Developing conservation strategies for endemic tree species when faced with time and data constraints: Boswellia spp. on Socotra (Yemen)
- First Online:
- Cite this article as:
- Attorre, F., Taleb, N., De Sanctis, M. et al. Biodivers Conserv (2011) 20: 1483. doi:10.1007/s10531-011-0039-7
- 183 Views
Many endemic tree species have important scientific, ecological and economic value but the scarcity of information about their biological and ecological features makes it difficult to develop conservation strategies for them. A four-step approach is presented to address this problem, based on the analysis of data collected in a limited-duration field study: (1) Data collected are used to analyse the ecological niche, population structure and regeneration status of the species in question. (2) Several IUCN Red List (RL) parameters, useful for assessing the species’ risk of extinction, are measured, including population counts, number of locations, extent and area of occurrence. (3) The IUCN RL parameters are used together with the other information gathered to set preliminary conservation priorities. (4) The analysis of utilization pattern is used to develop conservation actions specific to the environmental and socio-economic context. To test the applicability of this approach Boswellia spp. of Socotra island were analysed. Ground-rooted species (B. ameero, B. elongata and B. socotrana) were the most abundant and widespread and, according to the spatial analysis, were characterised by a geo-altitudinal zonation. However, the Weibull functions fitted on their stem diameters, and the absence or presence of only a small number of saplings highlighted a poor regeneration status. In the absence of conservation actions, these species will probably be subject to a progressive decline because of uncontrolled grazing. Of the four cliff-rooted species, which grow in sites that are less accessible to livestock, two (B. popoviana and B. dioscorides) were of lower conservation priority and may become the most numerically abundant. Conversely, the other two (B. nana and B. bullata), which may be threatened by stochastic events because of their reduced populations and small number of locations, were considered of very high priority. Different conservation actions were then identified for each species. In particular, for Boswellia species producing gum, the conservation-through-use action was discussed as a potential option.
KeywordsIUCN Red ListEcological niche modellingRandom ForestSoil parametersWeibull function
Local climate is influenced by large-scale weather phenomena, particularly the seasonally reversing monsoon: the NE (winter) monsoon, which lasts from November to March, and the SW (Indian or summer) monsoon, which blows from May to September. Mean annual rainfall and temperature are 216 mm and 28.9°C respectively, but the former varies across the island with exposure and elevation: on the coastal plains the annual precipitation may be as little as 125 mm or it may even be absent altogether, while in the mountains it can reach exceptional levels of 1,000 mm, of which preliminary measurements suggest that fog-derived moisture may constitute over two-thirds of the total (Scholte and De Geest 2010).
The long isolation of the Socotra archipelago has contributed to the creation of a unique flora (Miller and Morris 2004), and the sustainable traditional land practices have allowed such a “hot spot” of biological diversity to be conserved: Socotra hosts 308 endemic flowering plant species out of an estimated 825, a 37.3% level of endemism (Miller and Morris 2004). The semi-arid climate has favoured frankincense tree species belonging to Boswellia genus (Thulin and Al-Gifri 1998), which comprises 20 species distributed in the semi-arid regions of tropical Africa, the Arabian Peninsula and India (Gillet 1991; Vollesen 1989). Boswellia genus represents one of the most interesting examples of adaptive radiation on Socotra (Turelli et al. 2001) and this process has led to the evolution of seven endemic species, giving the relatively small island the highest concentration of frankincense tree species in the world (Thulin and Al-Gifri 1998).
Boswellia species of Socotra have two distinct growth forms (Miller and Morris 2004; Mies et al. 2000): the so-called cliff-rooted species (B. bullata, B. dioscorides, B. nana, B. popoviana); and the ground-rooted species (B. ameero, B. elongata, B. socotrana). Since early historical times, frankincense has been produced mainly from two ground species, B. elongata and B. socotrana, and then used for house cleansing and in traditional and veterinary medicine, or exported like other valuable natural products such as aloe and cinnabar, the resin of the well-known Dragon’s blood tree, Dracaena cinnabari.
Grazing by livestock, mainly goats, has been recognised as the main factor hindering the regeneration of Dracaena cinnabari (Adolt and Pavlis 2004; Attorre et al. 2007a; Habrova et al. 2009), and the same impact could be expected for Boswellia species. Moreover, foliage loping by herders in the dry months to feed their livestock can further stress the trees, making them less resistant to the predicted increasing drought (Miller and Morris 2004). The decline of Boswellia species may not only increase their risk of extinction, but also lead to the disappearance of a potential source of income for the rural population, an increase in soil erosion processes and a reduced appeal for eco-tourism.
The scientific, ecological and economic importance of these endemic tree species motivates conservation biologists to develop specific management strategies. However, such a pursuit can be severely hampered by the scarcity of information about their distribution, ecological characteristics, population structure, growth, sexual reproduction and dispersal mechanisms, genetic characteristics and isolation and past trends. The current generalised dearth of such detailed knowledge is primarily due to the relatively long time needed to collect relevant data. Trees, for example, account for the majority of plant species assessed in the IUCN Red List (RL), but this assessment has been carried out mainly according to the B Criterion on geographic range size (Newton and Oldfield 2008). The generalised tendency to gather information from herbarium data, which also explains the abovementioned bias towards trees, limits the possibility for analysis to carrying out mere preliminary evaluations on species’ conservation status (Lughadha et al. 2005).
In this context, Boswellia spp. of Socotra are of particular interest because their different growth forms and uses make comparative analysis a useful tool for testing the validity of a methodological approach for developing appropriate conservation strategies when time for data collection is limited.
Materials and methods
Analysis of the ecological niche (even spatially explicit), population structure, regeneration status of tree species.
Assessment of the extinction risk according to a set of IUCN parameters that are applicable when data about past population trends are unavailable: population counts, determination of the number of locations, measure of extent and area of occurrence.
Setting of conservation priorities by integrating the abovementioned information; this step is needed because of intrinsic limitations in the IUCN RL classification, namely while tree species can be classified under the same extinction risk category they may differ in actual conservation priority given their population structure and regeneration status (de Grammont and Cuaron 2006; Mace et al. 2008).
Identification of conservation actions that are specific to the ecological characteristics of each species and to the environmental and socio-economic context.
Moreover, the results of our analyses were compared with those obtained by several recent studies on the relationship between the spatial distribution, population structure and regeneration processes of Boswellia papyrifera and environmental factors and land use practices in the Horn of Africa (Ogbazghi et al. 2006a, b; Rijkers et al. 2006; Negussie et al. 2008).
In order to analyse the ecological niche and the population structure of Boswellia spp., all known areas characterised by the species were sampled in February–March 2007 using bibliographical information (Miller and Morris 2004; Mies et al. 2000; Thulin and Al-Gifri 1998) and indications provided by the local staff of the Environmental Protection Authority of Socotra. 108 sample plots, with a spatial resolution of 20 × 20 m for the cliff species and 50 × 50 m for the ground species, were identified using GPS coordinates. Dendrometric surveys were then carried out by measuring with a tape the circumference at breast height (1.30 m) of all the trees with a diameter equal to or over 2 cm. For B. nana which, due to the combined effects of strong wind and grazing, grows prostrate, stem diameter was measured at the base. The number of saplings was also recorded.
Limited to ground species, in each plot five soil samples were collected from the upper layer, one at the centre and four at random distances from the centre in different perpendicular directions. The soil samples were then pooled and stored in plastic bags for the chemical-physical analyses. In the laboratory, the samples were oven-dried, crushed in a mortar and passed through a 2-mm mesh sieve before being analysed using standard techniques (MIPAAF 2000): pH with the potentiometric method in a 1:2.5 (v/v) soil:water suspension; organic carbon with the Walkley–Black wet combustion method and organic matter, by multiplying values for organic carbon by the Van Bemmelen factor of 1.724; and total nitrogen (N) with the Kjeldahl method. To estimate available phosphorus (P), the Bray and Kurtz method was used for soils with pH < 6.5 and the Olsen method for soils with higher pH values. Exchangeable bases (sodium, potassium, calcium and magnesium) were measured by atomic absorption spectrophotometry, using ammonium acetate as the extractant. Electrical conductivity was determined in a 1:5 soil:water suspension. Finally, the particle size analysis was performed with the hydrometer method.
Abundance and spatial distribution
In order to investigate the effect of the environmental parameters on the abundance and distribution of Boswellia species two separate analyses were performed. In the first, limited to ground species, the relationship between abundance and environmental factors such as topographical, climatic and pedological parameters was investigated by means of a forward stepwise multiple linear regression. In the second, the most important environmental variables influencing the current spatial distribution of Boswellia species were identified in order to produce maps of their potential distribution. IVs were used as dependent variables, while climatic, topographical and geological data, in GRID format with a spatial resolution of 100 m, were used as independent variables. Climatic maps were obtained by interpolating precipitation and temperature data recorded in 10 manual meteorological stations and calculating the average data for the 2000–2008 period (Attorre et al. 2007b). Besides mean annual temperature and annual precipitation, a moisture index (Mi) was used. Mi is based on: Mi = P/PET, where P is the mean annual precipitation and PET is the potential evapotranspiration. A simplified geological map, including granitic, limestone, alluvial and sand substrata, was used as a surrogate for a pedological map, which still does not exist for the whole area.
For the statistical method a Random Forests (RF) model was used since it has proved to be efficient in predicting the spatial distribution of tree species (for technical details see Benito Garzòn et al. 2006; Prasad et al. 2006; Scarnati et al. 2009). Since RF requires records of absences, the use of presence-only data would bias the analysis and lead to overoptimistic predictions of the potential distribution. However, in the case of rare tree species such as Boswellia, absence data are difficult to obtain: a given location may be classified in the ‘absence’ set both when for historical reasons the species is absent even though the habitat is suitable and when the habitat is truly unsuitable; and only the latter is relevant for predictions. When no true absence data are available, one approach is to generate ‘pseudo-absences’ and to use them in the model as absence data for the species. In this case the method proposed by Engler et al. (2004) for rare and endangered species was chosen: first a suitability map was produced through the ecological niche factor analysis (ENFA) implemented in the BIOMAPPER package (Hirzel et al. 2002), then for each species a number of pseudo-absences equal to that of presences was randomly chosen in areas whose ENFA prediction was lower than the lowest one associated with observed presences.
To examine the population structure of Boswellia species a Weibull distribution curve was fitted on pooled stem diameters (DBH). This function has proved to be very flexible and suitable in fitting stem-size distribution data for ecological applications (Bailey and Dell 1973; Tanouchi and Yamamoto 1995) and it is popular with modellers dealing with uneven-age stands (Kamziah et al. 2000; Zhang and Liu 2006).
The scale parameter is approximately equal to the median DBH while the shape parameter controls the skewness of the distribution. When 0 < γ < 1, the curve approaches an inverse J-shape distribution, indicating abundant young individuals and a very good regeneration status of the species; for 1 < γ < 2.6, a positively skewed distribution with a right tail is expressed, meaning a few young individuals compared to the individuals of average age and indicating a medium regeneration status; for 2.6 < γ < 3.7, the coefficient of skewness approaches zero and the distribution approximates the normal with no tail, while for γ < 3.7 the distribution is negatively skewed to the left, indicating in both cases a poor regeneration status.
The conservation assessment was performed by measuring/calculating several IUCN RL parameters: population counts, number of locations, extent of occurrence (EOO) and area of occurrence (AOO).
The number of locations was calculated by applying the circular buffer method to the georeferenced plots. A radius of 1/10th of the maximum plot distance was used as suggested by Rivers et al. (2010) in a comparative study of the endemic species of the genus Delonix s.l. in Madagascar. Here the term locations rather than subpopulations was used because of the current lack of knowledge about the reproductive isolation and dispersal ability of Boswellia species.
EOO was estimated using the α-hull method (Burgman and Fox 2003), which provides a more detailed representation of the space contained by the sample plots than the traditional convex hull method. AOO, which is the part of EOO occupied by a taxon, was calculated using two methods. In the first it was estimated using a grid square of 2 km as recommended by IUCN (2010). In the second only the suitable area as estimated by the spatial modelling within EOO was considered. The use of spatial modelling to estimate range parameters is not new (see Sérgio et al. 2007) and in our study it is proposed as a way of avoiding the need to identify the appropriate resolution scale for AOO, which is a potential source of inconsistency and bias (IUCN 2010).
Abundance and spatial distribution
Altitudinal range, annual precipitation, mean annual temperature, number of plots, number of trees sampled, mean tree density and cover of Boswellia species
Altitude range (m)
Tot. prec. (Mm)
Mean temp. (°C)
No. of plots
No. of trees
Mean density n. tree/1002
Mean cover (m2/1002)
Physical and chemical soil properties of ground-rooted Boswellia species (B. ameero, B. elongata, B. socotrana)
Organic matter (%)
Available P (ppm)
Na (me/100 g)
K (me/100 g)
Tot exch. bases (me/100 g)
Soluble salts (me/100 g)
Results of the forward stepwise linear regression analysis to assess the importance of environmental factors on the abundance of ground-rooted species measured as IVs
The spatial modelling analysis was not performed for B. nana and B. bullata due to their small number of plots. For the other species, the potential distribution maps were produced according to the minimal predicted area (MPA), which is particularly useful when modelling the potential distribution of rare species. In fact, a model that predicts species presence everywhere might give the best evaluation results (because all presences would then be effectively predicted as presences), but such a map would be of no use for conservation purposes. On the contrary, consistently with the MPA criterion, the model should predict the smallest possible potential area while still covering a maximum number of species’ plots. To calculate the MPA, for each species the threshold of predicted IV scores was identified that encompassed 90% of species’ plots (rule of parsimony), below which predictions were set as zero.
It is worth noting that the three ground species show a stratified distribution linked to a geo-altitudinal gradient, with B. socotrana being potentially more widespread in the alluvial basal areas, B. elongata in the upper hill limestone plateau and B. ameero in the central mountain areas of the Hagghier characterised by a granitic substratum (Fig. 3a, d, g). Of the cliff species, B. dioscorides has the widest potential distribution while B. popoviana is more localised and restricted (Fig. 3c, f).
Variable importance predicted by RF for the Boswellia spp.
Population structure and regeneration status: descriptive statistics and Weibull distribution function parameters γ (shape) and α (scale) of Boswellia sp. stem diameters
No. of saplings
B. dioscorides had the largest number of saplings, while B. socotrana was the only ground species for which small trees and saplings were observed, albeit only in rocky fissures or under the protection of spiny shrubs.
AOO area of occurrence (km2) calculated with a 2 km grid resolution, AOO* area of occurrence considering only the potential area within the extent of occurrence (km2), EOO extent of occurrence measured with the α hull method (km2), No. of locations estimated with a buffer radius of 1/10th of the maximum sample distance
EOO α hull
Abundance, distribution and population structure
Despite being currently more abundant and widespread, ground species show signs of decline: B. ameero and B. elongata have a population structure characterised by a poor regeneration status. B.socotrana is the only ground species that shows a medium regeneration status with several small trees and saplings recorded, albeit solely in sub-vertical slopes or surrounded by thorny shrubs and thus inaccessible to livestock (Tables 1, 5 and Fig. 2a). This result supports the suggestion that, as for Boswellia papyrifera stands in the Horn of Africa (Negussie et al. 2008; Ogbazghi et al. 2006a), grazing is one of the most important human-induced factors affecting the population structure of ground Boswellia species of Socotra.
Grazing can have a significantly negative impact on the regeneration of ground species, leading to the removal of whole plants or parts of their vegetative or reproductive structures; in addition, trampling of the soil surface by animals can result in seedling mortality and soil compaction that limits recruitment and seedling establishment. Moreover, seedlings are vulnerable to grazing: being succulent and palatable, they are attractive to grazers and they also grow too slowly to escape them. Grazing by livestock has been taking place on Socotra for thousands of years and it is widespread throughout the island. The traditional semi-nomadic pattern, with livestock populations fluctuating markedly in accordance with periodic droughts, has probably allowed several populations of Boswellia ground species to survive by mass regeneration in the aftermath of droughts, when grazing pressure has remained low while rainfall has re-established (Scholte and De Geest 2010). This has also been hypothesised for the dragonblood tree based on the analyses of the fairly homogenous forest stands (Adolt and Pavlis 2004; Attorre et al. 2007a; Habrova et al. 2009). However, road construction and the spread of such practices as purchasing and stocking grain and fodder may lead to a reduction of the spatial and temporal grazing dynamics (Scholte et al. 2008), exacerbating the effects of grazing on the regeneration of ground species.
Land clearing and tapping trees for resin are the other factors that can cause the decline of Boswellia species, but they were negligible in Socotra.
Land clearing is the most extreme case as trees are removed from cultivated land, but it does not seem to have ever occurred on Socotra where agricultural activities are very limited and also in decline due to the increased access to food following the opening of the airport in 2001.
Tapping is reported to have a severe impact on the natural regeneration of B. papyrifera (Rijkers et al. 2006). In fact the hypothesised competition between the investment of carbohydrates in sexual reproductive structures and the synthesis of frankincense may cause over-exploited trees to produce fewer flowers, fruits and seeds than those that are exempt from tapping. On Socotra frankincense is produced mainly from B. elongata and B. socotrana, but tapping was observed only in one tree of the latter species. The reduced economic interest in this non-timber forest product has led to the collapse of traditional silvicultural activities; consequently, the harvest of frankincense gum is much less regulated and droplets of resin are collected freely and sold in the local market and to tourists (Miller and Morris 2004).
Of the cliff species, two show a geographical segregation: B. nana, which was found only in one location in the eastern Hamaderau mountain with only 23 individuals counted in three plots; and B. bullata, which was only recorded in three locations on the western part of the island (Table 6, Fig. 3b). The other two (B. dioscorides and B. popoviana) are more widespread and share the same habitat, their distribution being mainly limited to limestone cliffs and escarpments and determined by slope and potential evapotranspiration (Fig. 3c, f and Table 4).
All cliff species have a shape parameter (γ) of the Weibull function of between 1 and 2.6 and a clear right-skewed distribution curve and so can be attributed a medium regeneration status (Table 5, Fig. 2b). This result may be explained by the fact that they are able to germinate and grow in rocky fissures in sub-vertical escarpments inaccessible to goats. However, B. nana has unusual characteristics, including a limited distribution, a small number of individuals and no saplings, which make it much more vulnerable to extinction.
Conservation assessment and priorities
Assessment of the conservation status, priority for conservation and conservation actions for Boswellia spp.
Current IUCN RL category
Proposed IUCN RL category
Priority/actions for conservation
Very high/plant nursery, seed bank
VU B2ab(ii, iii)
Very high/plant nursery, seed bank
Very high/plant nursery, seed bank
EN B2 ab(v)
High/Enclosure programs conservation-trough-use, seed bank
High/Enclosure programs, conservation-trough-use, seed bank
Conservation priorities were then set by analysing all the above information. In particular, B. nana, B. bullata and B. ameero are considered of very high priority for conservation because they are more prone to threats from environmental factors and stochastic events. The first two species are characterised by small populations and very localised distributions, while B. ameero has a poor regeneration status, a more specialised environmental niche and the smallest AOO and EOO of the ground species. The two remaining ground species, B. elongata and B. socotrana, though classified respectively as endangered and vulnerable, are considered only of high priority because they have relatively large populations even though they are strongly affected by grazing. B. popoviana and B. dioscorides are ranked of medium priority because of their relative abundance, medium regeneration status and comparatively large number of saplings. This corroborates the hypothesis that, in the absence of any effective conservation actions, they are likely to become the most numerous Boswellia species on the island.
Specific conservation actions are ascribed to each of the species based on their particular ecological features, environmental context and economic value. These include the establishment and running of a seed bank, a plant nursery, enclosures and conservation-through-use programmes. Some such conservation actions have already been implemented and include carrying out a field study to collect seeds according to internationally agreed standards, growing saplings of all the species from cuttings in a plant nursery and implementing enclosure projects to verify the effect of grazing on the regeneration of B. elongata: one of these, implemented in 2008 in the Homil plateau by the GEF-UNDP “Socotra Conservation and Development Programme”, resulted in high seed production and germination during the rainy season, but none of the seedlings survived the dry summer monsoon (Nadim Taleb pers. comm.). This outcome, which was probably due to the excessive hardness of the eroded soil, was similar to that of Negussie et al. (2008), who recorded a high seedling mortality even after years of successfully establishing enclosures, and calls for additional types of intervention, from appropriate soil preparation to shading seedlings.
The conservation-through-use option is considered only for B. elongata and B. socotrana, the two ground species producing a quality of gum that is valued as a potential source of extra income for rural populations (Newton 2008). However, in order to effectively pursue this conservation option several activities need to be undertaken which, given the current limited skills of local agencies and institutions, require the support of the international scientific community. These include tapping and collection trials, chemical analyses of the extracts, application of agroforestry improvement techniques (time of seed collection, nursery practices, choice of appropriate planting sites and post-planting care), value-adding processing practices, international and domestic market analyses (Gebrehiwot et al. 2003).
Moreover, the progressive collapse of the traditional system of land management means that local communities need to be involved in forest protection and management activities; their participation can be ensured by providing incentives, including in the form of ownership/user rights and benefit-sharing mechanisms.
The methodological approach developed in the present study specifically addresses analytical difficulties that recur when the scarcity of information hinders the identification of conservation strategies and actions for endemic tree species. This applies particularly when, due to their relatively short duration, field studies aimed at improving the effectiveness of a given conservation action only allow a limited amount of information to be collected for use in a structured way.
Our operational context made it possible to sample all the known areas characterised by endemic Boswellia spp. and, therefore, to further knowledge about several important topics including their actual and current potential distribution, regeneration status and the environmental and human-induced factors influencing them. However, even in larger or more inaccessible areas where favourable conditions such as those found on Socotra cannot be met, this approach could be fruitfully applied if supported by a statistically defined sampling procedure.
Moreover, the data and information obtained in this study provide a strong basic framework for further research that can only be carried out at a later stage and with more time available, such as into pollination and dispersal mechanisms, phylogenetic relationship and genetic isolation of the identified populations. By increasing the knowledge of the ecological and biological characteristics of the species, this research could support the further refinement of the conservation strategies and options set out in the present study.
The results of the present study highlight the effectiveness of the method developed particularly when faced with time constraints, and suggest that it can be fruitfully consolidated through application to other endemic tree species and different contexts, particularly when preceded by cost-benefit analyses and factual estimates of the prospects of success versus logistical constraints.
Special thanks go to the personnel of the Environmental Protection Authority of Socotra for their logistic support and help during the field work, and in particular to Ahmed Adeeb, Ahmed Saed, Ahmed Issa, Mohammed Nageeb and Fahmi Bahashwan. We are also grateful to Edordo Scepi and Alessio Mastroianni for their invaluable help during field surveys, to Giuseppina Downgiallo for the soil analyses and to Laura Clarke for reviewing the text. This work was supported by the Socotra Conservation and Development Program funded by GEF-UNDP, DGCS the Italian Development Cooperation and the Government of Yemen.