Drought release and post-drought changes in herbaceous composition and diversity in two land uses subjected to selective bush control in a semi-arid Kalahari savanna

Land degradation due to bush encroachment and drought threatens the well-being of land users and forage production in the semi-arid savannas of southern Africa. This study aimed to analyse how herbaceous vegetation in the Kalahari savanna can recover from drought, and how this is affected by previously implemented bush control measures. We compared plant species composition and diversity of the herbaceous layer across a drought-release (2021) and post-drought (2022) year within two land use types (game and livestock) in bush-encroached and controlled areas, respectively. Diversity decreased significantly during the post-drought year. Annual species dominated in the drought-release year, which were replaced by perennial forbs and dwarf shrubs in the post-drought year. Selective bush control enhanced herbaceous species abundance and richness due to decreased competition. The species richness was higher in the game ranch compared to the cattle farm during the drought-release year. Herbaceous abundance, however, increased on the cattle farm. The management strategies, feeding behaviour and animal movement may have contributed to these differences in these two contrasting land use types. This study showed that herbaceous plant communities can recover rapidly after a drought period and in areas where bush control was applied.


Introduction
Drought release and post-drought changes in herbaceous composition and diversity in two land uses subjected to selective bush control in a semi-arid Kalahari savanna rehabilitate bush-encroached areas (Stafford et al. 2017).These efforts include the manual and/or chemical removal of indigenous woody species causing bush encroachment, such as Senegalia mellifera (Donaldson 1966), to enhance re-establishment of perennial grasses and forbs.Herbicides are also often used to clear woody plants, either unselectively by using aeroplanes, or selectively by hand (Harmse et al. 2016;Kellner et al. 2022).The effectiveness of these approaches, especially with respect to the regeneration of the herbaceous layer depends on a range of factors (e.g.different woody species, soil type, climatic impacts, etc.) and is of great interest to rangeland ecologists, policymakers and land users (Eldridge and Ding 2001).
Another threat to game and livestock farmers, especially in semi-arid savannas, is prolonged droughts.Recently, droughts have been shown to impact both woody and herbaceous vegetation in semi-arid savannas (Abbas et al. 2019;Wigley-Coetsee and Staver 2020).Furthermore, drought also impacts the well-being of commercial and communal landowners, including their economic challenges (Drysdale et al. 2021).Livestock farmers are often forced to sell animals to reduce livestock herds or buy fodder, whilst game ranchers often use professional hunters to reduce herds of wildlife or change herd compositions to match the grazing and browsing capacity of the veld (Walker et al. 1987;Kachergris et al. 2014).Future predictions show that drought events combined with higher daily temperatures will occur more frequently due to climate change impacts (Lawal et al. 2019;Mbokodo et al. 2020).It is therefore important to understand how drought affects herbaceous vegetation dynamics, and how vegetation recovers after prolonged drought events.Furthermore, little is known about how herbaceous vegetation (i) recovers from drought, (ii) is affected by the control of bush encroachment and (iii) how it varies between different land use types such as game ranching and cattle farming.An increased understanding of these aspects has the potential to enhance insights into how the sustainability of semi-arid Kalahari rangelands can be ensured in the long term through adaptive strategies.The aim of this study was therefore to compare herbaceous plant composition and diversity, identify plant indicator species across a droughtrelease year (2021) and a post-drought year (2022) within two land use types (cattle vs game) where selective bush control was applied ten to 15 years ago.We hypothesised that, in accordance with the study of Abbas et al. (2019), the herbaceous layer will be associated with rapid recovery (within one year after the drought), and may lead to an increase in palatable herbaceous species.Such a rapid recovery will therefore suggest drought resilience.In addition, indicator forb species will contribute to ecosystem resilience, such as nutrient cycling as well as forage stability during, and after dry spells (van Coller et al. 2021).Community shifts in the herbaceous layer will be evident across land use and where bush control treatments were applied (controlled vs uncontrolled).It is further postulated that annual plants will dominate during the drought-release year (2021), and that perennial plants will dominate in the post-drought year (2022).Responses of life forms (grasses and forbs) to drought and herbivory are therefore expected to be land use type-specific (cattle vs game) (Schmiedel et al. 2021), and also dependent on habitat preference, movement patterns and feeding behaviour of cattle and game (Searle and Shipley 2008;Marquart et al. 2023).

Study area
The area falls within the Eastern Kalahari Bushveld Bioregion of the Savanna Biome and is dominated by the Molopo Bushveld vegetation type (Rutherford et al. 2006).The dominant land use types include livestock farming (cattle, sheep and goats) and game ranching (Harmse et al. 2016). Soil characteristics (i.e. soil type, texture, topsoil, pH) in the region where the study was conducted are generally considered homogenous (Harmse et al. 2016).Soils are classified as deep, well-drained, sandy soils with low nutrient content (Donaldson and Kelk 1970) of the Hutton and Clovelly soil types (Rutherford et al. 2006).The soil surface consists mainly of the characteristic red, wind-blown sands, together with the occurrence of calcrete and silcrete (Rutherford et al. 2006).Soil pH ranges between 4 and 5 (Harmse et al. 2016;Kellner et al. 2022).The landscape is relatively level with an altitude that varies between 1 000 and 1 300 m (Rutherford et al. 2006).Temperatures can be extreme, especially during summer, with maximum daily temperatures that exceed 42 °C (Low and Rebelo 1996;Rutherford et al. 2006).In the winter, temperatures may fall below the freezing point leading to frequent frost events (Rutherford et al. 2006).The region is classified as semi-arid (Figure 1) with a highly erratic summer rainfall (Hagos and Smit 2005).The long-term mean annual precipitation (MAP) is approximately 333 mm (Rutherford et al. 2006).
Rainfall data were obtained from three farms, i.e.Harmonie (cattle farm), Omega (game ranch) and Bushy Flats (game ranch), located 20 km west of Omega, for which average rainfall from 2010 to 2022 was calculated (Figure 2).Vegetation surveys were conducted on one cattle farm (Harmonie) and one game ranch (Omega) in April 2021.A dry event preceded vegetation sampling during 2018/2019.Despite above-average rainfall being recorded for 2020-2021, the area was still visibly dry during a site visit (exploration trip) in December of 2020 (Figure 3a).Therefore, after sufficient summer rains recorded for the area from January to March 2021, field surveys were conducted in April of 2021 (Figure 3b).This year is therefore considered a drought-release year just after the dry spell was broken (Siebert et al. 2020a), and a slightly above-average rainfall was recorded during 2019-2020 (Figure 2).During May 2022 (Figure 3c), follow-up surveys were conducted within a year that received above-average rainfall.Hence 2022 was considered the post-drought year since recovery could occur one year after the drought-release (Ruppert et al. 2015).From the photographs, vegetation change occurred rapidly and drastically over a period of three years (Figure 3).

Field data collection
Vegetation surveys were conducted in April 2021 (droughtrelease year) and in May-June 2022 (post-drought year).Herbaceous vegetation was sampled on a game ranch and a cattle farm respectively during both survey years (Figure 1).A total of six transects were sampled per farm.were laid out in bush-encroached areas (uncontrolled) that had not been chemically treated in the past, whilst the other three transects were located within selectively bush-controlled sites (treated).On both farms, bush control was implemented at the sites between ten and 15 years prior to the study, using granular arboricides with the active ingredient tebuthiuron.All transects were 1 000 m in length and commenced at a water point to allow sampling of comparable spatial habitat use among the two land use types.On the cattle farm, one transect was established per fenced paddock (i.e. a maximum of four per water point), while on the unfenced game ranch, attempts were made to establish transects from water points with a minimum angle of 90 degrees.Four plots were established per transect at four distances, i.e. 150 m, 300 m, 600 m, and 1 000 m away from the watering point.Within each of these plots, a 50 m sub-transect was placed following the direction of the main water point transect.Thereafter, all rooted herbaceous plant species (i.e.grasses and forbs) were recorded within three nested 2 × 2 m sub-plots (quadrats) at 15 m intervals (Birnie et al. 2005).All recorded individuals were identified up to species level following the taxonomy and author citations according to Germishuizen and Meyer (2003).Systematic total counts of species within quadrats were used as presence and absence data (Rich et al. 2005), i.e. abundance data to eventually calculate species diversity indices and species composition.In the case where perennial stoloniferous or clonal plants were encountered, individuals (ramets) fragmented from the original/mother plant, were counted separately from the mother plant as an individual (Rich et al. 2005).

Data analysis
Herbaceous species diversity Diversity measures that were investigated were species richness, evenness, Shannon diversity and Simpson diversity.Simpson diversity indicated the important role of evenness since it combines species richness and evenness.In contrast Shannon diversity tends to be less sensitive to evenness and primarily considers abundance (Kent 2011).Furthermore, Simpson diversity can be considered as an indicator of heterogeneity or homogeneity in response to the tested variables of a study (Ellis et al. 2012;Püttker et al. 2015).Hence, both measures were included.
Diversity analyses were performed using R (Version 4.0.3;'Bunny-Wunnies Freak Out', R Core Team 2020) with the vegan package version 2.5-7 (Oksanen et al. 2019).The significance level of all test statistics was set at p = 0.05.Assumptions (homogeneity and normality of residuals) for linear mixed models were tested visually using respective plotting functions provided in R (QQ plots, residuals vs fitted values, etc.).Interaction effects of land use, selective bush control, and year on diversity measures were tested using linear mixed models with p-values derived using Kenward-Roger approximations for degrees of freedom.Abundance data were log 10 transformed and Simpson diversity was square transformed to meet the assumption of normality.To account for the hierarchical experimental design, we included water point transects, distances to water points and plots as random effects with plots nested within distances nested within transects.Maximal models included a three-way interaction between land use type, bush control and year.Interactions and fixed effects were removed from the maximal models in a stepwise backward manner according to AIC values to obtain minimal adequate models with the best goodness of fit.Models fit by maximum likelihood were used to compare AIC values, and the best-fitted models were refitted using restricted maximum likelihood (REML).REML is considered to provide better estimates for the random effects, but depends on which fixed effects are in the model, and thus two models fitted by REML are not comparable if the fixed effects differ.

Indicator plant species
Indicator species analysis was conducted to determine which species could be used as indicators for each year, bush control, and land use type combinations.The indicator value index (IndVal) was used to measure the relationship between a species and habitat resulting from treatment combinations.Following the method proposed by Dufrêne and Legendre (1997), the group with the highest association value was first identified and then the indicator value between species and each group was calculated.The statistical significance of this relationship was then tested using multi-level pattern analysis with group-equalised indicator values, excluding combinations of site groups.

Herbaceous composition
To examine differences in herbaceous species composition as a function of time in combination with selective bush control and land use type, Non-Metric Multi-Dimensional scaling (NMDS) was used to visualise similarity patterns.NMDS analyses were conducted using Bray-Curtis dissimilarity and fourth root transformation of community abundance data.Permutational Multivariate Analysis of Variance (PERMANOVA; permutations = 999; type III sum of squares) was used to test the effects of year, land use and selective bush control on species composition.Individual transects and plots nested in transects were included as additional blocking factors to account for the experimental design.NMDS was conducted using PRIMER 6 software (2012), while PERMANOVA was conducted using the Vegan: Community ecology package (version 2.9) in R version 4.0.3(R Core Team 2020).Similarity Percentage Analysis (SIMPER) was conducted on abundance data to identify herbaceous plant species responsible for compositional changes in plant communities using Paleontological Statistics Software (PAST) (Hammer et al. 2001).

Herbaceous species diversity
The results of the mixed models for all diversity measures are shown in Table 1.All diversity measures showed a significant decrease during the post-drought year, compared to the drought-release year after the dry spell (Figure 4).Species richness was lower on game ranches (9.3 ± 0.3) compared to cattle farms (12.1 ± 0.3) in the drought release year, while this difference was negligible during the post-drought year when overall richness was lower (6.3 ± 0.2).Selective bush control resulted in a slight increase in species richness of 0.7 species, independent of year and land use.Simpson diversity decreased from 0.77 ± 0.8 in 2021, to 0.64 ± 0.15 in 2022.In addition, land use type had a minor effect on Simpson diversity, with values on the game ranch averaging 0.03 units lower.Shannon diversity was similarly influenced by the year of sampling and land use type as Simpson diversity.Post-drought Shannon diversity was 0.5 units lower compared to the drought-release diversity (2021: 1.8 ± 0.02; 2022: 1.3 ± 0.03), while game ranching resulted in an additional decline of 0.13 units on average.Species evenness was only slightly affected by year, leading to a decrease from 0.78 ± 0.01 in 2021 to 0.73 ± 0.01 in 2022.Total herbaceous plant abundance was higher in the cattle land use type than in the game land use type.Similarly, herbaceous plants were more abundant in treated areas compared to untreated areas.These impacts were significantly more evident in the first year after the drought than in the second when total abundance of herbaceous plants was lower in general.While the difference in abundance between game and cattle averaged 25.5 individuals in 2021 (drought-release) (game: 64.4 ± 3.5; cattle: 89.8 ± 3.7), this difference decreased by 4.5 individuals in 2022 (post-drought) (game: 49.4 ± 1.9; cattle: 53.9 ± 2.1).

Indicator plant species
Drought-release (2021) and post-drought (2022) plant communities within both bush control treatments for the cattle farm and game ranch were associated with specific indicator species (Table 2) that varied in their abundance (Figure S1, Supplementary material).In total, ten indicator plant species were within the Fabaceae family across years, treatments and land use.Most of the indicator plant species in the drought-release year were annuals which were then replaced by perennials in the post-drought year, especially for the untreated cattle, untreated and treated sites at the game ranch when considering the ratio between annuals and perennial (Table 3).Interestingly, untreated sites at the cattle farm had two perennial grasses for each annual grass during the drought release year.This relationship increased during post-drought where there were three perennial grasses for each annual grass.Furthermore, an annual plant, Verbesina encelioides, was characteristic to the untreated sites at the cattle farm during the post-drought year (Table 2).With regards to life forms, indicator plant species in untreated sites at the cattle farm were mostly forbs (non-graminoids, i.e. monocotyledonous and dicotyledonous plants such as herbaceous forbs, dwarf shrubs and geophytes) for both years (Table 3).For every two grasses, there were three forbs during drought release year.Indicator species at the untreated sites on the cattle farm were only forbs.During drought-release, treated sites at the cattle farm had a total of eight indicator plant species, consisting of a combination of life forms, with the highest percentage associated with forbs (62%, Table 3).The grass:forb ratio decreased with regards to forbs between the two sampling years.In the drought release year there were seven indicator forbs for each indicator grass at the untreated sites on the cattle farm while for post-drought there was one forb for each annual grass.In contrast, during post-drought only two species were revealed for the treated sites at the cattle farm of which one was the highly preferred grass species, Schmidtia pappophoroides and a nitrogen-fixing forb, Indigofera heterotricha (Table 2).
Given that the game ranch revealed low species richness compared to the cattle farm, a low number of indicator plant species was to be expected as seen in the game ranch sites across both sampling years (Table 2).For the drought release year for both the untreated and treated sites at the game ranch, the life forms of indicator plants were limited to herbaceous forbs and grasses (Table 3).When considering the ratios between grasses and forbs at untreated sites at the game ranch, the ratio increased from one to two indicator grasses for each forb in the drought release and post drought year, respectively.Similarly, treated sites at the game ranch also showed an increase in the ratio.The drought release year had two indicator forb species for each grass species and three forbs for each two indicator grass species (Table 3).

Herbaceous composition
Separate clustering revealed by the NMDS analysis suggested that the drought-release (2021) and post-drought (2022) herbaceous plant community assemblages were floristically unique (Figure 5).Within these two main clusters, further clustering based on land use and bush control treatment was observed.PERMANOVA revealed that these groupings were highly significant (p < 0.001).SIMPER analysis of the pooled data (across years, bush control treatment and land use) identified 20 herbaceous plant species that contributed to 90% of the cumulative percentage (Table S1, Supplementary material).The top five species that contributed to 62% of the observed differences between the drought-release and post-drought assemblages were primarily perennial grasses which increased in abundance values during post-drought conditions.These included the perennial grasses Eragrostis lehmanniana, S. pappophoroides and Stipagrostis uniplumis.The annual grass, Melinis repens as well as the perennial forb, Tephrosia purpurea subsp.leptostachya showed a decline in the post-drought plant communities within bush control treatment and land use (Table S1).

Rainfall as a driver of vegetation change
The above-average rainfall in 2021 and 2022 was responsible for shifts in herbaceous species composition in this semi-arid Kalahari savanna in areas where the control of bush encroachment was carried out.This finding suggests these savanna herbaceous plant communities have the ability to recover rapidly and, hence are resilient to drought (Gil-Romera et al. 2010).Compositional shifts can be related to the recovery of the herbaceous layer that was initially evident by the abundance and density of annual life forms during drought release within both land use types (Riginos et al. 2018;Siebert et al. 2020a).Furthermore, annual plants recover more rapidly than perennial plants and hence, are known to have a faster recovery process after drought events (Ruppert et al. 2015).A turnover in species therefore occurred as annual plants were replaced by perennial plants in herbaceous communities in a short period of time (Skarpe 1992;Siebert et al. 2020a), irrespective of land use type and bush control.Similar to the findings of Swemmer et al. (2018), valuable palatable grazing grasses became more abundant than unpalatable grasses.Such species included the perennial E. lehmanniana, S. pappophoroides and S. uniplumis, while the annual Schmidtia kalahariensis grass declined during post-drought recovery.The presence of an annual invasive plant species, such as V. encelioides (golden crownbeard or wild sunflower), during the post-drought year (untreated sites at cattle farms) indicates the potential impact that climate change (e.g.low rainfall, higher temperatures) can have on the distribution of alien plants in plant communities (Verlinden et al. 2013;Moshobane and Esser 2022).Verbesina encelioides has been found to spread rapidly on sandy soils and is hence considered as a serious invader that is highly competitive and poisonous to livestock (Bromilow 2010).Due to its competitive and opportunistic nature, this plant species will be able to colonise suitable habitats much more rapidly than native species under favourable environmental conditions (Moshobane and Esser 2022).Therefore, it is called upon land users to eradicate this species through herbicides or mechanical removal of young individuals before seed production occurs (Bromilow 2010;Moshobane and Esser 2022), especially at the untreated sites on cattle farms.

Drought-release and its impacts on herbaceous diversity
Drought-release plant communities were characterised by higher diversity than those during post-drought across land-uses and bush control treatments (Figure 4).This could firstly be ascribed to the increased germination success and establishment of especially annual plants after the first higher rainfall occurrence and drought period (Veenendaal et al. 1996;Schmiedel et al. 2021).Additionally, annual plants are known to produce high numbers of viable seeds that remain in the seed bank until environmental conditions are favourable, which explains the high numbers of individuals just after the drought (Grime 1977;O'Connor 1991).In the absence or low abundance of perennial grasses, different life forms can flourish due to reduced competition for resources such as water, nutrients and light (Buitenwerf et al. 2011).This was particularly observed with respect to the different types of life forms that comprised the indicator plant species during the drought-release year at the untreated and treated sites on the cattle farms.A drought event accompanied by intense grazing, results in high mortality of perennial grass tufts, allowing annual forbs to colonise an area after a drought event (O'Connor 1995).As perennial grasses recovered, these annual forbs and grasses were then outcompeted as they are over-shaded by large tussock grasses due to competitive exclusion effects (Grime 1973;Jacobs and Naiman 2008).Van Coller et al. (2021) revealed that drought-tolerant annual forbs are usually characterised by prostrate growth forms.Annual forb indicator species in the drought-release year, such as Corbichonia decumbens, Euphorbia inaequilatera, Indigofera alternans, Limeum arenicolum and Limeum viscosum are prostrate growing.Prostrate growth forms also fulfil important ecological roles in plant communities, such as soil stabilisation by limiting soil erosion (wind and water), lowering soil temperatures and increasing nutrient input as plants die off at the end of their life cycle (Leys 1991;Jankauskas et al. 2008;Siebert et al. 2020b).
It was evident in this study that as herbaceous plant communities recovered after the drought, dominance shifts occurred (Huston 1979;van Coller et al. 2018).Indicator plant species that became more prominent were primarily perennials.Perennial forbs and dwarf shrubs can re-sprout from a woody rootstock or a persisting bud bank (Siebert et al. 2020a;van Coller et al. 2021).Therefore, perennial herbaceous plants do not only invest in a viable seed bank, but also in underground storage organs that provide them with competitive abilities contributing to their resilience in response to drought events (Pausas et al. 2018;Siebert et al. 2020a).In times of drought when other palatable sources are depleted, perennial forbs and dwarf shrubs may also serve as valuable sources of forage for cattle and game (Odadi et al. 2007;Siebert et al. 2020a), and hence contribute to forage stability (Siebert and Scogings 2015;van Coller et al. 2021).The lower diversity revealed for the post-drought plant communities, may be a result caused by increased competition as more perennial plants managed to establish, while the opposite is applicable under drought release where competition effects are weaker, resulting in a higher diversity (Levins 1968;Vandermeer 1970).During drought-release, subdominant plant species co-exist which contributes to the enhancement of evenness (Castillioni et al. 2020).Additionally, the post-drought declines in species evenness across land-uses, irrespective of the bush control treatment, also emphasise the shift in dominance in the herbaceous layer.

Importance of plant species indicators within the Fabaceae
Some genera within the Fabaceae family are known to have root nodules with nitrogen (N) fixing bacteria that live in symbiosis with the plant in its roots (Corby 1988).The N-fixing symbiotic ability of such species is considered to be an adaptive strategy in nutrient-poor habitats, such as this Kalahari savanna, that enable these plants to acquire N (Cornelissen et al. 2003;Sprent et al. 2013).Moreover, N-fixers also contribute to soil fertility in disturbed environments (Bradshaw 1997;Pedrol et al. 2014).
Grazing and drought play a role in soil carbon (C) dynamics, especially in semi-arid grasslands (Lucci 2019;Munjonji et al. 2020).Drought-release conditions may also be responsible for the loss of nutrients such as N, C and phosphorus (P) due to rapid re-wetting, the recharge of soil moisture after the occurrence of a drought through precipitation and infiltration (Mishra and Singh 2010;Lucci 2019).However, N-fixing herbaceous plants are considered to counteract the loss of N related to drought (Hofer et al. 2017).Since a total of ten indicator plant species are in the Fabaceae plant family that occurred in this study, it is suggested that these species have also contributed to nutrient cycling in this semi-arid Kalahari savanna during drought-release and post-drought conditions, irrespective of land use and bush control treatment.

Effects of bush control and land use
During this study, it was revealed that selective bush control affected the herbaceous vegetation, especially regarding abundance and species richness.When a large fraction of woody biomass is removed from the system, competition for nutrients and water is decreased (Smit 2005).As a result, opportunities for germination and increased survival of herbaceous species are created.Furthermore, dead woody biomass left above and below ground in the system can have a nursing effect on herbaceous species, providing them with favourable micro-habitats (Mangani et al. 2022).As many problematic woody species such as S. mellifera have thorny spines, branches left on the soil after treatment can lead to a form of 'grazing deterrent' protecting the establishment of palatable herbaceous plants from grazing animals (Mangani et al. 2022).In addition, the decomposition of roots and branches may also lead to an input of organic matter and an increase in soil nutrient availability (Dong et al. 2022).These positive effects of bush control on herbaceous vegetation were differently pronounced in the two years following the drought period at the study sites.In the drought-release year, especially herbaceous abundance in treated areas was higher compared to untreated areas, whereas this effect was less pronounced in the post-drought year.Since the drought had led to a strong decrease in herbaceous vegetation cover (Figure 3a), annuals and pioneer species dominated the treated areas in the drought release year, where water competition with more drought-resistant woody species was lower compared to the untreated areas (Van Aardt et al. 2020).This was also apparent when comparing the herbaceous composition and indicator species between treatments in the two years.
In the drought-release year, the composition in treated cattle managed areas was dominated by perennial forbs and dwarf shrubs.This emphasises the importance of forbs in livestock grazing systems, as forbs may provide a valuable role in the ecosystem, especially in providing forage stability during extreme events such as prolonged droughts (van Coller et al. 2021).This is true for species such as T. purpurea subsp.leptostachya, or Indigofera daleoides which have been recorded to be palatable and contribute to pasture value (Trytsman et al. 2019).On the other hand, species such as Senna italica and Urochloa panicoides were also associated with bush-controlled cattle managed areas and were indicators of disturbed and overgrazed areas (CABI 2022;Fouché et al. 2014).Likewise, in untreated cattle managed areas, indicator grass species such as M. repens are also commonly found in disturbed and degraded areas.Similarly, untreated game managed areas were associated with the annual grass species, such as Aristida congesta, an indicator for over-grazing and disturbed or degraded veld, which reflect the high grazing pressure and water stress during the drought period.Interestingly, the perennial grass species which were mainly associated with bush-controlled game managed areas was Stipagrostis obtusa, which has a high grazing value and is an indicator of good veld condition (van Oudtshoorn 2009).Thus, it can be assumed that the management and animal behaviour on the game ranch led to good conditions of the veld, which was even persistent after drought, especially in bush-controlled areas.In the post-drought year, however, a smaller number of species were representative of the herbaceous communities in treated and untreated areas overall.While in untreated areas on cattle farms, forbs and dwarf shrubs dominated, selectively controlled areas were mainly dominated by highly palatable grass species, such as S. pappophoroides.The latter can establish in a year and is a good indication of recovery from extensive grazing pressure during drought years.In the post-drought year on the game ranch, only perennial grasses and forbs were found to be indicator species.The two perennial grass species S. uniplumis and E. lehmanniana were mainly found in bush-controlled areas.These two palatable species are utilised by grazers, especially when young, and are very common in the area where the study was carried out.Our results illustrate that the herbaceous layer on a game ranch and cattle farm can recover within two years after a drought, especially in areas where bush control was carried out.

Conclusions
The results presented by this study, emphasise the influence of two contrasting land use systems (i.e. a cattle farm and game ranch) in a Kalahari savanna on herbaceous composition and diversity, especially after the control of bush encroachment in combination with erratic rainfall events and drought.
In response to a higher rainfall event, it was revealed that the herbaceous layer can recover more rapidly after an intense drought within a time frame of approximately one year.As a result of drought-release, annuals (of which some were unpalatable) dominated plant communities initially.As recovery occurred, palatable perennials established successfully during the post-drought season, after two years.Different life forms (graminoids, forbs, dwarf shrubs) responded dynamically to drought and herbivory within the two land use types (cattle and game).Forbs contributed to ecosystem functioning through soil stabilisation, soil nutrient input (those within the Fabaceae), and forage stability.Therefore forbs should be included in rangeland management strategies.This study emphasises the importance of the control of bush encroachment, a measure that is often considered costly and labour-intensive by land users.However, after the application of bush control strategies (treated areas), plant communities can recover faster than bush-encroached communities (untreated areas, uncontrolled) after a drought.
Future studies should aim to incorporate data on plant communities during pre-drought, as well as during and post-drought conditions over the long term to gain more insight regarding vegetation dynamics related to drought events in the Kalahari.To enhance understanding regarding ecosystem functioning and resilience, functional traits can also be measured.Long-term monitoring will add to more focused management strategies in these two different land use systems.

Figure 1 :
Figure 1: Location of the (a) study area within South Africa and (b) two farms (cattle farm -chequered block, game ranch -grey block) where vegetation sampling was carried out

Figure 2 :Figure 3 :
Figure2: Rainfall data obtained from landowners in the Bray region, where the study was conducted (where possible, averages were calculated to obtain the long-term mean rainfall for the study site)

Table 1 :
Test statistics of the intercepts and fixed predictor variables of the minimal adequate mixed models, analysing interactive effects of year, land use and bush control treatment on herbaceous diversity measures Results of species diversity measures for the drought-release (2021) and post-drought year (2022) within bush control treatments and land use types (dots indicate outliers; a data point is an outlier if it is less than the first quartile −1.5* interquartile range, or greater than the third quartile +1.5* interquartile range)

Table 2 .
Herbaceous plant species indicators listed for each sampling year, bush control treatment and land use.Species are listed based on IndVal values (high to low) and significance.An asterisk (*) after species names indicates alien invasive plant species

Table 3 .
Life forms and life histories of herbaceous plant species indicators across bush control treatments and land uses.Ratios were also calculated.Note that all non-graminoid life forms were grouped as forbs to determine the grass:forb ratios