Introduction

Climate change has caused forest communities to expand, contract or shift over the long term, they northwards as the climate warms and toward the south as it cools1,2. Global scientific studies and observations indicate that, due to climate change, plants will migrate to high altitudes and this will continue to occur over the next 100 years3,4,5. Climate change increases the risk of extinction of plant species by negatively affecting biodiversity and the geographical distribution of species6,7. The biological responses of species to climate change can result in three different situations6,8. First, the target species may migrate to another region by following ecological niches4,9; second, the species could adapt to the new climatic conditions of their current location and third, local extinction could occur10,11,12. To elucidate the specific impacts of climate change on species and reduce the negative effects of climate change on ecosystems and biodiversity, conservation strategies need to be integrated with species distribution modelling to identify suitable habitats for various species13,14,15,16. Species distribution models (SDMs) provide insights into future biogeographical attributes17,18. MaxEnt, based on the maximum entropy algorithm, is a tool used to model the suitable distribution areas of species and is based on a species distribution model19,20. The working principle of the MaxEnt algorithm is based on estimating potential distribution areas using data (sample points representing the natural distribution areas of the species) and bioclimatic variables (environmental variables)21,22.

Members of the Arbutus genus, which belongs to the Ericaeae family, are woody plants that do not shed their leaves in winter and are bushes and shrubs. A. unedo L. and A. andrachne L. grow naturally in Türkiye23. These two species are undoubtedly of great importance for forestry. This importance stems from the fact that the wood of the tree species belonging to this genus is used for decorative purposes and as fuel wood in many wood-based industries, with or without alterations to its structure; it also has a wide range of uses in packaging, chair-making and furniture-making. In addition to these areas of use, the fruits of these trees are widely consumed by humans and animals because they are rich in sugar and vitamin C24,25.

In this study, the current geographical distribution areas of Arbutus unedo in Türkiye and its potential geographical distribution areas in ~ 2050 and ~ 2090 according to the MIROC-ES2L model SSP2 and SSP5 scenarios were determined by using MaxEnt 3.4.1 software and other geographic information system (GIS) programs and tools. We aimed to predict and determine how the spatial distribution of the identified species will change. The MIROC-ES2L, Earth System Model (ESM) is the Model for Interdisciplinary Research on Climate (MIROC), which is known as Earth System version 2 and is used for long-term simulations26. The SSPs created for the models are based on five narratives that describe broad socioeconomic trends that may shape future societies, predicting how socioeconomic conditions will affect the implementation of climate policy. However, SSP2, which is used in the study, represents a moderate level of mitigation and adaptation, and SSP5 represents a high level of use of fuel and world resources.

Materials and methods

Study region and occurrence data

The occurrence data for Strawberry tree (Arbutus unedo) were obtained from the Flora of Türkiye, GBIF (global biodiversity advisory facility), TUBIVES (Turkish plant data service) and from literature reviews27,28,29,30,31,32,33,34. The data obtained from the Flora of Turkey30, TUBIVES33 and literature reviews were based on local field observations. The data obtained from GBIF25 were filtered with base maps, and points that were located in areas where the species would not show a natural distribution (agricultural land, settlements, dunes, etc.) were eliminated and not included in the study. A total of 43 points were selected as the sample area were processed in QGIS 3.24, and were converted to csv format to be input into the MaxEnt algorithm35. The sample points and the study area are shown in (Fig. 1), and the attributes of the sample points are shown in (Table 1).

Figure 1
figure 1

Input data for the maximum entropy area of Arbutus unedo at the study boundary—created using QGIS v3.2435.

Full size image
Table 1 Information on the attributes of Arbutus unedo L.
Full size table

The algorithm takes the projection of the sample points on climatic variables, yields the attribute information of the species in those areas and determines points suitable for the species based on the determined attributes by randomly assigning points in the background. Areas that have similar attributes in the obtained and future projection data are considered suitable distribution areas for the species. In this study, 10000 background points were used according to the area size.

Climate data

In this study, the model was first generated using all variables (Table 2). Then the impact levels of the bioclimatic variables were determined by using the jackknife (pull-out) test, and according to the results, the model was reconstructed using 9 high-impact variables36,37,38. The variables used in the newly created model were BIO 2, BIO 4, BIO 6, BIO 8, BIO 12, BIO 13, BIO 14, BIO 15 and BIO 18. Two different models were constructed for two shared socioeconomic pathways (SSP2-4.5 and SSP5-8.5) under greenhouse warming scenarios. All variable rasters were at a 30 arc-second resolution.

Table 2 Environmental variables; those used in the study are indicated in bold 22.
Full size table

Statistical model

MaxEnt modelling was carried out using the ‘auto features’ feature. The success of the model was measured by obtaining the AUC (area under the ROC curve) value extracted from ROC (receiver operating characteristic) analysis39,40. The AUC value is interpreted as the estimated probability of the existence of a randomly determined grid cell, and if the AUC value is > 0.5, the model performs better than a random guess41.

As the AUC approaches 1, the sensitivity of the model increases and more accurate predictions are made42,43,44,45. Five classifications were made for maps of potential distribution areas (0 is not suitable, 0–0.25 is slightly suitable, 0.25–0.50 is slightly suitable, 0.50–0.75 is suitable and 0.75–1.0 is very suitable). After the classifications were made, the maps obtained as raster were converted into vector format, and the areal coverage is presented in tables.

Results

In the model generated in this study, the AUC training data value was calculated to be 0.966 and the test data value was calculated to be 0.965 (Fig. 2). According to the jackknife (pull-out) test, the environmental variables with the five greatest gains are “Temperature Seasonality (BIO 4)”, “Mean diurnal range (BIO 2)”, “Annual precipitation (BIO 12)”, “Maximum temperature of the coldest month” temperature (BIO 6)” and “Precipitation of Wettest Month (BIO 13)” (Fig. 3).

Figure 2
figure 2

ROC curve and AUC of the model.

Full size image
Figure 3
figure 3

Predictive accuracy of the model variables the Jackknife test.

Full size image

The predicted potential distributions of the species displayed a significant degree of similarity with the sample points representing its current distribution. The area of highly suitable habitat was calculated to be 8205,5 km2 and the highly suitable habitat area was calculated to be 29560.1 km2 (Table 3). The map for the current potential distribution area of Arbutus unedo L. is shown in (Fig. 4), and the distribution areas for future periods generated according to the MIROC_ES2L climate change model are in Figs. 5, 6, 7, 8.

Table 3 Spatial distribution of the potential geographical distribution of Arbutus unedo L. in the present day and in the future according to the SSP2 4.5 and SSP5 8.5 climate scenarios (km2).
Full size table
Figure 4
figure 4

Potential distribution of Arbutus unedo—created using QGIS v3.2435.

Full size image
Figure 5
figure 5

Estimated distribution areas of Arbutus unedo according to the SSP2 4.5 scenario for the 2041–2060 period—created using QGIS v3.2435.

Full size image
Figure 6
figure 6

Estimated distribution areas of Arbutus unedo according to the SSP2 4.5 scenario for the 2081–2100 period—created using QGIS v3.2435.

Full size image
Figure 7
figure 7

Estimated distribution areas of Arbutus unedo according to the SSP5 8.5 scenario for the 2041–2060 period—created using QGIS v3.2435.

Full size image
Figure 8
figure 8

Estimated distribution areas of Arbutus unedo L. according to the SSP5 8.5 scenario for the 2081–2100 period—created using QGIS v3.2435.

Full size image

The existing data and the model created for the current study were consistent. The distribution area of Arbutus unedo is on the coast of the country, and its current distribution is far northwards on the northern coast of Marmara sea.

According to SSP2 4.5, in the models created for the distribution areas for the years 2041–2060 and 2081–2100, A. unedo will start to withdraw from the Mediterranean and Aegean regions, where it already has a limited distribution, and its distribution will shift to the northern parts of the Marmara region, where it is seen densely distributed. In the 2081–2100 scenario, the presence of very suitable distribution areas for the species in the northern areas will be severely restricted.

According to the distribution area model of Arbutus unedo in the 2041–2060 (~ 2050) and 2081–2100 (~ 2090) according to SSP5 8.5, the distribution of the species in year ~ 2050 will be almost the same as that in the year ~ 2090 under SSP2 4.5, except for local areas in Southern Marmara and around the Bafra district of Samsun Province near the Black Sea. Considering the conditions of the ~ 2090 period, it is unfortunately estimated that no suitable habitats for Arbutus unedo will exist. To reveal these differences, the results of the change analysis (loss and gain), in which the current distribution areas are compared with the predicted distribution areas from the model are shown in (Table 4), and corresponding maps are shown in Figs. 9, 10, 11, 12.

Table 4 Spatial changes between the current and predicted distribution areas of A. unedo under the SSP2 4.5 – SSP5 8.5 scenarios for 2050 and 2090 in Türkiye (km2).
Full size table
Figure 9
figure 9

Spatial changes between the current and predicted distribution areas of A. unedo under the SSP2 4.5 scenario for 2050—created using QGIS v3.2435.

Full size image
Figure 10
figure 10

Spatial changes between the current and predicted distribution areas of A. unedo under the SSP2 8.5 scenario for 2050—created using QGIS v3.2435.

Full size image
Figure 11
figure 11

Spatial changes between the current and predicted distribution areas of A. unedo under the SSP2 4.5 scenario for 2090—created using QGIS v3.2435.

Full size image
Figure 12
figure 12

Spatial changes between the current and predicted distribution areas of A. unedo under the SSP2 8.5 scenario for 2090—created using QGIS v3.2435.

Full size image

Discussion

Sarikaya and Orucu46, who studied Arbutus andrachne, which belongs to the same genus as A. unedo, modelled the present and future, ~ 2050 and ~ 2070 distribution areas of the species in their study. Their model indicated that the distribution of the species in the southern regions of Türkiye will decrease, but its distribution could expanded in the western and northern regions. The authors also noted that no potential habitat losses that would affect the distribution of A. andrachne would occur. Uzun and Sarikaya47 modelled the current and future distribution areas of Pistacia terebinthus using the random forest algorithm, and the results of their study indicated that very suitable distribution areas of the species with ethnobotanical uses will be completely eliminated, and suitable areas will decrease to 15%. Arslan et al.4 modelled the current and future distribution areas of Rosa canina, which has a wide distribution in our country, with MaxEnt; the model output suggested that the species distribution areas will shrink in the future.

According to the current and future distribution area models generated for Arbutus unedo, the potential distribution areas of the species in the future will gradually decrease. In the SSP5 8.5 model, which represents the highest level of world’s resources use this gradual decrease will reach its maximum level and there will be no suitable distribution area left for the species. This scenerio reveals that the species is endangered. Humans have a major impact on changes in the natural environment, especially with respect to vegetation, and are largely responsible for species distributions and habitat losses4,48.

Many previous studies have shown that climate change is one of the most important factors causing habitat loss and declines in biodiversity. The diversity of habitats and species that arise due to differences in topography makes the prediction of climate change difficult. On the other hand, research shows that some species will migrate north and that some endemic species may become extinct12,13,49,50,51,52,53,54.

Strawberry trees (Arbutus unedo) were included in the IUCN Red List in 2017 and are in the “Low Risk” but “Endangered” category55. Since the ethnobotanical and food use of this species is widespread in Türkiye, it is endangered, and as long as this level of use of the Earth’s resources continues, extinction will be inevitable for the Strawberry tree and many other species. Climate change is important from a wide variety of perspectives, from the distribution of plant species to their protection and use. In the present study, the potential distribution areas where A. unedo grows were identified. The MaxEnt model predicted significant fluctuations in A. unedo distribution areas in Türkiye over the next 50 years. Although the distribution areas of A. unedo are not significantly affected in the SSP2-4.5 projection from climate change, there is a significant decrease in the distribution area of this species in the SSP5-8.5 projection.

Sea level changes, which have profound effects on geomorphology, have undoubtedly also had great effects on human life56. Surrounded by sea on three sides, it is undoubtedly inevitable that Türkiye will be affected by global sea level changes. Coasts of country are home to approximately 30 million people and more than half of the country’s gross national products–around 60%. The areas that will be most affected by sea level rise on the Turkish coast are; It has been determined that there will be provinces with low areas such as coastal deltas57. In our study, it is seen that A. unedo is currently spreading in these risky provinces and will cease its spread in the future periods. According to the same study, it was concluded that Türkiye is among the countries with medium level risk58.

The rate of sea level rise is expected to increase, in part due to anthropogenic contributions to global warming. Additionally, other projected environmental changes, including changes in precipitation patterns, changes in the frequency of severe storms, and declining groundwater supplies of freshwater sources, may interact with sea level rise to affect the rate of change in terrestrial vegetation along coastlines59,60,61.

Ultimately, this study conducted at the geographical boundaries of Türkiye showed that the distribution of Strawberry trees in the coming years will gradually decrease and the species will become more threatened. Based on the maps, its distribution, which is dense in the coastal strip, will not be maintained in the Aegean strip in the future, and will decrease by shifting to the northern latitudes in the western Black Sea and Marmara regions. Based on this observation, it is possible that this species, like many other species, will shift its distribution to northern latitudes, not only in Türkiye but also in other Mediterranean countries. To avoid economic and ecological losses related to Strawberry trees, it is important to conduct studies on a global scale rather than on local scales to ensure the future of Strawberry trees.