Woody vegetation change over more than 30 years in the interior duneveld of the Kalahari Gemsbok National Park

Background and objectives: Long-term studies of woody plants in South Africa are scarce. This study, initiated in the late 1970s, therefore aids understanding of vegetation dynamics in the southern Kalahari by investigating woody vegetation change at and away from a watering point. Methods: At three sites, all woody individuals were counted by species in plots 0.5 or 1 ha in size. Seedlings were noted separately from the >0.2 m group of individuals. Results: Vachellia erioloba and shrub density decreased over time whereas dwarf shrub species’ numbers fluctuated markedly. Additionally, no increase in density of known bush encroaching species (e.g. Grewia flava , Rhigozum trichotomum and Senegalia mellifera ) was found in this large conservation area. Discussion and conclusion: The changes in density of the woody species seem to point to the importance of particular rainfall patterns or sequences of events over different years that are responsible for these changes in the southern Kalahari, and the evident lack of bush encroachment in this conservation area supports the notion that bush encroachment in arid savannas is driven primarily by land-use practices and not by elevated carbon dioxide levels that are sometimes provided as cause for encroachment.


Introduction
Detecting environmental change at multiple spatial and temporal scales through biotic and abiotic research is of crucial importance not only to policymakers and scientists, but also to natural resource managers (Haase et al. 2018;Parr et al. 2002;Peters et al. 2014).Long-term monitoring can be used as a tool to assess the past management of natural resources, but also provides an understanding of ecological patterns and processes, which can improve decisions on the management of ecosystem assets (Lindenmayer & Likens 2009).Lindenmayer and Likens (2010) identified three broad, long-term monitoring approaches, these are: (i) curiosity-driven or passive monitoring; (ii) mandated monitoring; and (iii) question-driven monitoring.The first type, passive monitoring, usually has no specific questions or underlying study design and has limited rationale other than curiosity, while the second type, mandated monitoring, is a requirement of government legislation or a political directive.Question-driven monitoring, the third type, generally tests predictions that are guided by a conceptual model and rigorous design.The categories are, however, not mutually exclusive.
Woody vegetation change over more than 30 years in the interior duneveld of the Kalahari Gemsbok National Park There are still relatively few biodiversity time-series that span decades (Magurran et al. 2010) and this is particularly true for woody species (Moustakas et al. 2008).In the southern hemisphere, observational networks and research investment are largely lacking, and therefore even relatively short-term current and archived historical data sets can provide baselines for investigating change (Chambers et al. 2016).
Numerous surveys were initiated in the Kalahari Gemsbok National Park (KGNP) [now part of the Kgalakgadi Transfrontier Park (KTP)] to document vegetation change from 1978 onwards.The aim of this paper is to report on the monitoring that was initiated at three sites after a number of years with above-average rainfall, and follows a passive monitoring approach.The principal questions asked at the time were: how does the woody vegetation change at and away from a watering point, and what is the rate of survival at each of the sites?Although the research was not hypothesis driven (Lindenmayer & Likens 2010), monitoring was conducted over nearly four decades and therefore the data warrant exploration and reporting.

Study area
The study was conducted in the KGNP in South Africa.The KGNP covers about 9 600 km² and is situated between 24°15' S and 26°30' S and 20°00' E, and 20°45' E in the southwestern corner of the Kalahari region (Van Rooyen et al. 2008).Monitoring was conducted at three sites in the vicinity of the Dankbaar watering point, which was opened in 1959 (Figure S1).The study sites were situated within the Acacia (Vachellia) erioloba-Schmidtia kalahariensis Low Duneveld (Van Rooyen et al. 2008).This open tree savanna occurs in the interior duneveld in the northern part of the park.The vegetation is dominated by V. erioloba and S. kalahariensis (grass), with other prominent woody species being Boscia albitrunca, Rhigozum trichotomum and Senegalia mellifera.
Rainfall is received mainly in late summer with a mean of 194 mm/yr recorded from 1976 to 2015 at Nossob Rest Camp (Figure S2).The rainfall is highly erratic (standard deviation ± 88.8 mm) and occurs primarily from January to April, with a peak in March.Temperatures show a large amplitude with winter lows reaching -10.3°C and summer highs reaching 45.4°C (Van Rooyen & Van Rooyen 1998).

Study sites
Three study sites were established in the interior duneveld between 1978 and 1982.The first site was 5 km to the east of the watering point (25°03'39.13"S and 20°07'17.37"E) (hereafter referred to as EoWP) and monitoring commenced in 1978.The second site, at the Dankbaar watering point, (25°04'10.90"S and 20°05'47.50"E) (hereafter referred to as WP), was monitored for the first time in 1980.At the third site, 5 km to the north of the watering point, (25°01'39.45"S and 20°05'0.30"E) (hereafter referred to as NoWP), monitoring started in 1982.At the time when monitoring commenced, WP was dominated by a dense stand of V. erioloba saplings; NoWP had an intermediate density of relatively young V. erioloba individuals; whereas EoWP was a mixed stand of mature V. erioloba and Vachellia luederitizii individuals at a low density.The plots at WP and NoWP were 0.5 ha in size, while the plot EoWP was 1 ha in size.In each plot, all woody individuals were counted by species.Additionally, seedlings (individuals <0.2 m tall) were noted separately from the sapling and mature group of individuals (>0.2 m individuals).All three plots were monitored intermittently, i.e.WP sampled 16 times, NoWP 14 times and EoWP 9 times, including the initial and 2016 surveys.

Statistical analysis
To visualise change in density over time at the different sites, we used linear regressions with the natural log (=ln) density (number of individuals/ha) per species over time (survey years), in the package Graphpad Prism (version 5, graphpad.com).Data for each site were analysed separately.Only species that occurred in the plots on four or more occasions were included in the analyses i.e. a minimum of four data points.
A linear regression of (a) seasonal; (b) annual; or (c) the cumulative annual rainfall of the previous three years against the annual change in tree density between two survey years was investigated.

Results
Generally, the density of tree species declined over the monitored period (Table 1).Furthermore, this decline was significant for Vachellia erioloba individuals (>0.2 m) at WP and NoWP (Figure 1a).Additionally, the seedlings (< 0.2 m) of V. erioloba decreased significantly at WP (Figure 1b).
No relationship between tree density and seasonal or annual rainfall or the cumulative annual rainfall of the previous three years was found.This does not exclude rainfall as driving factor, but seems to point towards the effect of a particular rainfall pattern or sequence of events over different years that is responsible for changes in tree densities.

Discussion
The authors acknowledge that this study was limited by the number of sites lacking replication and the possibility that individuals were missed during the field surveys, however the length of the study warrants exploration of the data set.
The most outstanding feature of this long-term data set is the significant decrease in density of V. erioloba (>0.2m) individuals at two sites (WP, NoWP) (Figure 1).The most prominent decline was for V. erioloba at WP from the early 1980s until the mid-1990s, a period when the mean annual rainfall was below average for most years (Figure S2).This dry period followed on a wet cycle in the late 1970s that preceded the surveys.The large numbers of seedlings could have been the aftermath of the wet cycle and their demise as a result of the following dry cycle in the 1980s.Numerous authors have reported large numbers of V. erioloba seedlings following above average rainfall periods (Barnes 2001a, Moustakas et al. 2008, Seymour 2008), however, survival of V. erioloba seedlings is very low (Barnes 2001b, Seymour & Milton 2003, Steenkamp et al. 2008, Van der Merwe et al. 2019) due to either drought, frost, herbivory or fire.The low survival of seedlings was supported by the results of this study.
When monitoring commenced, the density of the V. erioloba >0.2 m individuals was approximately fivefold higher at WP than to the north thereof and 36-fold higher than at EoWP.Such high densities of saplings are quite uncommon in the interior duneveld (pers.obs.N.v.R) and this raises the question as to whether the high densities are in some way associated with the watering point opened in 1959, 21 years prior to the commencement of vegetation monitoring.Vachellia erioloba

B
A is often perceived by landowners as an encroaching species in the Kalahari (Seymour & Milton 2003) and it could be speculated that this species encroached at the watering point after it was opened and before monitoring started.However, the speculated encroachment did not continue during the monitoring period.
In 2016, the density of V. erioloba at WP was 29% and NoWP 19% of the initial density.It is therefore unlikely that borehole drawdown (Shadwell & February 2017) led to the dieback of trees at WP, since a similar decrease also occurred away from the watering point.
The large decrease in density in young individuals at WP and NoWP could largely be ascribed to densitydependent self-thinning, whereby density decreases with a concomitant increase in biomass of the remaining individuals (Silvertown & Charlesworth 2001).
At the three sites shrub densities generally decreased significantly or remained relatively constant.Three of these tall shrubby species are often associated with bush encroachment on heavily utilised farmland in the Kalahari i.e.Grewia flava, Rhigozum trichotomum and Senegalia mellifera (Joubert et al. 2013, Moleele et al. 2002, Moore et al. 1988, Tews et al. 2004, Van Rooyen et al. 1994).However, densities of these known encroachers remained relatively constant (G.flava at WP and EoWP, S. mellifera at EoWP) or decreased significantly for R. trichotomum (NoWP) over the monitoring period at all sites.
The dwarf shrub densities showed large variability, as large as ten-fold, across the survey period.These fluctuations between high and low densities occurred at shorter time intervals than for the tree species.
Overall, the changes in density of the woody species seem to point to the importance of particular rainfall patterns or sequences of events over different years that are responsible for these changes in the southern Kalahari (Van Rooyen et al. 1990, 1994).Furthermore, the lack of evidence of bush encroachment in this conservation area at the three sites, one of which is situated at a watering point, supports the notion that bush encroachment in arid savannas is driven primarily by land-use practices and not by elevated carbon dioxide levels that are sometimes provided as cause for encroachment (Nackley et al. 2018).

Conclusions
Monitoring was initiated based on a simple passive monitoring question of 'How does vegetation change over time?'The authors acknowledge that such research may be limited in usefulness as cautioned by Lindenmayer and Likens (2010), however, due to the length of the data set, investigation was justified.A similar decline in density of V. erioloba tree seedlings and >0.2 m individuals was found at and away from the watering point.Furthermore, the density of known bush encroaching species at the watering point did not increase over the monitored period.

Figure 1 .
Figure 1.Linear regressions of the natural logarithm of the number of individuals per hectare, indicating significant trends in density of Vachellia erioloba over time.(a) >0.2 m individuals, and (b) seedlings.WP = Watering point, NoWP = North of watering point.

Figure 2 .Figure 3 .
Figure 2. Linear regressions of the natural logarithm of the number of individuals per hectare indicating significant trends in density of shrub species over time.(a) Ehretia alba; (b) Lycium bosciifolium; and (c) Rhigozum trichotomum.WP = Watering point, NoWP = North of watering point, EoWP = East of watering point.