The study was carried out in West Africa including Senegal, Guinea, Sierra Leone, Liberia, Ivory Coast, Ghana, Burkina Faso, Togo, Benin, and Nigeria (Fig. 1A). In terms of biogeography, four major phytochoria with specific vegetation types can be broadly distinguished: (i) Sahelian savannah (S), (ii) Sudanian savannah (SZ), (iii) Guinea savannah (SG), (iv) Guineo-Congolian (GC) forest with closed semi-deciduous forests and evergreen forests (Fig. 1A). The African lowland rain forest has three regions, namely Congolia, Upper Guinea, and Lower Guinea, the latter two being separated by the Dahomey Gap (White 1983; Poorter et al. 2004; Leal 2004). The current forest range originated from the dry period in the region during the Quaternary, especially the Pleistocene and the Holocene (Salzmann and Hoelzmann 2005). The Dahomey Gap has been identified as a major ecogeographical impediment to the dispersal of rainforest species in this region and is, as such, of crucial significance for their distribution patterns in West Africa (van Bruggen 1989; Martin 1991; Jenik 1994). The Dahomey Gap experienced a decline in annual precipitation from more than 2000 mm in the rainforest areas of Nigeria and Ivory Coast to about 1200–1000 mm in the savanna region. The GC region is relatively dry and receives between 1600 and 2000 mm of precipitation annually, while the SZ and S zones receive 500–1400 mm and 150–500 mm of annual precipitation respectively.

Detarium senegalense is currently present in a wide geographical range across the four ecological zones: GC, SG, SZ, andS (Fig. 1A). The species is mainly found in Senegal, Guinea, Sierra Leone, Liberia, Ivory Coast, Ghana, Burkina Faso, Togo, Benin, and Nigeria (WFO 2024). The species is a tree with a large leafy crown reaching 10–40 m in height (Fig. 1B) (Akoègninou et al. 2006). The trunk is generally greyish and fairly smooth but covered with fine cracks. The wood is dark reddish brown, hard, and fine-grained. The leaves are bright green, pinnate and have 10 alternate leaflets, which are oval to elliptical, thinly coriaceous, or papery, usually 4–6 cm long and 2.5–3 cm wide, with rather few translucid gland-dots (Fig. 1C). The inflorescence is a panicle or loose axillary fascicle, dense and lax. The apetalous flowers have a calyx with 4 sepals, creamy white, in bud glabrous or sparsely pubescent outside, and 10 stamens. The ovary is ellipsoid, ca 2 mm long, densely hairy, 1-celled, style 3–4 mm long, arched. The fruit is a globose or subglobose drupe, slightly flattened about 5–6 cm in diameter, thick, fleshy (Fig. 1D), and it is divided into three main parts: (i) dark green epicarp, hard for immature fruits, light green tending to brown and brittle for ripe fruits, cracks at maturity, (ii) greenish mesocarp (pulp) intertwined with fibres inserted on the core, (iii) voluminous, woody core covered with fibrous meshes and enclosing a single ovoid and flattened seed of dark brown colour. Flowering occurs from February to May, and fruits take about five to six months to mature. The fruits are dispersed mainly by humans, monkeys, rodents, and parrots (Kouyaté and Lamien 2011). Detarium senegalense also holds a high potential to shoot forth suckers (Fig. 1E).

Occurrence data were gathered from two sources: online repositories and previously published data. Online data included the records from the Global Biodiversity Information Facility (GBIF.org 2021) and RAINBIO (Dauby et al. 2016). These occurrences were cleaned by removing duplicates and records without geographic coordinates. In addition, older occurrences collected before 1980 were removed to reduce the effects of temporal bias and match occurrence to the climate datasets (Idohou et al. 2017). As for previously published data, occurrences were extracted from two papers, namely Dangbo et al. (2019) and Diop et al. (2010). These coordinates were obtained by georeferencing the maps and digitizing the datapoints. To avoid occurrence bias, the global gazetteer v.2.3 (http://www.fallingrain.com/world) was used to assign digital coordinates to the occurrences (Hemami et al. 2020). We projected these coordinates on Google Earth v.7.1 to ensure they matched the target localities. Finally, all occurrences were combined, and we used the occurrence rarefy tool of Spatial Analyst in ArcGIS v.10.8 to reduce autocorrelation among the occurrences. Thus, we kept only one point within each grid cell (5 × 5 km) to be consistent with the environmental layers. Overall, after the cleaning process, 220 occurrences were used to run the models across West Africa: 175 from GBIF, 36 from RAINBIO, and 9 from the two papers (Supplementary material 1).

A total of three sources of environmental layers (bioclimatic, edaphic, and topography) were used. Past, current, and future climatic data were obtained at the spatial resolution of 2.5 minutes (~5 km at the equator) from WorldClim v.2.1 (Fick and Hijmans 2017). The edaphic (Leenaars et al. 2014) whereas elevation variables were taken from the digital elevation model (ASTER DEM) (Fick and Hijmans 2017). Slope data was generated from Google Earth Engine (Gorelick 2013) using a West African boundary. The current variables layers were derived from 1950–2000 climate data (Hijmans et al. 2005). For the past projection, bioclimatic variables layers were obtained from the MIROC-ESM model for the Mid-Holocene (MH, about 6000 years ago) and the Last Glacial Maximum (LGM, about 22,000 years ago). As for the future projection, the MIROC6 model was used to be consistent with the past projection. MIROC are advanced climate models developed by the Atmosphere and Ocean Research Institute (AORI) of the University of Tokyo (Tatebe et al. 2019). The benefit of MIROC6 models is that they have improved climate projections (temperature and precipitation), advanced physical processes, and comprehensive earth system modelling (Tatebe et al. 2019). Past and future layers were also downloaded from the WorldClim website (Fick and Hijmans 2017). The suitable areas were projected under two climate scenarios, ssp245 and ssp585, for the year 2050 (average for 2041–2060). The downloaded variables cut across WA and were resampled at 5 km. We removed four bioclimatic variables (bio8, bio9, bio18, and bio19) because of discontinuities between neighbouring pixels for the African continent (Montoya-Jiménez et al. 2022). The bioclimatic data consisted of 15 bioclimatic variables (Supplementary material 2). Then, we performed the Pearson correlation and jackknife tests (Supplementary materials 3 and 4) using the R package usdm v.2.1-6 (Naimi 2017). These two tests were used to exclude highly correlated variables with the coefficient of correlation (r) value of |0.7| and to highlight those with high contribution. The final candidate factors were selected after reviewing the literature to determine which variables were ecologically significant for the species.

We modelled the past, current, and future distributions of D. senegalense in West Africa. Four machine learning methods were used for this purpose, and the results were averaged for the predictive models. These methods included generalized additive models (GAM) (Wood 2006), generalized linear models (GLM) (Hastie and Tibshirani 1990), maximum entropy (MaxEnt) (Phillips et al. 2006), and random forest (RF) (Breiman 2001). These models were chosen for their effectiveness and robustness (Rong et al. 2020; Daï et al. 2023). The RF is a learning technique that constructs several decision trees during training and outputs to the class. GAM is a statistical model that is used to link response variables to predictor factors (Wood 2006). The GLM model describes the relationship between the dependent and independent variables (Hastie and Tibshirani 1990). The MaxEnt technique is a probabilistic model that aims to find the probability distribution with the highest entropy (or the least uncertainty) given specific restrictions (Phillips et al. 2006, 2017). All models were run with the R package sdm v.1.1-8 (Naimi and Araújo 2016). Environmental factors were used as independent variables, whereas point locations (geographical coordinates) were used as dependent variables. We also used 10³ pseudo-absence points since we do not have the real absence data. Furthermore, the occurrences were split into two samples where 70% were used to train the models and 30% to test them. To increase the models’ accuracy, each model was replicated 10 times with subsampling and bootstrapping. We assessed the accuracy and the performance of the models using the area under the receiver operating characteristic curve (AUC) and the true skills statistic (TSS) (Allouche et al. 2006). The AUC provides the probability that the predictive power of a model is better than random prediction (AUC = 0.5) (Swets 1988). A model with an AUC value close to 1 (AUC ≥ 0.75) is considered to be a good fit. As for TSS, values range from -1 to 1 and values close to 1 indicate the high performance of the models.

We evaluated the dynamics of the suitable areas and the direction of range shifts using SDMtoolbox v.1.0b (Brown 2014). This toolbox is a python-based GIS software that is capable of describing the magnitude and direction of the predicted change over time by summarizing the core shifts in the species distribution reducing the distribution variation to a single centroid (central) point and generating a vector file (Brown 2014). In our case, the range shifts, as well as range dynamics, were assessed by comparing Mid-Holocene (MH)-current, Last Glacial Maximum (LGM)-current, current-ssp245, and current-ssp585. We also evaluated the dynamics of the suitable areas using Distribution Changes Between Binary SDMs (Brown 2014). The process was the same as described above but here the tool generated a single layer in TIF format and an Excel sheet that held the statistics. Overall, four classes of habitats were automatically generated: contraction, expansion, areas of no change or stability, and no occupancy areas or unsuitable.

The conservation status of D. senegalense was evaluated based on the IUCN criteria, especially B. We determined the extent of occurrence (EOO) and the area of occupancy (AOO) with the R package ConR v.1.3.0 (Dauby and Lima 2023). The EOO is a polygon that can be drawn to encompass all the known present occurrences of a species (IUCN 2012). The AOO is related to the area covered by the number of occupied cells of a grid for a given resolution (IUCN 2012). In this study, we considered grid cells of 2 km², as recommended by IUCN guidelines (IUCN 2024). In addition, the diverse threat levels were measured by estimating the number of locations in protected areas. For this purpose, the occurrence of the species within the West African boundary (https://geoportal.icpac.net) and within protected areas (World Database on Protected Areas, UNEP-WCMC 2023) was used.

Overall, five variables were identified to influence the distribution range of D. senegalense (Fig. 2A). These variables were related to temperature and precipitation. Isothermality (bio3) was the variable that contributed the most (41.10%). The occurrence probability of the species reached its maximal value when isothermality values were close to 100 (Supplementary material 5). This variable was followed by the Precipitation of Wettest Month (bio13, 21.50%) and the Precipitation Seasonality (bio15, 19.23%). For bio13, the occurrence probability of the species increased when the bio13 values increased. In contrast, species occurrence decreased when the bio15 values varied by around 35%. The Temperature Annual Range (bio7, 10.60%) decreased the occurrence probability of the species with bio7 values below 75 mm (Supplementary material 5). Even though bio7 had a minor impact on species distribution, the occurrence probability of the species tended to increase between values of 10°C and 40°C (Supplementary material 5). Precipitation of Driest Month (bio14) showed the lowest contribution (7.57%). The performance and accuracy of the models showed an average area under the curve (AUC) of 0.92 and a TSS value of 0.73 (Fig. 2B), indicating that the models are of excellent quality.

The extent of suitability shown for the four classes (expansion, unsuitable, stable, contraction) for the five time periods (LGM, MH, current, ssp245, ssp585) is documented in Supplementary material 6. In the past, notably at the LGM, the stable ranges of D. senegalense distribution, estimated at 1,183,095.65 km² or 21.30% of the study area, mainly occurred from Nigeria to Sierra Leone, and from the Guinean savannah to the evergreen forest ecological zones (Fig. 3A). Overall, D. senegalense was found across the SZ to GC zones of Nigeria and Benin, in the SG and GC zones of Togo, and from the SG to the evergreen forest zone of Ivory Coast. The species was also found from the GC to the evergreen forest zone of Guinea and Sierra Leone and the evergreen forest zone of Liberia. A very slight range expansion of 241,458.47 km² or 0.16% was observed during this period, mainly in the extreme west of Guinea. Meanwhile, the main part of the study area (3,803,962.10 km² or 66.15%) was unsuitable (Fig. 3A, F). The species lost an extent of 807,233.78 km² or 12.39% from the LGM to the current day.

The MH period exhibited a constant favourable range of 1,362,492.32 km² or 21.50% with an unsuitable range of 3,885,769.34 km² or 60.10% (Fig. 3B). If we compare the MH to the current period, there was no discernible rise in the stable distribution range. The increase was estimated at 62,067.80 km² or 0.03% of the total study area (Fig. 3B, F). The contraction range of the species’ distribution area was estimated at 725,420.54 km² or 18.37%.

At present, D. senegalense occurs in a large geographical range across Senegal, Guinea, Guinea-Bissau, Sierra Leone, Liberia, Gambia, Ivory Coast, Ghana, Burkina Faso, Togo, Benin, and Nigeria, and spreading across four ecological zones: Guineo-Congolian (GC), Guinea savannah (SG), Sudanian savannah (SZ), and very lower part of the Sahelian savannah (S) zones (Fig. 3C). The species occupies an extent of 1,048,558.40 km² in West Africa. These suitable areas or stable areas for D. senegalense distribution under current climatic conditions are located between 5°00’00”N–11°00’00”N and 10°00’00”W–5°00’00”W in the GC and SG zones (Fig. 3C). Countries such as Ivory Coast, Nigeria, Benin, Togo, Ghana, Liberia, and Sierra Leone showed the largest stable ranges. The countries where D. senegalense was less represented are Burkina-Faso, Guinea-Bissau, Guinea, Gambia, and Senegal. Overall, 3,954,573.16 km² of the total West Africa was not suitable for the species under current conditions.

Regarding the potential distribution of D. senegalense in the future, the stable ranges were 506,147.78 km² or 17.70% under ssp245 and 157,558 km² or 13.98% under ssp585 by horizon-time 2050 (Fig. 3D, E). Results also showed that the unsuitable range predicted for the future appeared to be higher in both scenarios (ssp245 and ssp585) than in the LGM and MH periods respectively. These unsuitable areas were 5,342,985.84 km² or 78.70% and 5,551,460.57 km² or 78.13% under ssp245 and ssp585 respectively. However, these two scenarios showed much smaller contraction ranges than the two previous periods (LGM and MH), with 181,355.14 km² or 2.93% for the ssp245 scenario and 322,893.51 km² or 7.29% for the ssp585 scenario. At the same time, some slight expansions occurred estimating around 5,261.25 km² or 0.67% for ssp245 and 3,837.92 km² or 0.60% for ssp585. In terms of spatial distribution, suitable areas are expected in the SZ to the GC zones of Nigeria and Benin, in the SG and GC zones of Togo, and from the SG to the evergreen forest zone of Ivory Coast. However, the species’ suitable areas have shrunk on both sides of the Ivory Coast-Ghana border, with a more pronounced loss in the ssp58 scenario.

The centroid of the suitable area for D. senegalense revealed different patterns of range change shift direction (Fig. 4). In past, especially during LGM, range shift directions tended to be more longitudinal rather than latitudinal, with a north-eastern shift for LGM and a south-western shift for MH. However, both future scenarios showed range shifts to the south-east (Fig. 4). The current centre of distribution of D. senegalense was predicted to be in the extreme east of Ivory Coast (Fig. 4).

All PAs located between 6°N–13°N and 12°W–10°E were currently suitable for the species (Fig. 5). The models indicated that all PAs in Ivory Coast and Liberia are suitable for the species. In Benin, the PAs suitable for D. senegalense were those in the SG and GC zones. There were no suitable areas in the SZ, including the Pendjari and W parks. Similar observations showed that the PAs in Nigeria, Togo, and Ghana that are currently suitable for D. senegalense were located in the SG and GC zones, but more predominantly in the central east and south of the GC zone and in the west of the SG zone of Nigeria, in the south-east of the SG zone and throughout the GC zone of Ghana, and from halfway the SG zone to the GC zone of Togo. In Guinea, the suitable PAs were located in the south-east of the country in the SG and GC zones. Only the PAs along the coast of Senegal, Guinea-Bissau, and Gambia were capable of hosting the species in the present day. In addition, the PAs in the SG zone of Burkina Faso were suitable for the species under current conditions.

According to the future scenario (ssp245), most of the protected areas that are suitable under current conditions should also be suitable for the species by 2050. However, some unsuitable protected areas are predicted under this scenario. These are located mainly in the south-west and north-west of Ghana and the south-east and north of Ivory Coast, in the SG and GC zones respectively. In addition, some unsuitable PAs are predicted to a slight extent in the SZ zone to the east of the GC zone in Nigeria and in the east of the SG and GC zones in Guinea. The extent of suitability related to each class within the protected areas is provided in Supplementary material 6. The number of protected areas by country as well as their extent were also quantified. Overall, the models indicated that in the present conditions, only 202,569.67 km² or 22.90% of the 2011 PAs of West Africa are suitable for the species. In contrast, 698,554.08 km² or 77.10% of the total extent (i.e. 901,123.75 km²) is not suitable. Moreover, 98,915.92 km² or 21.01% and 27,573.05 km² or 14.60% could be suitable under ssp245 and ssp585 by 2050 respectively. Significant contraction was observed under the ssp585 scenario with 21,7141.24 km² or 8.61%. The contraction predicted under ssp585 was 145,798.37 km² or 2.50%. Suitable areas or stable habitats were found especially in Liberia, central and southern Sierra Leone, in the GC zone of southern Nigeria, in the SG zone of Benin and Togo and the south-west of Ivory Coast, and in a few protected areas in central and south-eastern Ghana. The expansion of habitat was estimated at around 413.18 km² or 0.51% and 468.27 km² or 0.35% under ssp245 and ssp585 respectively. As for unsuitable areas, the estimations were 698,554.08 km² or 77.10% under current conditions. Future scenarios were also high with 655,996.27 km² or 75.98% for ssp245 and 655,941.18 km² or 76.44% under ssp585 by 2050.

The status of D. senegalense based on IUCN criteria across West Africa was documented in Fig. 6. The results showed an EOO of 1,958,696 km² (> 20,000 km²) and an AOO of 668 km² (< 2,000 km²). The large EOO points towards Least Concern, while the AOO could indicate Vulnerable (VU). However, the results also indicated that 30.8% of the 220 unique occurrences are located within 44 protected areas. Across the study area, we found 130 subpopulations (considering a 2 km² radius) and 151 locations (considering a grid size of 10 km²). Therefore, the species can be considered as Least Concern in West Africa.

Environmental factors play an important role in forest establishment and persistence (Greve et al. 2011). As a result, under climate change, a set of factors that favour the distribution of a given plant species, particularly multipurpose trees such as D. senegalense, may help elucidate its resilience. In this study, climatic factors (temperature and precipitation) were the major variables that explained the species distribution in West Africa consistent with the ecological requirements listed by Diop et al. (2010). However, this finding differs from that of Agbo et al. (2019) who found that soil factors (cation exchange capacity) were important for the distribution of the related species D. microcarpum in Benin. Even though soil properties are recommended in plant modelling (Hengl et al. 2017; Zuquim et al. 2020), their incorporation is meaningful in local contexts in contrast to regional ones, as is the case in this study (Pearson and Dawson 2003).

The distribution of Detarium senegalense throughout West Africa was mainly affected by isothermality (bio3). Overall, isothermality refers to how similar the average daily maximum and minimum temperatures are throughout the year (O’Donnel and Ignizio 2012). Here, values were close to 100, indicating that the species preferred areas with consistent diurnal temperatures throughout the year. This factor is a determinant of plant distribution in tropical areas including West Africa (Huang et al. 2021). In natural ecosystems, the species is primarily distributed along rivers in gallery forests with an annual precipitation of 800 to 1800 mm, in contrast to D. microcarpum, which is more commonly found in open savannah (Dangbo et al. 2019; Houénon et al. 2022). Our findings also showed that species occurrence increased when the precipitation during the wettest month (bio13) increased, which explains why species are more common in rainy ecosystems. In contrast, species occurrence probability decreased when the precipitation of the driest month (bio14) was around 75 mm. Similarly, the occurrence probability decreased for precipitation seasonality (bio15) values of around 35%. This explains how a slight variation in precipitation in a season, particularly during the dry seasons, could influence species seed dispersal, which is facilitated by water and animals. The annual temperature range (bio7) had a minor impact on species distribution, but the occurrence probability of the species increased for temperatures between 10°C and 40°C. These findings suggested that the species may have adapted to the wide range of temperatures under climate change but more research is needed to understand how the species might respond to drought, including germination rates, growth rates, and overall plant health. Moreover, the model performance revealed an AUC value of 0.92, indicating the robustness of the models (Idohou et al. 2017; Agbo et al. 2019). Therefore, they could be considered steady and adequate for constructing the potential ecological niche and conservation ranges of D. senegalense through its geographical distribution (Favi et al. 2022).

The current study assessed the dynamics of D. senegalense distribution from the past to the predicted future suitable areas. In general, our findings revealed that the suitability range was relatively stable from the last glacial maximum to the present day. In addition, suitable habitats occurred from Nigeria to Sierra Leone, and from the Guinean savannah to the evergreen forest ecological zones. However, a study of a population in Togo indicated that adult individuals are rare and that there is a low regeneration rate, and thus a predominance of small-diameter trees (Dangbo et al. 2019). Similar results were recently found in Benin by Houénon et al. (2022) and this shows that modelling results need to be considered with caution. Indeed, human pressures on tree species in the Sub-Saharan region, such as illegal cutting and agricultural expansion, have increased ecosystem vulnerability to climate change, affecting their spatial distribution (Gross et al. 2018).

Moreover, the suitable areas of the species shifted from low latitudes to high latitudes from the Last Glacial Maximum to the Mid-Holocene. These findings are consistent with those of Saupe et al. (2019) about the high-latitudinal migration of tropical species as temperatures rise over geological ages. The Dahomey-gap, which was established during the Holocene, created a unique climate condition in the Guinean Gulf, allowing plant species to spread across a variety of climate zones (Salzmann and Hoelzmann 2005).

According to Trisos et al. (2022), West Africa is one of the most vulnerable zones to climate change. Since the Industrial Revolution, the average annual temperature has risen above 1.1°C, affecting already the growth and productivity of plant species. Global average annual temperature is expected to reach 5°C by 2100 under worse scenarios (Tokarska et al. 2020; Chai et al. 2021). In this context, the models predicted a moderate range decrease for D. senegalense, especially under the pessimistic scenario ssp585. In terms of spatial distribution, suitable areas are expected in the SZ to GC zones of Nigeria and Benin, in the SG and GC zones of Togo, and from the SG to the evergreen forest zone of Ivory Coast. Previous studies predicted that the congeneric species D. microcarpum could expand in suitable areas (Agbo et al. 2019). Indeed, co-occurring species may react differently in the same envelop spaces, contracting or expanding their suitable areas in response to global climate change (Hällfors et al. 2016). This could be supported by the acclimatization or phenotypic plasticity of species to escape extinction (Silva et al. 2019). Given the increasing warming climate conditions predicted in Africa, most native plant species will experience a decline in the future, e.g. Parkia biglobosa (Jacq.) R.Br. ex G.Don (Ayihouenou et al. 2016), and Borassus aethiopum Mart. (Salako et al. 2019). Therefore, the hypothesis of a warming climate inducing a decline in the distribution range of a plant species seems to be well-supported (Li et al. 2019; Imorou 2020). Moreover, suitable habitat tends to shift to the south-east (wettest areas or low latitudes) in response to future climate scenarios. Detarium senegalense usually occurs in West African riparian forests (Arbonnier 2009), from Upper and Lower Guinea to the Dahomey Gap, and in Central Africa (Akoègninou et al. 2006; WFO 2024). This region of the continent has a climate with an average annual precipitation of around 1,800 mm, which is ideal for D. senegalense, which has moderate water requirements.

Ecological niche modelling or species niche modelling is a powerful and widely used tool to predict the occurrence of many taxa based on environmental factors (Guisan and Zimmermann 2000). In this study, ENMs were used to have a clear picture of the past, present, and future distribution of D. senegalense.

The study also addressed the conservation status of the species under current conditions. Our models predicted the suitable areas for the conservation of the target species were located in the GC zones of Sierra Leone, Liberia, Ivory Coast, Ghana, Togo, Benin, and Nigeria. Furthermore, most protected areas that are suitable under current conditions should be suitable for the species by 2050, except for the south-west and north-west of Ghana and the south-east and north of Ivory Coast, in the SG and Guinean climatic zones, respectively. Based on these findings, we recommend both in situ and ex situ conservation, as the species’ suitable habitats are expected to decrease both within and outside of protected areas. According to BGCI and IUCN (2019), the species is of Least Concern. In addition to this statement, they indicated that there are no significant current or future threats to this species. Our findings also classify the species as Least Concern. However, only 30.8% of individuals were found within protected areas. The most important populations are distributed outside protected areas where they are affected by the encroachment due to the agriculture expansion and tree cutting (Dassou et al. 2023). Previous studies also revealed that the commercial exploitation of timber and medicinal products, associated with the growing regional trade of fruits and seeds, and a low potential for natural regeneration of seeds constitute significant threats to D. senegalense (El-Kamali 2011; Houénon et al. 2021, 2022). Nowadays, traditional agroforestry systems are another way to preserve tree species within agricultural lands (Sharma et al. 2016). It is also possible to introduce the species in well-protected community and sacred forests in the GC zone. However, for conservation to be effective, greenhouse experiments must be conducted first, as the species has a low natural regeneration. The outputs of these experiments will be a mainstay for the publication of technical manuals for nursery management. The vision would be to share these techniques with various conservation organisations (NGOs, forestry departments) to facilitate seedling production for the National Days of the Tree in its distribution range and to promote the installation of plantations of the species. Implementing these participatory management policies across local communities and introducing them into agroforestry systems will safely foster species sustainability (Peters and Simaens 2020). The introduction of the species into the botanical gardens must be also envisaged.

Given the usefulness of species distribution models in conservation research (Stockwell and Peterson 2002), the findings of this study are important since they improve insights into the potential changes in D. senegalense distribution. The success of model predictions in this study was likely due to the sample size and wider range of data collected, reducing inaccurate predictions caused by incomplete data (Wang et al. 2021). Even though the models showed interesting performance, models can be improved including physiological data, seed dispersal capacity of the species, and local adaptation data (Engler et al. 2012; Adoukonou-Sagbadja et al. 2023). Recently, some new approaches have emerged incorporating intraspecific variation data into the models. For instance, in Benin, the response of the population of Pterocarpus erinaceus Poir. to climate change was assessed using this method (Biaou et al. 2023).