Introduction
Reintroductions, defined as attempts to establish a species in areas within its historical range where it has gone extinct, have become an accepted intervention in conservation (Seddon et al. Reference Seddon, Armstrong and Maloney2007, Armstrong and Seddon Reference Armstrong and Seddon2008). Such programmes are often accomplished via the release of captive-bred individuals (IUCN 1998, Moorhouse et al. Reference Moorhouse, Gelling and Macdonald2009, Bernardo et al. Reference Bernardo, Lloyd, Olmos, Cancian and Galetti2011) and post-release monitoring is necessary to confirm the validity of reintroduction as a cost-effective tool. Such monitoring can provide (1) information to assess the success of the release operation (IUCN 1998); (2) unique opportunities for clarify the species’ niche requirements because they are likely to colonise the most suitable habitats first (Hirzel et al. Reference Hirzel, Posse, Oggier, Crettendand, Glenz and Arlettaz2004); and (3) fill gaps in knowledge (population biology, community ecology and conservation) of the species involved (Burnside et al. Reference Burnside, Carter, Dawes, Waters, Lock, Goriup and Székely2012).
The Double-spurred Francolin Francolinus bicalcaratus (Linnaeus 1766) is found in tropical West Africa and also in Morocco where an isolated subpopulation occurs as a local resident (Thévenot et al. Reference Thévenot, Vernon and Bergier2003). Francolinus bicalcaratus ayesha is a subspecies endemic to Morocco. While Francolinus bicalcaratus has a conservation status of ‘Least Concern’ in the IUCN Red List (BirdLife International 2012, IUCN 2012), the Moroccan subspecies ayesha is reported as ‘Critically Endangered’ by Thévenot et al. (Reference Thévenot, Vernon and Bergier2003), McGowan et al. (Reference McGowan, Dowell, Carroll and Aebischer1995) and El Agbani et al. (Reference El Agbani, Qninba, Radi, El Hamoumi, Cherkaoui, Himmi, Bouajaja and Dakki2011). Currently, this subspecies occurs in the hinterland of Rabat-Casablanca in localities near Sidi Yahia des Zaër, Sidi Bettache and Ben-Slimane (Thévenot et al. Reference Thévenot, Vernon and Bergier2003). However, it seems to have vanished from Souss (south-west Morocco) where it was common in the early 1920s and for which the most recent record was of 1–2 birds in 1987 (Thévenot et al. Reference Thévenot, Vernon and Bergier2003). Its diet consists mainly of grains, seeds, berries and insects (Alaoui Reference Alaoui, Kassinis and Panayides2001). The population of F. b. ayesha has been reduced due to hunting and habitat destruction (Thévenot et al. Reference Thévenot, Vernon and Bergier2003). Given this situation, the reestablishment of a viable breeding population is desired to restore its status. Consequently, as part of its strategic efforts to strengthen wild populations, Morocco’s High Commission for Water, Forests and Desertification Control, in collaboration with the Royal Moroccan Federation of Hunting and the Captive Breeding Centre ‘Domaine la Gazelle-Gibiers’ has carried out a Double-spurred Francolin reintroduction programme.
Studying the habitat use of released animals can provide valuable insights to their responses in a new environment (Attum et al. Reference Attum, Otoum, Amr and Tietjen2011). This refers to the way in which an individual or species uses habitats to meet its survival needs (Block and Brennan Reference Block, Brennan and Power1993). This kind of study is also the most effective way to assess the ecological requirements of a species and to set management guidelines to assist in conservation of populations. Indeed, knowing habitat factors is an important step in the development of effective conservation strategies, particularly for endangered species (Conway and Martin Reference Conway and Martin1993, Pasinelli Reference Pasinelli2000).
In Morocco, the habitat preferences of the Double-spurred Francolin are poorly known. Historically, a variety of habitats are reported to have been occupied by the subspecies ayesha (De la Perche Reference De la Perche1992, Thévenot et al. Reference Thévenot, Vernon and Bergier2003), among them, dense matorral (thickets) of wild olive Olea europaea and lentisc Pistacia lentiscus, open woodlands of thuja Tetraclinis articulata, and holm and cork oak Quercus ilex and Q. suber, but no clear pattern has emerged. Moreover, there is a complete lack of quantitative data relating to the habitats used by this ‘Critically Endangered’ subspecies.
We therefore studied data from a north-west Moroccan woodland, known to have supported a wild Double-spurred Francolin population in the past. Occupied/random data were recorded during planned surveys, as these are clearly preferable to occupied-only data in habitat modelling studies aimed at delineating niche boundaries (Franklin Reference Franklin2009).
Since habitat quality at a release site is recognised as being critical to the success of a species reintroduction program (Bennett et al. Reference Bennett, Doerr, Doerr, Manning, Lindenmayer and Yoon2012), our major aim was to identify and quantify those characteristics of forest structure associated with the habitats chosen by F. b. ayesha. In this way we sought to provide reliable recommendations for future reintroductions adequate to meet the conservation needs of the subspecies. We anticipate that, in future, the results of this investigation will serve as an invaluable basis for conservation and management of this critically endangered subspecies.
We generated two related hypotheses: (1) that the Double-spurred Francolin will use forest areas with a high density of cork oak Quercus suber and a high shrub cover, and (2) that released birds will remain around the release point provided that the habitat is suitable; i.e. the habitat used both where a bird was released and on its habitat requirements.
Methods
Study area
Double-spurred Francolins were reintroduced and monitored at a game reserve at Sidi Allal Al Bahraoui (SABGR) (34°00’52”N, 6°28’18”W; Figure 1), which is located within the Ma’amora forest, Morocco. This forest is classified as a Biological and Ecological Interest Site by the Moroccan Protected Areas Study (AEFCS 1996). SABGR is a fenced area of about 493 ha. It is situated near the city of Sidi Allal Al Bahraoui, at 210 m asl, in the canton C of Ma’amora where annual precipitation averages 450 mm, and monthly temperatures vary from 12°C (January) to 25°C (July–August). This region is characterised by hot summers and mild winters (a semi-arid bioclimate). The study site is managed and livestock, such as sheep that are common in the region, are not permitted.
Figure 1. Map showing the location of the game reserve of Sidi Allal Al Bahraoui (SABGR) in north-west Morocco. Black points represent the location of francolins and white delta points represent the location of random points in the wooded matorral.
The vegetation of this part of Ma’amora is characterised by Araceae, Arecaceae, Cistaceae, Fabaceae, Fagaceae and Lamiaceae. The tree layer consists solely of cork oak and the shrub layer is dominated by needle-leaved broom Teline linifolia, Mediterranean dwarf palm Chamaerops humilis, sage-leak rock rose Cistus salviifolius and Spanish lavender Lavandula stoechas. The forest landscape is dominated by two habitats: (1) matorral (269 ha) and (2) wooded matorral (224 ha), which is characterised by cork oak in association with the shrub species mentioned above.
Game management in the study area is aimed mainly at Double-spurred Francolin conservation. The Royal Moroccan Federation of Hunting has set up forest trails, in which water points have been established as well as food supplements mostly composed of bread wheat Triticum aestivum.
Release and population source
As stipulated in IUCN guidelines (1998), the SABGR was selected as a suitable site for Double-spurred Francolin reintroduction because: (1) there were no longer any Double-spurred Francolins in the Ma’amora forest (Thévenot Reference Thévenot, Villemant and Fraval1991, Cherkaoui et al. Reference Cherkaoui, Dakki, Selmi, Rguibi Idrissi and Thévenot2007, Reference Cherkaoui, Selmi, Boukhriss, Rguibi Idrissi and Dakki2009); (2) it is within the historic range of the species (Thévenot et al. Reference Thévenot, Vernon and Bergier2003); and (3) it contains suitable habitats for the species (De la Perche Reference De la Perche1992, Thévenot et al. Reference Thévenot, Vernon and Bergier2003). It is known that the success of such reintroduction programmes is often dependent upon the suitability of habitats in the immediate area of the release site (Ewen and Armstrong Reference Ewen and Armstrong2007, Bennett et al. Reference Bennett, Doerr, Doerr, Manning, Lindenmayer and Yoon2013).
Altogether 300 Double-spurred Francolins were released at SABGR. They came from the livestock area, called “Domaine la Gazelle-Gibiers” (DGG). The centre managers have been authorised to collect eggs from a relict natural population at Ain Sferjla Royal Reserve. The eggs were incubated and hatched at the DGG. Releases took place in early winter (November 2011). Birds were soft-released after keeping them for two weeks in a 2.5 ha acclimatisation pen at SABGR, which contained 20 aviaries of 20 x 10 m each. Water and food were provided in the aviaries. After release, the birds were provided with food and water at several points within the SABGR. The francolins were released in groups, aviary by aviary, with a time interval of one hour. All 300 francolins were subject to a health check and found to be free from parasites and other diseases.
Sampling design
Determination of Double-spurred Francolin territories
We conducted breeding season surveys in March 2014 and as the Double-spurred Francolin is territorial (the male sings to attract the female and to defend its territory), we used the point count method enhanced by playback (Hanane and Qninba Reference Hanane and Qninba2014). Nowadays this technique is commonly used to increase the detection of many secretive bird species (Conway et al. Reference Conway, Eddleman, Anderson and Hanebury1993, Zuberogoitia and Campos Reference Zuberogoitia and Campos1998, Brambilla and Rubolini Reference Brambilla and Rubolini2004), such as Galliformes (Evans et al. Reference Evans, Redpath, Leckie and Mougeot2007, Ponce-Boutin Reference Ponce-Boutin1992, Kasprzykowski and Goławski Reference Kasprzykowski and Goławski2009, Jakob et al. Reference Jakob, Ponce-Boutin, Besnard and Eraud2010, Reference Jakob, Ponce-Boutin and Besnard2014, Fuller et al. Reference Fuller, Akite, Amuno, Fuller, Ofwono, Proaktor and Ssemmanda2012). Ten permanent transects were established (1.2–1.5 km), on each of which 4–5 points were identified as being sufficiently far apart (0.2–0.3 km) to avoid double counting. Counts were conducted at these points five times during March 2014. Each survey started between dawn and 10h00. We played Double-spurred Francolin calls from a notebook via a VLC media player and two speakers (5W each). One cycle of Double-spurred Francolin male territorial calls was played for 10 s, and any response noted in the ensuing 60 s. At each point, this process was repeated three times in each cardinal direction, to locate calling males. We discovered that the released francolins almost exclusively use wooded matorral in the game reserve (Hanane and Qninba Reference Hanane and Qninba2014), and therefore we decided to restrict our counting to there.
Less than half of the francolins recorded were located visually by finding a roost site or by directly observing a breeding pair. Such visual locations give the strongest indication of the core of each territory. When we heard francolin calls, often in response to playback, and we could not located it visually, usually because of dense vegetation, we used triangulation to determine its location with the help of three observers surrounding the calling bird and taking bearings. All Double-spurred Francolin positions were geo-referenced using a portable GPS (Magellan eXplorist XL) and then reported in an Open Source GIS (Quantum GIS v1.7.3). When a male was heard, calling either without using playback or immediately after playback was first used, it was assumed that it was actively defending its territory, and thus its location was indicative of the territory core (F. b. ayesha has distinct, far-carrying vocalisations). We performed point counts only in good weather conditions (not excessively hot and no wind or rain) for two reasons: (1) to make it easier to hear birds at a greater distance; (2) to facilitate collecting the most accurate localisations of birds and random points through using GPS instruments in the best conditions (Trimble GNSS Planning online tools).
Selection of random points
As a first step, we excluded all points at which Double-spurred Francolins were recorded. In the second step, points were selected by drawing 40 random points using the QGIS random selection tool. The random and the presence points were equal in number to give a balanced design.
Measurement of habitat characteristics
Forest structure was quantified either at the locations where francolins were recorded (hereafter ‘Francolin points’) or for random points in circular plots with an 11.3 m radius (0.04 ha). We assumed that this radius was sufficient to characterise the francolins’ habitat use because (1) the study area is well protected and is not affected by such human activities as grazing, clearing and logging, which are known to affect the habitat structure (Kie et al. Reference Kie, Baldwin and Evans1996, Fimbel et al. Reference Fimbel, Robinson, Grajal, Fimbel, Grajal and Robinson2001, Hanane Reference Hanane2014); (2) the minimum distance recorded between the singing Double-spurred Francolin males was 13 m, and (3) a lesser radius was used for characterising habitat selection for other francolin species such as Handsome Francolin Francolinus nobilis (Ssemmanda and Fuller Reference Ssemmanda and Fuller2005), Grey Francolin Francolinus pondicerianus (Kidwai et al. Reference Kidwai, Sankar, Qureshi and Khan2011, Kidwai Reference Kidwai2013) and Black Francolin Francolinus francolinus (Kidwai et al. Reference Kidwai, Sankar, Qureshi and Khan2011).
We considered: 1) geomorphological variables such as altitude, (Alt using a GPS), slope (%_Slo using a clinometer, ± 0.05 m) and distance (m) to the nearest small valley (D_sva) with QGIS; 2) vegetation variables: tree cover (%_tree with visual estimation), height (m) of the tallest tree (H_tree using a clinometer), diameter (cm) at breast height [DBH with a measuring tape (± 0.01 m) of the tallest tree; tree density (D_tree) by counting the number of trees within the 11.3 m radius; shrub cover (%_mat using Gayton (2003) method); average height (m) of the shrub (H_mat using a clinometer); herbaceous cover (%_her using Gayton [2003] method), and average height (m) of the herbaceous layer (H_her using a clinometer); and 3) variables related to the only man-made structures present at SABGR (release enclosures, tracks and water points) as closest distance (m) to the release site (D_rea ); distance (m) to the closest track (D_tra), and distance (m) to the closest water point (D_wat), through the application of Geographic Information Systems (QGIS, v1.7.3) to measure distances. Climatic factors were not taken into account due to the small study area (493 ha), which means that we assume similar values for temperature, precipitation, and humidity across the studied landscape.
Statistical analysis
Before performing statistical analyses, we checked for normality and homogeneity of variance of all the variables. Variables that did not conform to the requirements for parametric tests were log-transformed prior to all analyses (Zar Reference Zar1984, Underwood Reference Underwood1996, Quinn and Keough Reference Quinn and Keough2002). We also checked for possible correlations among variables by using Pearson’s rank correlation (r) index. We collapsed habitat structure variables into independent vectors using Principal Component Analysis (PCA), since this allowed us to: (i) reduce the dimensionality of the set of variables to a smaller number of ‘representative’ and ‘uncorrelated’ variables (n = 14); (ii) investigate multicollinearity; and (iii) describe dominant ecological gradients (Legendre and Legendre Reference Legendre and Legendre1998). For each PCA, a varimax normalised rotation was applied to the set of principal components with eigenvalues > 1.0, to obtain simpler and more interpretable gradients (Legendre and Legendre Reference Legendre and Legendre1998). We interpreted the biological meaning of the principal components, which explain the greatest amount of combined variation within the habitat structure data, by examining the component loadings of each variable (McGarigal et al. Reference McGarigal, Cushman and Stafford2000).
After this first stage of analysis, we turned to modelling the occupancy probability of the Double-spurred Francolin as a function of the orthogonal predictor factors of habitat structure using the Generalised Linear Model (GLM) with binomial error (logistic regression: presence vs. random). In order to select the best GLM models, we developed an all-inclusive design by using multi-model inference (Burnham and Anderson Reference Burnham and Anderson2002): 31 possible combination models were tested and only the best ones (10) were reported (Table 3). For each model, Akaike Information Criteria (AICs) were calculated from the general formula AIC = -2 (log likelihood) + 2K, where K is the number of parameters. The model with the lowest AIC was selected as the best fitting model. We corrected AIC for small sample size (n = 80) using AICc (Burnham and Anderson Reference Burnham and Anderson2002).
To assess whether the residuals of the best model are normally distributed, and thus acceptable, we tested the goodness-of-fit of the best model using the Le Cessie and van Howelingen test (1991). The error rate of the best model was assessed using a receiver operating characteristic (ROC) procedure.
One of the assumptions of parametric statistics is that observations are independent of each other. This assumption is often violated with spatial data. As a result, it is important to test for and subsequently address spatial autocorrelation in data prior to data analysis. Spatial autocorrelation was tested on the residuals of the best model in terms of AIC value of the second step. We used Moran’s I correlogram with 10 lags of 100 m each. We assessed the significance of the values for each lag with a Monte-Carlo test of 999 permutations. A correlogram was significant if at least one lag resulted in P < 0.05.
All statistical analyses were performed in R 3.1.0 software (R Development Core Team 2013). We used the package “ade4” for Principal Component Analysis (Dray and Dufour Reference Dray and Dufour2007) and “spdep” for Moran’s I autocorrelation index (Paradis et al. Reference Paradis, Claude and Strimmer2004). After model selection, we also used a two sample t-test to determine which of the habitat variables differed significantly (P < 0.05) between presence and random plots.
Results
The t-tests conducted on the characteristics of francolin-present and random points demonstrated that distance to the nearest small valley (D_sva), shrub cover (%_mat), tree cover (% _tree), shrub height (H_mat), herbaceous layer height (H_her), tree density (D_tree), distance to the closest point of release (D_rea), distance to the closest water point (D_wat) and slope (Slp) were significantly different while altitude (Alt), tree height (H_tree), tree DBH, herbaceous cover (%_her) and distance to the closest track (D_tra) were not (Table 1).
Table 1. Sample means and standard errors (SE) for variables measured at Francolin and random points, at SABGR, Morocco, 2014. An asterisk (*) indicates that data differed significantly (two-sample t-tests, P < 0.05).
The PCA summarised the 14 original variables into five axes (PC) with eigenvalues > 1, accounting together for 76.6% of the variance in the original dataset (Table 2). The first PC (PCman-made structures) represented a gradient of increasing distance to man-made structures (water points, tracks and release site). The second PC (PCtree size) depicted a gradient of increasing height of trees with high DBH. The third PC expressed the slope (PCslope) and high herbaceous height. The fourth and fifth PC axes, respectively described gradients of increasing shrub cover (PCshrub cover) and cork oak tree density (PCcork oak density) (Table 2). They can also be interpreted as concealment factors.
Table 2. Results of the principal component analysis showing the loadings of the habitat variables within each of the principal components.
These five orthogonal factors were used as independent explanatory variables in a logistic regression model to assess their significance in predicting Double-spurred Francolin occupancy probability. Thirty candidate models were obtained with these five factors. In accordance with the ΔAICc values, the most parsimonious model of Double-spurred Francolin habitat occupancy included a GLM fit to PCman-made structures, PCslope, PCshrub cover and PC cork oak density (Tables 3 and 4).
Table 3. Models with the number of parameters used (k), the Akaike information criterion for small simples size (AICc), the difference between each selected model and the best model (ΔAICc), and the Akaike weight (AICwi). Only the 10 best models are shown (out of 31 examined).
Table 4. Parameters and standard errors (SE) of the GLMs to explain Double-spurred Francolin occupancy probability using habitat PCman-made structures, PCSlope, PCShrub cover and PCCork oak density as predictors.
The goodness-of-fit test indicated acceptable fit (z = 0.512, P = 0.108). The error rate based on the ROC function was 0.16. The model explained 60% of the deviance in the Double-spurred Francolin occupancy and 61% of their variance.
In addition, we did not find evidence of spatial autocorrelation in model residuals between plots. The correlogram of residuals from the top AICc ranked GLM shows no significant spatial autocorrelation (Figure 2), suggesting that the results of this non-spatial GLM model were not biased by possible spatial covariance in the data.
Figure 2. Correlogram of the residuals of GLM model of predicted occupancy probability as function of geomorphological variables, vegetation structure, and human activities.
Francolin occupancy probabilities were negatively related to PCman-made structures and PCslope (Figure 3a,b) and positively related to PCshrub cover and PCcork oak density (Figure 3c,d). These results are consistent with those of student’s t-tests (α = 0.05) (Table 1).
Figure 3. Occupancy probability of Double-spurred Francolins according to man-made structures (a), slope (b), shrub cover (c) and cork oak density (d) at SABGR, north-west Morocco, 2014.
Discussion
The aim of the study in SABGR was to explore a relatively wide spectrum of environmental factors and identify variables responsible for habitat occupation by Double-spurred Francolin in a sample of the cork oak forest of north-west Morocco. Through this work, we have been able to present the most detailed data available to date on the habitat use of Double-spurred Francolins in Morocco.
As expected, reintroduced, captive-bred Double-spurred Francolins showed a preference for topographically flatter land (a gentle slope < 15%), a dense and high shrub layer (respectively 61% and 2.53 m), dense cork oak trees (up to 12 trees within 0.04 ha), and a proximity to man-made structures especially water points and the release point (aviaries). This supports previous observations that Double-spurred Francolins are associated with densely vegetated valleys (Thévenot et al. Reference Thévenot, Vernon and Bergier2003) and with cork oak forest with dense undergrowth (Alaoui Reference Alaoui, Kassinis and Panayides2001). We also established that the predicted distribution of the subspecies is unrelated with the size of cork oak trees. This shows that the cover of both cork oak trees and shrubs is crucial for this galliform subspecies in this protected area. This is not surprising to the extent that vegetation cover is known to be very useful to the birds as it functions as both a refuge for prey and as concealment from predators (Carrascal and Alonso Reference Carrascal and Alonso2006, Rantanen et al. Reference Rantanen, Buner, Riordan, Sotherton and Macdonald2010). Both ground and avian predators are common throughout the Double-spurred Francolin’s range in Morocco and this forces them to remain hidden in dense vegetation to minimise the predation risk. It is worth noting that in Benin, for instance, Double-spurred Francolins use similar vegetation cover (51–90%) and shrub-layer height (1.0–2.7 m) (Codjia et al. Reference Codjia, Ékué and Mensah2003). Other francolin species, such as Redwing Francolin Francolinus levaillantii (Jansen et al. Reference Jansen, Robinson, Little and Crowe2001) and Grey Francolin (Hussain et al. Reference Hussain, Nisa and Khalil2009) also use similar habitat. In a study of the ecology of Black Francolin on the Iranian plain of Sistan, Heidari et al. (Reference Heidari, Arbabi, Noori and Shahriari2009) highlight the effect of a reduction of vegetation cover in leading to a decrease in the species’ population size and distribution. It is however well known that to some extent increasing vegetative cover leads to increased thermal problems (Trautman Reference Trautman1982, Gatti et al. Reference Gatti, Dumke and Pils1989, Gabbert et al. Reference Gabbert, Leif, Purvis and Flake1999, Novoa et al. Reference Novoa, Dumas and Resseguier2006) as well as increasing predation risk by mammals (Meriggi et al. Reference Meriggi, Montagna and Zacchetti1991, Tapper et al. Reference Tapper, Potts and Brockless1996). Conversely, decreased vegetative cover increases avian predation, and reduces the camouflaging values and thermal protection of the habitat (Subramanian et al. Reference Subramanian, Sathyanarayana and Kambarajan2002).
At SABGR, Double-spurred Francolins occupied the wooded matorral (with the vegetation characteristics mentioned above) near places where water was available and the release point. Indeed, the majority of francolin locations were within 100 m (90% of birds) of water points and 600 m from the release point (80%). In fact, it is recognised that bird populations are likely to be more affected at short distances from man-made structures (Madsen Reference Madsen1985, Benítez-López et al. Reference Benítez-López, Alkemadea and Verweij2010). So, why do Double-spurred Francolins in SABGR occupy wooded matorral located close to man-made structures? This pattern is undoubtedly related to water points, which are most often located beside paths where supplementary food left by humans can sometimes be found as wells. An alternative explanation could be that wooded matorral close to man-made structures is used as a refuge, which can be reached quickly once a threat is detected. This hypothesis is consistent with Martin (Reference Martin1993), who stated that the most common factor determining habitat use in birds under predation risk is distance to a potential refuge. Moreover, we also cannot exclude the hypothesis that francolins might have become adapted to the low frequency of human disturbance (one car passing per week) in this area. Such a low level of anthropogenic disturbance seems to be acceptable to these birds. Montgomery et al. (Reference Montgomery, Roloff and Millspaugh2013), in studying animal response to roads, have also recorded habitat selection near trails (tertiary roads) under low levels of anthropogenic disturbance.
Overall, it seems that Double-spurred Francolin habitat selection would be a trade-off between the need for water and the risk of predation. Thus, by making the most of the complex habitat structure of the matorral forest of SABGR, the francolins may reduce the risk of predation. This is consistent with previous studies (Warfe and Barmuta Reference Warfe and Barmuta2004, Chalfoun and Martin Reference Chalfoun and Martin2009), who suggest that predation risk can be affected by the attributes of the habitat occupied by the prey.
Our results have led us to understand and become aware of the role of predation in the selection of habitat by Double-spurred Francolins. This is mainly by raptors such as Booted Eagle Hieraaetus pennatus, Long-legged Buzzard Buteo rufinus, Black Kite Milvus migrans and Common Kestrel Falco tinnunculus that occur in the same area (Cherkaoui et al. Reference Cherkaoui, Dakki, Selmi, Rguibi Idrissi and Thévenot2007) and to a lesser extent by mammals (e.g. common genet Genetta genetta, least weasel Mustela nivalis and red fox Vulpes vulpes; pers. obs.). This argument has already been proposed by Martin (Reference Martin1993), who stressed the implication of predation as a major factor in habitat selection and in the evolution of life history strategies of birds.
Finally, it seems that Double-spurred Francolins released at SABGR begin to adapt to local conditions as evidenced by the first recorded nests in 2014 (authors’ unpubl. data). This finding should be explored by further studies.
Implications for future releases
The investigation of habitat-animal associations relies on a manager’s ability to understand the scale at which wildlife responds to, and interacts with, the environment (Pastor et al. Reference Pastor, Mladenoff, Haila, Bryant, Payette, Mooney, Cushman, Medina, E Sala and Schulze1996, Guerena et al. Reference Guerena, Castelli, Nichols and Williams2014). It is for this reason that the results of this study should be taken into account in planning future releases of the F. b. ayesha. In addition to the desirability of choosing a release site that is within the historical range of the subspecies (IUCN 1998), it is also important to choose a site that offers sufficient cover of cork oak trees and shrubs (e.g. needle-leaved broom). This appears to be essential in helping to protect Double-spurred Francolins from both aerial and terrestrial predators. Setting up water points and food supplements are also key prerequisites for successful release programs and conservation (Whittingham and Evans Reference Whittingham and Evans2004). According to our results, both water points and the release site should be near or within suitable wooded matorral. Such conservation prescriptions would enable the birds to adapt rapidly to prevailing conditions within release sites.
To our knowledge, this is the first study that highlights the association of Double-spurred Francolins with water sources and gentle slopes. It will therefore be interesting to continue monitoring the Double-spurred Francolins at SABGR, especially to discover more about their ecology and population dynamics. Further multi-scale studies will also valuable in improving our understanding of the effects of biotic factors (especially predation) on the processes of habitats selection. As suggested by Winnard et al. (Reference Winnard, Di Stefano and Coulson2013) knowledge of habitat suitability thresholds at predator-free locations will also be important for selecting appropriate reintroduction sites in the future. Finally, in view of the fact that francolins have already begun to nest in SABGR, we believe it is important to continue our work in order to enhance our knowledge on habitat use, nest habitat selection, productivity and survival of Double-spurred Francolin.
Acknowledgements
We are grateful to Humphrey Sitters for his efforts to improve an earlier draft of the manuscript. We thank the two anonymous reviewers and the editor of Bird Conservation International for their comments and advice. We also thank L. Becha and Ahmed for helping during the field work. This study was supported by the Forest Research Centre, High Commission for Water, Forests and Desertification Control, Morocco.
