Global distribution of Sapindus habitats under current and future climate change scenarios

Sapindus (Sapindus L.) is a widely distributed economically important tree genus that provides biodiesel, biomedical and biochemical products. However, with climate change, deforestation, and economic development, Sapindus germplasm resources have been lost. Therefore, utilising historical environmental data and future climate projections from the BCC-CSM2-MR global climate database, we simulated the present and future global distributions of suitable habitats for Sapindus using a Maximum Entropy (MaxEnt) model. The estimated ecological thresholds for critical environmental factors were: a minimum temperature of 0– 20°C in the coldest month, soil moisture levels of 40–140 mm, a mean temperature of 2–25°C in the driest quarter, a mean temperature of 19–28°C in the wettest quarter, and a soil pH of 5.6–7.6. The total suitable habitat area was 6059.97 × 104 km 2 , which was unevenly distributed across six continents. As greenhouse gas emissions increased over time, the area of suitable habitats contracted in lower latitudes and expanded in higher latitudes. Consequently, surveys and conservation should be prioritised in southern hemisphere areas which are in danger of becoming unsuitable. In contrast, other areas in northern and central America, China, and India can be used for conservation and large-scale cultivation in the future. Our results indicated that the size and distribution of suitable habitats for Sapindus varied under different climatic scenarios, suggesting that climate change had an uncertain and region-specic effect on the distribution of these habitats. Under future climate scenarios, we estimate that suitable habitats for Sapindus will expand and contract signicantly across continents. These habitats will signicantly contract in lower latitudes and expand in higher latitudes. Our model showed that as greenhouse gas emissions increased and time passed, the area of suitable habitats contracted sharply, peaking in the second half of the 21st century. In terms of the northern and southern hemispheres, the Sapindus suitable habitats for Sapindus in the northern hemisphere predominantly expanded, while in the southern hemisphere they predominantly contracted. Jayasinghe and Prevéy also found a declining trend in the area of suitable habitats for Camellia sinensis (L.) O. Kuntze [63] and huckleberry [54] in the face of global warming, while He [64] found an expanding trend in the highly suitable habitat area for Xanthoceras sorbifolia Bunge under global warming in China. The Sapindus genus consists of drought-tolerant and hardy species, but they are more sensitive to extreme temperatures. With future global warming, parts of the high latitudes of the northern and southern hemispheres may be transformed into suitable habitats for Sapindus due to an increase in extreme low temperatures, while parts of the lower latitudes may experience persistent extreme heat or drought as the climate continues to warm. This poses a major problem for Sapindus survival. In He’s study [64] , the distribution of X. sorbifolia was observed to be localised in areas of the Loess Plateau in China, which is at a higher latitude. Thus, a trend toward the expansion of suitable cultivation areas in the face of global warming is already evident. Our results indicated that Sapindus will also experience some localised suitable habitat expansions in the northern hemisphere at higher latitudes, but,


Introduction
Sapindus (Sapindus L.) is a genus containing 13 evergreen and deciduous tree species in the Sapindaceae family. It is globally distributed across warm-temperate and tropical regions in Southeast Asia and North and South America [1] . S. saponaria L. is the type species of the Sapindus genus, and it is widely distributed in North and South America, with some localised distributions in Africa and Australia. S. mukorossi Gaertn. is the second most widespread species and is found in East and Southeast Asia. Sapindus seed oil is high yielding (26.15-44.69 %) with a high medium-chain monounsaturated fatty acid content [2] . Crude extracts from the fruit pericarps of Sapindus are also rich in triterpenoid saponins (4.14-27.04 %) and sesquiterpenoids [3] , such as Saponin A, mukurozisaponin G, and sapinmusaponin K [4] , which exhibit excellent surface activity as well as antibacterial [5] , elution [6][7][8] , antibacterial [9,10] , insecticidal [11] , pharmacological [12] , and physiological [13] effects. Saponin serves as an e cient natural surfactant in commercial soaps, shampoos, and cosmetic cleansers [14] . Therefore, Sapindus species are regarded as economically important sources of biodiesel, as well as biomedical and multi-functional products [1,15] . Sapindus germplasm resources are generally scattered in the form of single plants or extremely small populations. With global deforestation and rapid economic development, Sapindus germplasms have been persistently damaged or lost [1] . Modern cultivation of Sapindus species has only begun recentlyand lacks support from relevant research. Therefore, it is still in a state of low yield with elite varieties lacking and severe germplasm destruction ongoing [1,4,16] . Therefore, the protection and management of the core Sapindus distribution areas should be strengthened, and natural populations at risk of destruction should be protected through in situ or ex situ conservation efforts.
Niches are habitats with the minimum thresholds necessary for survival [17] . The forest niche is strongly affected by the environment, and it changes or moves with environmental change. The Intergovernmental Panel on Climate Change (IPCC) estimates that a 0.2 °C temperature increase will occur in each future decade that is subject to greenhouse gas emissions [18] . Temperatures will rise by a maximum of 2.6-4.8 °C or a minimum of 0.3-1.7 °C in the 21st century [18] . In the face of upcoming rapid climate change, forest trees will unlikely be able to adjust their range with su cient speed to colonise suitable areas. Species extinction rates would subsequently increase, and warmer temperatures may impact plant growth and yield [19][20][21] . Hence, there is an urgent need to understand the extent of climate change in the coming decades, and the use of alternative methods to assess its impact on forest tree habitats will be helpful for designing conservation and cultivation plans in the future [22] Species distribution modelling is an emerging research eld based on niche theory. Its principle is to infer the ecological needs of species through mathematical models based on occurrence data and environmental variables, and to create a statistical or mechanism model of its potential distribution [23][24][25] . At present, the commonly used niche models are GARP [26] , MaxEnt [27] , Bioclim [28] , Random Forest [29] , and the Boosted Regression Tree [30] . Many model intercomparison studies have reported that the MaxEnt model, which is based on the principle of maximum entropy [24,27,31,32] , typically outperforms other species distribution models (SDMs) in terms of high tolerance and high predictive accuracy [33][34][35] . Over the past 10 years, worldwide research teams have achieved excellent results in the study of rare animal and plant diversity protection [36][37][38][39] , invasive species risk prediction [40,41] , marine ecosystem protection [42,43] , disaster distribution prediction [44] , and disease propagation [45,46] using the MaxEnt model.
Zhang et al. [47] predicted the distribution of suitable habitats for Cinnamomum camphora (L.) Presl in China under future climate change scenarios. Peng et al. [48] simulated the distribution of suitable areas for the new oil crop Paeonia ostii. Li et al. [49] predicted suitable area distributions for three Coptis herbal plant species under current and future climate change scenarios. It is noteworthy that most previous studies have focused on the prediction of suitable habitats in local regions with models that basically follow the default parameters, resulting in suboptimal model accuracy. At present, there are no reports or studies, to our knowledge, that focus on current and future suitable habitat projections on a global scale for Sapindus.
Based on 5674 global occurrence data records for Sapindus, historical environmental factors, and future climate models combined with ArcGIS 10.5 and MaxEnt modelling, we predicted the current global distribution of suitable habitats for Sapindus and its response to future climate change scenarios. This study had three objectives: (1) to evaluate the main environmental factors affecting the distribution of Sapindus, (2) to explore the distribution of suitable habitats for Sapindus under current environmental conditions, and (3) to predict the redistribution pattern of potential Sapindus habitats in response to future climate change scenarios in the 21st century, and (4) to identify hotspots of habitat degradation and expansion to facilitate climate change-adaptive biological conservation recommendations.

Modelling evaluation and selection
We implemented model performance screening using the SDMtoolbox data analysis package in ArcGIS. After simulating regularisation multipliers of 0.5, 1, and 2, as well as ve feature type (linear, product, hinge, quadratic, and threshold) model parameters, model performance was assessed based on the AUC values. According to the evaluation results, we found that the MaxEnt model with a regularisation multiplier of 1 and a feature type of LQ parameters (linear and quadratic) was the most successful. Therefore, we applied this parameter model to conduct further analysis. The MaxEnt model for current Sapindus suitable habitats provided satisfactory results, with an AUC value of 0.835, which was higher than that of a random model (0.5). The MaxEnt model for future Sapindus suitable habitats also achieved satisfactory results, with AUC values ranging from 0.838 to 0.848.

Current suitable habitats for Sapindus
The MaxEnt model's internal jackknife test of factor importance showed that the minimum temperature of the coldest month (Bio6, 26.0 % of variation), soil moisture (Sm, 14.1 % of variation), mean temperature of the driest quarter (Bio9, 13.1 % of variation), mean temperature of the wettest quarter (bio8, 11.6 % of variation), and soil pH (Sph, 10.5 % of variation) were the major contributors to the Sapindus distribution model, with a cumulative contribution of 75.3 % (Table 1). According to the MaxEnt results and environmental factor response curves, the ecological thresholds for the critical environmental factors were the minimum temperature of the coldest month (0-20 °C), soil moisture (40-140 mm), mean temperature of the driest quarter (2-25 °C), mean temperature of the wettest quarter (19-28 °C), and soil pH (5.6-7.6).   (ssp245, 2021-2040) to 1267.23 × 10 4 km 2 (ssp370, 2081-2100). Overall, there was a signi cant contraction in the size of suitable habitats with increasing greenhouse gas emissions. It is noteworthy that the contraction was most pronounced in the second half of the 21st century in the southern hemisphere, including South America, Central Africa, and Oceania.
There was signi cant expansion and contraction of suitable habitats for Sapindus in all the future climate scenarios and these expansions and contractions differed signi cantly between continents (Table S1) (Table S1). The contracted areas were mainly located in Brazil, Peru, Bolivia, Paraguay, and Argentina in South America; the Congo, Mozambique, and Madagascar in Africa; and eastern Australia in Oceania (Fig. 3).

Discussion
There are 13 species of Sapindus worldwide, all of which are scattered in ecosystems as either individual plants or small populations [1,16] . With global deforestation, rapid expansion of industrial, agricultural, and urban land use, increasing human interference, and over-exploitation of groundwater in the last century, Sapindus germplasm resources have been persistently damaged or lost [3] . Therefore, it is imperative to sustainably exploit and conserve Sapindus germplasm diversity by predicting the distribution of current and future suitable habitats through ecological niche modelling.
This study analysed projections of current and future distributions of suitable habitats for Sapindus using the MaxEnt model. Habitat maps were created for Sapindus based on occurrence data sets with mean AUCs of 0.835 and 0.843 for the present and the future, respectively. Therefore, we consider that our model performance is robust and adequate for construing the overall suitable habitat distribution of Sapindus. To our knowledge, this is the rst study to analyse the suitable habitat distribution of Sapindus for the present and future using the MaxEnt model.

Suitable habitat distribution patterns of Sapindus
After a long period of natural selection, anthropogenic disturbance, and geographic isolation, genetic variation in the Sapindus genus has increased greatly, resulting in patterns of geographic variation in both phenotype and ecological adaptation [3,50,51] . According to the response curves of ecological factors, the relationships between the probability of species occurrence and the main ecological factors can be determined, and it is generally considered that if the probability of species occurrence is greater than 60 %, the corresponding ecological factor thresholds are suitable for the survival of this species [52][53][54] . In our study, the MaxEnt results and environmental factor response curves indicated that the critical environmental factors affecting suitable habitats for Sapindus were temperature, soil moisture, and soil pH. Similarly, Adeyemi found that the minimum temperature of the coldest month was the most critical environmental factor in determining the suitable habitat distribution of Sapindaceae Juss. species in West Africa [55] . Sun et al. [56,57] also found that the annual minimum temperature played an important role in Sapindus saponin variation, but high precipitation levels were found to inhibit saponin synthesis [56,57] [56,57] . Liu found that the precipitation of the warmest quarter and isothermality were the critical environmental factors that determine the distribution of Sapindus habitats in China.
Sapindus are highly thermophilous [3,58,59] and xerophilous tree species [1,16] that prefer areas with abundant heat and sunlight resources [3] . They do not tolerate excessive cold and humidity. Moderate rainfall and soil conditions are more conducive to their survival. On a global scale, we found that the critical environmental factor for Sapindus survival was the minimum temperature of the coldest month. Breaking dormancy requires low-temperatures. However, extremely low temperatures are also detrimental to the survival of Sapindus [60] . This indicates that the suitable habitats for Sapindus are likely to be mainly distributed in tropical and subtropical regions between the 50° north parallel and the 40° south parallel latitudes. This is consistent with the ndings of Liu, who found that the northernmost margin of Sapindus distribution in China did not go beyond the Qinling Mountains or the Huai River (for the most part). Moderate rainfall can maintain soil moisture within an appropriate range. Sapindus are deep-rooted tree species [61] . Therefore, soil moisture ranges of 40-140 mm and weakly acidic or neutral soil types are more conducive to their survival.
Our results indicated that the total suitable habitat area for Sapindus globally was 6059.97 × 10 4 km 2 . Of this, the proportion of low suitability areas accounted for 46.91 %, suitable areas accounted for 34.54 %, and high suitability areas accounted for 18.55 %. The total suitable habitats for Sapindus were found to be unevenly distributed across six continents, with high suitability areas concentrated in the Mississippi Plain, the Florida Peninsula, and Mexico in North America; La Plata Plain in South America; and the North China Plain, Middle and Lower Yangtze Plain, Sichuan Basin, Yunnan-Guizhou Plateau, and Southern Himalayas in Asia. These results are consistent with a previous study, which found that Sapindus are mainly distributed in America and Asia [1,62] . High suitability areas are more conducive to cultivation and the conservation of Sapindus diversity than low suitability and suitable habitats. Therefore, it is recommended that ex situ conservation, cultivation, and breeding of Sapindus species be implemented in high-suitability areas.
Response of suitable habitats for Sapindus to future climate change Climate plays a signi cant role in de ning species distributions, and changes in the distribution of species are also the most clear and direct response to climate. Global warming may substantially change the structure and function of terrestrial ecosystems, resulting in signi cant changes in the extent and distribution of biological habitats [17,21] .
Our results indicated that the size and distribution of suitable habitats for Sapindus varied under different climatic scenarios, suggesting that climate change had an uncertain and region-speci c effect on the distribution of these habitats. Under future climate scenarios, we estimate that suitable habitats for Sapindus will expand and contract signi cantly across continents. These habitats will signi cantly contract in lower latitudes and expand in higher latitudes. Our model showed that as greenhouse gas emissions increased and time passed, the area of suitable habitats contracted sharply, peaking in the second half of the 21st century. In terms of the northern and southern hemispheres, the Sapindus suitable habitats for Sapindus in the northern hemisphere predominantly expanded, while in the southern hemisphere they predominantly contracted. Jayasinghe and Prevéy also found a declining trend in the area of suitable habitats for Camellia sinensis (L.) O. Kuntze [63] and huckleberry [54] in the face of global warming, while He [64] found an expanding trend in the highly suitable habitat area for Xanthoceras sorbifolia Bunge under global warming in China. The Sapindus genus consists of drought-tolerant and hardy species, but they are more sensitive to extreme temperatures. With future global warming, parts of the high latitudes of the northern and southern hemispheres may be transformed into suitable habitats for Sapindus due to an increase in extreme low temperatures, while parts of the lower latitudes may experience persistent extreme heat or drought as the climate continues to warm. This poses a major problem for Sapindus survival. In He's study [64] , the distribution of X. sorbifolia was observed to be localised in areas of the Loess Plateau in China, which is at a higher latitude. Thus, a trend toward the expansion of suitable cultivation areas in the face of global warming is already evident. Our results indicated that Sapindus will also experience some localised suitable habitat expansions in the northern hemisphere at higher latitudes, but, generally, Sapindus habitats will predominantly contract in Although the MaxEnt model is the most commonly used model in recent years for predicting species distribution change and has greater prediction performance than other species distribution prediction models, it still has some limitations [53] . Despite climatic, soil, topographical, and solar radiation factors being taken into account in this model, species distributions can also be constrained by other factors such as adaptive capacity, interspeci c interactions, human activities, land use. When all factors are considered together, we estimate that the area of suitable habitats for Sapindus will further reduce and the contraction in habitat area in the face of future global warming may be more signi cant. However, importing all variables into the model may lead to di culties in variable screening, the effects of key variables may be weakened, and the resulting simulations may not necessarily be more accurate than the current model. Therefore, our ndings on the distribution pattern of suitable habitats for Sapindus and its response to future climate change, based on the MaxEnt model, provide an important theoretical basis and valuable recommendations for the conservation and sustainable exploitation of Sapindus genetic diversity.

Conclusions
Our modelling study showed that temperature, soil moisture, and soil pH may play an important role in determining the global distribution of suitable habitats for Sapindus. The ecological thresholds for critical environmental factors affecting Sapindus were the minimum temperature of the coldest month (0-20 °C), soil moisture (40-140 mm), mean temperature of the driest quarter (2-25°C ), mean temperature of the wettest quarter (19-28 °C), and soil pH (5.6-7.6). We found that the total suitable habitat area for Sapindus was 6059.97 × 10 4 km 2 , which was unevenly distributed across six continents. High suitability areas were found to be concentrated in the Mississippi Plain, Florida Peninsula, and Mexico in North America; La Plata Plain in South America; and the North China Plain, Middle and Lower Yangtze Plain, Sichuan Basin, Yunnan-Guizhou Plateau, and Southern Himalayas in Asia. By integrating simulations from four future climate scenarios from 2020 to 2100, the size and distribution of suitable habitats for Sapindus were found to vary signi cantly across different continents depending on the climatic scenario. As greenhouse gas emissions increased and time passed, the overall area of suitable habitats contracted to an increasing degree, resulting from a signi cant contraction in lower latitudes and a slight expansion to higher latitudes. Suitable habitats for Sapindus in the northern hemisphere predominantly expanded, while in the southern hemisphere they contracted. Therefore, germplasm surveys and resource conservation should be prioritised in areas of the southern hemisphere that will likely become unsuitable for Sapindus in the future. In contrast, the Mississippi Plain, the Florida Peninsula, Mexico, Central America, the Middle and Lower Yangtze Plain, the Yungui Plateau, and central India can be used as a base for resource conservation, large-scale cultivation, and utilisation of Sapindus in the future.

Materials And Methods
Design framework Based on global Sapindus occurrence data combined with historical environmental factors and future climate models, we explored the current and future potential habitats of Sapindus by utilising the MaxEnt (MaxEnt version 3.4.1) model and ArcGIS (ArcGIS 10.5). The speci c process is shown in Fig. 4.

Sapindus occurrence records
We collected 5674 world-wide Sapindus occurrence records from well-designed surveys, including: (1)  were some spatial clusters of occurrence records, especially in southern North America and Southeast Asia. When these spatial clusters exist, models are often over-tted in terms of environmental biases, and model performance values are in ated [65,66] . Therefore, we employed the Spatially Rarefy Occurrence Data tool in SDMtoolbox 2.0 [67] and set the spatial interval at 10 km to eliminate spatial clusters of occurrences. After elimination, 2245 occurrence records were used in the model (Fig. 5).

Environmental parameters
There were 32 environmental factors used in this projection, including bioclimatic, topographical, UV-radiation, and soil parameters, to model the potential suitable habitat for Sapindus (Table 3). The bioclimatic factors for current climate conditions were extracted from the 2.5 minute resolution historical climate (average for 1970-2000) database (https://www.worldclim.org/data/worldclim21.html). Topographical factors were obtained from the Harmonized World Soil Database v1.2 from the Food and Agriculture Organization of the United Nations (http://www.fao.org/soils-portal/soil-survey/soil-maps-anddatabases/). Soil factors were extracted from the Center for Sustainability and the Global Environment dataset (https://nelson.wisc.edu/sage/). UV-radiation variables were obtained from the glUV dataset (https://www.ufz.de/gluv/) [68] .
Future climate projections were extracted from the 2.5 minute resolution Shared Socioeconomic Pathways (SSPs) scenarios available from the BCC-CSM2-MR global climate (2021-2040, 2041-2060, 2061-2080, and 2081-2100) database of the Coupled Model Intercomparison Projects 6 (CMIP6) (https://www.worldclim.org/). This included the SSP1-2.6 (ssp126), SSP2-4.5 (ssp245), SSP3-7.0 (ssp370), and SSP5-8.5 (ssp585) scenarios. These scenarios were employed in our climate modelling and research to predict four possible future climates, all of which were considered possible depending on the quantity of greenhouse gases emitted in the near future. All environmental factors were statistically resampled to a 2.5 minute resolution using ArcGIS. However, there were several environmental factors that may have exhibited collinearity. Strong collinearity between these factors may arti cially in ate the accuracy of a model [69] . To avoid the potential problem of multicollinearity between environmental factors, all these factors were subjected to the Remove Highly Correlated Variables tool in SDMtoolbox. Set 0.9 as maximum correlation allowed value, we retained 21 environmental factors for subsequent modelling, including 15 bioclimatic factors, 2 UV-radiation variables, 3 soil variables, and 1 topographical variable.

Model application
The MaxEnt model utilises the maximum entropy principle, applying ve different feature constraints (linear, product, hinge, quadratic, and threshold) to environmental variables to calculate the potential geographic distribution probability of a species [53,70] . We used the SDMtoolbox to run the MaxEnt model on the ArcGIS platform. We set 25 % of the occurrence data as testing data and 75 % of the occurrence data as training data. We analysed the contribution of each environmental factor using jackknife analysis. To calibrate and validate the robustness of the MaxEnt model evaluation, receiver operating characteristic curve (ROC curve) analysis was used, and an area under the receiver operating curve (AUC) was employed to estimate the accuracy of the model predictions [71,72] . Model performance was classi ed as failing (0.5-0.6), poor (0.6-0.7), fair (0.7-0.8), good (0.8-0.9), and excellent (0.9-1) [73] .
AUC values closer to 1 indicated more successful models.
In the future distribution model, the future climate scenarios of four greenhouse gas emission models were used to simulate the global migration and change of suitable habitats for Sapindus from 2020 to 2100. Topographic and soil factors were set as stabilising variables in our model since topographic and soil factors are largely unaffected by climate change. In order to visualise the expansion and contraction of suitable habitats for Sapindus under different future climate scenarios, we created habitat maps using ArcGIS and compared the current suitable habitats with those in each future climate scenario.
Declarations Figure 1 Map of suitable habitat areas for Sapindus under current climate conditions. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 2
Distribution area of Sapindus habitats on six continents.

Figure 3
Changes in the distribution of potential suitable habitats for Sapindus under the ssp126 (A), ssp245 (B), ssp370 (C), and ssp585 (D) scenarios from years 2020 to 2100 compared with the current distribution of suitable habitats. Note: Red areas indicate habitat expansion, yellow areas indicate no change, and blue areas indicate habitat contraction. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 4
Design framework of the suitable habitat distribution model for Sapindus.

Figure 5
Global distribution of occurrence records for Sapindus. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Supplementarymaterial.docx