Predicting the Potential Distribution of Stachys fontqueri Pau (Lamiaceae), a Strictly Endemic Medicinal Species of the Moroccan Rif, Under the Effects of Climate Change for Sustainable Conservation

Abstract

Stachys fontqueri is a strict endemic species of the Moroccan Rif that depends on specific ecological conditions. To understand the effect of climate change on the projected change in potential distribution under current and future climatic scenarios, ecological niche modelling was performed using MaxEnt algorithm based on 70 occurrence records, and 5 bioclimatic variables at a spatial resolution of 30 arc-seconds. To model the effect of climate change, four climatic scenarios were used, namely CSM2-SSP1-2.6, CSM2-SSP5-8.5, MIROC6-SSP1-2.6, and MIROC6-SSP5-8.5, for the period 2061–2080, and the Maximum Test Sensitivity Plus Specificity threshold was used to distinguish suitable from unsuitable habitats. The results demonstrated high model performance, with an AUC ranging from 0.921 to 0.930 and a TSS from 0.81 to 0.83. Three bioclimatic variables contributed significantly to determining the suitable potential distribution area of the species, namely Precipitation Seasonality (Bio15), Temperature Annual Range (Bio7), and Annual Mean Temperature (Bio1). The suitable area covered 3244 km² under the current climate and is projected to decrease by 23.25% to 29.59% under future climate scenarios. This contraction of suitable habitat due to climate change could be exacerbated by human activities, thereby requiring urgent in situ and ex situ conservation measures to ensure the species’ resilience.

1. Introduction

Climate change is one of the most significant environmental challenges of the 21st century, leading to an increase in the frequency and intensity of extreme climatic events such as heat and cold waves, droughts and floods. It exerts increasing threats to ecosystems and their resources [1,2,3,4,5,6], and altering the distribution and physiological performance of the species [7]. The Mediterranean basin, including northern Morocco (classified as a “Hotspot”), is among the most sensitive to and affected regions by climate change due to rapidly rising temperatures and decreasing precipitation [6,8,9,10].

To address these environmental challenges, ecological niche modeling (ENM) is one of the most important analytical tools in ecology and biodiversity conservation biology. This approach enables the prediction of species distributions by relating occurrence data to environmental variables and has proven effective in identifying suitable habitats under current and future environmental conditions [11,12]. These models also make it possible to understand the climatic and environmental factors that control species distribution [5,13], and to predict potential shifts in their range in response to climate change [5,12]. ENM employ a variety of approaches to predict species distributions from environmental variables. the commonly used include Generalized Linear Models (GLMs), Generalized Additive Models (GAMs), Random Forest (RF), Boosted Regression Trees (BRT), and Maximum Entropy (MaxEnt), each with specific strengths and limitations depending on the characteristics of the data and the objectives of the study [14]. Among these various ENM techniques, MaxEnt is one of the most widely used algorithms for modelling species distributions from presence-only data due to its high predictive performance, stability, and sensitivity when applied to limited occurrence dataset. However, its performance depends on factors such as data quality, sampling design, and model parameterization [3,4,11,15,16,17,18,19].

Medicinal and aromatic plants constitute a plant resource of considerable health and economic importance, given their widespread use in traditional medicine and their content of secondary metabolites with therapeutic properties used in the pharmaceutical and aromatic industries [20,21,22,23]. Studies have shown that these plants are highly sensitive to climate change, as variations in temperature, carbon dioxide, and ozone concentrations, as well as the increase in drought periods, affect their growth, biological production, and geographical distribution [23,24]. Furthermore, the persistence of these environmental pressures, combined with the degradation of natural habitats, threatens many wild medicinal species by altering or reducing their range, particularly in sensitive areas such as the Mediterranean basin [5,21,22,25,26,27]. Consequently, endemic medicinal species with restricted distributions, such as S. fontqueri, represent high conservation priorities because they are simultaneously exposed to ecological and anthropogenic pressures.

Stachys fontqueri Pau is an endemic plant species of the Moroccan Rif Mountains [28,29]. It belongs to the genus Stachys L., one of the largest genera within the family Lamiaceae, specifically the subfamily Lamioideae. The genus comprises approximately 275–300 species of herbaceous and shrubby plants with a nearly worldwide distribution [30,31]. S. fontqueri is distinguished by several morphological characteristics, including densely pubescent stems, a deeply divided upper corolla lip, and prominently exserted stamens [30]. The species occurs in restricted and fragile mountain habitats, where it depends on specific environmental conditions for its survival [26,29]. Its narrow geographical distribution further increases its vulnerability to multiple threats, including habitat degradation, unsustainable resources exploitation, water drainage, Cannabis cultivation, and the increasing impacts of climate change in the Mediterranean region [26,29]. Owing to these pressures and its restricted distribution, S. fontqueri was classified as Vulnerable (VU) at the national level by Fennane (2018) [28].

Given the numerous threats facing S. fontqueri, together with its non-representation in botanical garden collections worldwide [26,29], and its limited ex situ conservation, with only a single accession currently preserved in the gene bank of the Scientific Institute of Rabat (Morocco) [32], as well as its promising agro-alimentary, ornamental and medicinal potentials [31,33,34], the conservation of this species is of growing importance. This necessitates monitoring the effects of climate change on the species’ potential distribution within its natural habitat, alongside the adoption of ex situ conservation strategies combined with enhanced in situ protection.

Several studies have demonstrated the effectiveness of the MaxEnt model in predicting potential future changes in the geographic distribution of medicinal plant species in Morocco under various climate change scenarios (e.g., [5,35,36,37,38]). To date, no study has evaluated the potential impacts of climate change on the potential distribution of S. fontqueri. Furthermore, with the exception of research focusing on its ex situ conservation status and the potential agro-alimentary and ornamental values [29,33,34], little attention has been given to the in situ conservation of this endemic species or to the combined impacts of climate change and human pressures on its potential distribution in Morocco.

Despite its endemic status, restricted geographic distribution, specialized ecological requirements, and conservation importance, the effects of climate change on the current and future distribution of S. fontqueri remain poorly understood. This knowledge gap limits the development of effective in situ conservation and management strategies necessary to ensure the long-term survival of the species. Given the restricted geographic range and ecological specialization of S. fontqueri, we hypothesize that future climate change will lead to a contraction of its climatically suitable habitat. We further hypothesize that temperature and precipitation will constitute the principal environmental drivers shaping its distribution, consistent with the species’ dependence on the cool and humid mountain ecosystems of the Moroccan Rif. To test these hypotheses, this study aims to (i) model the current potential distribution of S. fontqueri in the Moroccan Rif, (ii) assess the effects of climate change on its distribution using ecological niche modeling, (iii) identify the most influential environmental factors shaping its distribution, and (iv) provide recommendations for its conservation and management.

2. Materials and Methods

2.1. Study Area

The Moroccan Rif (Figure 1) reaches a maximum elevation of 2440 m and is characterized by a Mediterranean climate [39]. According to the Köppen–Geiger Climate Classification System (2026) [40], this climate is classified as Csa (temperate with hot, dry summers) at low and mid elevations and as Csb (temperate with warm, dry summers) at higher elevations. Across the sampled occurrence localities, the mean annual precipitation was 909.2 ± 134.6 mm, while the mean annual temperature was 14.675 ± 2.54 °C [41] (https://www.worldclim.org/, accessed on 28 August 2023).

The Moroccan Rif is a geographically distinct region of remarkable uniqueness within both Morocco and the Mediterranean basin. It represents one of the eleven national biogeographic divisions identified by Fennane et al. (1999) [42]. As part of the Mediterranean biodiversity hotspot, the Rif region harbors a significant proportion of Morocco’s plant diversity and is particularly rich in endemic species, many of which are classified as endangered or threatened due to their restricted distribution ranges. These taxa may be exposed to pressures related to overexploitation and/or illegal trade [29,43].

Numerous habitat types were identified within the study area, including coniferous forests dominated by Cedrus atlantica (Endl.) Carrière, Abies marocana Trab., Pinus halepensis Mill., Pinus nigra subsp. mauretanica (Maire & Peyerimh.) Heywood, Tetraclinis articulata (Vahl) Mast., and Taxus baccata L. Deciduous and evergreen oak forests are represented by Quercus ilex L., Q. canariensis Willd., Q. faginea Lam., Q. pyrenaica Willd., and Q. suber L. The study area also includes Mediterranean matorral communities dominated by Quercus coccifera L., Pistacia atlantica Desf., P. lentiscus L., Myrtus communis L., Cistus spp., Calicotome villosa (Poir.) Link, Chamaerops humilis L., … [5,25,26,27,36].

Stachys fontqueri Pau (Lamiaceae) is a narrow endemic (stenendemic) taxon restricted to the Western Rif Mountains of northern Morocco. It is one of the 17 Stachys species found in the country and is characterized by a greenish appearance with light to moderate hairiness, rarely becoming whitish-tomentose (Figure 2) [44].

Within the Rif biogeographical subdivision, S. fontqueri occurs as small, fragmented populations inhabiting matorral communities dominated by Pistacia lentiscus and Quercus coccifera, Chamaerops humilis steppes, and coniferous forests, where it frequently coexists with other endemic or threatened species such as Abies marocana and various orchid species. The species can grow on both calcareous and acidic substrates and occurs at elevations above 280 m a.s.l. [25,26,45].

2.2. Methods

After removing duplicate, outlier, and/or erroneous records, the final dataset included 70 occurrences of Stachys fontqueri. The initial dataset of occurrences were compiled from field surveys (5 occurrences) and presence data downloaded from the Global Biodiversity Information Facility (GBIF; 65 occurrences) [46]. To guarantee data quality and reliability, the dataset was subjected to manual screening to ensure the availability of precise geographic coordinates within the species’ distribution range, along with an accurate recording date for the occurrence site. Thus, occurrences with missing or invalid geographic coordinates, obvious georeferencing errors (e.g., records located outside the species’ known distribution or in marine environments) were removed. The occurrence data were spatially filtered using a 1 × 1 km grid, retaining a single occurrence per grid cell. The 1 × 1 km grid was chosen to match the spatial resolution (30 arc-seconds, approximately 1 km²) of the environmental variables used in the MaxEnt model, thereby ensuring consistency between occurrence records and predictor variables while minimizing the effects of spatially clustered sampling and autocorrelation [5,47,48,49].

The Maxent model (Version 3.4.3, November 2020) was used to predict the ecological niche of S. fontqueri. The climatic data used for the modeling were extracted from the WorldClim database [41] and handled using ArcGIS software (v. 10.8). Initially, a set of 19 environmental variables and elevation were used to build a preliminary MaxEnt model to simulate the current potential distribution of the species, following the approach of Worthington et al. (2016) [50], Wei et al. (2018) [51], and Mechergui et al. (2025a) [3]. Following this initial modeling, four variables were excluded from further analyses due to their lack of contribution to the model (percent contribution = 0) (Table S1).

The remaining variables were then selected based on a combination of preliminary MaxEnt performance, ecological relevance, and statistical independence. First, the ecological importance of each variable was assessed through a review of the relevant literature, focusing on studies of Mediterranean and endemic plant species with ecological characteristics similar to S. fontqueri [5,38,47]. Subsequently, pairwise Pearson correlation coefficients were calculated for all environmental and topographic variables across the species occurrence points using IBM SPSS Statistics 23. Variables exhibiting strong correlations (|r| > 0.8) were considered collinear and excluded to reduce multicollinearity and minimize model overfitting [1,51]. For each pair of correlated variables, the predictor with the higher contribution in the preliminary MaxEnt model and the greatest ecological relevance was retained, whereas its correlated counterpart was discarded [3]. Following this selection procedure, five bioclimatic variables were retained for the final model, namely: Annual mean temperature (Bio1), Temperature annual range (Bio7), Precipitation seasonality (Bio15), Precipitation of the warmest quarter (Bio18), and Precipitation of the coldest quarter (Bio19) (Table S2).

To assess the potential impacts of climate change on the distribution of S. fontqueri, the same set of selected environmental variables was projected onto future climate conditions derived from General Circulation Model (GCMs). Future climate data were obtained from the BCC-CSM2-MR and MIROC6 models under two Shared Socio-economic Pathways (SSP) scenarios: SSP1-2.6, representing a low-emission pathway, and SSP5-8.5, representing a high-emission pathway, for the period 2061–2080. The use of these two models provides complementary projections of future climatic conditions for assessing the potential distribution of the species [52]. All climatic datasets were processed at a spatial resolution of 30 arc-second, which is considered suitable for species distribution modeling and provide reliable estimates of habitat suitability [47]. To account for uncertainty in future climate projections, the SSP1-2.6, and SSP5-8.5 scenarios, under two GCMs, were chosen to represent a range of plausible socioeconomic and greenhouse gas emission pathways, spanning from low- to high-emission futures [53].

The MaxEnt model configuration was set to a maximum number of 500 iterations, with a convergence threshold of 0.00001. To generate a different random partition of test data and background points for each model run, the “random seed” option was enabled. Model performance was assessed using a 10-fold cross-validation procedure, and predictions were generated in a logistic output format [54,55]. For the feature classes, the “Auto” setting was applied, while the regularization multiplier was maintained at its default value of 1. This modeling approach, which improves the robustness of performance estimates, is particularly suitable for species with limited occurrence records [54].

Model performance was evaluated using the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) [56]. This parameter reflects the probability that a randomly selected presence site is assigned a higher suitability score than a randomly selected background or absence site [57]. A value of 0.5 indicates a random prediction, whereas a value of 1.0 represents perfect model discrimination [58]. In addition to the AUC, model performance was evaluated using the True Skill Statistic (TSS), a threshold-dependent metric that combines sensitivity and specificity, with −1 < TSS < 1. Values greater than 0.8 indicate excellent model performance [5]. The TSS was calculated based on Maximum training sensitivity plus specificity Logistic Threshold, using formula: TSS = Sensitivity + Specificity − 1. Omission rate is another evaluation metric that quantifies the proportion of observed presence records incorrectly predicted as absent and provides a threshold-dependent measure of model predictive performance. Lower omission rates indicate greater model reliability and better predictive ability [59,60].

To compare habitat suitability under climate scenarios, MaxEnt outputs were imported into ArcGIS 10.8 and analyzed using the Reclassify tool within the Spatial Analyst extension. MaxEnt predicts habitat suitability for each grid cell on a continuous scale ranging from 0 to 1. To distinguish suitable from unsuitable habitats, the Maximum Test Sensitivity Plus Specificity (MTSPS) threshold, a robust and widely adopted criterion in species distribution modeling, was applied [5,61,62]. Habitat suitability maps were reclassified into two categories: suitable habitat (p > MTSPS) and unsuitable habitat (p ≤ MTSPS) [5].

Response curves generated by MaxEnt for all environmental variables, which describe the relationship between each predictor and the predicted habitat suitability were examined. Their interpretation was based on their biological relevance, showing that variations in environmental variables generally produced smooth and predictable responses in the species’ probability of presence, without extreme fluctuations [3,4,62].

3. Results

3.1. Model Evaluation

The high values of the AUC of the five models developed to assess the current and future potential distribution of Stachys fontqueri demonstrated excellent predictive performance. The mean AUC values were 0.921 ± 0.027, 0.930 ± 0.027, 0.929 ± 0.044, 0.929 ± 0.035, and 0.926 ± 0.039 for the current climate model and the CSM2-SSP1-2.6, CSM2-SSP5-8.5, MIROC6-SSP1-2.6, and MIROC6-SSP5-8.5 scenarios, respectively. The consistently high AUC values, approaching 1, together with the low standard deviations (0.027–0.044), indicate strong model performance and stability. In addition, the TSS values varied from 0.81 to 0.83, further confirming the high predictive accuracy and reliability of the models. However, the test omission rates obtained for the current and future climate scenarios were consistently low, ranging from 1.43% to 8.10%, indicating good predictive performance. These results further support the robustness and reliability of the MaxEnt models beyond the AUC and TSS values.

The models predicted suitable habitats that closely match the currently known geographic distribution of S. fontqueri in northern Morocco. Nevertheless, the four projected scenarios identified additional areas of high habitat suitability in the Ketama–Targuist, Bouhachem, Bab Taza, and Tangier regions, where no occurrences of the species have been documented to date. Although these areas were identified as potentially suitable by the MaxEnt model, field surveys are required to confirm their suitability and determine whether S. fontqueri is present.

3.2. Bioclimatic Factors

The potential distribution area of S. fontqueri in Moroccan Rif were mostly affected by Precipitation Seasonality (Bio15), which contribute 61.6%, 43.8% and 38.1% for current, MIROC6-SSP5-8.5, and CSM2-SSP1-2.6 future climatic scenarios respectively. However, the Temperature Annual Range (Bio7) that have the greatest effect on habitat suitability under CSM2-SSP5-8.5 and MIROC6-SSP1-2.6, which contribute 55.8% and 43.6% respectively (

Table 1. Permutation importance of the bioclimatic variables used to model the potential geographic distribution of Stachys fontqueri in Moroccan Rif.

Bioclimatic Variables	Current (%)	CSM2-SSP1-2.6 (%)	CSM2-SSP5-8.5 (%)	MIROC6-SSP1-2.6 (%)	MIROC6-SSP5-8.5 (%)
Annual Mean Temperature (Bio1)	17.9	19.9	16.7	16.2	11.9
Temperature Annual Range (BIO5-BIO6) (Bio7)	13.7	27.3	55.8	43.6	32
Precipitation Seasonality (Bio15)	61.6	38.1	9.6	28.9	43.8
Precipitation of Warmest Quarter (Bio18)	1.7	5.2	1.9	4.8	8.1
Precipitation of Coldest Quarter (Bio19)	5.1	9.6	16	6.5	4.1

).

3.3. Response Curves

Under current climate (Figure S1), the response curves derived from the ecological niche modeling of S. fontqueri indicated a high probability of species occurrence for Precipitation Seasonality (Bio15) values ranging from 77 to 88 mm, an Annual Mean Temperature (Bio1) below 12.8 °C, and a Temperature Annual Range (Bio7) below 17.4 °C and between 26.4 and 28.4 °C. Although less influential than the three principal predictors, suitable conditions were also associated with Precipitation of the Warmest Quarter (Bio18) values below 10 mm and between 17 and 25 mm, and Precipitation of the Coldest Quarter (Bio19) between 410 and 490 mm. Although the response curves for Bio7 and Bio18 exhibited a bimodal pattern, this should be interpreted cautiously, as such complex responses may result from interactions among environmental predictors or model structure rather than representing distinct ecological preferences.

Similar response patterns were observed under the CSM2-SSP1-2.6 scenario (Figure S2), with habitat suitability remaining primarily associated with annual mean temperature (Bio1), precipitation seasonality (Bio15), and temperature annual range (Bio7), although slight shifts in the optimal ranges were predicted relative to the current climate. Under this scenario, the response curves indicated Bio18 values below 9 mm and between 21 and 27 mm, while optimal Bio19 values shifted to between 470 and 560 mm.

Under the CSM2-SSP5-8.5 scenario (Figure S3), Temperature Annual Range (Bio7) emerged as the most influential factor determining habitat suitability, with optimal values ranging from 28.4 to 30.3 °C. This was followed by Annual Mean Temperature (Bio1) below 16.1 °C and Precipitation of the Coldest Quarter (Bio19) between 460 and 540 mm. In contrast to the other scenarios, Bio18 was also associated with a broader optimum, ranging from 16 to 40 mm. In addition, Precipitation Seasonality (Bio15) ranged from 78 to 92 mm. Compared with the current climate, the model predicted a shift toward slightly higher values of these climatic variables.

Under both the SSP1-2.6 and SSP5-8.5 scenarios projected by the MIROC6 model (Figures S4 and S5), the highest probabilities of occurrence for S. fontqueri were associated with similar bioclimatic variables and relatively comparable value ranges. Under the SSP1-2.6 scenario, habitat suitability remained primarily associated with Temperature Annual Range (Bio7) between 28.4 and 30.2 °C, Precipitation Seasonality (Bio15) between 77.5 and 88.5 mm, and an Annual Mean Temperature (Bio1) below 15 °C. Bio18 was associated with values below 10 mm and between 17 and 23 mm, whereas Bio19 ranged from 375 to 450 mm. In contrast, under the SSP5-8.5 scenario, habitat suitability remained primarily associated with Bio15 values between 77.5 and 88 mm, Bio7 values between 28.6 and 30.6 °C, and Bio1 values below 17.2 °C. The corresponding response curves indicated Bio18 values between 15 and 23 mm and Bio19 values between 340 and 410 mm.

Overall, except under the CSM2-SSP5-8.5 scenario, Bio18 and Bio19 showed relatively weak contributions to habitat suitability despite exhibiting identifiable response ranges. Under the CSM2-SSP5-8.5 scenario, however, Bio19 became one of the dominant predictors, indicating that winter precipitation may play a more important role in determining the future climatic suitability of S. fontqueri under more extreme warming conditions.

3.4. Potential Current Distribution Area of Stachys fontqeuri

The modeling of the potential distribution of S. fontqueri under current climatic conditions indicates that the species could occupy relatively extensive, although fragmented, suitable habitats concentrated in the mountainous massifs of the Moroccan Rif, particularly in Jbel Moussa, Jbel Kelti, and the Talassemtane National Park. However, the predicted suitable habitat also extends to surrounding areas, including Bouhachem Natural Park, the Tangier region, Bab Taza, Ketama, and Issaguen, covering a total suitable habitat area of approximately 3244 km² (

Table 2. Probabilities of the suitable and unsuitable habitat area of Stachys fontqueri in Moroccan Rif predicted by the models. Habitat contraction indicated as “% of loss” compared to the current suitable area.

Suitability	Current km2	CSM2-SSP1-2.6	CSM2-SSP5-8.5	MIROC6-SSP1-2.6	MIROC6-SSP5-8.5
MTSPST < p < 1	3244.3877 km2	2287.1922 km2 (Loss: −29.5%)	2284.313 km2 (Loss: −29.59%)	2377.154 km2 (Loss: −26.73%)	2490.146 km2 (Loss: −23.25%)
p < MTSPST	17,507.31 km2	18,464.51 km2	18,467.39 km2	18,374.55 km2	18,261.55 km2

Caption: Values of MTSPST: under current climate; 0.2999, under CSM2-SSP1-2.6 scenario; 0.3329, under CSM2-SSP5-8.5 scenario; 0.3294; under MIROC6-SSP1-2.6 scenario; 0.3161, under MIROC6-SSP5-8.5 scenario; 0.3423.

; Figure 3).

3.5. Effect of Climate Change on the Potential Distribution Area of Stachys fontqueri

Compared to the current potential distribution, the potential distribution area of S. fontqeuri under all future climate scenarios, was projected to decrease in the study area (23.25% to 29.59% of suitable habitat loss). The CSM2-SSP5-8.5 seems to be the most unfavorable climatic scenario for the endemic species (Table 2).

When considering climatic factors alone and disregarding human influences, S. fontqueri is projected to face a risk of habitat contraction in the Moroccan Rif. Future climate projections indicate a marked decline in the extent of suitable habitats for the species throughout Moroccan Rif by 2070 (Figure 4).

Comparison between the current and future habitat suitability maps revealed distinct patterns of habitat contraction, expansion, and stability across all climate scenarios (Table 2; Figure 4). Habitat contraction consistently exceeded habitat expansion, indicating a net reduction in the climatically suitable range of S. fontqueri. Habitat contraction ranged from 818.30 km² under MIROC6-SSP5-8.5 to 999.67 km² under BCC-CSM2-MR-SSP5-8.5, whereas habitat expansion remained limited, varying from only 7.85 km² under BCC-CSM2-MR SSP1-2.6 to 64.00 km² under MIROC6 SSP5-8.5, representing less than 2% of the current suitable area. Despite these projected changes, a substantial proportion of the current suitable habitat remained stable, ranging from 2244.72 km² (69.2%) under BCC-CSM2-MR-SSP5-8.5 to 2426.09 km² (74.8%) under MIROC6-SSP5-8.5.

Spatially, habitat contraction was mainly concentrated along the margins of the species’ current distribution, whereas stable habitats were primarily retained within the core mountain areas of the Moroccan Rif. Habitat expansion was restricted to a few localized areas adjacent to the current distribution, particularly in the Ketama-Targuist, Bouhachem, Bab Taza, and Tangier regions. However, these newly identified areas represent potentially suitable habitats predicted by the model and require field surveys for validation before being considered occupied or suitable for conservation planning.

4. Discussion

Climate change is one of the major drivers of biodiversity decline in the Mediterranean Basin [2,6]. Its impacts are particularly severe on endemic and threatened species, especially when combined with human pressures [5]. As an endemic taxon of forest clearings and calcareous scrublands, Stachys fontqueri contributes to local biodiversity and plays an important role in maintaining habitat integrity in the Moroccan Rif [29]. Therefore, modelling the potential habitat changes and the response of the species to climate change are essential to develop effective conservation strategies.

Bioclimatic variables provide valuable predictors for understanding plant distribution patterns and for modeling the potential effects of climate change on habitat suitability, thereby enabling the assessment of current and future species distribution patterns [38,63]. However, the results of ecological niche models should be interpreted with caution, as they do not fully account for important ecological processes such as biotic interactions, dispersal capacity and barriers, evolutionary history, and human activities [5,63,64,65,66,67].

The models applied to analyze the ecological niche of S. fontqueri showed high predictive performance, with AUC values ranging from 0.921 to 0.930 and TSS values between 0.81 and 0.83. These performance metrics are in close agreement with those reported by El Haddouti et al. (2026) [5] and Jaouani et al. (2025) [38] for species distribution models in northern Morocco, further confirming the robustness and reliability of the modeling approach [57]. However, these metrics should not be interpreted as definitive evidence of model robustness or as guarantees of accurate future projections. AUC and TSS primarily assess the model’s ability to discriminate between suitable and unsuitable environmental conditions based on the available occurrence and environmental data, but they do not account for all sources of uncertainty associated with species distribution projections, including sampling bias, model parameterization, and uncertainty in future climate scenarios [59]. Although additional evaluation metrics, such as the continuous Boyce index, can provide complementary information on model performance and reliability [68], these metric was not generated during the original modelling workflow. Future studies should incorporate complementary evaluation metrics, together with calibration statistics, optimized parameter tuning, and spatially structured validation, to provide a more comprehensive assessment of model predictive performance.

The resulting habitat suitability maps indicate that most of the currently known populations of S. fontqueri are located within the core of the potential distribution area predicted by the models (Figure 3). However, the absence of S. fontqueri from several areas predicted to be suitable, including Jbel Bouhachem, Bab Taza, Ketama, Issaguen, and the Tangier region, deserves attention. This discrepancy may be explained by the influence of additional factors that were not incorporated into the models but are known to shape the species’ realized distribution, including local abiotic conditions and human disturbances such as overgrazing, wildfires, deforestation, and illegal plant collection [5,69,70]. Alternatively, the apparent absence of the species may simply reflect insufficient botanical investigations, underscoring the need for field validation of habitats identified as suitable by the ecological niche models [5].

Ecological niche modeling of S. fontqueri revealed that its potential geographic distribution is primarily associated with three key bioclimatic variables: Precipitation Seasonality (Bio15) and the two temperature-related variables, Temperature Annual Range (Bio7) and Annual Mean Temperature (Bio1). These variables consistently showed the greatest relative contribution within the fitted MaxEnt models under both current climatic conditions and the four future climate scenarios. However, these contribution values should be interpreted as model-dependent measures of predictor importance rather than direct evidence of causal ecological relationships, as they may vary according to model structure and correlations among environmental variables. The ecological significance of these variables is further supported by the response curves and by the known environmental requirements of the species. In contrast, Annual Mean Temperature (Bio1) was identified as the most influential predictor of the potential distribution of Stachys inflata Benth. in Iran [47]. Similarly, in northern Morocco, Bio1 and Bio7 have been reported among the principal environmental factors associated with the distribution of two Lamiaceae species, Origanum compactum Benth. and O. elongatum (Bonnet) Emb. & Maire [38].

Although Precipitation Seasonality (Bio15) exhibited the highest relative contribution in the current MaxEnt model (61.6%), this result should be interpreted as indicating its importance within the fitted model rather than as definitive evidence of its ecological dominance. Nevertheless, the prominence of Bio15 is biologically consistent with the ecology of S. fontqueri which is endemic to the western Rif, a region characterized by a Mediterranean climate with pronounced seasonal variability in precipitation. Seasonal rainfall patterns are known to influence water availability and may affect the growth, persistence, and distribution of Mediterranean mountain plant species [37,38]. In this context, Precipitation Seasonality (Bio15) represents an indicator of seasonal water stress, as increasing contrasts between the wet and dry seasons may reduce habitat suitability [71]. Similarly, recent research has shown that variation in soil water availability can strongly influence plant functional traits and productivity, highlighting the fundamental role of water availability in shaping plant responses to changing climatic conditions [72]. Under future climate scenarios (CSM2-SSP1-2.6 and MIROC6-SSP5-8.5), Bio15 remained among the principal predictors in the MaxEnt models, although its relative contribution decreased compared with the current climate (Table 1). This pattern suggests a shift in the relative importance of predictors within the models under future climatic conditions, with temperature-related variables contributing more strongly to habitat suitability predictions. However, this should not be interpreted as direct evidence that temperature will necessarily become the dominant ecological driver of the species’ distribution. This interpretation is consistent with previous studies showing that climate change may alter the relative influence of climatic predictors in species distribution models [73,74].

The prominent contribution of Precipitation Seasonality (Bio15) to the ecological niche modeling of S. fontqueri differs from the findings of Hosseini et al. (2024) [18] for Thymus species in Iran, where the most influential environmental variables varied among species. Altitude was identified as the most significant predictor of the distribution of Thymus fedtschenkoi Ronniger and T. pubescens Boiss. & Kotschy ex Čelak., whereas precipitation-related variables, particularly Precipitation of the Driest Month (Bio14), also played an important role, especially for T. fedtschenkoi. In contrast, Shaban et al. (2023) [47] identified a combination of climatic variables, including Annual Mean Temperature (Bio1), Mean Temperature of the Wettest Quarter (Bio8), and Annual Precipitation (Bio12), as the principal determinants of the distribution of Stachys inflata, with relative contributions of 41.1%, 39.06%, and 37%, respectively. This variability indicates that no single environmental variable consistently dominates species distribution models; rather, the relative importance of environmental predictors is determined by the specific ecological requirements of each species and the environmental conditions prevailing within its natural habitat [75].

Under the future climate scenarios CSM2-SSP5-8.5 and MIROC6-SSP1-2.6, Temperature Annual Range (Bio7) emerged as the most influential predictor, with relative contributions of 55.8% and 43.6%, respectively. This shift in the dominant climatic drivers suggests that, under accelerated climate warming, annual thermal constraints are likely to replace precipitation-related constraints as the primary factors regulating habitat suitability for S. fontqueri. Similar transitions have been reported for other Mediterranean plant species that are particularly sensitive to climate change [3,38]. The shift described above indicates that thermal variability plays a key role in regulating the physiological performance and reproductive capacity of the species. Large annual temperature fluctuations may influence the species distribution range and have been reported to affect biological processes, such as flowering phenology and fruit production in plants [71]. However, these physiological responses were not assessed in the present study. Furthermore, all four future climate scenarios predicted an increase of approximately 1–2 °C in Temperature Annual Range (Bio7) relative to current climatic conditions. Such an increase is expected to exacerbate environmental aridity and intensify thermal stress, thereby imposing greater physiological constraints on the species [7].

However, the apparent bimodal response of Bio7 and Bio18 should be interpreted with caution, as complex response curves in MaxEnt may arise from interactions among correlated predictors, model structure, or limited sampling rather than reflecting true ecological preferences. Consequently, these modeled relationships should be regarded as hypotheses requiring independent ecological validation through field observations or experimental studies [14,59].

Except the CSM2-SSP5-8.5 scenario, Annual Mean Temperature (Bio1) ranked as the third main influential bioclimatic variable determining the potential distribution of S. fontqueri in the Moroccan Rif under future climate scenarios, with relative contributions ranging from 11.9% to 19.9% (Table 1). In contrast, this thermal variable was identified as the primary predictor of the potential distribution of S. inflata in Iran, contributing 41.10% to the MaxEnt model [47], and of Heteromera philaenorum Maire & Weiller in North Africa, contributing 28.9% [4]. Annual Mean Temperature is a fundamental predictor in ecological niche modeling because it provides an estimate of the total energy input available within an ecosystem throughout the year, which can strongly influence the metabolic rates, growth, and physiological performance of the species [71].

Ecological niche modeling under current climate revealed that the predicted suitable habitat of S. fontqueri is mainly distributed across the Rif Mountains, covering an estimated total area of approximately 3244 km². The most suitable habitats are located around Talassemtane National Park, Jbel Moussa, and Jbel Kelti. This occurrence in the Rif mountains may be attributed to suitable climatic conditions, particularly an Annual Mean Temperature (Bio1) below 12.8 °C, and Precipitation Seasonality (Bio15) ranging from 77 to 88 mm. Together, these conditions provide a cool and relatively humid environment that is conducive to the growth and persistence of the species [76,77,78].

Future climate projections predicted a contraction of the suitable habitat of S. fontqueri under all four climate scenarios evaluated, with habitat losses ranging from 23.25% to 29.59% under the high-emission scenarios MIROC6-SSP5-8.5 and CSM2-SSP5-8.5. These findings are consistent with the trends commonly reported for endemic and geographically restricted species in the Mediterranean region under the influence of climate change [2,38]. The projected range contraction occurs primarily along the southern and eastern margins of the species’ current potential distribution, suggesting that climatically suitable habitats may become increasingly concentrated in the western Rif, where relatively cool and humid conditions prevail [79]. It is noteworthy that the low-emission scenarios also predicted substantial habitat losses, with reductions of 29.50% and 26.73% under the CSM2-SSP1-2.6 and MIROC6-SSP1-2.6 scenarios, respectively. These results highlight that, even under ambitious greenhouse gas mitigation pathways, climate change is expected to exert considerable pressure on this narrow endemic species [53].

Overall, the projected contraction of the potential distribution of S. fontqueri is consistent with the findings of Hosseini et al. (2024) [18], who reported substantial reductions in suitable habitats for several Thymus species by 2050 and 2070 as a consequence of increasing thermal and water stress under climate change. Similarly, Shaban et al. (2023) [47] predicted a marked decline in the most suitable habitats for S. inflata under future climate scenarios. Collectively, these studies indicate that climate warming not only leads to a reduction in the extent of suitable habitats but also alters the environmental drivers governing habitat suitability, with temperature-related variables becoming increasingly important under future climate scenarios [80,81].

At the regional scale, previous studies have reported a widespread decline in biodiversity across the Moroccan Rif, driven by both local environmental conditions and increasing human pressures [5,25,26,43,82]. In addition to the impacts of climate change projected in the present study, our field surveys revealed several ongoing threats to the natural habitats of S. fontqueri, including livestock grazing, cannabis cultivation, and water abstraction, even within protected areas such as Talassemtane National Park. Furthermore, the proximity of some populations to urban centers, particularly the cities of Tetouan and Fnideq, may expose the species to increased risks of habitat fragmentation, loss of connectivity among populations, overexploitation, and even complete habitat destruction associated with urban and infrastructure development [83,84].

Given the restricted distribution and threatened status of S. fontqueri, the highest conservation priority should be the implementation of in situ conservation measures to prevent further habitat degradation and population decline. These measures should include the effective enforcement of existing legislation, such as Law 29-05 of 2 July 2011, which protects wild species from illegal collection and trade [85], the extension of protected areas where appropriate, and the restoration of degraded habitats. The in situ conservation of this endemic species and its natural habitat, particularly in the Tetouan region, would contribute to safeguarding regional biodiversity while supporting ecosystem restoration and environmental education [83]. In parallel, effective conservation plans should address the principal anthropogenic threats affecting the species, notably overgrazing, forest fires, illegal plant harvesting and trade, and habitat degradation [43,82].

As a complementary conservation strategy, ex situ conservation should be strengthened to safeguard the species’ genetic diversity. Although a previous study reported the presence of only a few seed lots of S. fontqueri conserved in the seed bank of the Scientific Institute (Rabat, Morocco) [32], additional efforts are needed to collect and conserve a larger and more representative quantity of seeds, with priority given to Moroccan seed banks. Ex situ collections should also support germination, propagation, and reintroduction trials in restored habitats, in accordance with the recommendations of the Global Strategy for Plant Conservation (GSPC) [86].

Finally, community engagement should complement both in situ and ex situ conservation efforts. Raising awareness among local communities about the ecological, medicinal, socioeconomic, and cultural importance of S. fontqueri would promote the sustainable management of its natural habitat, reduce human pressures on wild populations, and improve the long-term success of conservation initiatives.

This study constitutes a valuable contribution to understanding the effects of climate change on the potential distribution of S. fontqueri in the Moroccan Rif. Nevertheless, it presents certain limitations, particularly regarding the use of 65 occurrence records obtained from GBIF, as publicly available biodiversity databases can contain uncertainties associated with taxonomic misidentification, georeferencing errors, and temporal inconsistencies [87,88]. Although the occurrence dataset was carefully cleaned and validated prior to modelling, residual uncertainties may remain and could influence model predictions. Nevertheless, GBIF is one of the most comprehensive and widely used sources of biodiversity data for species distribution modelling, particularly for rare and endemic species for which field observations are limited [89]. Future studies should increase field sampling to improve the completeness and accuracy of occurrence datasets and to further validate the predicted distribution of S. fontqueri. The use of a limited number of bioclimatic variables for modeling the species’ ecological niche may constitute another limitation of this study. Although the five bioclimatic variables were selected for their contribution and statistical independence, they may not adequately represent all the ecological processes shaping the species’ geographic distribution. Beyond occurrence data and climate, other elements could affect the survival of the species, namely habitat fragmentation, pastoralism, land-use change, and fires [5,25,26]. These factors may constrain the colonization of newly suitable areas projected by the models. Furthermore, areas predicted to be climatically suitable do not necessarily correspond to the species’ current distribution, as potential climatic suitability does not always reflect realized occupancy. This mismatch highlights the importance to conducting additional field surveys to validate the predicted suitable areas and of incorporating additional ecological, biotic, and human factors into future models to enhance predictive accuracy [5,47].

Another limitation of the present study is the use of only two CMIP6 General Circulation Models (BCC-CSM2-MR and MIROC6). Although these models provide complementary climate projections [52], they do not encompass the full range of uncertainty associated with future climate conditions. Therefore, future studies should incorporate a larger ensemble of CMIP6 models to better quantify climatic uncertainty and increase confidence in projections of the future potential distribution of S. fontqueri.

A further limitation is that the MaxEnt model was implemented using the default regularization multiplier (1) and the “Auto” feature class setting without systematic parameter tuning. Although these settings have been widely applied and generally provide reliable predictions [59,60], optimizing model complexity through the selection of appropriate regularization multipliers and feature-class combinations can improve predictive performance and reduce overfitting, particularly for species with limited occurrence records. Therefore, future studies should optimize MaxEnt parameterization using tools such as ENMeval prior to model calibration [90].

Another limitation of this study is that model performance was evaluated using random 10-fold cross-validation rather than spatially structured cross-validation. Although spatial filtering was applied by retaining one occurrence record per 1 × 1 km grid cell to reduce sampling bias and spatial autocorrelation, random partitioning of occurrence records may still overestimate model performance when geographically proximate records share similar environmental conditions [91]. Future studies should therefore incorporate spatial cross-validation approaches, such as those implemented in the blockCV package [92], to improve the robustness and reliability of model evaluation.

A further limitation concerns the use of a single thresholding method to convert continuous habitat suitability predictions into binary suitability maps. The Maximum Test Sensitivity Plus Specificity (MTSPS) threshold was selected because it balances omission and commission errors by maximizing the sum of sensitivity and specificity, making it one of the most widely recommended thresholding methods for species distribution models [93,94]. Nevertheless, different thresholding methods can produce substantially different estimates of suitable habitat area and, consequently, may influence conservation assessments. Future studies should therefore evaluate the sensitivity of model predictions to alternative thresholding methods to improve the robustness of habitat suitability estimates.

Another important limitation of this study is that the ecological niche models were developed primarily using climatic variables and therefore predict the potential climatic suitability of S. fontqueri rather than its realized distribution. Although climate is a major determinant of species distributions at broad spatial scales, other ecological processes, including biotic interactions (e.g., competition, pollination, and herbivory), dispersal limitations, and anthropogenic pressures such as habitat degradation, land-use change, overharvesting, and cannabis cultivation, may strongly influence the occurrence and long-term persistence of populations. Consequently, the climatically suitable areas identified by the models should be interpreted as potentially suitable habitats rather than definitive future distributions.

Finally, although the models showed high predictive performance, high AUC and TSS values should not be interpreted as definitive evidence of model robustness or accurate future projections. Model performance metrics primarily assess the ability to discriminate between suitable and unsuitable environmental conditions based on the available occurrence and predictor data, but they do not account for all sources of uncertainty associated with future species distribution projections.

Overall, addressing these limitations by integrating larger ensembles of climate models, optimized model parameterization, spatially structured validation, multiple thresholding methods, and additional ecological and anthropogenic predictors into future modelling frameworks would improve the robustness and ecological realism of predictions, thereby providing stronger scientific support for the conservation and management of this endemic and vulnerable species.

5. Conclusions

This study successfully modelled the current potential distribution and evaluated the potential impacts of climate change on the climatic suitability of Stachys fontqueri in the Moroccan Rif using ecological niche modelling. The current distribution model accurately identified the principal areas of habitat suitability, which are concentrated within Talassemtane National Park, Jbel Moussa, and Jbel Kelti, highlighting these areas as priorities for conservation. Our results identified three key bioclimatic variables as the primary determinants of the species’ potential distribution: Precipitation Seasonality (Bio15), Temperature Annual Range (Bio7), and Annual Mean Temperature (Bio1). Future climate projections consistently predicted a contraction in climatically suitable habitats under all climate scenarios considered, with habitat losses ranging from 23.25% to 29.59% by 2070. Nevertheless, new potentially climatically suitable habitats may emerge in parts of the eastern Rif, particularly around Ketama and Issaguen. These projections represent potential climatic suitability rather than actual future occupancy and should therefore be interpreted with caution, as they do not account for biotic interactions, dispersal limitations, land-use change, or other anthropogenic factors that may influence the realized distribution of the species.

Considering the combined effects of climate change and the numerous anthropogenic pressures identified during our field surveys, the contraction of potentially suitable habitats for S. fontqueri could be even more severe than predicted if existing environmental regulations are not effectively implemented and if in situ and ex situ conservation measures remain insufficient. Future research integrating morphological and genetic studies aimed at identifying the most resilient populations, together with investigations of seed germination requirements to support natural regeneration, would improve our understanding of the adaptive capacity of the species and contribute to the development of effective long-term conservation strategies. In addition, botanical surveys should be intensified in areas predicted to be suitable but not yet inventoried, including Jbel Bouhachem, Bab Taza, Ketama, Issaguen, and the Tangier region, to validate the model predictions and identify key habitat corridors. Overall, the findings of this study provide a scientific basis for prioritizing conservation actions and long-term management of S. fontqueri. Priority should be given to strengthening in situ conservation through habitat protection and restoration, complemented by ex situ conservation of genetic resources, long-term monitoring of natural populations and land-use dynamics, sustainable resource management practices, and community-based awareness initiatives. Together, these measures would help mitigate the combined impacts of human activities and climate change while enhancing the long-term resilience of the species in the Moroccan Rif.