A 40-year Evaluation of Drivers of African Rainforest Change

Background: Tropical forests are repositories of much of the world’s biodiversity and are critical for mitigation of climate change. Yet, the drivers of forest dynamics are poorly understood. This is in large part due to the lack of longitudinal data on forest change and changes in drivers. Methodology: We quantify changes in tree abundance, diversity, and stand structure along transects rst enumerated in 1978 and resampled 2019 in Kibale National Park, Uganda. We tested ve predictions. First, based on the purported role of seed dispersal and herbivory and our quantication of changes in the abundance of frugivores and herbivores, we tested two predictions of how faunal change could have inuenced forest composition. Second, based on an evaluation of life history strategies, we tested two predictions concerning how the forest could have changed following disturbance that happened prior to written history. Finally, based on a 50-year climate record, we test the possible inuence of climate change on forest dynamics. Results: More trees were present on the assessed transects in 2019 (508) than in 1978 (436), species richness remained similar, but diversity declined as the number of dominant species increased. Rainfall increased by only 3 mm over the 50 years but this effect was not signicant. Annual average monthly maximum temperature increased signicantly by 2.2°C over 50 years. The abundance of frugivorous and folivorous primates and elephants increased over the 50 years of monitoring. The predictions that as the abundance of seed dispersing frugivores increases the abundance of their preferred fruiting tree species would increases and that as the abundance of folivorous would cause a decline in their preferred species were both not supported. Since Kibale was disturbed prior to historical records, we predicted that light-demanding species would decrease in abundance, while shade-tolerant species would increase - this was supported. Finally, while temperature increased over the 50 years, we found no means to predict a priori how individual species would respond. Conclusions: Our study revealed subtle changes in the tree community over 40 years, sizable increases in primate numbers, a substantial increase in the elephant population and an increase in local temperature. Yet, a clear picture of what set of interactions impact the change in the tree community remains elusive. Our data on tree life-history strategies and frugivore/herbivore foraging preferences suggest that trees species are under opposing pressures. change on specic tree species independently of changes in rainfall. Thus, we were unable to test how specic tree species are affected by the observed increase in temperature. We present a 40-year record of change in a tropical tree community, and some of the longest and most detailed records of tropical forest mammal populations dynamics ever accumulated, and site-specic information on tree life-history strategies, climate change, and on forest disturbance that occurred prior to written history. With respect to primates and elephants, we have decades of observational data and ecological studies upon which to examine the inuence of foraging on plant species sorting. Our study revealed subtle changes in the tree community between 1978 and 2019, sizable increases in primate numbers, and a substantial increase in the elephant population. Yet, a clear picture of what set of interactions impact the change in the tree community remains elusive. Our data on tree life-history strategies and frugivore/herbivore foraging preferences suggest that species are under opposing pressures. For example, both C. durandii and D. were predicted to decrease because they have been senescing after an anthropogenic disturbance 200-400 years ago. However, since the frugivores that disperse seeds of these tree species have increased in abundance, any inuence from the prior anthropogenic disturbance may have been obscured.


Introduction
Tropical forests are repositories of much of the world's biodiversity. Covering only 7% of the world's land surface, tropical forests account for 60% of the world's biodiversity (Bradshaw, Sodhi and Brook 2009).
These forests are critical to successful mitigation of climate change. For example, tropical forests and wetlands are estimated to contribute 23% of the mitigation needed to limit global warming to 2°C by 2030 (Griscom et al. 2017; Wolosin and Harris 2018). Yet, these forests are increasingly threatened. Globally, ~60 million ha of tropical primary forest were lost from 2002 to 2019, with most forest loss occurring in Brazil (24.5 Mha), Indonesia (9.5 Mha), and the Democratic Republic of the Congo (4.8 Mha)(Weisse and Gladman 2020). To put this in perspective, an area of old-growth tropical forest larger than Madagascar was lost over 18 years. Restoring tropical forest is a necessary part of mitigating the effects of climate change and requires an understanding of what naturally drives tropical forest dynamics and the ecological processes that are affected (Ma et al. 2016). Surprisingly, the drivers of forest dynamics are poorly understood, due in large part to the lack of suitable longitudinal data spanning decades. Many of the species involved in structuring tropical forest ecological systems have generation times stemming from years to many decades (trees Swaine Further complicating our understanding of the drivers of tropical forest dynamics is the need to consider the synergistic interaction of multiple drivers. Important interacting processes include pollination, seed dispersal and predation, herbivory, disease, competition, disturbance regimes, and climate. All of these are perturbed by human actions and the legacy effect of human imposed disturbance that occurred decades or centuries earlier has to be considered (Richards 1996). Furthermore, normative ecosystem response is obscured by stochastic events like droughts (Condit et al. 2017). Thus, it is hardly surprising that our understanding of driving factors/processes has often been judged by examining the strong signal produced by extreme events. For example, Harrison et al. (2013) provided detailed tree census data 15 years after intensive hunting eliminated most large frugivores. They documented a consistent decline in tree diversity but found no evidence of reduction in above-ground biomass (see also Chapman et al. 2003;Poulsen, Clark and Palmer 2013). Their study clearly illustrates the importance of frugivores in maintaining tree diversity but does not contribute to an understanding of the relative importance of drivers of forest composition under less extreme conditions.
Here, we quantify changes in tree abundance, diversity, and stand structure (species rank abundance, and size class structure) and mammal abundance of ten species in Kibale National Park, Uganda (hereafter Kibale) between 1978 and 2019. We consider ve drivers of change and how they have affected the forest tree community over 23-50 years.
Mounting evidence suggests bottom-up processes, like seed dispersal and herbivory, are dominant drivers of tropical forest communities (Crawley 1989 . For example, a reduction in populations of large-bodied seed-dispersing primates corresponds with lower seedling densities of large-seeded forest trees species (Chapman and Onderdonk 1998;Pacheco and Simonetti 2000) and higher seedling aggregations around parent trees (Pacheco and Simonetti 2000).
Similarly, by foraging on trees, elephants (Loxodonta africana) can convert forest ecosystems to grasslands (Laws 1970 2013b). Based on the purported role of seed dispersal and herbivory in structuring tropical forests, we advance two predictions. First (Prediction 1), an increase in the abundance of seed dispersing frugivores is expected to correspond to an increase in the abundance of fruit bearing trees prominent in their diet and vice versa. Second (Prediction 2), increases in arboreal herbivore/folivore abundance is expected to correspond to a decrease in the abundance of their preferred foods.
Research since the 1980s has shown that many forests traditionally considered "pristine" were disturbed by people relatively recently (i.e., between 100 and 4000 years ago (Clark 1996) (Coley 1983;Hubbell et al. 1999;Dalling et al. 2012) are most adapted for a particular location and time following the disturbance. For example, light-demanding species are better adapted to recruit in gaps following disturbances and use new resources for growth, while shade-tolerant species tend to recruit into the system slowly over decades and invest more in their wood density, roots, and defensive mechanisms (e.g., plant toxins) so that they are not disrupted by herbivory during their establishment (Grubb 1977;Richards 1996;Wright 2002;Chave et al. 2009). In addition, some species are adapted to recruit after catastrophic disturbances that create extremely large clearings and these large clearings can be made naturally or through human actions (Chazdon 2003). For example, mahogany (Swietenia macrophylla) recruits in areas dramatically disturbed by hurricanes (Snook 1996) or in areas of erosion or in forests killed by ooding (Gullison et al. 1996). With respect to Kibale, the rainforest was disturbed by people prior to written or oral history, thus we make the following predictions. First (Prediction 3), we predict that lightdemanding species should decrease in abundance between 1978 and 2019, while shade tolerant species increase. Second, (Prediction 4) tree species that recruit in areas typically disturbed by human clearance (i.e., larger than a single tree fall gap) should decline in abundance over the 40 years.
Plants respond to slight shifts in temperature and rainfall associated with climate change (van Vliet and Schwartz 2002; Walther et al. 2002). For example, the average rst owering date of 385 British plant species has advanced by 4.5 days over the past decade compared to the previous four decades (Fitter and Fitter 2002;Wolkovich et al. 2012). In Panama, ower and seed production increased during El Niño years (Wright and Calderón 2006). In Kibale, annual fruiting varied over 3.8-fold between 1998 and 2013 and fruiting was positively in uenced by temperature, rainfall, and solar radiation. As we have documented such relationships among climate and phenology patterns in Kibale (Chapman et al. 2005; Chapman et al. 2018b), we propose the following prediction, Prediction 5 shifts in the composition of the tree community correspond to directional change in the climate at Kibale. We develop scenarios of tree community response to climate change based on habitat preferences (e.g., trees that typically occur in wet valley bottoms will increase in abundance if the climate gets wetter).

Study Site and Vegetation
Our longitudinal study of vegetation was conducted in Kibale National Park, Uganda spanning the period from the rst assessment of forest composition conducted in December 1978 to the resampling of the same plots completed in May 2019 -40 years and 5 months apart. The 795 km 2 park is in western Uganda (0° 13' -0° 41' N and 30° 19' -30° 32' E) near the foothills of the Rwenzori Mountains (Struhsaker 1997;Chapman and Lambert 2000). Kibale is dominated by mid-altitude (920 -1590 m), moist-evergreen forest that receives a mean annual rainfall of 1655 mm (1970 -2020).
Rainfall data were collected immediately adjacent to the study area. The daily rainfall data were summarized per month. The collection of these meteorological data was maintained through rebel intrusions into the park and the COVID19 pandemic and data for only 8 of a total 612 months were incomplete and thus not included. For the missing 8 months, we tted an ARIMA time series model with Fourier terms for seasonality to interpolate these values using all other values. Temperature data (daily minimum and maximum) were collected over the same period. However, thermometers had to be replaced several times, and they were relocated twice ( rst by a distance of ~ 1 km, and then by only 30 m). An analysis of the temperature data from 1970 until 2020 indicated that these changes in thermometer and location (hereafter sources) had impacts on measured temperature that were challenging to control for. For example, the magnitude of the difference between minimum and maximum temperature appears to vary with the source (i.e., some thermometers show higher max. temperature, hereafter T max , and lower min. temperature, hereafter T min ). Therefore, we used the TerraClimate dataset  (Table 1).
We categorized species as light-demanding or shade-tolerant from a statistical assessment of stem distribution among habitats described in Zanne and Chapman (2005)  early successional (pioneer) species (i.e., species that die within 20-40 years after they colonize a disturbance). These four species often become canopy level trees in old-growth forest: -Celtis africana, Celtis durandii, Diospyros abyssinica, and Funtumia latifolia. The lifespan of these trees is unknown, but it is likely that they live at least a few hundred years. To further verify if these species typically recruit after large anthropogenic disturbances, we established seven 200 by 10 m plots in the study area and seven similar plots in a large disturbed area immediately adjacent to the study area (Nyakatojo 86.2 ha). This disturbed area was an anthropogenically derived grassland, dominated by elephant grass (Pennisetum purpurem), but between 1967 and 1968 the area was converted to a pine plantation (Kingston 1967;Struhsaker 1975 removed after they were counted to ensure that they were not repeatedly counted. The tracks and dung of the two duiker species can be distinguished when the sign is of good quality, but quality declines over time and depends on the season and environment. Thus, it was not always possible to distinguish the species, so we report a combined duiker value. Censuses of duiker, bushbuck, and bushpigs in Kibale are available from prior to 1996 (Nummelin 1990;McCoy 1995;Struhsaker 1997;Lwanga 2006); however, there are methodological differences among studies (Struhsaker 1997) that make comparisons problematic.
To examine Prediction #1 that changes in the abundance of seed dispersing frugivores results in a corresponding change in the abundance of fruit-bearing tree species, we determined the 10 most frequently used fruiting tree species for blue monkeys (Rudran 1978), redtail monkeys (Stickler 2004 only in the K30 area), and mangabeys (Olupot 1998 data from 1992 and 1993). These species often eat fruits from the same species and this comparison produced 17 tree species that were examined for changes in their abundance ( Table 2). Prediction #2 was evaluated for folivorous primates and the tree species most likely to be killed by colobine foraging (2013a) were monitored for their change in abundance from 1978 to 2019 ( insu ciently known to permit predictions of how they may affect forest composition change. However, we report on changes in the abundance of these species so that evaluations may be made in the future.

Analysis
We estimated sampling saturation or completeness and species richness of the tree community using the estimator of sample coverage in the R package 'iNEXT' (Hsieh, Ma and Chao 2013). Because species richness is not sensitive to species abundances and gives disproportionate weight to rare species, we measured tree species diversity with Hill's numbers (Jost 2006 Climate change in uences forest plant community composition and structure, either directly (e.g., causing tree or seedling death) or indirectly (e.g., causing the disruption of processes such as pollination). We decomposition. For rainfall, T min , and T max , we applied a "Seasonal and Trend decomposition using Loess" (STL) in the 'fabletools' package for R. We applied linear models to the trend component from these decompositions as the outcome variable and date as the predictor variable.

The Forest in 1978 and 2019
More trees were present on the sampled transects in 2019 (508) than in 1978 (436), but species richness remained similar, decreasing by only two species (Table 5)

Changes in Animal Abundance
While the number of groups of frugivorous seed-dispersing monkeys detected per km walked uctuated slightly over time, there was no signi cant change in the relative abundance of groups over the last 50 years (Table 6). The only exception to this trend was a decrease in abundance of blue monkeys. This decrease has been monitored and is occurring park-wide (Butynski 1990 (Table 6) and thus primate density in the area actually increased.
Similarly, the abundance of folivorous primates -red colobus and black-and-white colobus -groups in the area varied slightly between 1970 and 2019, but with no overall change in group density. However, again group sizes increased and thus population density (number of individuals per km walked) increased (Table   6).
In general, the abundance of elephants, duiker, and bushbuck increased between 1996 and 2005 and has remained relatively stable since. In contrast, bushpig abundance increased from 1996 until 2008 and declined thereafter ( Table 6).

Evaluation of the Predictions
Prediction 1: as the abundance of seed dispersing frugivores increases, the abundance of their preferred fruiting tree species increases. Of the 17 preferred species in the diet of the frugivorous primates, seven increased in abundance as predicted, three declined, and for seven there was no change in abundance ( Table   2). Given that greater rates of seed dispersal with increasing frugivore numbers would take time to be represented as fruit-bearing trees, we examined if there was an increase in the abundance only in the smallest size classes. Considering only those stems between 10 and 15 cm DBH, four species increased as predicted, ve decreased, and for the remainder there was no change in abundance. Thus, Prediction 1 was not supported.
Prediction 2: as the abundance of folivorous primates increases, the abundance of heavily defoliated tree species declines. Of the 13 tree species frequently used by colobus, seven occurred in the sample area. Of these (90 trees across both sampling years, 6.9 individuals per species, range 1-40), four species support the prediction, two species increased contrary to the prediction, and for one species there was no change (Table   3). Considering only the two species with ≥10 individuals (Dombeya mukole, Markhamia lutea), both species declined in abundance in accordance with the prediction.
Also, with respect to Prediction 2, we expected that as elephants increased in abundance, the tree species that elephants preferentially fed on would decline. There were twelve highly preferred elephant food species (Table 4). Of those in the area, 56% increased in abundance, the opposite to what was expected, 33% decreased as expected, 11% remained the same (Table 4). Considering only those species with ≥10 individuals, all three increased in abundance. Thus, Prediction 2 as it applies to elephants was not supported. Consistent trends were observed for all species and the ten most abundant species (Table 5). Comparing rank abundance curves there was a decrease in the dominance of light demanding species and an increase for shade-tolerant species (Figure 2). Comparing the size-frequency distributions of light-demanding and shade-tolerant species, they follow the expected J-shaped curve (Figure 3). The increase in shade-tolerant species was particularly marked in the smaller DBH size classes (<40 cm DBH) and shade-tolerant species dominated the 10-19.9 cm DBH size class ( Figure 3). Thus, Prediction 3 was supported.

Discussion
There were no major disturbances in the study forest in recorded history and correspondingly the changes in the forest structure that we documented between 1978 and 2019 were subtle. We recorded slightly more trees in the second enumeration, but species richness was similar and the tree community assemblage became more evenly distributed. To gain further insights to what might be driving changes in tree community composition, we examined changes in seed dispersal, herbivory, human induced disturbance, and climate on putative changes to the tree community over the last 40 years.
Animal populations increased in density, which we predicted would affect the tree community. We found no evidence that observed increases in the abundance of frugivores led to an increase in the trees whose seeds they disperse. Similarly, despite ve decades of observation to determine which tree species the folivorous primates damage and kill through overuse, only slightly more than half of the tree species examined followed the predicted pattern. For elephants, we found that only 33% of preferred tree species in the diet of elephants decreased in abundance. The effects of frugivory and herbivory do not appear to be strong enough to affect forest composition over the time period and spatial scale evaluated.
It is possible that other biotic factors obscured the effect of herbivory and frugivory. For example, the increase in the number of frugivores may have resulted in more seeds being dispersed, but seed predators increased during this same period, masking recruitment trends such as observed for Monodora myristica in Kibale (Balcomb and Chapman 2003). A site with greater frugivore density had more seeds dispersed, but this did not result in more saplings. Alternatively, it is possible that 40 years of monitoring is insu cient to detect change as tropical trees have slow growth. For example, Chrysophyllum sp. seedlings and saplings grow extremely slowly in the shaded understory, with their mean height doubling only every 27 years (Connell and Green 2000). Thus, a 20 cm seedling could take almost 60 years to reach a meter in height, if it survived that long in the understory and did not have the growth advantage of a light gap (see also Kalbitzer, McInnis and Chapman 2019). Clearly a long-term perspective is needed to examine the cascading effects of one change, such as the gradual decline in seed disperser abundance in a forest.
With respect to anthropogenic changes -century old human-induced disturbance and climate change -our ndings are mixed. Forest disturbance that occurred prior to written history was still affecting changes in the forest tree species composition, but no consistent pattern is revealed. As Prediction 3 suggests, lightdemanding species decreased in abundance over time, while the abundance of shade tolerant species increased. Presumably, the light demanding species became abundant in the forest following a historical disturbance and these trees are now senescing, dying, and not being replaced. However, we also predicted that species that typically recruit into large disturbed areas should decline between 1978 and 2019 and this was not supported by our data. We could not make apriori predictions of the effect of temperature change on speci c tree species independently of changes in rainfall. Thus, we were unable to test how speci c tree species are affected by the observed increase in temperature.
We present a 40-year record of change in a tropical tree community, and some of the longest and most example, both C. durandii and D. abyssinica were predicted to decrease because they have been senescing after an anthropogenic disturbance 200-400 years ago. However, since the frugivores that disperse seeds of these tree species have increased in abundance, any in uence from the prior anthropogenic disturbance may have been obscured.
Our exploration illustrates the challenges that must be faced to understand and predict change in terrestrial plant community dynamics. Of critical importance to addressing forest dynamics are longitudinal data and the interactions among important variables/processes over long time frames (Franklin et al. 2016). To understand the nature of the interactions, emphasis should be given to forests that have not recently experienced major disturbances, both in terms of the forest structure and animal populations. Disturbance to these interactors may have cascading effects on the forest community that take decades to return to a typical state -if there is a typical state at all (Pickett 1980). However, such studies can act as a comparison point to build a framework for future efforts. Within such a framework the scienti c community can address whether and when forests will be in uenced by novel biogeochemical conditions (e.g., CO 2 enrichment or N deposition) and novel assemblages of plants and animals, including invasive species or where diseases or human actions cause dramatic declines in populations (Franklin et al. 2016). Given the current global conditions it will be important to explore how interactions between climate and disturbance regimes lead to shifts among vegetation types, with special attention given to thresholds. Modeling efforts will be required to integrate plant physiology, demography, and biogeography, past forest history, and future climate and land use change (Franklin et al. 2016). A signi cant challenge will be to predict how forest communities have been in uenced by past human impacts and how they will respond to future policy changes. By meeting these challenges researchers will have the information to convince policy makers of the appropriate actions that must be made to most effectively conserve the rich biodiversity of tropical forests.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable. Tables   Table 1: The density (trees per ha) of the ten most common tree species >10 cm DBH in an old-growth section    2 Data from Kasenene (1980;   Descriptions of the tree community in 1978 and 2019 in an old-growth section of forest in Kibale National Park, Uganda. 1D is the exponential of Shannon's entropy and is interpreted as the number of 'common' species in the community, which weights each species according to its frequency in the community, and 2D is the inverse Simpson concentration, which favors abundant species and is therefore interpreted as the number of 'very abundant' or 'dominant' species in the community.  Patterns of rainfall, maximum temperature (°C, Tmax) and minimum temperature (°C, Tmin) between 1970 and 2020 for the area near Makerere University Biological Field Station in Kibale National Park, Uganda. For details, see text.

Figure 2
The shade-tolerant and light-demanding species enumerated in plots ranked in order of abundance.
Sampling was conducted approximately 40 years apart (1978 and 2019) in an old-growth area of Kibale National Park, Uganda.