Prediction of distribution areas of the rare Iris kuschakewiczii B.Fedtsch. species across its range using the MaxEnt model

Abstract

Iris kuschakewiczii B.Fedtsch. is a rare ornamental plant endemic to the Northern Tien Shan mountains. The analysis of the species’ distribution range is important given the declining habitat area. This study employed MaxEnt to project the distribution of I. kuschakewiczii under current and three future climate change scenarios (SSP126, SSP245, SSP585) for two time periods (the 2050s and the 2090s). Prediction models showed high validity (AUC = 0.997), strong accuracy, and reliability. Moreover, key environmental variables affecting the distribution of I. kuschakewiczii were identified. It included Weight and Percentage of Sand (70%), Temperature Seasonality (~ 2300), and Elevation (700–1300 m above the sea level). Highly suitable distribution areas were mainly found in the Chu-Ili Mountains (Southeastern Kazakhstan). The analysis showed that only 26.7% of the total highly suitable area is protected by reserves and nature parks, underlying the need for conservation. The recently founded Kurty Valley Important Plant Area in Kazakhstan provides a strong foundation for the scientific conservation of I. kuschakewiczii and other rare and endangered plants. This research provides insights into the potential dynamics of habitat distribution for the endangered I. kuschakewiczii and offers a basis for strengthening cooperative international conservation and sustainable management strategies in the face of climate change.

Introduction

Climate change is reshaping the distribution of plant species worldwide at an unprecedented rate^1–6. Rising temperatures, shifting precipitation patterns, and increasing climatic variability are forcing species to track suitable conditions across latitudes and elevations, often faster than natural dispersal allows. For rare and narrowly distributed species, these pressures are particularly acute, as shrinking habitat windows and fragmented landscapes leave little room for adaptation^3,4. In biodiversity hotspots such as Central Asia^7–10, where warming is progressing faster than the global average and endemic species are concentrated in ecologically sensitive mountain systems^11–13, understanding how climate change will affect species distributions is an urgent conservation priority.

Iris kuschakewiczii B.Fedtsch. is a perennial herbaceous plant belonging to the subgenus Scorpiris within the genus Iris Tourn. ex L. It commonly grows on gravelly, fine-earth, and clayey slopes of foothills, plateau-like peaks, and hills, and on foothill plains of the Northern Tien Shan (Kazakhstan, Kyrgyzstan). The plant is a 5–15 cm tall, early-flowering bulbous iris (juno) that holds ornamental and melliferous value^14,15. The lectotype of I. kuschakewiczii, preserved in the Herbarium of the Komarov Botanical Institute (LE 00050068)¹⁶, was collected in Zaitzewka (now Shelek Village), Almaty region, by Kushakevich in 1877 and described in 1905 by Fedtchenko¹⁷. In Kyrgyzstan, the earliest documented record is reported near Frunze (now Bishkek) by Nikitina in 1929¹⁸. The species grows in open communities dominated by grasses and low shrubs, often on south-facing slopes that warm quickly in spring. As a bulbous geophyte, it spends most of the year underground, making it particularly sensitive to soil conditions, especially drainage, during dormancy. These ecological characteristics make I. kuschakewiczii highly dependent on a narrow set of habitat conditions, which partly explains its restricted distribution and vulnerability to both land use change and climate-driven habitat shifts. Kokoreva et al.¹⁹ comprehensively studied the morphology of species in different ontogenetic states and described the ecological characteristics of I. kuschakewiczii. More recently, the distribution range was specified in accordance with the floristic regions of Kazakhstan²⁰, and its chloroplast genome was characterized²¹. To date, these are the only studies published on this species, notwithstanding the inclusion of I. kuschakewiczii in the Red Data Book of Kazakhstan under Category I¹⁴. The given category indicates that the species is endangered, with a declining geographic range, and that its continued existence is impossible without the implementation of specific protection measures²².

As a narrow endemic confined to the Northern Tien Shan¹⁴, one of the most botanically significant yet climatically vulnerable mountain systems in Central Asia, it exemplifies the category of species for which SDM is most urgently needed but least applied. Moreover, the increasing impacts of negative anthropogenic factors, such as the popularization of ecotourism in vulnerable species’ communities, land-use policies, and aridification, place the naturally small populations of I. kuschakewiczii at great risk of extinction and of interest to conservation. Although its inclusion in the Red Data Book provides legal protection, the species remains vulnerable to extinction without effective, targeted conservation strategies.

To address this issue, species distribution models (SDMs) could be employed to predict species’ geographic distributions based on occurrence records and environmental data^23,24. One widely used modeling tool is the Maximum Entropy (MaxEnt) model^3,13,25,26. MaxEnt works best for a small number of occurrence sites^27,28, which is ideal for projecting the distribution areas of rare and endangered species. Therefore, the implementation of SDM provides a robust framework for assessing the potential distribution of I. kuschakewiczii, identifying high-priority areas for protection, and contributing to the development of evidence-based conservation strategies for this endangered species.

In this study, we used the MaxEnt program to model the current and future distribution of I. kuschakewiczii. Future projections were based on three climate change scenarios across two time periods and incorporated bioclimatic, edaphic, and solar radiation variables. We assessed habitat suitability, identified high-priority areas for conservation, and proposed applicable protection measures. The findings of this study constitute a valuable scientific resource, guiding the development of rational policies aimed at the long-term conservation and controlled cultivation of I. kuschakewiczii.

Materials and methods

Species occurrence data

The geographical distribution data for I. kuschakewiczii were mainly obtained from field surveys, the Global Biodiversity Information Facility (GBIF)²⁹, and the Herbarium Fund of the Institute of Botany and Phytointroduction (AA, Almaty, Kazakhstan, accessed December, 2024). The photo of I. kuschakewiczii in nature and its habitat were provided in Fig. 1. To minimize spatial clustering, the R package ‘spThin’³⁰ was used to remove samples within 1 km². The thinning distance was set below the grain of the environmental predictor layers, using a predictor-resolution thinning distance to reduce spatial autocorrelation while retaining the maximum number of records³¹. As a result, from the total of 190 occurrence points, 51 unique sites were compiled into a CSV file for further modeling analysis (Supplemetary File).

Fig. 1.

Photo of the I. kuschakewiczii species and its habitat.

Environmental variables

This study used 35 environmental variables, including 19 bioclimatic (WorldClim)³², 9 edaphic (Harmonized World Soil Database, HWSD v2.0)³³, 6 global UV-B radiation (gIUV)³⁴, and the human footprint index³⁵. Bioclimatic data sourced from WorldClim at 2.5 arc-minute resolution was used for projecting conditions for 1970–2000 (current) and future time periods of 2041–2060 (2050s), and 2081–2100 (2090s) under three Shared Socioeconomic Pathways (SSP) by sixth IPCC assessment report³⁶ SSP126 (the sustainable pathway), SSP245 (the medium pathway), and SSP585 (fossil-fueled development). The MIROC6 climate model from the Coupled Model Intercomparison Project Phase 6 (CMIP6) was selected for future scenarios because of its reliable simulations of Central Asia^37,38.

Environmental variables were selected through a sequential three-step procedure. First, all 35 candidate variables were assessed for ecological relevance to I. kuschakewiczii by overlaying species occurrence records with each variable layer and retaining only those with a plausible mechanistic link to the species’ habitat requirements, including soil texture, thermal stability, elevation, seasonal precipitation, and UV-B radiation³⁹. Second, pairwise Pearson correlation analysis (|r| ≤ 0.7) was performed on the ecologically screened subset using the R package ‘stats’⁴⁰ to identify and remove redundant variables. Third, variable contribution and permutation importance were evaluated across multiple preliminary models representing different SSPs and time periods to ensure the stability of the final predictor set across scenarios. Seven non-collinear environmentally relevant variables were retained for all subsequent modeling (Table 1, Supplementary File).

Table 1.

Environmental variables used to model the suitable habitat area of Iris kuschakewiczii using MaxEnt.

Variable	Name	Unit	Database
SAND	Weight and percentage of sand	%	HWSD
ELEV	Elevation	m	WorldClim
BIO4	Temperature seasonality	/	WorldClim
BIO14	Precipitation of driest month	mm	WorldClim
BIO19	Precipitation of coldest quarter	mm	WorldClim
UVB2	UV-B seasonality	J/m2/day	gIUV
UVB1	Annual mean UV-B	J/m2	gIUV

All environmental layers were converted from GeoTIFF to ASCII format to ensure full compatibility for MaxEnt, harmonized by normalizing to a common spatial extent and resolution using the R software package ‘terra’⁴¹.

Species distribution modeling of I. kuschakewiczii on MaxEnt

MaxEnt v.3.4.4 was used to model 51 geographic records of I. kuschakewiczii and 7 environmental variables under three SSPs and two time periods. The parameters included 75% of the distribution points as training data, the remaining 25% as testing data, and cross-validation was repeated 10 times to enhance model precision. The maximum number of iterations was set to 1500 to increase the gain. Other parameters were kept at the default. The model was configured to produce logistic output, yielding suitability values between 0 and 1 that correspond to the probability of species occurrence.

Model performance was evaluated based on the area under the receiver operating characteristic curve (AUC). The value of AUC ranges from 0 to 1; 0.8 to 0.9 indicates good performance; 0.9 to 1.0 represents excellent performance of the model with all threshold values^1,3. Thus, the AUC is directly proportional to the accuracy of the model’s predictions. Jackknife test, Percent contribution, Permutation importance by MaxEnt were employed to estimate the value of environmental layers.

For subsequent analysis of habitat suitability, the output logarithmic (0–1) distribution file from MaxEnt was classified into four levels of suitability on QGIS v.3.40.8: unsuitable (0-0.2), low suitability (0.2–0.5), moderate suitability (0.5–0.7), high suitability (0.7-1.0)^1,42. Further, the area of each suitability category was calculated using PyQGIS.

Centroid migration analysis

Centroids of potential habitats were calculated using the package ‘terra’⁴¹ in R. Centroids represent the area-weighted geographic centers of suitable habitat areas (suitability ≥ 0.5). Consequently, habitat suitability classes representing moderate and high suitability were converted to binary raster layers, then converted to polygons, and centroid coordinates were extracted. The centroid shift between time periods was quantified as the distance between the centroids. This analysis allows us to assess the trajectory of habitat suitability changes and interpret potential migration trends of the species under future climate conditions.

Conservation status of I. kuschakewiczii

To evaluate the conservation status of I. kuschakewiczii, whose distribution area is shrinking, Specially Protected Natural Territories (SPNTs) of Kazakhstan were downloaded from www.oopt.kz, and the protected areas of adjacent countries, projected in MaxEnt models, were derived from www.protectedplanet.net. Protected areas included National and Regional Reserves, State Nature Reserves (SNRs), State National Nature Parks (SNNPs), and Other Effective Area-based Conservation Measures (OECMs). Highly suitable distribution areas (suitability ≥ 0.7) of all periods were merged and overlaid with protected area territories. Suitable habitat areas were spatially overlaid with protected natural areas to assess the degree of overlap^43,44.

Results

Model validation and contribution of the environmental variables

A total of 35 variables’ correlation was analyzed, and variables with a correlation coefficient value lower than 0.7 were selected for further investigation. Accordingly, variables with coefficients above the threshold were removed to avoid multicollinearity (Fig. 2). The lowest correlation was observed between SAND and BIO14 (r = 0), followed by SAND and BIO19 (r = 0.02). The highest coefficient, 0.69, is observed between UVB3 and BIO4, followed by 0.65 between BIO19 and BIO14.

Fig. 2.

The heatmap of Pearson correlation among environmental variables used to construct MaxEnt. Red – positive correlation, blue – negative correlation.

The MaxEnt models were constructed to predict suitable habitats for I. kuschakewiczii using 51 occurrence records and 7 environmental variables across three climate change scenarios. Predictive performance of the constructed models was evaluated using the AUC metric. Models demonstrated an average AUC value of 0.997 for selected environmental variables (Fig. 3), with a standard deviation of 0.001. This indicates that the models have high predictive reliability.

Fig. 3.

Receiver operating characteristic (ROC) curve and AUC values of the MaxEnt model. The red curve indicates the mean AUC, the blue indicates the standard deviation, and the black line indicates random prediction.

Among the environmental variables, SAND (41.1%), BIO4 (18.8%), and ELEV (14.4%) showed significant contributions. The total contribution of those three variables comprises 74.3% (Table 2). The weights of the environmental factors used in MaxEnt models were determined using the jackknife method (Fig. 3). SAND, UVB3, and UVB2 contributed the most to the jackknife test of regularized training gain, respectively.

Table 2.

Percent contribution of environmental variables.

Environmental variable	Name	Percent contribution, %
SAND	Weight and percentage of sand	41.1
BIO4	Temperature seasonality	18.8
ELEV	Elevation	14.4
BIO19	Precipitation of driest month	13.1
BIO14	Precipitation of coldest quarter	5.6
UVB2	UV-B seasonality	4.2
UVB3	Mean UV-B of the highest month	2.7

The environmental variable with the highest individual gain is SAND, suggesting that it provides the most useful information in the models (Fig. 4). The results are concordant for the jackknife tests using AUC values and test gain. This infers that weight and percentage of sand are the most important factors affecting the potential distribution of I. kuschakewiczii.

Fig. 4.

Jackknife test of variable importance based on regularized training gain.

The response curves for the four most important environmental variables are depicted in Fig. 5. These illustrations could serve as a useful basis for identifying environmental conditions under which species achieve optimal distribution. The response curve results showed that the optimal value of SAND is above 70% (Fig. 5A). Among the curves, the response of BIO4 (Fig. 5B) is particularly notable for its wide standard deviation, likely linked to its correlation with another UV-B-associated variable – UVB3 (Fig. 2). The probability of I. kuschakewiczii occurrence is high on elevation range 700–1300 m above the sea level (Fig. 5C). BIO19 showed the highest probability at low values, suggesting that I. kuschakewiczii is adapted to habitats with relatively dry winters (Fig. 5D).

Fig. 5.

Response curves of the key environmental variables affecting the distribution of Iris kuschakewiczii: (A) – Weight and Percentage of Sand (SAND); (B) – Temperature Seasonality (BIO4); (C) – elevation (ELEV); (D) – Precipitation of driest month (BIO19).

Predicted current suitable distribution areas for Iris kuschakewiczii

The suitability area map was constructed based on MaxEnt results and colour-coded into three suitability classes proposed by Thapa et al.⁶⁰(Fig. 6). The white area is the unsuitable area (< 0.2), the yellow area indicates low suitability (0.2–0.5), the green area stands for a moderately suitable area (0.5–0.7), and the red area represents the highly suitable area (0.7-1.0) for the growth of I. kuschakewiczii. The current total suitable distribution area comprises about 266,5 × 10³ km² (Table 3), and predominantly occupies Southeastern Kazakhstan. Among the suitability classes, the low-suitability area has the highest share, covering 65.3% of the suitable distribution area, whereas the moderate and high-suitability areas account for 19.2% and 15.5%, respectively.

Fig. 6.

Current predicted suitable habitat areas for I. kuschakewiczii.

Table 3.

Predicted current and future suitable distribution area of I. kuschakewiczii.

Climate scenario	Predicted suitable distribution area, km2
	Low suitability	Moderate suitability	High suitability	Total suitable habitat
Current	173,981.88	51,158.82	41,355.12	266,495.82
SSP126
2050s	148,122.87	50,288.28	53,642.69	252,053.84
2090s	218,526.34	62,784.52	66,404.64	347,715.5
SSP245
2050s	202,648.87	62,898.00	76,012.77	341,559.64
2090s	189,368.13	62,532.13	79,013.81	330,914.07
SSP585
2050s	183,379.14	59,751.42	70,167.49	313,298.05
2090s	247,708.51	47,093.40	56,595.65	351,397.56

Predicted future suitable distribution

As shown in Fig. 7, the potentially suitable area for I. kuschakewiczii was modeled under three SSPs representing future emission levels: SSP126, SSP245, and SSP585. Generally, future distribution range has shown an increase in the northeastern boundary of the currently suitable areas. The suitability map for the 2090s under the SSP126 scenario showed an increase in highly suitable habitat along the northwestern boundary, including Karatau. Under the SSP245 scenario, both moderate and highly suitable predicted distribution areas achieved the highest results, with remarkable increases in the northwestern direction towards Karatau and Talas Alatau, while along the southern borders of the Fergana Range in Kyrgyzstan (Table 3). On the contrary, under the SSP585 scenario, a shrinkage was observed in the projected moderate- and high-suitability areas, with a decrease in highly suitable areas in the northwestern territory (Karatau and Talas Alatau) (Fig. 7). The total suitable area reached a peak of 351,397.56 km² in the SSP585 scenario. In general, the most suitable areas are in the Almaty and Zhambyl regions of Kazakhstan. The results show that the suitable distribution areas for I. kuschakewiczii remain stable, with only small fluctuations in the Karatau and Talas Alatau Mountain ridges, despite future climate change scenarios.

Fig. 7.

The distribution of predicted suitable habitat areas of I. kuschakewiczii under shared socioeconomic pathways (SSP): SSP126, SSP245, and SSP585 for the 2050s and 2090s.

Centroid migration analysis under climate change

The centroids of I. kuschakewiczii for three scenarios were identified to represent changes in the highest-concentration points of suitable habitat zones across different time periods, as well as the distances between them (Table 4). Subsequently, the GPS coordinates of the centroids were identified and visualized on a map (Fig. 8).

Table 4.

Changes in distribution centroids of I. kuschakewiczii under three distinct climate change scenarios.

Scenario	Period	Longitude (°E)	Latitude (°N)	Elevation, m	Migration distance, km
	Current	74.48	43.46	1118
SSP126	2050s	75.89	43.67	1325	116.41
SSP126	2090s	75.55	43.78	963	29.93 (2050s →2090s)
SSP245	2050s	75.29	43.61	912	68.21
SSP245	2090s	74.87	43.42	1202	40.45 (2050s →2090s)
SSP585	2050s	75.66	43.66	841	98.55
SSP585	2090s	77.26	44.32	834	147.17 (2050s →2090s)

Fig. 8.

The migration roads of suitable habitat area (˃0.5) centroids for I. kuschakewiczii under (A) SSP126, (B) SSP245, (C) SSP585 climate change scenarios..

As mentioned, centroids represent the centers of suitable habitats (< 0.5), which we can interpret as priority areas for subsequent analysis and conservation measures. The centroid position is at 74.48 N, 43.46E in the present time period (Fig. 8). Further, to understand spatiotemporal alterations in the center of mass, the distance between centroids of I. kuschakewiczii under different climatic scenarios over time periods was measured (Table 4).

The common northeastward shift in the current centroid distribution to the 2050s and higher longitudes is observed in all scenarios. Under the SSP126 scenario, the migration in the 2090s moves to a lower longitude by 29.93 km northwest of the 2050s centroid position (Table 4), whereas under SSP245, the centroid of the 2090s follows the southwestern trajectory. Only the centroid of the 2090s of SSP245 shifts to lower latitudes by 40.45 km. The farthest migration line is between 2050s and 2090s of SSP585 (147.17 km), moving in a northwestern direction. As for the elevation, it remains in the range across all scenarios, but under SSP126 and SSP585 shows a downward shift. In general, the distribution range of I. kuschakewiczii might shift towards higher longitudes in a northwestern direction under future climate change. Centroids are concentrated primarily along the boundary between Almaty and Zhambyl in the Chu-Ili Mountains, highlighting that conservation measures should be implemented initially at those sites.

Assessment of conservation status of I. kuschakewiczii

The total area of the merged highly suitable habitat (yellow) is 101,395.80 km². Protected areas of Kazakhstan, Kyrgyzstan, and Uzbekistan, along with the highly suitable areas, were overlaid (Fig. 9). The intersection area between the predicted highly suitable habitat and protected territories is 27,116.79 km² (red hash), which comprises only 26.7% of the high-suitability area. Of them, 27,071.02 km² lies in Kazakhstan, and the other 45.76 km² in Kyrgyzstan.

Fig. 9.

The intersected area between the specially protected natural territories of Kazakhstan and the highly suitable habitats of I. kuschakewiczii.

The analysis shows that the intersected area comprises 20 SPNTs, of which 16 are in Kazakhstan, 3 in Kyrgyzstan, and a common transboundary protection site among Kazakhstan, Kyrgyzstan, and Uzbekistan. The vast majority of the overlaid area lies within the following national protected areas in Kazakhstan: Zhusandaly SNR, Altyn-Emel SNNP, Korday-Zhaysan SNR, and Charyn SNNP. Kazakhstan, Kyrgyzstan, and Uzbekistan share the Western Tien Shan, a World Heritage site (https://whc.unesco.org/en/list/1490/). In which the distribution was predicted to have 507.65 km² of highly suitable habitat. Only the current centroid of the priority areas for conservation of I. kuschakewiczii lies within the Korday-Zhaysan SNR of local significance. This indicates that protection is needed for the future shift.

Discussion

Model performance and validity

The escalating global warming is profoundly influencing plant species distribution and biodiversity^45,46. Climate change heightens the urgency of applying effective conservation strategies, particularly habitat modeling, to protect rare and endangered plant species^47–49. Moreover, according to the Red Data Book of Kazakhstan¹⁴, I. kuschakewiczii is listed as an endangered species with a shrinking distribution. Until recently, however, limited data availability led to the misconception that the species was endemic to Kazakhstan. The expansion of digital biodiversity records has since revealed a broader distribution. The current predicted distribution area is consistent with other studies, which report that the species predominantly occurs in the Chu Ili Mountains^19,50,51. Also, distribution extends to the west and northeast of the Chu Ili Mountains and to the gravelly slopes of the Karatau, Talas Alatau, and Kirghiz Alatau mountain ridges^14,21,50. This study used MaxEnt to build current and future suitable habitat area models for I. kuschakewiczii across its range under three climate change projections. It also identified territories with high conservation priority (Fig. 7). The results obtained could be applied to biodiversity conservation action strategies.

This MaxEnt-based study of the distribution of I. kuschakewiczii reveals that the total suitable area generally remains stable despite future climate change and is concentrated mainly in the Almaty and Zhambyl regions (Fig. 7). The results also show high validity, with a firmly valid AUC (0.997), indicating robust predictive efficiency of predictive models^46,52. All response curves for environmental variables largely followed the same trend, with a narrow standard deviation (Fig. 5). The consistent trend in predictor-response curves across the current climate and different SSPs suggests that these relationships capture fundamental niche characteristics that remain stable under varying levels of stress.

Current and future distribution patterns under SSP scenarios

SSPs define the extent of future climate change, incorporating factors such as technological development, economic growth, and human population. SSPs were reported in 2023 and are combined with the equivalent Representative Concentration Pathways (RCPs) from the previous IPCC Fifth Assessment Report (2014)^53,54. SSP126 is a low-emission scenario that combines the SSP1 socioeconomic pathway with the RCP2.6 radiative forcing target, representing a sustainable future. SSP246: middle-of-the-road pathway with intermediate greenhouse gas (GHG) emissions, the combination of SSP2 and RCP4.6. SSP5 and RCP8.5, in collaboration, give fossil-fueled development, resulting in the highly emissive SSP585 (IPCC6). Nowadays, SSPs are extensively used in habitat suitability studies^3,54.

Predicted current distribution of I. kuschakewiczii refines its western border range (Fig. 6), extending it beyond the geographic points represented by thinned occurrence records. The occurrence dataset is concentrated predominantly in the Chu-Ili Mountains of Southeast Kazakhstan. However, the model predicts suitable conditions extending further westward into the foothill zones of the Western Tien Shan, including areas around the Aksu Zhabagly Nature Reserve in South Kazakhstan. Rather than representing a model artefact, this westward prediction is corroborated by botanical literature documenting the species in these regions^56–59, suggesting that the current occurrence dataset underrepresents the true western extent of the range due to limited field sampling in remote foothill areas. This highlights a broader challenge in SDM studies of rare Central Asian endemics, where occurrence data are inherently sparse and geographically biased toward more accessible localities. The model, therefore, complements and extends the documented range, identifying undersampled western areas as priorities for targeted field verification. Future surveys confirming presence in these localities would not only validate model predictions but also substantially improve the accuracy of subsequent distribution modeling and conservation assessments for this species.

Under future climate models, compared with the current, the extent of highly suitable areas extends southward, encompassing Northwest Kyrgyzstan. Based on the calculated suitability area, it can be assumed that among climate change scenarios, SSP245 provides the greatest moderate habitat suitability in the 2050s and the highest suitable habitat for the species by the 2090s. The possible reason for an inclination towards these time periods could give us the range of estimated warming degrees within which the I. kuschakewiczii distribution occurs. Under SSP245, by the 2050s, the global temperature increases by 2.0 °C, while preferable SSP245-2090s cumulative effect rises the global temperature by 2.7 °C³⁶. According to our findings, the distribution of I. kuschakewiczii reaches peak values at temperature elevations of 2.0–2.7 °C. However, according to the latest revision by Hausfather⁶⁰, warming by 2100 is estimated at 2.9–3.7 °C. This value exceeds the preferred temperature-increase range for the future distribution of I. kuschakewiczii. Therefore, actions to conserve plant communities with I. kuschakewiczii should be given the utmost priority.

Key environmental drivers and their ecological interpretation

The most influential environmental variable among the models is SAND, indicating that the distribution of I. kuschakewiczii is strongly influenced by the weight and percentage of sand in the soil (Table 2). The response curve of the variable SAND shows a strong positive relationship between sand content and habitat suitability (Fig. 5). Predicted suitability remains negligible at low sand percentages, increases sharply beyond 70–80% sand, and reaches optimal values at very high sand content. Soil can act as a determinant of plant species spatial distribution⁶¹. In our case, sandy soils might be preferable for habitat due to their ability to retain water⁶², and they are described as plant communities with I. kuschakewiczii. These results suggest that soil aridification does not represent a limiting factor for the growth of I. kuschakewiczii. There are currently no available datasets in soil databases for future timelines applicable to species modeling. Consequently, the SAND variable was held constant across all future scenarios, representing a simplification that may affect the accuracy of future habitat suitability projections.

The strong positive response of I. kuschakewiczii to sand content exceeding 70% is ecologically interpretable in the context of its life form as a tunicated bulbous geophyte. Sandy soils offer rapid drainage, a condition critical for bulb survival during summer dormancy⁶³. This interpretation is corroborated by the BIO19 response curve, which shows the highest habitat suitability at low winter precipitation values, consistent with adaptation to dry dormancy conditions. These two independent model outputs thus converge on a single ecological explanation: I. kuschakewiczii is physiologically constrained to substrates that remain well-drained during its dormant period.

The second predictor, with high contribution across most models, is BIO4, which shows the range of temperature seasonality. The response curve is strongly negative, implying that I. kuschakewiczii avoids seasonal temperature shifts and adapts to stable thermal conditions (Fig. 5). Elevation contributed by 14.4% (Table 2), and the response curve of ELEV shows consistent results with centroid migration analysis (Fig. 5; Table 4). The findings indicate that habitats located between 800 and 1300 m above sea level are most suitable for the growth of I. kuschakewiczii. Precipitation-related variables, BIO19 and BIO14, which specify the moisture preference of modeled species, contributed 13.1% and 5.6%, respectively (Table 2). Our findings indicate that precipitation is not a limiting factor for I. kuschakewiczii. The least influential predictors in the prediction models are UVB2 and UVB3. The species requires moderate seasonal variation in UV-B, with suitability peaking at intermediate levels (1.5 × 10⁵ J/m²/day) rather than extreme values. It also shows high habitat suitability only in areas with low UV-B exposure during the month with the highest UV-B exposure, indicating sensitivity to high ultraviolet radiation. To sum up, the principal drivers of the distribution of I. kuschakewiczii are SAND, BIO4, and ELEV.

Centroid migration and its conservation implications

The centroid shifts to lower elevations (834–912 m) under SSP245 and SSP585 represent areas where future climate conditions may become physiologically tolerable for I. kuschakewiczii, but several factors may prevent actual distribution. First, lower elevations in Southeastern Kazakhstan and Northern Kyrgyzstan are subject to intensive agricultural cultivation, permanent settlement, and unregulated livestock grazing, the same anthropogenic pressures already documented as the primary drivers of population decline in the species¹⁴. Climatic suitability in these zones is therefore largely theoretical in the absence of habitat integrity. Second, I. kuschakewiczii is a bulbous geophyte with no documented long-distance dispersal mechanism. Natural downslope colonization of new habitat patches would likely occur over timescales far exceeding the 2050s and 2090s projection windows modeled here, even if suitable climate conditions materialized. Third, no field surveys have documented I. kuschakewiczii at elevations below approximately 700 m, and our elevation response curve shows a sharp decline in suitability below this threshold under current conditions. The projected centroid shifts to lower elevations, therefore reflecting a change in the center of mass of suitable habitat, not evidence of immediate range expansion into lowland areas.

We therefore interpret the lower-elevation centroid positions as a signal that conservation planning should account for potential future habitat at intermediate elevations (700–1000 m) in less anthropogenically disturbed terrain, rather than as a prediction that the species will colonize lowland agricultural zones. Field surveys across the elevation gradient in the Chu-Ili Mountains are needed to confirm these projections.

Conservation gap analysis and recommendations

The study shows that the area of overlap between the protected natural territories of Central Asia (Kazakhstan, Kyrgyzstan) and the highly suitable habitats of I. kuschakewiczii is 27,116.79 km² (26.7%). The other 73.3% of highly suitable habitat requires protection. Protected areas have been proven effective for protecting endangered species^51,64. However, under climate change scenarios, their distributions could shift outside protected areas. The limiting factor in the natural population decline of I. kuschakewiczii is anthropogenic factors, such as unregulated livestock overgrazing and land cultivation¹⁴. During our field investigations, specimens of I. kuschakewiczii grazed by livestock were documented. The species is only protected in Aksu-Zhabagyly SNR¹⁴, though I. kuschakewiczii is absent in the latest list of vascular plants of the reserve (www.kazmab.kz/index.php/). Therefore, mitigation of distribution area shrinkage should include previously defined centroids, as these represent core priority areas for conservation. Enhancing international collaboration on biodiversity conservation and ecological security across Central Asian countries and regions, and supporting joint protection initiatives, is critically important today.

The priority area for conservation identified by centroid migration analysis is only partially protected by the Korday-Zhaysan SNR, which is of local significance. As mentioned above, most (73.3%) of highly suitable areas are outside protected areas (Fig. 8). This means that most of the distribution and high-priority areas of I. kuschakewiczii are not protected from environmental and anthropogenic influence. Consequently, conservation in a single SPNT is insufficient; establishing small protected areas and wildlife corridors within high-priority conservation areas is recommended. Most importantly, the population dynamics of I. kuschakewiczii should be monitored at monitoring sites on a long-term basis, using cultivation methods such as ex situ (botanical gardens, seed and DNA banks) and in situ (national parks, wildlife sanctuaries, and biosphere reserves), as well as integral quasi-in situ. The latter quasi in situ was proposed by Volis and Blecher as a conservation method within an ex situ-in situ cultivation strategy in highly suitable legally protected natural sites and showed promising results with Iris atrofusca Baker^65,66. We suppose that the application of OECMs in vulnerable populations is most applicable to the case of rare and endangered I. kuschakewiczii.

While expanding formal protected areas remains a long-term goal, the extensive pastoralist communities across Southeastern Kazakhstan and Kyrgyzstan make this approach socially and politically complex. Within this regard, OECMs offer a pragmatic complementary pathway recognized under the Kunming-Montreal Global Biodiversity Framework (Target 3)⁶⁷. As defined by the CBD (2018), OECMs are areas outside formal protected areas, which achieve sustained positive outcomes for in situ biodiversity conservation and associated ecosystem and cultural values through their governance and management⁶⁷. Also, previously emerged terms, such as Important Plant Areas (IPAs) that fall within the definition of OECMs, are included in it, contributing to the development of conservation policies’ management^67,68. In the past decade, 31 IPAs in the Northern Tien Shan botanical-geographical subprovince^69,70 of Kazakhstan have been identified^51,64. IPAs are key biodiversity conservation sites established by evaluating species distribution and botanical richness, with an emphasis on threatened plants and habitats (https://www.plantlife.org.uk/international). Recently, Kurty Valley IPA was described in 2024⁷¹. The new IPA includes habitats of Tulipa regelii Krasn. and I. kuschakewiczii. The identification of IPAs is particularly relevant under the Kunming-Montreal Global Biodiversity Framework⁶⁷, which encourages the development of national strategies. IPAs have been proposed as the primary tool for identifying priority areas for vegetation protection^72–74. Therefore, establishing the legal status and common formation criteria for IPAs is required to protect rare and endangered plants.

To address anthropogenic influences on I. kuschakewiczii, livestock breeders and farmers should emphasize sustainable land use and minimize interactions with wildlife. In pastoralist communities in the range of I. kuschakewiczii, OECMs can align conservation with traditional land stewardship, reducing land-use conflicts while maintaining habitat connectivity for I. kuschakewiczii. Collaborative agreements with livestock breeders, already mentioned in our conservation recommendations, could serve as a foundation for OECM designation in high-priority centroid areas. However, a critical distinction must be drawn between climatic suitability and realized distribution.

These findings are crucial for identifying the conditions most suitable for conserving plant communities with I. kuschakewiczii. In the present study, a suitable distribution habitat was modeled for the current timeline and under future climate change using MaxEnt, incorporating bioclimatic, UV-B, and edaphic factors. Although the research demonstrated strong reliability of the results (AUC = 0.997) and included various types of variables, biotic factors such as I. kuschakewiczii species interactions with other plant species and animals, as well as human activities, should be studied. Plant species growth is affected by many additional factors not included in the study, indicating that there remains considerable potential to investigate future population projection of I. kuschakewiczii by incorporating other variables and alternative prediction models.

Limitations

The MaxEnt forecast included several biotic factors affecting the distribution of I. kuschakewiczii, including edaphic and radiation factors. Appropriate conservation pathways were proposed to conserve species in the high-priority areas identified by the center-of-mass shift analysis. Centroids could not distinguish northern and southern populations, as they represent the geographical center of the suitable habitat. Therefore, future studies should delineate northern and southern population groups as separate units and integrate genetic and demographic data to determine whether these groups constitute distinct conservation units with differentiated vulnerability to climate change. Such population-level analyses would substantially improve the resolution of conservation recommendations for this species.

The life history of I. kuschakewiczii, including dispersal mechanisms, pollinator specificity, and soil biotic interactions, remains largely unstudied at the species level. Our model identifies habitat associations but cannot explain the biological mechanisms underlying them. We therefore call for targeted autecological research on this species, particularly in the high-priority centroid areas identified in the Chu-Ili Mountains, where field access and monitoring infrastructure could realistically be established.

Conclusion

Using MaxEnt modeling, we projected the current and future distribution areas of I. kuschakewiczii under three climate change scenarios. The results showed perfect validity, with a mean AUC of 0.997. Among the variables used, key environmental factors influencing the distribution of I. kuschakewiczii include weight and sand percentage, temperature seasonality, and elevation. In the future, the distribution area of I. kuschakewiczii would remain consistent, with only minor losses, under climate change conditions. The estimated temperature increase range under which I. kuschakewiczii reaches high distribution is 2.0–2.7 °C. Centroid shift analysis identified a high-priority conservation area for I. kuschakewiczii in the western Almaty and Zhambyl regions of Kazakhstan. For I. kuschakewiczii, an endangered species, distribution areas and conservation measures were suggested, such as the establishment of protected areas, monitoring sites, and the strengthening of the legal status of Important Plant Areas. Given the ongoing trend of global warming, the findings of this study could serve as a valuable theoretical basis for devising efficient policies to conserve plant communities with I. kuschakewiczii.

Author contributions

Conceptualization – A.A., Sh.A.; Data curation – S.A., Y.T.; Formal analysis – A.A.; Funding acquisition – Sh.A.; Investigation – A.A.; Methodology – A.A., Sh.A.; Project administration – Sh.A., Y.T.; Resources – S.A., Y.T.; Software – A.A.; Supervision – S.A., Y.T.; Validation – Sh.A., S.A.; Writing – original draft – A.A., Sh.A.; Writing – review and editing – S.A., Y.T. All authors reviewed and approved the final manuscript.

Funding

This research has been funded by the Science Committee of the Ministry of Science and Higher Education of the Republic of Kazakhstan (Grant No AP23483825).

Data availability

The datasets analysed during the current study are available in the Zenodo repository, [https://doi.org/10.5281/zenodo.18847551](https:/doi.org/10.5281/zenodo.18847551).

Declarations

Competing interests

The authors declare no competing interests.