Evidence for climate-induced range shift in Brachystegia ( miombo ) woodland

AFFILIATIONS: 1School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Johannesburg, South Africa 2South African Environmental Observation Network, Phalaborwa, South Africa 3School of Geography, Archaeology and Environmental Studies, University of the Witwatersrand, Johannesburg, South Africa 4School of Agricultural, Earth and Environmental Sciences, University of KwaZulu-Natal, Pietermaritzburg, South Africa *Current address: Global Change and Sustainability Research Institute, University of the Witwatersrand, Johannesburg, South Africa


Introduction
At regional and global scales, climate broadly limits the distribution of plant taxa, 1,2 and the response of species to changing environments is consequently likely to be largely determined by population responses at range margins. 3owever, few studies have examined climatic determinants of range shift at the trailing margin of a savanna species in the southern hemisphere.Understanding the historical and present-day spatial dynamics of vegetation plays a crucial role in our ability to predict the likely community responses and biodiversity consequences of future global change. 4achystegia spiciformis Benth. is the dominant component of miombo, the colloquial term used to describe sub-tropical woodlands dominated by Brachystegia, Julbernardia and Isoberlinia -three closely related genera in the family Fabaceae and subfamily Caesalpinioideae.5 Miombo encompasses the woodland-dominated savanna ecosystems 6,7 which cover (Figure 1) an estimated 2.7 million km 2 of south-central Africa 8 and is coincident with White's 9 Zambezian Phytochorion, the largest regional centre of endemism in Africa 10 .The dynamics of miombo woodland are largely determined by the woody component which, apart from climate, is predominantly influenced by people and fire.An estimated 75 million people inhabit areas covered by, or formerly covered by, miombo woodland; an additional 25 million urban dwellers rely on miombo wood or charcoal as a source of energy.11 Much of the woodland has been, and continues to be, modified by people.Changes in vegetation structure occur both directly as a result of woodland cover removal, and indirectly from changes in fire regime brought about by higher grass production under a more open tree canopy.12 Most of the dominant miombo canopy species, including B. spiciformis, are considered to be fire-tender species, which decline in abundance under regular burning and increase under complete fire protection.5 Mean annual precipitation and mean annual temperature throughout the miombo region range from 650 mm to 1400 mm and 15 °C to 25 °C, respectively.The majority (>95%) of precipitation falls during the summer season which prevails from October to March. 5 However, sporadic dry periods during the onset of the precipitation season may cause large fluctuations in soil moisture and temperature.The resulting water stress during this concurrent and limited germination phase accounts for high seedling mortality in miombo woodlands. Brachystegia are likely under-represented in pollen spectra because of the relatively low pollen production typical of entomophily. Evidence suggests that the central plateau vegetation has undergone significant flux over this period, ranging from cool upland grassland during the Last Glacial Maximum (LGM, 19 000 years BP), to Brachystegia-dominated savanna during the warm interglacial period of the Mid-Holocene (6000 years BP).5,16 Although literature regarding climate-induced range shifts at the genus level is limited for the subcontinent, Eeley et al. 17 concluded that the distribution of the forest biome in KwaZulu-Natal, South Africa, receded during the colder and drier conditions of the LGM, while warmer and wetter conditions during the Mid-Holocene were conducive to forest expansion.,18,19 Sediment cores containing pollen from the genus Brachystegia have been recovered from Pretoria, Mookgophong and Tate Vondo (Figure 2) in South Africa, 15 the dates of which stand as evidence of a much more widespread distribution during the recent past (~6000 to less than 1000 years BP). Tese records support the presence of the genus at least up to 450 km southwest of the present-day distribution limit.This most recent and abrupt range retraction suggests either a sudden change in climate, for which there is no evidence, 19 or that minor shifts in temperature and moisture regimes have triggered marked changes in Brachystegia population dynamics.5 A change in geographical distribution from the simultaneous migration of populations throughout their range is unlikely.Instead, change is generated by the establishment of new, often discontinuous, populations at the leading edge of a species distribution and the coincident death of individuals and extirpation of populations at the trailing edge.20 These retractions are often not complete, but instead leave behind fragmented populations that persist as relicts, in isolated enclaves of favourable environmental conditions within an inhospitable regional climate.21 Globally, numerous relict populations have resulted from species range shifts experienced after the LGM. 21For example, the present-day distribution of Neotropical seasonal dry forest formations is considered to comprise fragmentary remnants of the once extensive forests that characterised the dry climatic maxima of the Pleistocene.22 Rutherford et al. 23 suggested that B. spiciformis, which at the time was known only from north of the South African border, could find a suitable ecological niche under future global climate change conditions along the high rainfall savanna-grassland biome interface of northeast South Africa.However, they concluded that range expansion through long-distance dispersal of the species into South Africa, across the Limpopo River Valley, was unlikely to occur unassisted given that seed dispersal distance is limited to less than 6 m, 13 and that regeneration takes place predominantly through coppice regrowth and root suckers, rather than seed.24 Startlingly, the discovery of the isolated B. spiciformis woodland (~15 ha, Figure 2) in the eastern Soutpansberg of South Africa 25 placed the species within the region predicted by Rutherford et al. 23 This isolated population, occurring some 200 km south of the continuous miombo woodlands of the subcontinent, suggests a trailing edge refugium persisting from a previously wider historical range dating prior to the LGM or during the Mid-Holocene.This woodland relict (known locally as Gundani) shares characteristics of the extensive B. spiciformis woodlands of Zimbabwe and southern Mozambique, including the open structure, medium canopy height, poorly developed lower strata and high dependence on root suckering and coppicing for regeneration.26 The presence of a second Brachystegia species within the Gundani woodland, limited to a single B. utilis individual, further suggests that this is likely a climate relict of a vegetation type from a time when, at least, the genus Brachystegia dominated the Soutpansberg massif in northeastern South Africa.27 Trailing range edge studies reflect a bias (86%) towards high-latitude range margins and temperate vegetation communities, 3 which have minimum temperature-related constraints.Conversely, this rare situation provides a unique opportunity to explore climatic constraint at lowlatitude range margins for an ecologically significant savanna species.
In this climate specific study, we used a predictive modelling approach to determine (1)

Species data
A total of 914 mature B. spiciformis specimens were considered (717 in-situ observations and 197 from regional and online herbaria).Their locations were subsequently converted to quarter-degree square (QDS) centroid points to accommodate numerous herbaria records.After duplicates were excluded, a total of 76 QDS centroid points (n=76) remained.QDS is a commonly used format for general species distribution mapping in South Africa. 28These species data include the vicariant B. spiciformis population from South Africa, as the inclusion of relict populations has been shown to improve the performance of model-based projections. 20urce: Adapted from White's 9 map of African vegetation.

Predictor variables
Nineteen regional bioclimatic variables and altitude ('BioClim') grid data were downloaded from the 'WorldClim' 31 database.The bioclimatic variables are derived from monthly temperature and rainfall recorded worldwide (period 1950-2000) and are often used in ecological niche modelling (for examples see Blach-Overgaard et al. 32 and Penman et al. 33 ).A soil variable (dominant soil group) was added from the Harmonised World Soil Database. 34ENM Tools v.1.3 35was used to produce pairwise Pearson correlation coefficients to eliminate spatially correlated variables (r>0.85)from the modelling process, with preference given to variables with higher influence on species presence probability.Summer precipitation variables were also prioritised during the elimination process as Trouet et al. 14 suggest they have a strong influence on B. spiciformis growth performance.The 14 variables used in all models were Bio2-6, Bio8-9, Bio15-19, altitude and soil (see the Appendix for a description of the variables).
There are currently at least 24 coupled atmosphere-ocean general circulation models (AOGCMs) used to project climatic changes for more than 10 different greenhouse-gas emission scenarios. 36We selected three commonly used AOGCMs 37,38 -HadCM3, CGCM2 and CSIRO-MK2 -to forecast the impact of climate change on B. spiciformis distribution.The A2a emission scenario, an intermediate scenario representing regional development and slow economic growth, was used across all three AOGCMs.

Model settings
The spatial resolution of all variables and landscape grid cells was resampled to QDS resolution.MaxEnt iterations were set to 5000 and accuracy was evaluated by constructing the model using 75% of presence records as training points, with the remaining 25% used in validation.The accuracy of the present-day ecological niche model was inferred from the area under the receiving operating characteristic curve (AUC) score that varies from 0 to 1, with 0 being the lowest and 1 being the highest probability of matching the species distribution. 31ubsequently, the present-day ecological niche model was applied to the three selected future (2050) AOGCMs.The increase in AUC score was used as a test metric to determine the most important bioclimatic variables explaining B. spiciformis woodland distribution, when each variable was used in isolation.

Ecological niche comparison
The fundamental niche 39 was represented by a box plot, with the interquartile range (box with median present) representing 50% of the total data values and the upper and lower whiskers representing 25% of the values, respectively.The interquartile space is interpreted as the 'most suitable' range of climatic or environmental conditions for the species, while the lower and upper whiskers span less suitable conditions, reflecting lower and upper tolerance limits, respectively.Extreme values, beyond the 1.5 coefficient value, were plotted as individual open circles.
The grid value at each point was used to create box plots representing the present-day ecological niche (Figure 3).
Seasonal data (precipitation and temperature) from Mid-Holocene ocean-atmosphere-vegetation models were used to intersect three palaeohistorical Brachystegia distribution points using Geospatial Modelling Environment.These grid values were subsequently used to construct Bio16 (precipitation of the wettest quarter) and Bio4 (temperature seasonality, the standard deviation of mean monthly temperature in degrees Celsius) as per BioClim.These two bioclimatic variables were selected as they achieved the highest increase in AUC score.This methodology was repeated across all available Mid-Holocene ocean-atmosphere-vegetation models (ECBILTCLIOVECODE, ECHAM53-MPIOM127-LPJ, FOAM, MRI-CGCM2.

Limiting factors map
A limiting factor map (Figure 4) was produced as per Elith et al. 30 Using Geospatial Modelling Environment and implementing the method described above, box plots were created from randomly selected QDS centroid points (n=76) within the area immediately adjacent to the current distribution of the species, for which a specific bioclimatic variable was shown as limiting.Box plots and paired sample t-tests were used to compare B. spiciformis ecological niche and limiting factor dimensions per selected bioclimatic variable (Bio4, 5, 6, and 16).

Model performance and variable contribution
The present-day ecological niche model for B. spiciformis achieved an AUC value of 0.923, which is considered very good. 43The model was subsequently applied to three future AOGCMs.Precipitation of the wettest quarter (Bio16) and temperature seasonality (Bio4) were identified as the two most important bioclimatic variables explaining B. spiciformis woodland distribution in southern Africa (Table 1).This result was consistent across the present-day and three future models.

Present-day ecological niche
The ENM (Figure 2) predicts the presence of isolated B. spiciformis populations in South Africa, one of which is coincident with the location of the Gundani relict discovered in 2001.The model further indicates suitable ecological conditions for the species elsewhere at disjunct locations along the high rainfall savanna-grassland interface of northeastern South Africa.However, the species does not presently occur outside of the relict population. 44Besides the fact that White's map of African vegetation 9 does not indicate miombo woodland presence in South Africa, the distribution is coincident with the present-day ENM projection for B. spiciformis in southern Africa.
Precipitation of the wettest quarter (Figure 3) suggests a relatively narrow range of optimal suitability of 422-576 mm, with a median of 507 mm.Similarly, the 'most suitable' range for temperature seasonality (Figure 3) is constrained between 2.6 °C and 3.0 °C, with a median of 2.8 °C.The value of the former variable at the Gundani relict population (401 mm) falls within the lower tolerance limit of conditions experienced by the species, whereas the value for the latter variable at the relict population (2.6 °C) falls within the 'most suitable' range for B. spiciformis in southern Africa.

Past ecological niche
A crude palaeohistoric distribution of Brachystegia, as inferred from three Mid-Holocene sediment core pollen records from South Africa (Figure 2), supports past average climate values of 529 mm and 591 mm (two records) for precipitation of the wettest quarter, and temperature seasonality averages of 4.3 °C and 4.8 °C (two records) for that period.For precipitation (Figure 3), these averages fall within (or nearly within) the 'most suitable' range identified under present-day conditions (422-576 mm).Alternatively, the historical temperature seasonality values exceed even the upper tolerance limit (a maximum of 3.4 °C) for this species across its present-day southern African distribution (Figure 3).

Future ecological niche
The future ecological niches for all three models (Figure 5) are consistent with a predicted decrease in the continuous B. spiciformis woodland distribution of Zimbabwe and southern Mozambique (30.6% to 47.3%) and a decrease in the overall ecological niche (including South Africa) of between 16.9% and 32.2% by 2050 (Table 2).These figures reflect net change, with a decreased niche potentially resulting from larger retraction at present-day distributions combined with comparatively smaller range increases at newly favourable distributions elsewhere.(3.9-4.7 °C), but exceeds the maximum 'most suitable' temperature seasonality of 3.0 °C experienced by the species across its present-day range (Figure 3).Although there is overlap between the lower quartile of the limiting factor range and the upper tolerance limit of the current distribution, the mean of the limiting factor range (4.2 °C) is significantly higher (p<0.01)than the mean of the current distribution (2.8 °C).

Present-day limiting factors
Maximum temperature of the warmest month in areas to the north and northwest of the Zimbabwean population, and to the south and east of the Mozambican population (Figure 4), is tightly constrained between 35.9 °C and 37.2 °C (Figure 3).This maximum exceeds the upper limit of the 'most suitable' range of 35.2 °C for this variable established from the present-day distribution of the species.Despite substantial overlap between the upper tolerance of the current distribution range and the upper-and even the interquartiles for the limiting factor range, the means (36.6 °C versus 33.1 °C) differed significantly (p<0.01), with temperatures too high to support B. spiciformis woodland.
Correspondingly, precipitation seasonality (Bio15) and minimum temperature of the coldest month (Bio6) may restrict present-day distribution of the fragmented B. spiciformis ecological niche in South Africa (Figure 4).However, there is no significant difference (p=0.5499) between the limiting factor mean of adjacent minimum temperature of the coldest month (8.4 °C) and that of the 'most suitable' range (8.7 °C) for B. spiciformis (Figure 3).Consequently we rejected minimum temperature of the coldest month as a valid limiting factor for this species.Alternatively, precipitation seasonality's range is tightly constrained (70-78 mm) and falls short of the minimum precipitation seasonality value of 87.8 mm which bounds the present-day 'most suitable' range of the species (Figure 3).Although there is overlap between the interquartile range of the limiting factor and the lower tolerance limit of the current distribution range, the mean of the limiting factor range (74.3 mm) is significantly lower (p<0.01)than that of the current distribution (98.2 mm).

Present-day ecological niche
The fragmented ecological niche suggested for B. spiciformis in South Africa is a refinement of the Rutherford et al. 23 model, which predicted a similar distribution for this species under future global change conditions.Moreover, the presence of a population within one of the modelled ecological niche fragments suggests a refugium.Defined on climatic grounds, refugia are physiographical settings that can support a population once prevalent regional climates have been lost (or are being lost) as a result of climate shifts. 45The isolated B. spiciformis woodland relict at Gundani has therefore likely persisted in a refugium created by the varied topography of the Soutpansberg massif in South Africa.This topography has led to the formation of an isolated pocket of habitat that experiences the suitable climatic conditions (precipitation of the wettest quarter and temperature seasonality) that define the distribution of the species across the remainder of its current southern African range, and which prevailed at sites where Brachystegia occurred prior to the LGM and during the warmer and wetter conditions of the Mid-Holocene in the recent past.
Precipitation seasonality values in areas immediately adjacent to this isolated refugium are significantly lower than those for the ecological niche dimension of B. spiciformis woodland in the rest of southern Africa (Figure 3).Consequently, this bioclimatic variable reflects climatic constraints on the present-day ecological niche and isolated relict population in South Africa, potentially preventing the distribution of the species outside of the relict population.Similarly, high temperature   seasonality and maximum temperature of the warmest month (Figure 3) place climatic constraints on the extremities (southwest and northeast, respectively) of the continuous B. spiciformis woodlands of Zimbabwe and southern Mozambique (Figure 4).Although most savanna plants can tolerate maximum temperature extremes, Chidumayo 46 showed that some savanna trees are sensitive to seasonal highs.

Past ecological niche
According to the precipitation of the wettest quarter bioclimatic variable, conditions were considered suitable for the presence of B. spiciformis at the three Brachystegia pollen sites during the Mid-Holocene.However, temperature seasonality 6000 years BP was unfavourably high according to the present-day ecological niche dimensions of B. spiciformis woodland in southern Africa, casting doubt on the value of this climatic metric in controlling the distribution of B. spiciformis in its current range (from ENM), and in limiting its distribution north of South Africa's border (from limiting factors).However, two assumptions underlie most analyses of past climate using proxies and models.The first is that climate sets the boundaries to vegetation types, and therefore vegetation types are in equilibrium with climate except during the most rapid periods of climate change.Under this assumption, pollen-climate transfer functions can therefore provide reliable estimates of past climate.The second assumption is that, on long time scales, climate changes are driven by solar insolation changes. 47 should therefore be considered that precessional insolation in the southern hemisphere reached a minimum during the Early Holocene ~9000 years BP. 48Hence, there were reduced seasonal cycles and lower temperature seasonality.Subsequently, a steady increase in summer insolation, and therefore temperature seasonality, between 9000 and 6000 years BP may already have reached an unfavourable range for B. spiciformis as the Mid-Holocene bioclimatic layer would suggest.The presence of pollen at these sites indicates that vegetation type boundaries are not in equilibrium with climate during rapid response phases to climate change.Adult trees may persist in an area of previously suitable climate during extended periods of climatic constraint (the storage effect), particularly if populations are capable of adaptive dynamics such as clonal regeneration.Brachystegia may therefore be subject to an extensive period of persistence between vegetationclimate equilibrium and local extirpation.

Future ecological niche
In keeping with the approximate northeastward retraction of the species distribution over the past 6000 years, a further ecological niche retraction of between 30.6% and 47.3% of the continuous B. spiciformis woodland in Zimbabwe and southern Mozambique is predicted by 2050.The decreasing suitability of habitat at the periphery of the continuous miombo population is supported by climate projections for the subcontinent.
It has been suggested that southern Africa will become hotter (temperature will increase by up to 3 °C) and drier (precipitation will decrease by ~10%) by 2060 and 2100, respectively. 49,50The projected increase in heat waves and the number of days above 35 °C50 over the continuous miombo woodland region will subsequently amplify variability in mean monthly temperatures.This temperature variability is likely to stress the tolerance of B. spiciformis further by pushing maximum temperature of the warmest month and temperature seasonality into the upper tolerance limit of the species, and closer to the limiting factor mean (Figure 3).
1][52] The associated decrease in precipitation of the wettest quarter over central Zimbabwe 50 may therefore place climatic constraint on the future ecological niche as this variable, which is considered the most important in determining B. spiciformis distribution, will become increasingly unsuitable.
In contrast, the suitable B. spiciformis niche in South Africa is modelled to experience very large and inconsistent spatial shifts by 2050, varying between range retractions of 70.6% to a range expansion of 376.5% (Table 2).The large discrepancy between models likely relates to the coincidence with a topographical heterogenous area on the northeastern escarpment, where the central highlands of South Africa abruptly give way to low-lying coastal plains.Fine-scale topographical effects on climate, which seem necessary for refugia, are not well captured within downscaled AOGCMs, contributing to model uncertainty.
Nevertheless, future climate change projections for northeast South Africa, which include the B. spiciformis woodland relict, suggest hotter (temperature increase of ~0.9 °C) and wetter (precipitation increase of ~11%) conditions by 2100. 53Although subcontinental projections indicate increases in temperature seasonality, 50 we suggest that future change will be buffered by the varied topography of the Soutpansberg massif, as has occurred historically.We suggest that the subsequent increase in precipitation of the wettest quarter (~8% 53 ) will shift local climatic conditions at Gundani, from the present-day lower tolerance limit, into the 'most suitable' range identified for the species (Figure 3).However, the concurrent regional decrease (~6% 53 ) in precipitation seasonality may cause conditions to become less tolerable as rainfall shifts away from the 'most suitable' range of B. spiciformis woodland (Figure 3).
Considering that there may be greater uncertainty with regard to the impact of climate on the ecological niche of B. spiciformis at the distribution edge, as compared to the continuous range, we should consider all three possible responses of populations to climate change: migration, adaptation and extinction.

Migration
It is unlikely that B. spiciformis woodland is capable of unaided range expansion through long-distance dispersal events, particularly given the very limited dispersal distances reported for the genus.There are limited untransformed or protected areas that could facilitate the establishment of new populations within the suggested ecological niche.In addition, current land-use practices have led to large-scale fragmentation which does not allow natural corridors for population expansion through clonal regeneration.Subsequently, any predicted range expansion of B. spiciformis woodland in South Africa along the eastern escarpment is considered unlikely as a result of anthropogenic constraints.

Adaptation
The persistence of populations as relicts may result from pockets of environmental suitability within the landscape, or may be the consequence of some adaptive dynamics by which species manage to overcome the climatic constraints posed by a changing climate.For example, plant ecology studies indicate climatic constraints on seed production. 54Consequently, rather than relying on pollination and seed production, many climate relicts rely more strongly on vegetative or clonal reproduction.The well-documented reliance of B. spiciformis on coppice regrowth and root suckering, rather than seed germination, both at the Gundani relict 27 and in the continuous woodland 6 suggests that even in the event of failed recruitment through climatic intolerance, the populations would persist though vegetative reproduction and the longevity of genets.Relict populations can furthermore improve their survival prospects by enlarging their climatic tolerance through phenotypic plasticity or micro-evolutionary adaptation. 55Once the capacity for phenotypic adjustment is exceeded, genetic adaptation remains the only option. 22wever, several characteristics of relict populations imply that their potential for micro-evolutionary adaptation may be limited.Firstly, strong phenotypic divergence, compared with conspecifics from other parts of the range, does not appear to be a common phenomenon despite the presumed long-term exposure of relict populations to relatively strong selective pressures. 56Secondly, within-population genetic variation tends to be low in many climate relicts as a consequence of small population size, past bottlenecks 3 and the occurrence of clonality.Thirdly, the small size of many relict populations, in combination with strong selection pressure, is likely to substantially elevate their risk of extinction as a result of demographic or environmental stochasticity before they can effectively adapt. 56

Extinction
Although broadscale changes in species distribution can reasonably be forecast, 21 our understanding of the environment and ecology of most climate relicts remains too poor to adequately anticipate their persistence or demise. 22The historical resilience of such populations demonstrates that we cannot simply assume that they will be extirpated rapidly after climate change. 22The storage effect suggests that individuals of the species may be present in an area long after the climate has become marginally tolerable, or completely unsuitable, depending on the life stage on which the climate control may act.The isolated B. spiciformis woodland plays an important role in the natural heritage of the local people and has been placed under a traditional woodland management regime.The tree is not currently used ethnobotanically and the direct threat of anthropogenic extinction is considered unlikely.
With future climate-induced migration ruled out and opportunities for adaptation beyond phenotypic adjustment limited, what then is the fate of South Africa's only miombo woodland?We suggest that the medium-term persistence of the relict is plausible based on (1) future climate change projections (2100), which suggest that local climatic variables will remain within (or nearly so) the 'most suitable' range of B. spiciformis, (2) the longevity of genets and the species' ability to regenerate vegetatively and (3) its historical resilience.
The suggested ecological niche retraction of the continuous B. spiciformis woodland in Zimbabwe and southern Mozambique may, however, be less exaggerated.Bioclimatic models are limited in their ability to determine the full extent of ecological interactions, especially in savannas in which the direct impact of climate on species distribution may be variable.One such example is that of future increases in atmospheric CO 2 .Species with large below-ground carbon sinks, like miombo, benefit from CO 2 fertilisation 57 which increases tree water use efficiency. 58This tree water use efficiency is consequently likely to contribute towards a greater tolerance of suggested decreased precipitation of the wettest quarter.

Conclusion
Through this unique study, we have identified the likely climatic determinants responsible for range shift at the trailing distribution margin of B. spiciformis in southern Africa.
Given the suggested divergent future responses of the continuous woodlands of Zimbabwe and Mozambique relative to that of the South African relict, an understanding of population dynamics and demographics across the subcontinent will prove critical in validating our predictions concerning the climate response phases (migration, adaptation, persistence) of B. spiciformis and elucidating the lifehistory stages during which climatic constraints may be imposed.
Understanding both historical and contemporary climatic determinants of a species range is the first step towards assessing the vulnerability of extant climate relicts under global climate change.Theoretical responses must then be supported by in-situ monitoring of the populations.
Climate relicts have value as instructive models and natural laboratories for investigating how populations react to ongoing climatic change. 59eyond the scope of this study exists an opportunity to explore genetic variation within the B. spiciformis relict population and investigate potential micro-evolutionary adaptation since isolation.Furthermore, development of long B. spiciformis and B. utilis tree-ring chronologies from the relict would allow for the investigation of temporal El Niño Southern Oscillation (ENSO) variability and predictions of future regional effect.Considering that ENSO effects on precipitation variability are strongest in southern Africa during the wettest quarter, 14 the investigation may assist to adequately anticipate the persistence or demise of the relict population.

Figure 1 :
Figure 1: Distribution of miombo woodland in Africa.

Figure 2 :
Figure 2: Ecological niche of Brachystegia spiciformis woodlands in southern Africa under present-day climate conditions.Location of the Gundani climate relict population (black circle) and three Mid-Holocene pollen records (black triangles) are included.
Climate values falling outside of the present-day fundamental niche or suitability range of B. spiciformis woodland in southern Africa place a climatic constraint on the species, and can therefore be considered as limiting factors to its geographical footprint.Three bioclimatic variables were identified as potential limiting factors for B. spiciformis.Temperature seasonality (Bio4) and maximum temperature of the warmest month (Bio5) are suggested to be responsible for restricting distribution of the continuous B. spiciformis woodlands of Zimbabwe and southern Mozambique (Figure 4).The temperature seasonality range in areas to the southwest of the present-day distribution in Zimbabwe is broad Research Article Climate-induced range shift in Brachystegia (miombo) woodland Page 4 of 9 The box plot represents the fundamental niche, with the interquartile range (box with median present) representing 50% of the total data values and the upper and lower whiskers representing 25% of the values, respectively.The interquartile space is interpreted as the 'most suitable' range of climatic or environmental conditions for the species, while the lower and upper whiskers span less suitable conditions, reflecting lower and upper tolerance limits, respectively.Extreme values, beyond the 1.5 coefficient value, are shown as open circles; palaeohistoric (solid triangle) and relict population (solid circle) values are included for comparison.

Figure 3 :
Figure 3: Comparative box plots for the present-day range (76 quarterdegree-square (QDS) centroid points) of the ecological niche of Brachystegia spiciformis woodland in southern Africa and a limiting factor (76 adjacent, randomly selected QDS centroid points) for five bioclimatic layers (a) precipitation of wettest quarter, (b) temperature seasonality, (c) maximum temperature warmest month, (d) minimum temperature coldest month and (e) precipitation seasonality.

Figure 4 :
Figure 4: Limiting factors of the present-day ecological niche of Brachystegia spiciformis woodland in southern Africa (excluding Lesotho and Swaziland).

Figure 5 :
Figure 5: Predicted future (2050) ecological niche for Brachystegia spiciformis woodland in southern Africa (dark grey) under the A2a scenario across three general circulation models: (a) HadCM3, (b) CGCM2 and (c) CSIRO-MK2.The present-day ecological niche (light grey) is included for comparison.

Table 1 :
Area under the curve (AUC) scores for top-performing bioclimatic variables explaining current and three future distributions for Brachystegia spiciformis in southern Africa

Table 2 :
Percentage ecological niche change by 2050 for Brachystegia spiciformis woodland in southern Africa according to three different climatic models under the A2a scenario