 2A, B), while Hua gabonii had the smallest leaves (Fig. 2A). The highest SLA was observed in Tabernaemontana penduliflora (212.28 ± 114.44 cm².g^-1) and the lowest in Uvariopsis solheidii (144.94 ± 51.88 cm².g^-1) (Fig. 2C).

There was no significant difference in stomatal size measured for leaves collected at different positions in the crown (Table 2). Indeed, leaf stomata had lengths around 23.43 ± 6.38 µm regardless of the position of the leaves in the crown (Fig. 2D), whereas their widths oscillated around 15.87 ± 4.82 µm on average (Fig. 2E). Overall, the largest stomata were measured on specimens of U. solheidii and the smallest on specimens of H. gabonii. The stomatal pore length of all species studied was distributed around 15.54 ± 4.3 µm (Fig. 2F). Similarly, the stomatal pore width of all species studied was distributed around the mean value 7.17 ± 2.26 µm (Fig. 2G).

Hua gabonii had a considerably higher stomatal density compared to other species, with a mean of 695 ± 40 stomata.mm^-2. The species T. penduliflora had the lowest stomatal density with a mean of 155 ± 65 stomata.mm^-2 (Fig. 2I).

Several leaf traits of understory woody species changed significantly between the periods pre-1960 and 2019–2022 in the Yangambi Biosphere Reserve (Table 3).

The surface area of leaves collected in 2019–2022 (133.62 ± 60 cm²) was significantly larger than the one of specimens collected in pre-1960 (104.18 ± 55 cm²; Table 3; Fig. 3A). The leaf mass was not different between the two time periods, but there was a significant interaction effect with species (Table 3). The leaf mass of H. gabonii was higher in 2019–2022 than in pre-1960, while the leaf mass was lower for C. canephora, S. thonneri, and U. solheidii collected in 2019–2022. Leaves collected in 2019–2022 showed a significantly larger specific leaf area (182.19 ± 79.1 cm².g^-1) when compared to those collected in pre-1960 (137.21 ± 28.51 cm².g^-1; Table 3; Fig. 3C).

The length and width of stomata of the five studied species in the YBR understory did not change over the past 60 years, nor was there a significant interaction effect (Table 3; Fig. 3D, E). In contrast, pore length did change over this time period (Table 3; Fig. 3F). Leaves collected in 2019–2022 had shorter pores (15.54 ± 4.3 µm) than leaves from pre-1960 (16.88 ± 6.49 µm). In addition, there was an interaction effect with a more pronounced decrease in pore length for S. thonneri, T. penduliflora, and U. solheidii. Pore width showed no significant difference between the two time periods (Table 3; Fig. 3E).

There was a significant interaction between collecting period and species for stomatal density (Table 3). Four out of five woody species studied showed a decrease in the number of stomata per unit area during the 2019–2022 period (320 ± 203 stomata.mm^-2) compared to pre-1960 (339 ± 239 stomata.mm^-2; Fig. 3I). Only C. canephora showed an increase in stomatal density from 244 ± 40 stomata.mm^-2 in pre-1960 to 298 ± 60 stomata.mm^-2 in 2019–2022.

The stomatal density values found in our study are consistent with those found by other authors in other tropical regions (Hultine and Marshall 2000; Hetherington and Woodward 2003). In Côte d’Ivoire, Djinet et al. (2016) recorded 50 stomata.mm^-2 on leaves of adult Elaeis guineensis Jacq., while Camargo and Marenco (2011), studied 35 tropical forest tree species, reported that stomatal density ranged between 110 stomata.mm^-2.in Neea altissima Poepp. & Endl. and 846 stomata.mm^-2 in Qualea acuminata Spruce ex Warm. Overall, we found a negative relation between stomatal density and stomatal size across the five species, which was more pronounced in recently collected leaves. Such a negative correlation between density and size of stomata has been observed in many other studies (Franks et al. 2009; Fanourakis et al. 2015; De Boer et al. 2016; Tian et al. 2016; Kafuti et al. 2020). The Specific Leaf Area in the species studied ranged from 42.74 cm².g^-1 for Hua gabonii to 778.45 cm².g^-1 for Scaphopetalum thonneri. These values are similar to those reported in a study on Brazilian forests (Hoffmann et al. 2005). Generally, a low SLA reflects the nutrient poverty of the environments (Tian et al. 2016; Gao et al. 2022).

The uniformity of leaf traits in different positions of the crown can be explained by the fact that all leaves of the species from the understory are exposed to rather uniform light conditions. This is very much unlike leaves from trees in the canopy, whose degree of exposure to light differs depending on whether the leaves are found at the top, middle, or base of the crown (Bauters et al. 2020). One of the main issues with using herbarium specimens from canopy trees to study climate change effects is the fact that it is very often unknown which position in the crown, and therefore sun exposure, the herbarium leaves had at the moment of collecting (e.g. Bauters et al. 2020). Given that light affects stomatal initiation (Royer 2001; Kouwenberg et al. 2007) as well as other functional traits (Poorter et al. 2006), light exposure should be accounted for in studies of leaf traits of trees.

Very high chlorophyll levels accompanied by high photosynthetic capacity have been reported for leaves fully exposed to the sun compared to those in the forest canopy that are not sun exposed (Hollinger 1989). The results found in this study are similar to those found for Ficus benjamina L. in Côte d’Ivoire, where the Specific Leaf Area and stomatal density did not vary with the height of leaf removal (Koffi et al. 2014). In contrast, a decrease in leaf dry mass from the top to the base of the tree crown was observed in canopy trees in tropical deciduous forests (Ellsworth and Reich 1993). In evergreen forest species, empirical data have shown an increase in leaf dry matter concentration from the base to the crown of canopy trees (Lewandowska and Jarvis 1977; Hollinger 1989). The photosynthetic capacity of the leaves, which is positively correlated with the nitrogen concentration of the leaves (Reich et al. 1995), is in the opposite direction: it is greater at the top than at the base in the forest canopy (Ellsworth and Reich 1993).

Although physiological and biochemical traits were not measured in this study, it is possible they also vary little with leaf position in the crown of trees in the understory. However, given the strong correlation between specific leaf area and photosynthetic capacity (Tian et al. 2016), it would be interesting to check whether there is any variability in physiological (e.g. stomatal conductance) and biochemical traits (e.g. nitrogen and phosphorus concentration) in understory leaves depending on their position in the crown. We realise that other sources of variation in leaf traits, such as ontogenic stage, season, soil nutrient content, and water content (Fortunel et al. 2020; Schmitt et al. 2022) were not accounted for in this study. However, we did make a maximal effort to reduce this variation by sampling individuals in similar conditions in the tropical forest understory and we do not expect that these sources of variation have a major influence on the current conclusions.

In general, our results confirm that studying effects of recent climate change events on tropical forest understory species using herbarium specimens according to modern and standardised protocols (Pérez-Harguindeguy et al. 2013) is more reliable when using tropical forest understory species instead of canopy trees (Peñuelas and Matamala 1990).

In this study, the main objective was to analyse the variation in leaf traits of woody species in the understory of the Yangambi Biosphere Reserve and to check whether there was a potential link between this variation and climate change. Our data showed several changes over time in foliar traits of five species in the understory of the YBR. These changes are potentially linked to climatic changes that have occurred over the past decades.

The main climatic change that has taken place in the Congo Basin is an increase in the temperature, due to globall increased CO₂ levels, while the annual rainfall has remained constant. For the period 1960–1992 a temperature increase of 1.6°C has been reported throughout the Democratic Republic of Congo (Kazadi and Fukunyama 1996). Recordings from the meteorological institute at INERA Yangambi showed a rising temperature trend between 1961 and 2021 (Fig. 1A). This result is consistent with that of Likoko et al. (2019), who observed that annual mean temperatures increased from 25.23°C to 25.78°C for the two previous decades (2000–2010 and 2010–2018), compared to the mean temperature recorded in Yangambi (Alongo et al. 2013). Furthermore, it corroborates the findings of Kazadi and Fukunyama (1996) who observed an increase in temperature throughout the Congo Basin, including the Yangambi region. In contrast to Kazadi and Fukunyama (1996) who also observed a considerable decrease in rainfall throughout the Congo Basin, the average rainfall remained stable in the Yangambi region between 1960 and 2021, although changes in rainfall seasonality cannot be excluded.

For tropical forest understory species from the Congo Basin, the stomatal density decreased significantly over the past 60 years for four out five species. This is consistent with the observations of Bauters et al. (2020) who analysed herbarium specimens of canopy trees in the Congo Basin. However, no significant change in stomatal density was observed in herbarium specimens collected over one century of two northern Amazonian tree species (Bonal et al. 2011). A decrease in stomatal density, as we found in our study, has been attributed to the response of plants to environmental conditions such as a temperature increase and the increase of air CO₂ concentration (Woodward 1986; Koffi et al. 2014). A decrease in stomatal density, which is tightly related to stomatal conductance, can be considered a key response to increasing CO₂ levels, potentially resulting in higher intrinsic water use efficiency (iWUE) (Bonal et al. 2011). However, Bauters et al. (2020) found a decrease in iWUE over the past 60+ years in the Congo Basin, despite decreasing stomatal density. Isotope measurements, providing insight in leaf physiological changes, would be very interesting to determine whether such a decrease in iWUE is also found in Congo Basin tropical forestunderstory species. Other factors besides CO₂ levels may also explain changes in stomatal density. An effect of increasing temperature on the decrease in stomatal density, for example, has been experimentally demonstrated (Beerling and Chaloner 1993; Hovenden 2001). An increase in stomatal density with a spatial temperature decrease at higher altitudes has been observed for Quercus kelloggii Newb. and Nothofagus solandri (Hook.f.) Oerst. species in New Zealand (Kouwenberg et al. 2007). In other studies, it was also clearly demonstrated that stomatal density was positively correlated with altitude (Woodward 1986; Hovenden 2001; Woodward et al. 2002), although there are some exceptions in terms of species and regions (Hultine and Marshall 2000).

The fact that recently collected leaves have higher SLA values may indicate that tropical forests have been enriched with nutrients, and especially atmospheric CO₂, leading to an overall gain in biomass (Lewis et al. 2009; Hubau et al. 2019), as a result of the increased photosynthetic capacity of the leaves. However, in other tropical regions, no increase in plant growth was observed while water use efficiency increased with atmospheric CO₂ enrichment for both understory and canopy species (van der Sleen et al. 2014; but see Bauters et al. 2020). Bonal et al. (2011) did not observe changes in LMA with changing CO₂ levels in two neotropical tree species, but they sampled material over a much wider geographical area leading to potential confounding effects of soil and climate. Generally, it is known that larger leaves have a higher light absorption capacity as well as a higher photosynthetic capacity (Tian et al. 2016). Variation in leaf area was observed along an altitudinal gradient in tropical forests for one bamboo species with a decrease in values with increasing altitude (Guo et al. 2018). The same trend has been observed in New Zealand (Kouwenberg et al. 2007). This spatial trend of increasing leaf size with increasing temperature at lower altitudes is in line with the temporal trend we found as the average temperature in the Yangambi region had risen by 1°C between 1961 and 2021. However, further studies are required to disentangle to which extent the different environmental factors that have changed over the past 60 years have contributed to changes in SLA in our study area.

The results of this study show that these understory species of the YBR may have responded to the changed temperature conditions in the Yangambi region. Although rainfall remained stable in Yangambi, an increased temperature may have resulted in higher evaporation and drier soil conditions. Such drier conditions may also explain the decrease in stomatal density, as a lower stomatal density is generally observed in plants that are drought resistant (Franks et al. 2009; Bertolino et al. 2019).