Dongting Yan?Wei Chen?Li Liu?Jing Li?Lin Liu?Yiling Wang

Introduction

Conservation of biodiversity is highly important in the face of changing climatic conditions(IPCC 2001,2014).And major considerations are changes in species distributions and extinction rates,as have been observed in floral biodiversity(Iverson and Prasad 2001;Thompson et al.2009).However,few studies have assessed the impact of climate change on biodiversity in China.Predicting current and future species geographic distribution is necessary to understand the interaction between a species and its environment under various climate change scenarios(Baldwin 2009;Adhikari et al.2012).Understanding the future impacts of climate change will help to prioritize biodiversity conservation strategies on a local scale.Predictions of the potential geographic distribution of species are fundamental to ecological engineering strategies and to design of sustainable climate change adaptation management practices to conserve biodiversity(Zai et al.2009;Gelviz-Gelvez et al.2015).

There are many species distribution models(SDMs)available for the estimation of species geographic distribution,such as BIOCLIM,DOMAIN,GARP,GLMs,GAMs,and MaxEnt,and these models differ by input parameters and criteria(Busby 1991;Carpenter et al.1993;Stockwell and Peters 1999;Elith and Leathwick 2009).Most of these SDMs use species presence and absence data,while MaxEnt can run with presence data alone(Phillips et al.2006).The MaxEnt model is a maximum entropy-based machine-learning program developed to estimate the probability distribution for a species’occurrence based on environmental data.Two data sets are input to the program,viz.GPS points that demarcate locations occupied by the species of interest and environmental layers.Based on these data,the software can analyze the possible distribution of species in target areas(Phillips et al.2006)and can be a useful tool to predict the geographic distribution of species.

U.lamellosa(Ulmaceae)is a deciduous,wind-pollinated tree with bisexual flowers and key fruit(Wu et al.1999).That reproduces mainly by clonal reproduction via rhizomes(Bi et al.2002).It is a perennial herb currently distributed in Hebei,Henan,Jiangsu,Liaoning,Inner Mongolia,Shanxi,and Beijing in China.It is usually distributed in semi-humid to semi-arid open forest at 1100–1650 masl(Bi et al.2002;Gu and Du 1994)or on cinnamon soil,brown forest soil or mountain meadow soil(Ru et al.2007).Previous studies of U.lamellosa mainly focused on population structure(Ru et al.2007),ecological characteristics of communities(Bi et al.2002),ecological niche(Bi et al.2003)and genetic diversity(Liu et al.2016).Over time,the habitat of U.lamellosa has shrunk due to climate change,environmental degradation,and human destruction due to economic development.This species has been included in the List of Class II State-Protected Endangered Plant Species in China(Bi et al.2002;Ru et al.2007).U.lamellosa seeds contain abundant decylic acid,which is an important medicinal and light industrial raw material(Chen and Huang 1998).However,over-exploitation and promotion of blind cultivation has resulted in loss of a valuable wild resource and a decline in the quantity and quality of U.lamellosa.

In this study,the MaxEnt model was used to estimate the current and future potential geographic distributions of U.lamellosa in China.The aim was to identify the key environmental variables that influence the species geographic distribution and discuss the reasons leading to changes in climatic suitability.Based on our results,we formulated the conservation and management measures presented here.

Materials and methods

Species occurrence data

We compiled a total number of 36 U.lamellosa occurrences(Table 1).Of these,20 records were from field surveys and 16 were from literature databases(e.g.,CNKI,Springer,Wiley InterScience,and ScienceDriect),the Chinese Virtual Herbarium(http://www.cvh.org.cn/),and the Global Biodiversity Information Facility(GBIF:http://www.gbif.org.cn/).We used Google Earth v7.0(http://earth.google.com)to assign latitude and longitude coordinates to those entries that did not include this information.

Climatic data and environmental variables

Data for nineteen bioclimatic variables from 1950 to 2000(i.e.,the present time)were downloaded from the Global Climate Databases with 2.5-min resolution(http://www.worldclim.org).These 19 data layers are produced by interpolating the average monthly climate data from weather stations across the globe on a 30-arc-second(approximately 1 km2)resolution grid(Hijmans et al.2005).Future climate data(years 2030,2040,2050,2060,and 2070)were downloaded from Climate Change,Agriculture and Food Security(CCAFS)(http://www.ccafs-climate.org).We used these data to model the habitat suitability distributions for U.lamellosa in RCP4.5 scenario.

Species distribution models with many potentially relevant variables may lead to over-fitting and unreliable prediction performance(Li et al.2012;Franklin et al.2006;Guyon et al.2010).Variables with a cross-correlation value greater than 0.85 were eliminated(Yang et al.2013),leaving six,environmental variables that were selected as inputs to MaxEnt.

Predictive modeling

The 19 environmental variables and 36 occurrences were used as inputs for MaxEnt.Of these records,25%were randomly selected to be test data,and the remainder were used as training data.MaxEnt uses a set of features(e.g.,linear,quadratic,product,threshold and hinge),that are functions of environmental variables that constrain the geographic distribution of a species.Cross validation was maintained in the replicate run,and the number of iterations were fixed at 500.To avoid over-fitting the test data,we set the regularization multiplier value at 0.1,following Phillips et al.(2004).All other model parameters were kept as the defaults.The model was replicated 10 times.We refer to the 6 bioclimatic variables as environmental factors.Area under the ROC(receiver operating characteristic)curve(AUC)values were used to evaluate model performance,where values vary from 0 to 1(Fielding and Bell 1997).An AUC value of 0.50 indicates that the model did not perform better than random,whereas a value of 1.0 indicates perfect model fit;we used values＞0.9 to indicate high performance,following Swets(1988).The Jackknife procedure was used to assess the importance of the bioclimatic variables in the modeling analysis.Using MaxEnt-generated response curves,we related habitat suitability for U.lamellosa to bioclimatic variables.The estimate of habitat suitability is made up of values that range from 0 to 1,with 0 indicating low and 1 indicating high suitability.

Grid layers were manipulated in ArcGIS 10.2.2.Administrative maps at 1:1,000,000 scale weredownloaded from the National Geomatics Center of China(NGCC)to display the prediction map for U.lamellosa.

Table 1 Occurrences data locations of U.lamellosa in China

Results

Model performance and environmental variable contribution

The AUC values generated by MaxEnt for U.lamellosa are shown in Fig.1.The training data and test data yielded AUC values of 0.948 and 0.879,respectively,indicating that this model was a good fit for U.lamellosa data.

The relative importance of different environmental variables showed that temperature seasonality(bio4),precipitation of wettest month (bio13),precipitation of warmest quarter(bio18),mean temperature of the driest quarter(bio9),isothermality(bio3),and annual mean temperature(bio1)were the most important in predicting the geographic distribution for U.lamellosa in China(Table 2).The cumulative contribution of these six environmental variables was 97%,suggesting that U.lamellosa distribution is strongly influenced by these six environmental variables.From the response curve(Fig.2),U.lamellosa had the highest probability of habitat suitability when temperature seasonality reached 9500,mean temperature of driest quarter reached -5°C, and isothermality reached 27(Fig.2a,d,e).The habitat suitability of a region for U.lamellosa increased sharply with annual mean temperature and precipitation of wettest month,when annual mean temperature reached 21°C and precipitation of wettest month reached 150 mm(Figs.2b and f).Probability of habitat suitability for U.lamellosa increased with an increase in precipitation of warmest quarter but decreased sharply when it reached 170 mm.

Mapping the geographic distribution of U.lamellosa

The suitable geographic distribution of U.lamellosa(including high to low suitability)was mostly projected in northwestern and north China(Fig.3).Some regions in east,northeast,and central China were also suitable for U.lamellosa.

In northwest China,U.lamellosa was predicted to be in Xinjiang and Gansu provinces,as well as in the central and northern areas of Shaanxi and Ningxia provinces.In north China,climatically suitable areas were predicted to occur in Tianjin,Hebei,central and northern Shanxi,and southern and western Inner Mongolia provinces.In east China,U.lamellosa was predicted to occupy most of Shandong province.In northeast China,highly suitable habitat was predicted to occur in Liaoning and Jilin provinces.In central China,U.lamellosa was predicted to occupy most of Henan province.Central areas of Xinjiang,Inner Mongolia,and western Liaoning provinces were predicted to be highly suitable regions for U.lamellosa.

Fig.1 The receiver operating characteristic(ROC)curve for U.lamellosa.Area Under ROC curve(AUC)values were calculated.These values range from 0.5 to 1.0,with 0.5 indicating a no better than random fit to the data,1.0 indicating perfect model performance,and values＞0.9 indicating high performance.The AUC values ranges from 0 to 1

Table 2 Environmental variables used to create the species distribution model and their percent contribution to model prediction

Model projections for 2030 showed less suitable climatic conditions for U.lamellosa(Fig.4).The suitable regions shrank in Gansu,Ningxia and Henan provinces,and new suitable regions appeared in Jiangsu,northern Anhui,southwestern Heilongjiang,and northwestern Hubei provinces.Areas of highly suitable habitat shifted to east China.New highly suitable region appeared in parts of northern Liaoning,along the eastern Hebei-Liaoning boundary,in central Jilin,and central Shandong provinces.New areas did not,however,include the central parts of the Xinjiang and Inner Mongolia provinces.

The estimated distribution of U.lamellosa in 2040 were not different from 2030.The suitable regions declined in area in southern Shaanxi province,and new regions appeared in central Ningxia,Jiangsu and northern Anhui provinces.Highly suitable regions were mainly in northern Shanxi,Shaanxi,central Hebei,and western Liaoning provinces,and along the Inner Mongolia-Liaoning boundary.

In 2050,the estimated distribution and suitable regions of U.lamellosa were again similar,except the estimated distribution slightly expanded in Gansu,central Shandong,and eastern Hebei provinces.

In 2060,the suitable regions were considerably smaller in Heilongjiang,Jilin,Liaoning,and Shandong provinces,while suitable habitat increased in Shaanxi province.The regions with the highest probability of suitable habitat were in the central-south parts of Shaanxi province and in the central part of Liaoning province.

In 2070,the model projected a slightly increased area of distribution.Suitable regions were smaller in Shaanxi and Hebei provinces but were larger in Heilongjiang,Henan,Shandong,and Jilin,and Liaoning provinces.The areas with highly suitable habitat were in central and southern Shanxi,central and southern Hebei,northern Shandong,and northern and western Liaoning.Overall,the predicted geographic distribution of U.lamellosa gradually declined in area and was concentrated in north,northeast and northwest China,with a small portion in east(Shandong)and central(Hubei)China.Highly suitable regions were concentrated in northeast and north China.

Fig.2 Relationships between six main environmental variables and the probability of presence of U.lamellosa in China:a ‘bio4’is temperature seasonality;b ‘bio13’is precipitation of wettest month(mm);c ‘bio18’is precipitation of warmest quarter(mm);d ‘bio9’’is mean temperature of driest quarter(°C);e ‘bio3’is isothermality;f ‘bio1’is annual mean temperature(°C)

Change in parameters due to future climatic change

The climate conditions and distribution of U.lamellosa are shown in Figs.5,6 and Table 3.Temperature seasonality and precipitation were projected to increase from contemporary values through the mid-twenty-first century and then decline from midcentury through the end of the twenty-first century.

From current to 2070,both temperature seasonality(bio4)and precipitation of the wettest month(bio13)showed an irregular fluctuating trend.Compared to 2050 also precipitation of the wettest month(bio13)decreased until 2070 but remained above current values.Compared to 2050 temperature seasonality(bio4)was predicted to decline until 2070 and remain below current values.Precipitation of the warmest quarter(bio18)showed a slowly increasing trend from today through 2070.

Fig.3 The estimated geographic distribution of U.lamellosa in contemporary(1950–2000)China.Blue to red colors indicate suitability of areas from 0(completely unsuitable)to 1(optimal).White points indicate training locations and purple points indicate test locations

Discussion

The estimated current geographic distribution of U.lamellosa was concentrated in the central areas of Xinjiang,Inner Mongolia,and western Liaoning provinces.Within these areas,light and heat resources are abundant,while the condition of water resources is unbalanced:there is a hot and rainy season in the summer,and a cold and dry season in the winter(Wu et al.2011).The average annual rainfall is below 1000 mm,and the rainfall is higher in the south than in the north.U.lamellosa prefers to grow and spread in this kind of environment.

Climate change is a dominant factor affecting the geographic distribution of plants(Pearson and Dawson 2003;Kelly and Goulden 2008).The distribution of U.lamellosa was estimated to shrink from the present time through 2050.The highly suitable regions of U.lamellosa shifted from the central areas of Xinjiang,Inner Mongolia,and western Liaoning provinces to north and northeast China.The possible reasons are as follows.First,temperature is estimated to increase over the next 30 years in the northwestern region,especially in the middle-eastern part of Gansu and Xinjiang provinces(Zhao et al.2008).Precipitation might also slightly increase in the northwest region,while drought trends would increase severely in other parts of China.Water shortage and drought could seriously prevent the growth of U.lamellosa influencing its distribution.In this study,precipitation was estimated to increase by approximately 72 mm from the present time to 2050(Table 3),but that is not sufficient to compensate for increasing evaporation due to warmer air temperatures(Wu et al.2000);additionally,temperature seasonality would add 1240.5°C,which would be unfavorable to the growth of U.lamellosa.However,the increase in minimum air temperature and increased precipitation would provide a wetter environment suitable for the growth of U.lamellosa in Liaoning.Second,over-exploitation and blind cultivation will lead to destruction of large areas of vegetation resulting in cessation of flows and future drought that will fragment the habitat of U.lamellosa.Third,low seed fertility of U.lamellosa and clonal reproduction dependence on rhizomes(Bi et al.2002)might result in reduced genetic diversity in U.lamellosa,which is detrimental to population expansion.

Fig.4 The estimated geographic distribution of U.lamellosa in 2030–2070 China

Fig.5 The values of the three main climate variables for 1950–2000 and 2050(averaged from 2041 to 2060).A jackknife test suggest that most of the variable contribution(69.4%)comes from the top three variables:temperature seasonality(bio4),precipitation of wettest month(bio13),and precipitation of warmest quarter(bio18).Note:Different colors correspond to the values of the figure.Pink to red brown indicates values from low to high

Fig.6 The values of the three main climate variables for 2070(averaged from 2061–2080).A jackknife test suggest that most of the variable contribution(69.4%)comes from the top three variables:temperature seasonality(bio4),precipitation of wettest month(bio13),and precipitation of warmest quarter(bio18).Note:Different colors correspond to the values of the figure.Pink to red brown indicates values from low to high

Table 3 Environmental variable values based on the RCP45 scenario from the geographic distribution of U.lamellosa

From year 2050 to 2070,the estimated distribution of U.lamellosa was predicted to continue shrinking.The highly suitable regions became concentrated in the south of north China and a small part of northeast China(i.e.,central Shanxi,southern Hebei and northwestern Liaoning).Temperature seasonality was estimated to decrease 2820.2°C,and suitable regions for U.lamellosa were projected to shift to southern north China as a result.U.lamellosa,a mesophyte,usually grows in semi-humid to semi-arid open forest.In future,temperature and precipitation in north China are estimated to increase(Liu et al.2007),which could provide a better habitat for U.lamellosa.

The predicted contemporary geographic distribution of U.lamellosa was greater in area than its actual distribution.For this study we were able to compile only 36 records for U.lamellosa.The MaxEnt model typically predicts a potential geographic distribution that over-estimates the actual distribution area.Because the MaxEnt model considers only niche-based presence data,it predicts the species fundamental niche rather than its realized niche(Pearson 2007;Kumar and Stohlgren 2009).

Climate change will reduce the area of suitable habitat for U.lamellosa and this will cause local extinctions.This phenomenon has been documented for other endangered species in China,such as Sinopodophyllum hexandrum and Amygdalus mongolica(Guo et al.2014;Ma et al.2014).Thus we recommend that future climate scenarios should be included in the designs for restoration and conservation strategies for the protection of medicinally and economically important species to avoid future extinction.

AcknowledgementsThe authors thank Xin Zheng for her assistance with field sampling.

Adhikari D,Barik SK,Upadhya K(2012)Habitat distribution modeling for reintroduction of Ilex khasiana Purk.,a critically endangered tree species of northeastern India.Ecol Eng 40:37–43

Baldwin RA(2009)Use of maximum entropy modeling in wildlife research.Entropy 11:854–866

Bi RC,Zhang J,Su JX(2002)Ecological characters of rareendangered plant Ulmus lamellosa in Shanxi Province.J Plant Res Environ 11:45–50(in Chinese)

Bi RC,Yi WB,Wang YN(2003)Study on niche of population of Ulmus lamellosa in the south area of Shanxi province.Acta Bot Boreal Occident Sin 23:1266–1271(in Chinese)

Busby JR(1991)BIOCLIM—a bioclimate analysis and prediction system.In:Margules CR,Austin MP(eds)Nature conservation:cost effective biological surveys and data analysis.CSIRO,Canberra,pp 64–68

Carpenter G,Gillison AN,Winter J(1993)DOMAIN:a flexible modelling procedure for mapping potential distributions of plants and animals.Biodivers Conserv 2:667–680

Chen HY,Huang CJ(eds)(1998)Fagaceae,Ulmaceae and Rhoipteleaceae.In:Flora of China.Science Press,Beijing,pp 334–350(in Chinese)

Elith J,Leathwick JR(2009)Species distribution models:ecological explanation and prediction across space and time.Ann Rev Ecol Evol Syst 40:677–697

Fielding AH,Bell JF(1997)A review of methods for the assessment of prediction errors in conservation presence/absence models.Environ Conserv 24:38–49

Franklin J,Wiser SK,Drake DR,Burrows LE,Sykes WR(2006)Environment,disturbance history and rain forest composition across the islands of Tonga,Western Polynesia.J Veg Sci 17:233–244

Gelviz-Gelvez SM,Pavo′n NP,Rangel PI,Barrera CB(2015)Ecological niche modeling under climate change to select shrubs for ecological restoration in Central Mexico.Ecol Eng 74:302–309

Gu S,Du C(1994)Relationships between geographical distribution of Ulmus lamellosa and climate in China.J Inner Mong For Coll 16:28–32(in Chinese)

Guo YL,Wei HY,Lu CY,Zhang HL,Gu W(2014)Predictions of potential geographical distribution of Sinopodophyllum hexandrum under climate change.Chin J Plant Ecol 38:249–261(in Chinese)

Guyon I,Saffari A,Dror G,Cawley G(2010)Model selection:beyond the bayesian/frequentist divide.J Mach Learn Res 11:61–87

Hijmans RJ,Cameron SE,Parra JL,Jones PG,Jarvis A(2005)Very high resolution interpolated climate surface for global land areas.Int J Climatol 25:1965–1978

IPCC (2001)Climate change 2001:impacts,vulnerability and adaptation.Contribution of working group II to the third assessment report of the intergovernmental panel on climate change.IPCC working group 2.Cambridge University Press,Port Chester,New York

IPCC (2014)Climate change 2014:impacts,vulnerability and adaptation.Contribution of working group II to the fifth assessment report of the intergovernmental panel on climate change.IPCC working group 2.Cambridge University Press,Port Chester,New York

Iverson LR,Prasad AM(2001)Potential changes in tree species richness and forest community types following climate change.Ecosystems 4:186–199

Kelly AE,Goulden ML(2008)Rapid shifts in plant distribution with recent climate change. Proc Natl Acad Sci USA 105:11823–11826

Kumar S,Stohlgren TJ(2009)Maxent modeling for predicting suitable habitat for threatened and endangered tree Canacomyrica monticola in New Caledonia.J Ecol Nat Environ 1:94–98

Li GQ,Liu CC,Liu YG,Yang J,Zhang XS,Guo K(2012)Effects of climate,disturbance and soil factors on the potential distribution of Liaotung oak(Quercus wutaishanica Mayr)in China.Ecol Res 27:427–436

Liu YX,Wu TW,Guo YF,Yan JH(2007)Prediction research of climate change trends over north China in the future 30 years.Acta Meteorol Sin 65:45–51(in Chinese)

Liu L,Chen W,Zheng X,Li J,Yan DT,Liu L,Liu X,Wang YL(2016)Genetic diversity of Ulmus lamellosa by morphological traits and sequence-related amplified polymorphism(SRAP)markers.Biochem Syst Ecol 66:272–280

Ma SM,Nie YB,Geng QL,Wang YX(2014)Impact of climate change on suitable distribution range and spatial pattern in Amygdalus mongolica.Chin J Plant Ecol 38:262–269(in Chinese)

Pearson RG(2007)Species’distribution modeling for conservation educators and practitioners.Synthesis.American Museum of Natural History.http://ncep.amnh.org

Pearson RG,Dawson TP(2003)Predicting the impacts of climate change on the distribution of species:are bioclimate envelope models useful?Glob Ecol Biogeogr 12:361–371

Phillips SJ,Miroslav D,Schapire RE(2004)Maxent software for species distribution modeling.http://cs.princeton.edu/～scha pire/Maxent/

Phillips SJ,Anderson RP,Schapire RE(2006)Maximum entropy modeling of species geographic distributions.Ecol Model 190:231–259

Ru WM,Zhang GP,Bi RC,Zhang JT(2007)Population structure and pattern of endangered Ulmus lamellosa in Shanxi.Chin J Appl Environ Biol 13:14–17(in Chinese)

Stockwell D,Peters D(1999)The GARP modelling system:problems and solutions to automated spatial prediction.Int J Geogr Inf Sci 13:143–158

Swets JA(1988)Measuring the accuracy of diagnostic systems.Science 240:1285–1293

Thompson I,Mackey B,McNulty S,Mosseler A(2009)A synthesis of the biodiversity/resilience/stability relationship in forest ecosystems.Forest resilience,biodiversity,and climate change.Secretariat of the convention on biological diversity,Montreal,technical series no.43,p 67

Wu HZ,Xu SW,Cui SC(1999)The four new records of plant species in Shanxi province.Shanxi For Sci Tech 2:20–21(in Chinese)

Wu JD,Wang SL,Zhang JM(2000)A numerical simulation of the impacts of climate change on water and thermal resources in northeast China.Resour Sci 22:36–42(in Chinese)

Wu ZT,Dijkstra P,Koch GW,Penuelas J,Hungate BA(2011)Responses of terrestrial ecosystems to temperature and precipitation change:a meta-analysis of experimental manipulation.Glob Change Biol 17:927–942

Yang XQ,Kushwaha SPS,Saran S,Xu J,Roy PS(2013)MaxEnt modeling for predicting the potential distribution of medicinal plant Justicia adhatoda L.in Lesser Himalayan foothills.Ecol Eng 51:83–87

Zai XM,Qin P,Wan SW,Zhao FG,Wang G,Yan DL(2009)The application of beach plum(Prunus maritima)to wasteland vegetation recovery in Jiangsu Province,China:seedling cloning and transplantation.Ecol Eng 35:591–596

Zhao CY,Nan ZR,Cheng GD,Zou SB,Zhang YZ(2008)Prediction of the trend of the future climate change in Northwestern China by statistical down scaling.J Lanzhou Univ(Nat Sci)44:12–18(in Chinese)