Prediction of potential survival areas of Smooth cordgrass (Spartina alterniflora) in China

(1)

Prediction of potential survival areas of Smooth cordgrass (Spartina alterniflora) in China

Jinghan Zhang

Degree project in biology, Master of science (2 years), 2012 Examensarbete i biologi 45 hp till masterexamen, 2012

Biology Education Centre, Uppsala University, and Life Science Center at Nanjing University, China

Supervisors: Shuqing An and Peter Eklöv

(2)

ABSTRACT

Smooth cordgrass (Spartina alterniflora) is one of the harmful quarantine weeds in China. Since its first introduction in China in 1979, this alien species has spread rapidly and damaged local ecological environments. Research to predict a suitable new area is an important step for management of the species and to prevent a further spread. In this study, Spartina alterniflora’s ecological niche was modeled using the application MAXENT. Analysis was based on species’ current distribution.

The investigations of this study were two-fold. First, a large-scale global

investigation (outside China) was conducted to predict suitable areas in China by comparing global and Chinese records of the species. In the second set, the combined records were used to predict suitable areas in the Jiangsu Province. The model’s accuracy was evaluated by Receiver Operator Characteristic (ROC) curve.

The areas under the ROC curve (AUC value) were all over 0.95, which indicated high predictive ability of this model. In the large scale prediction, Shanghai, Zhejiang, Fujian, Guangzhou, Guangxi and southern part of Wuhan, Jiangsu and Anhui were all potentially endangered by S. alterniflora invasion. On the smaller scale, the prone to invasion areas were mostly concentrated on southern part and some coastal areas of Jiangsu Province, where the precipitation and temperature were appropriate for this grass. Because of S. alterniflora has high dispersal ability and human induced history, the potential distribution areas in China are

considerable and it may invade more areas, in result spreading faster in the future.

To prevent further invasion and spread, an early eradication program should be adopted in the newly invaded areas. Meanwhile, the monitoring programs should also need to be applied in potential survival areas, especially in coastal harbors, airports, and tourism areas which are highly vulnerable to S. alterniflora invasion.

INTRODUCTION

Biological invasion is a major environmental problem in the world, affecting agriculture, forestry, fisheries, human health and natural ecosystems (Drake et al.

1989, Gewin 2005). This threat received great attention from ecologists and environmentalists (Gewin 2005, Mooney et al. 2005). Spartina alterniflora,

commonly known as smooth cordgrass, is a world-wide, notorious invasive species that has colonized large areas in coastal China. This perennial rhizomatous grass is native to the Atlantic and Gulf coasts of North America, occurring from Quebec (Canada) and Newfoundland (Canada) to Florida (America) and Texas (America). It is usually found in intertidal wetlands, especially estuarine salt marshes (An et al.

2007). In 1979, for the purpose of erosion control, soil melioration and dike

protection, smooth cordgrass was intentionally introduced to China. With the help

of a suitable climate, an exceptional adaptability and reproductive ability, this

(4)

species dispersed extensively and caused deleterious impacts on native species. The Chinese government has explored various ways to eliminate this invasive species, including burning, cutting, waterlogging and reaping young ramets as well as applying biomass harvesting, diking and aquaculture in the control and

management program (An et al. 2007, Li et al. 2007, Guo et al. 2007, Lin et al.

2011).

Species records

Smooth cordgrass (Spartina alterniflora) is a perennial salt tolerant grass, which normally grows 1 to 1.5 m tall with smooth, hollow stems. The roots are

well-developed and can expand from 30 to 100 cm deep into the soil. S. alterniflora is very fertile and like most of Gramineae, it can reproduce both sexually and asexually (Metcalfe et al. 1986, Partridge 1987, Riggs 1992, Simenstad &

Thom1995, Deng et al. 2006). Moreover, S. alterniflora can tolerate a wide range of environmental conditions including: inundation up to 12 hours a day, pH levels from 4.5 to 8.5, and high salinity (Xu & Zhuo 1985, Chung-Hsin 1993, Wang et al.

2006, Guo et al. 2007). These characters have helped this species to expand rapidly into new habitats, and human control methods reducing the rate of spread are limited. The initial idea was to introduce S. alterniflora as an ecologically and economically beneficial for the coastal areas. In the view of the physical aspect, it has provided new land resources by sedimentation, stabilized shorelines and protected intertidal zones, improved soil quality by decreasing soil salinity, extracting heavy metals in soils (Salla et al. 2011) and increasing its biological fertility. The expansion of S. alterniflora has also increased vegetative cover, thereby potentially restrained the greenhouse effect (Tang & Zhang 2003). From an economic perspective, it can be used as biofuel and food, and can even be applied in the health care for CVD (cardiovascular disease) since the BML (bio-mineral liquid) and TFS (total flavonoids) of Spartina alterniflora can prevent and control cardiovascular disease (Cai et al. 1996, Tang & Zhang 2003). Even though S.

alterniflora owns the advantages mentioned above, its expansion is out of control, and it has negative impact on animals and their habitats (Wang et al. 2006, Li et al.

2007, Guo et al. 2007). One of the reasons for its success is that S. alterniflora is a much stronger competitor than native plants. In Willapa Bay and San Francisco Bay, S. alterniflora successfully outcompeted native species Zostera marina L.,

Salicornia virginica L., Triglochina maritimum L., Jaumeacarnosa A .Gary, Phragmites australis, Scirpus mariqueter and Fucusdistichus L. (Simenstad &

Thom 1995, Daehler & Strong 1996, Baumel et al. 2003). Furthermore, S.

alterniflora may also hybridize with the native Spartina species (Daehler & Strong 1994) which can cause a threat to the survival of native Spartina species. In both New Zealand and San Francisco Bay, S. alterniflora have invaded mudflats and channels and converted those habitats to marshland (Partridge 1987, Daehler &

Strong 1994). Consequently, foraging habitat for numerous residential as well as

migrating shorebirds and waterfowl was affected by the loss of their habitats. In the

(5)

4

first stage of invasion, Spartina alterniflora alternates properties of the environment, such as soil physical and chemical properties, physiognomy, hydrology, intertidal zones character and ecosystem processes. Then it threatens native plants and waterway traffic directly or alternate microbial community structure, decrease native biodiversity, degrade the intertidal ecosystems, and eventually lead to serious public issues, such as fishery and tourism (Fig.1). In China, Spartina alterniflora has expanded over 112000 hectares from Beihai (21°36′N,109°42′E) in the south to Tianjin (38°56′N,121°35′E) (Guan & An 2003, An et al. 2007) and has directly caused millions of dollars in economic loss per year (Chen et al. 2004).

Figure 1. The potential impacts of Spartina alterniflora invasions in the invaded regions. Cited from Wang Qing et al. 2006,with permission from the publisher.

Study aims

A great numbers of researchers focus on invasive species’ ecology, dynamic of invasion and its consequences (Li et al.2007). In the attempt to control smooth cordgrass, Chinese researches have mainly focused on physical, chemical and biological measures. A management approach that determines its current and potential distribution and prevents them from getting into new areas was proposed (Wang et al. 2006) and has successfully helped to eradicate the invasive S.

alterniflora in Humboldt Bay (Daehler & Strong 1996). In consequence, finding new endangered areas is the first step and the most effective way for the

management of the species.

(6)

The objective of this study was to predict new potential areas prone to invasion of Spartina alternifora in China, especially in the Jiangsu province. This was done by combining large and small spatial-scale modeling. A large-scale habitat mapping used all the location of S. alterniflora (outside China) and the basic environmental factors (e.g. temperature and precipitation) to explore the maximum potential survival areas of S. alterniflora in China. This result was then compared with available records (distribution of S. alterniflora in China) to see if they were

congruent. In the next step, the Chinese data was added as the background data. The tidal range was used as a specific environmental factor to predict the possible spread of this grass in Jiangsu province (China).

MATERIALS AND METHODS

Modeling methods and evaluation

Ecological niche models are widely employed to predict potential geographic distribution of species; they have provided an important tool to quantify the risks imposed by invasive alien species, impacts of climate change, and spatial patterns of species diversity (Fideling & Bell 1990, Phillips et al. 2004, Elith et al. 2006, Ward 2007, Wang et al. 2007b). In general, survival prediction models can be divided into correlative models and mechanistic model (Wang et al. 2007b). The correlative model can be further divided into group discrimination model and profile model. In the discrimination model, two kinds of species’ distribution data are required, presence species and absence species data. As in the profile model, only presence/absence species data was used (Table 2). Different modeling methods may give different predictions and a number of recent papers have provided an overview to ecological niche models, or a comparison of modeling methods (Fideling & Bell 1990, Ward 2007).

Table 2. Modeling algorithms families, data requirements, and selected references (Continuous variables take arbitrary real values which correspond to measured quantities such as altitude, annual precipitation, and maximum temperature. Categorical variables take only a limited number of discrete values such as soil types or vegetation type (Phillips and Dudik 2008))

Categ ory

Subcateg

ory Model Species data Environmental

data Reference

Correl ative model s

discrimina tion model

GLM( Generalized linear model)

Presence /Absence

Continuous and/or categorical

McCullagh & Nelder 1989

GAM(Generalized additive model)

Presence/Absenc e

Hastie & Tibshirani 1990，Yee & Mitchell 1991, Pearce and Ferrier 2000

(7)

6

I used the MAXENT (maximum entropy) model (Phillips et al. 2004) as the main tool. The MAXENT model utilizes the information from presence-only data together with background layers of the environment (e.g. Phillips et al. 2004, Elith et al. 2006, Wu &Li 2009, Cao et al. 2010). The idea is to find the probability distribution of maximum entropy; that is closest to uniform in order to estimate the target probability distribution. It is implemented in the software MAXENT, a machine-learning technique based on a probability density estimation method (Phillips et al. 2008). MAXENT relies on data of species’ abundance (such as from natural history museums and herbaria) rather than presence/absence survey, which allows for modeling many more species. Although other techniques are available for this as well, Steven Phillips (2006) emphasized, their model has a clear mathematical formulation with explicit assumptions which allow linkage to ecological theory. In addition, it is a user-friendly and free software

(http://www.cs.princeton.edu/~schapire/maxent).The MAXENT was run with three suits of environmental variables: bio-climatic variables, ENFA (Ecological niche factor analysis) selected variables and Jackknife selected variables. For the smaller scale, tidal range was added in the environmental factors. When running the software, I used the default parameters proposed by the software. The model

BRT(Boosted decision trees)

Presence/Absenc e

Friedman, Hastie

&Tibshirani 2000 MARS(Multivariate

adaptive regression splines)

Presence/Absenc e

Continuous and/or

categorical Friedman 1991& 1993

GARP（Genetic Algorithm for Rule-set Production）

Presence(generat es

pseudo-absences internally)

Continuous and/or

categorical Stockwell & Peters 1999

CART( Classification and regression tree)

Presence/Absenc e

Continuous and/or

categorical Breiman et al. 1984 MAXENT(Maximum

entropy) Presence Continuous and/or

categorical

Phillips et al. 2004 &

2008

profile model

BIOCLIM (Boxcar

environmental envelope) Presence Continuous Busby 1991 DOMAIN(Gower’s

distance) Presence Continuous and/or

categorical Carpenter et al. 1993 ENFA(Ecological niche

factor analysis) Presence Continuous Hausser et al.2002, Hirzel

et al. 2002& 2006 Mahalanobis (Mahalanobis

distance) Presence Continuous Farber and Kadmon 2003

Mech anistic model

CLIMEX Presence Continuous Sutherst 1995

SDM No species

distribution data

(8)

created 70% species' points by random as training set; the remaining 30% data were set aside for testing the resulting model. All background environmental layers were restricted to the maximum sampling depth. In order to define the probability distribution and evaluate model predictions, in this study, MAXENT used 10,000 random background points in the study area. The output format was ASCII grid layer, which can be projected into GIS (Geographic Information System) software showing the potential geographic distribution of S. alterniflora. Natural breaks (Jenks & George 1967) were used to divide the risk of invasion into four-group, from probabilily to negligible chance of invading. The habitat suitability maps identify areas (1) where invasive species may actually be present (maybe not yet detected) or the most probable areas for the species; and (2) where species may disperse to in the future or minor suitable habitats for S. alterniflora, thus providing assistance for planning and prioritizing areas for surveillance (Ward 2007).

Fivefold cross-validation and the area-under-curve value of Receiver Operator Characteristic (ROC) plots (Wang et al. 2007a) were both used to estimate

predictive accuracy and model performance. By setting 5 random partitions rather than a single one, the average behavior of the algorithms can be accessed. Usually, there are two types of prediction errors in presence/absence models. The first type is over-prediction, which is also called false positives (commission error), results in areas being classified as climatically suitable when they are not; the second type is under-prediction, which results in areas being classified as climatically unsuitable when they are not, it is also named false negatives (omission error) (Wang et al.

2007a). A ROC curve was created by plotting sensitivity against the corresponding

proportion of false positives (equal to 1-specificity), the area under the ROC curve

(AUC) was calculated by presence vs. background data (Phillips et al. 2006) and

measures the ability of a model to discriminate between sites where a species is

present versus those where it is absent (Fideling & Bell 1990). It was not dependent

upon a particular threshold by providing a single measure of overall accuracy

(Fideling & Bell 1990). Presence locations were generated in two parts, training

dataset and testing dataset. The first one was used to predict with model and further

draw the Habitat Suitability map (HS map) while the latter would act as a test part

(Fig.2). AUC value ranges from 0 to 1, with the highest value (1) indicates a perfect

discrimination, otherwise, with a lower value (e.g. 0.5) suggests the discrimination

is no better than random. For example, a value of 0.7 for the AUC means that there

is a 70% probability that a random selection from the presence records will have a

model score greater that a random selection from the absence records. In the

prediction results, MAXENT would draw the ROC curve and provide the AUC

value of the model directly. The software used in this study is presented in Table 3.

(9)

8

Figure 2. General procedure of ROC curve analysis. Cited from Wang et al. 2007a,with permission from the publisher.

Table 3. The software used in this study and their origin

Software Version Function Source

ArcGIS 9.3 Geo-information system

application http://support.esri.com/en/

MAXENT 3.3.3.e MAXENT model

application

http://www.cs.princeton.edu/~schapire/

maxent

Bio-mapper 3.2 ENFA factor analysis http://www2.unil.ch/biomapper/

BLM-shipping 2.0 Record tidal value http://shipping.boloomo.com/us/treaty_

cn.html

Hawths tool bag Extracting the points http://arcscripts.esri.com/

Av2idrisi Transforming the format of

layers From Biomapper

Species occurrence data

In order to use an ecological niche model to predict potential areas prone to invasion by S. alterniflora, collecting as much as possible information about species’

distribution is one of the most important tasks. However, in most cases, it is

impossible to have systematic investigation for the target species, hence, distribution data is mainly derived from related references, specimen records in the museums or special databases, e.g. GBIF (Global Biodiversity Information Facility). In this study, 737 species-locality records of the occurrence of S. alterniflora were obtained from online database GBIF (Global Biodiversity Information Facility). The database represented records from early 1900s to 2011 throughout the world (except China).

The records were collected by scientists from Universities (e.g. Yale University,

(10)

University of Washington Burke, and University of Connecticut) and government institutions (e.g. The New York Botanical Garden). The records contained

information on data publisher, locality, year of collection, collector name and number, identifier, latitude and longitude. After selecting and removing useless and duplicate records, 464 presence localities remained (Fig.3). Unfortunately, none of these points was located in China. Therefore Chinese distributions of S. alterniflora were all obtained from published papers where 82 sites were derived (Fig.4). In the large scale prediction, 464 records were used as the background data to explore the

potential survival areas of smooth cordgrass in China. In the small-scale, all Chinese records were added as the background data, and tidal factor was also added in the environmental set. Totally, 546 points were used to find the potential survival areas for S. alterniflora in Jiangsu Province (China).

Figure 3. Occurrence records for S. alterniflora (green triangles, 464 records) in the world (expect China) used in this study. Data are derived from GBIF (Global Biodiversity Information Facility) (Biodiversity occurrence data published by: Field Museum of Natural History, Museum of Vertebrate Zoology, University of Washington Burke Museum, and University of Turku) (Accessed through GBIF Data Portal, data.gbif.org, 2007-02-22)

Figure 4. Occurrence records for S. alterniflora (red triangles, 82 records) in China used in this study.

Data derived from published literatures.

Environmental variables

Another issue in species distribution modeling is that the number of species and

(11)

10

climatic variables vary in the models. The choice of variables used in the model not only influences distributions of species, it also affects the degree of model

generality. Climatic variables, especially temperature and humidity play a huge role in determining the distribution of Spartina species. For the large-scale,

environmental data were collected from WORLDCLIM (version1.4, http://www.worldclim.org/ ; Hijmans et al. (2005). The dataset contains 56

environmental variables in three categories: elevation, temperature and precipitation.

These values were obtained by interpolation of climate stations during 1950 to 2000 at four different spatial resolutions: 30 seconds (0.93 x 0.93 = 0.86 km

²

at the equator), 2.5, 5 and 10 minutes (18.6 x 18.6 = 344 km

²

at the equator (Hijmans et al.2005)). Considering the computation and relative precision of the species’

distribution data, 5-min resolution was selected. The data has two formats: 'BIL' for generic grids and 'ESRI' for ESRI grids (Hijmans et al.2005). Both of them can be applied into Arc-INFO directly. Climatic variables, especially temperature, rainfall and humidity, played important roles in determining the distribution of species.

Three approaches were used in choosing the appropriate environment variables.

First, the 56 environment variables were already divided into 6 factor sets automatically: altitude above sea level (m), precipitation (mm), average monthly minimum and maximum temperature, average monthly mean temperature, and bioclimatic which variables derived from the monthly temperature and precipitation.

Bioclimatic variables represent annual trends (e.g. mean annual temperature, annual precipitation) seasonality (e.g. annual range in temperature and precipitation) and extreme environmental factors (e.g. temperature of the coldest and warmest month, and precipitation of the wet and dry quarters) (Hijmans et al. 2005). I used the whole bio-climatic set which contains 19 predictor variables (Table 4) as the first environmental set. ENFA (ecological niche factor analysis) within Bio-mapper software was used as the second tool to examine the similarity between the climate variables. To ensure the majority (90%) of the original variables variation degree were represented by a few variables. The eigenvalues from ENFA gave an indication of how much variance was explained by the factors. Both marginality and specialization coefficients for each EGV (Eco-geographic variable) were calculated for each variables. Positive values indicated the focal species preferred this variable and vice versa for negative values (Wang et al. 2007a). The third

method for choosing environment factors was via Jackknife (in MAXENT software) which calculate the contribution of each variable itself. Initially, all the variables were put into the model as environment factors and in turn excluded. Then the top 10 variables which contributed the most were selected.

Table 4. Explanation of environment variables in bioclimatic set (Cited from

http://www.worldclim.org/bioclim). For further detail see (http://www.worldclim.org, or Hijmans et al.

2005)

Name Description

BIO1 Annual Mean Temperature

BIO2 Mean Diurnal Range (Mean of monthly (max temp - min temp))

(12)

BIO3 Isothermally (BIO2/BIO7) (* 100)

BIO4 Temperature Seasonality (standard deviation *100) BIO5 Max Temperature of Warmest Month

BIO6 Min Temperature of Coldest Month BIO7 Temperature Annual Range (BIO5-BIO6) BIO8 Mean Temperature of Wettest Quarter BIO9 Mean Temperature of Driest Quarter BIO10 Mean Temperature of Warmest Quarter BIO11 Mean Temperature of Coldest Quarter BIO12 Annual Precipitation

BIO13 Precipitation of Wettest Month BIO14 Precipitation of Driest Month

BIO15 Precipitation Seasonality (Coefficient of Variation) BIO16 Precipitation of Wettest Quarter

BIO17 Precipitation of Driest Quarter BIO18 Precipitation of Warmest Quarter BIO19 Precipitation of Coldest Quarter

Apart from the basic environmental factors, the degree of tidal inundation was also considered as a significant factor controlling the distribution of plant species in the intertidal zone (Chapman 1940). Since it was difficult to obtain the whole world’s tidal range, this specific factor was only applied in the small-scale prediction. The tidal data was collected from BLM-shipping software. This software records daily tidal value along Chinese coastal harbors of which values (4 records per day) were used to calculate the daily tidal range by the formula: DTR (Daily tidal range)

=HTL –LTL, where HTL was the highest tidal level and LTL was the lowest tidal level. Later the data were converted into ESRI grids which can be used into the model and Arc-INFO. The map of China was downloaded from the national fundamental geographic information system (http://nfgis.nsdi.gov.cn).

RESULTS

Selection of environmental variables

The correlation tree of the 56 environmental variables is presented in Fig.5. All of the environment variables were not independent; especially some of them like bio1-tmin4, bio11-bio6, tmax1-tmax12, bio13-bio16, and bio14-bio17 are highly correlated that their correlation coefficient are all over 0.99.

The coastal line from Jiangsu to Fujian and Taiwan (the red part) displayed high

tidal range, especially Shanghai, Zhejiang and Taiwan coastal areas (Fig. 6).

(13)

12

Environmental variables were selected by bio-climatic, ENFA and Jackknife selection procedures (Table 5). Most of the environmental variables were related to precipitation.

Figure 5. The correlation tree of the environment variable (derived from Bio-mapper) (bio1-19 refer to Table 4. Alt means altitude (elevation above sea level), tmax 1-12 refer average monthly maximum temperature from January to December, tmin 1-12 refer average monthly minimum temperature from January to December, tmax 1-12 refer average monthly maximum temperature from January to December, prec 1-12 refer average monthly precipitation from January to December.)

(14)

Figure 6.Tidal range from high to low along Chinese coast. The red indicates a large range and blue shows small range. Dots mean location of Chinese port. (As it is calculated from arc-GIS, the upper side where should impossible to own the tidal range also show the range of tidal)

Table 5. Selected environmental variables. Tidal in bracket would be added in the small-scale prediction (pre1-12 refer average precipitation from January to December, alt means altitude that elevation above sea level, tmax 11 means average monthly maximum temperature in November, bio1-19 refer to Table 4). A~F refer to the graph number below.

Methods Results

Selection of environmental variables Large-scale map Small-scale map

Bio-climatic Bio1-bio19 (tidal) A D

ENFA

Bio6,bio7, bio12, bio14,bio17, bio19,prec1, prec2, prec11, prec12 (tidal)

B E

Jackknife

Alt, bio14,bio17,bio19,prec1, prec2,prec3,prec5, prec12, tmax11 (tidal)

C F

Habitat suitability maps and model performance

(15)

14

Large-scale

Habitat suitability maps with presence-only data provided relative index of

suitability (Ward 2007). All of the three results illustrated that the potential survival areas of S. alterniflora was mostly concentrated in southeast part in China (Table 6

& Fig. 7). The mean AUC value of these three environmental sets by MAXENT were 0.959, 0.954 and 0.962 respectively (ROC-plot in Appendix Fig. 2). Regions predicted to be suitable for S. alterniflora, were particularly located in the coastal and southern part of China.

A: Bio-climatic B: ENFA C: Jackknife

Figure 7. Predicted potential geographic distribution for Spartina alterniflora in China based on 464 occurrence records and three sets of the climatic variables. Environmental variables were given in the selection of Bio-climatic (A), ENFA (B) and Jackknife (C). Prediction strength is colored as either blue, light blue, yellow or red. While blue color represents areas that are least suitable, red colored represent areas that is most suitable. Light blue and yellow regions represent lower and intermediate habitats.

Table 6. Predicted potential geographic distributions for Spartina alterniflora in China based on 464 occurrence records and three sets of the climatic variables.

Degree Bio-clmatic ENFA Jackknife

High

Shanghai, Southern part:

Jiangsu, Jiangxi, Hunan Northern part: Zhejiang, Hainan, Taiwan,Guangxi (middle), Fujian & Guangzhou (part of coastal area)

Shanghai, Zhejiang(north), Fujian(east), Hunan(middle and south), Guangxi

(northeast), Guangzhou (north and middle), Middle part:

Jiangxi, Taiwan

Shanghai,Jiangsu(south and middle),

Zhejiang(north)

Median

Anhui (north part), Jiangsu(middle south), Jiangxi(east and west), Hunan(east), Wuhan (middle south), Hainan (middle)

Jiangsu (south), Wuhan (south), Fujian(west), Guangzhou (middle), Jiangxi (south and middle)

Middle part: Jiangxi, Guangzhou, Anhui;

Hunan (east)

Low

Fujian,Hunan(west), Middle part: Jiangsu, Wuhan, Taiwan

Anhui(south), Guizhou(east),et al.

Jiangsu(north), Wuhan (southeast), Guangxi (east), Fujian(northwest)

None Rest of China (e.g. Yunnan, Tibet, North east part)

Rest of China (e.g. Yunnan, Tibet, North east part)

Rest of China (e.g.

Yunnan, Tibet, North east part)

(16)

Small-scale

Nantong and the coastal line of Binhai and Sheyang, northern coastal part of Qidong displayed the most suitable areas for S. alterniflora in the whole province (Fig. 8). Lianyungang district (most northern part of Jiangsu province) and the southern part of Jiangsu are the second vulnerable places for S. alterniflora’s invasion; the other parts theoretically had few or no appropriate regions for S.

alterniflora. ENFA result illustrated a geographical range of suitable areas for S.

alterniflora from north to south Jiangsu. Lastly, Jackknife result demonstrated southern part of Jiangsu such as Wuxi, Suzhou, Nantong and Haian areas have the largest number of suitable habitats for S. alterniflora, indicating that the

precipitation there is particularly fit for this plant. Middle south part, middle north and northern part showed median, low and none of suitable habitats respectively.

The AUC values of MAXENT by these three environmental sets were 0.966, 0.952 and 0.966 independently (ROC-plots in Appendix Fig. 3).

D: Bio-climatic E: ENFA F: Jackknife Figure 8. Predicted potential geographic distribution for Spartina alterniflora in Jiangsu Province (China) based on 546 occurrence records and three sets of the climatic variables. Environmental variables are given in the selection of Bio-climatic (D), ENFA (E), Jackknife (F) with tidal range.

Four colors are used to indicate the strength of the prediction for each individual map pixel.

Prediction strength is colored as either blue, light blue, yellow or red. While blue colored represent areas that are least suitable, red colored represent areas that is most suitable. Light blue and yellow regions represent lower and intermediate habitats.

DISCUSSION

In the large-scale maps, the potential survival areas were mostly concentrated to the southeast part of China where the latitude is similar to the specie's origin (An et al.

2007). In the small-scale, the maps produced by ENFA and Jackknife were

approximately similar. Both of them indicate that south parts of Jiangsu Province

are the most suitable areas for S. alterniflora. It seems like, the temperature and

precipitation is more appropriate for this species in this area. The areas under the

(17)

16

ROC curve (AUC value) were all over 0.95, indicating this model had a highly predictive ability.

Selection of environmental variables

Theoretically, with more environmental factors involved, the results would be more precise. Nevertheless, redundant factors not only increase the model’s running time, but also may lead to incorrect predictions. Each of the 56 variables were not

independent and the result indicated that some of the variables were closely related (Fig. 5). Particularly adjacent months displayed extreme relevance, such as bio 11-bio 6, tmax 1-tmax 2, tmax 12-tmax 2, tmin 12-tmin 2, tmin 1-tmin 12, tmin 1-tmin 12, tmin 1-tmin 2, tmax 1-tmax 12. Those correlation coefficients were all above 0.995, meaning that, redundant information existed among the variables. Fig.

5 illustrates that there are differences between the variables, yet the same variables grouped together. Hence, it was necessary to reduce dimensions in order to find the appropriate variables.

In the ENFA analysis, one of its main objectives was to deal with the dimension of the environmental variables. If the species was distributed randomly throughout the study area, the eigenvalues would all be close to 1; marginality would be close to 0 and tolerance would be close to 1. In ENFA, the overall marginality value is 2.231>>1, and the specialization value of 4.969>>1 indicated that S. alterniflora had a strong environmental preference. The factor composition (Appendix Table 1) showed that the contribution coefficient of bio17 (Precipitation of Driest Quarter), bio14 (Precipitation of Driest Month) and bio19 (Precipitation of Coldest Quarter) were all larger than 0.2.Especially bio17 (Precipitation of Driest Quarter) and bio14 (Precipitation of Driest Month), both of them were up to 0.31, indicating S.

alterniflora would be more affected by these factors. From the correlation analysis above, it also can be shown that the two variables bio14 and bio17 were extremely relevant. In addition, all variables were correlated with precipitation, which means that S. alterniflora displays a tendency of preference to high humid areas, such as wetlands. Furthermore, the coefficient of the environmental variable bio15 (Precipitation Seasonality) was -0.22, illustrating S. alterniflora distributes in the areas with lower variation in precipitation. All of these were coincided with the characteristic of S. alterniflora as a hydrophyte which usually lives in intertidal areas. In previous research (e.g. Daehler & Strong 1996, An et al. 2007, Guo et al.

2007), Spartina alterniflora were always investigated in estuaries or intertidal areas.

Mckee & Patrick(1988) demonstrated that in Atlantic and Gulf coasts the growth range of S. alterniflora is positive correlated with mean tidal range. This again confirms the fact that Spartina alterniflora prefers humid areas.

Large-scale habitat mapping

The potential survival areas were mostly concentrated from the coastal line to the

inland in the southeast part of China. According to the different environmental

variables that I used, the three habitat maps were slightly different because of

(18)

various predictor variables. Bio-climatic analyses and ENFA demonstrated more suitable areas than Jackknife. This was due to selection of environmental variables.

In Jackknife, all 10 variables were selected by MAXENT without considering the ecology of the species. As an intertidal plant, S. alterniflora usually grows in intertidal zone or coastal areas ( Mckee & Patrick 988, Daehler & Strong 1996, An et al. 2007, Guo et al. 2007). Therefore, altitude may have little effect on its distribution. This was clear in the ENFA analysis showing that the other variables mostly related with precipitation were selected. All the three maps were mostly in congruence with the actual presence records of S. alterniflora (Fig. 4) showing a distribution along the coastal line from Jiangsu Province to Guangxi Province. This range has the same latitude as the origin areas (Atlantic and Gulf coasts of North America) and belongs to the warm temperate zone and subtropical monsoon climate zone (An et al. 2007). The humidity, temperature and even the beach characteristics here are almost the same as in the origin regions of S. alterniflora (Li et al. 2007, Zuo et al. 2009). However, the distribution of S. alterniflora also covers the upper northeast part of Shandong, Tianjin and partial of Jining, which were demonstrated to be the unsuitable areas. In theory, more environmental factors would create higher accuracy; yet this is based on natural distribution. The distribution of S.

alterniflora in China were mostly introduced by humans, while prediction models always only found the basic niche for the species, that would explain the over distribution in the upper north east part. On the other hand, due to other biotic interactions (e.g. grazing, inter-specific competition), or geographic barriers, the species realized niche may also be smaller than its fundamental niche. An ecological niche model indicated that S. alterniflora could spread in inland areas (Wuhan, Jiangxi) while actually its survival possibility are extremely low, because soil salinity (Bradley & Morris 1992) and tidal effect (Daehler & Strong 1996, Mckee & Patrick 1988) , both are unsuitable in these areas.

Small-scale habitat mapping

At the small-scale, tidal range was added as a specific environmental factor in order to produce a more precise prediction for Jiangsu Province. The results of the three maps (Fig. 8) were slightly different. Both ENFA and Jackknife analyses

investigated more climatically suitable areas than by Bio-climatic analyses. The bio-climatic analyses, found the coastline of Sheyang (Yancheng) to be relatively sensitive to invasion by S. alterniflora. In fact, Yancheng coastal wetland (Jiangsu, China) is the largest S. alterniflora distributed areas in China. Since S. alterniflora first introduction in 1982, it had spread over 1.25×10

⁴

hm

²

. Currently, in Jiangsu coastal line, this species had occupied areas about 5×10

⁴

~6.5×10

⁴

hm

²

(Liu et al.

2009). Fig. 9 shows the landscape change of Yancheng (Jiangsu) coastal wetland in

1992, 1999 and 2007. The red parts indicate the areas of S. alterniflora. In 15 years,

S. alterniflora spread to both north and south (e.g. Suaeda heteroptera Kitog). Red

crowned Crane Nature Reserve (Sheyang, Yanchen) located in Sheyang is the

largest coastline wetland nature reserve in China. Its coastline is 587 kilometers and

the whole area is 453 thousand hectares. S. alterniflora is found in the nature

(19)

18

reserve (Liu et al. 2009) and due to its fast rate of spreading, it has invaded over 144.92 km

²

(Liu et al. 2009). This has caused a serious threat to the organisms inside the nature reserve.

Figure 9. Landscape pattern dynamic change of Yancheng (Jiangsu) coastal wetland in 1992, 1999 and 2007.The red part indicated the areas of S. alterniflora. Cited from Liu et al. 2009,with permission from the publisher.

However, invaded areas in Jiangsu: Dafeng, Dongtai, Ganyu, Lianyungang and Qidong have in contrast suitability which is predicted to be low or even none. If environmental conditions at the occurrence localities constitute samples from the realized niche, a niche-based model would thus represent an approximation of the species’ realized niche (Elith et al. 2011). Even a model captures a species’ full niche requirements; areas of predicted presence would be larger than the species’

realized distribution. Despite the climatically suitable areas S. alterniflora mostly occur and grow in salty and coastal or river side areas (Mckee & Patrick 1988, Bradley & Morris 1992, Daehler & Strong 1996, Zuo et al. 2009).The prediction model investigated theoretical climatically suitable habitats for the species. Besides, those unsuitable invaded areas were human induced initially. Other factors like inter-specific competition may also restrict smooth cordgrass survival and successful spread. In conclusion, the results emphasized the significance of the southern part of Jiangsu (e.g. Wuxi, Yixing, and Suzhou) with high risk for S.

alterniflora. Because the temperature and precipitation were suitable here for the growth of S. alterniflora, it was predicted to spread in southern part of Jiangsu.

Drawbacks

Museum records have great potential for ecological researches, conservation issues, and in the study of invasive species (Loiselle et al. 2003, Suarez & Tsutsui 2004).

Museum records are particularly useful because the records contain individual

point-locality information, which can be put into species distribution models

directly without transformation. However, museum record represent presence-only

(20)

data and since absent data is also a necessary component of many modeling

methods, it is more useful in determining the realized niche. In addition, errors may exist in the occurrence localities, due to transcription errors, lack of sufficient geographic detail, or species misidentification. In addition, the number of occurrence localities may be too low to reliably estimate the parameters of the model. Similarly, the set of available environmental variables may not be sufficient to describe all the parameters of the species’ fundamental niche that is relevant to its distribution. Global environmental change alters the spatial distribution of physical conditions, habitats and species on earth (Chapin et al. 2000). As all the

environmental factors were downloaded from world-climate, representing the period from 1950 to 2000, with an increased climatic change, environmental conditions should have been adjusted before inserted into models (Thuiller et al.

2008). Besides, the environmental set of 19 bio-climatic factors reflects an annual average. If the focal objects only survive in a period of time of a year, for example only occurs in the spring or summer, the annual statistical data becomes incorrect.

Meanwhile, choosing the right climate variables based on the biology of the study species also plays a significant role in robust modeling. Apart from temperature and precipitation, other factors such as salinity, nitrogen, phosphorus and sulfur, which are also the most significant nutrients influence growth and reproduction of S.

alterniflora (Dai et al. 1997, Stribling et al. 1997, Craft et al. 1999, Gao et al. 2007, Yuan & Shi 2008), may have contributed to the variation in the distribution of S.

alterniflora. Especially, salinity is a potentially important factor which could vary substantially among marsh locations (Nestler 1977, Mckee & Patrick 1988).

Regrettably, due to time limitation and data resources, I was not able to consider this. In the small scale approach, if only Chinese environmental data had been used instead of whole world climate information, maybe the results would be more precise.

Undoubtly, model selection is an important part in species distribution research.

According to Elith et al. (2006), when the purpose is to predict species, ecological niche model should be preferred over other models. However, model predictions based on machine learning such as MAXENT are difficult to understand and to further explain ecologically (Elith et al. 2006& 2011, Wang et al. 2007a &2007b).

In order to choose the most optimal model, comprehension and application ability should be combined with biological knowledge of the species (Anderson & Lew 2003). In the application to a specific species, it must be clear that each model has its special assumptions and constraints. It is also worth to mention that all models give theoretical predictions only and cannot replace actual field and laboratory investigations.

I advocate map-based analysis of ecological coherence as an efficient tool in

adaptive management, both for assessing the relative strengths and weaknesses of

evolving MAP networks, and for visualizing and communicating the results to

stakeholders and policy makers. It is believed that a better knowledge of species

(21)

20

and models would create a better investigation of this kind of study.

CONCLUSION

S. alterniflora is a potentially significant threat to Jiangsu district, thus, continued studies of this species is warranted, particularly the potential distribution of the species. In this study, potential spread of S. alterniflora in China was predicted for the first time. Hopefully, it will help the local government to manage this invasive species in the future more efficiently. I also suggest that a cost-effective way to prevent the transformation of unique mudflat and saltmarsh communities into introduced Spartina-dominated marshes is to survey the vulnerable sites frequently and eliminate introduced Spartina spp. propagules before they spread. On the other hand, people could improve the exploit of S. alterniflora, for example, the huge biomass of S. alterniflora could be treated as a tremendous biological resource that can be employed in renewable energy. In that case, this invasive species may turn into wealth.

ACKNOWLEDGMENTS

This study was conducted with my supervisor professor An Shuqing of Nanjing University (China) and co-supervisor Peter Eklöv from Uppsala University (Sweden). I am especially grateful to my two supervisors for their precious time, guidance and advice. My sincere thank goes to the PhD student Junchen Lei (Nanjing Forestry University) who has given me advice and assistance with GIS. I also would like to thank Shenglai Yin for helping entering the tidal datum. Finally, I really appreciate all the websites and softwares that I used in this study, without them, it would have been impossible for me to accomplish this study. Lastly my special appreciation goes to my family and friends for their love and supporting me to finish my study.

REFERENCE LIST

An S.Q, Gu B.H, Zhou C.F, Wang Z.S, Deng Z.F, Zhi Y.B, Li H.L, Chen L, Yu D.H, Liu Y.H.

2007.Spartina invasion in China: implications for invasive species management and future research. European Weed Research Society 47: 183-191.

Anderson R.P, Lew D. 2003. Townsend Peterson. Evaluating predictive models of species’

distributions: criteria for selecting optimal models. Ecological modeling 162: 211-232.

(22)

Baumel A, Ainouche M.L, Misset M.T, Gourret J-P and Bayer R.J.2003. Genetic evidence for hybridization between the native Spartina maritime and the introduced Spartina alterniflora (Poaceae) in South-West France: Spartina ×neyrautii re-examined. Plant Syst. Evol. 237:87-97.

Bradley P.M. & Morris J.T.1992. Effect of salinity on the critical nitrogen concentration of Spartina slterniflora Loisel. Aquatic Botany 43: 149-161.

Breiman L, Friedman J.H, Olshen R.A and Stone C.J. 1984. Classification and regression trees.

Wadsworth International Group. Belmont, California, USA.

Busby JR. 1991. BIOCLIM—a bioclimate analysis and prediction system. In: Nature

Conservation: Cost Effective Biological Surveys and Data Analysis (eds Margules CR, Austin MP), pp. 64-68. CSIRO, Melbourne.

Cai M, Qin P, Zhang K.X, Xie M. 1996. Effect of TFS and BML on blood glucose in mice.

Ocean science 4: 12-13.

Cao X.F, Qian G.L, Hu B.S, Liu F.Q. 2010. Prediction of potential suitable distribution area of Flaveria bidentis in China based on niche models. Chinese Journal of Applied Ecology 21:

3063-3069.

Carpenter G, Gillison AN, Winter J .1993. DOMAIN: a flexible modeling procedure for mapping potential distributions of plants and animals. Biodiversity and Conservation, 2:

667-680.

Chapin F.S.I. 2000. Consequences of changing biodiversity, Nature 405:234-242.

Chapman V.J. 1940. Studies in salt marsh ecology; section VI and VII. Comparisons with marshes on the east coast of North America. Journal of Ecology 28:118-151.

Chen ZY, Li B, Zhong Y, Chen JK.2004. Local competitive effects of introduced Spartina alterniflora on Scirpus mariqueter at Dongtan of Chongming Island, the Yangtze River estuary and their potential ecological consequences. Hydrobiological 528: 99-106.

Chung C.H. 1993. Thirty years of ecological engineering with Spartina plantations in China.

Ecological Engineering 2:261-289.

Craft C, Reader J, Sacco J.N, Broome S.W.1999. Twenty-five years of ecosystem development of constructed Spartina alterniflora (Loisel) marshes. Ecological Applications 9:1405-1419.

Daehler C.C, Strong D.R. 1994. Variable reproductive output among clones of Spartina Alterniflora (Poaceas) invading San Francisco Bay, California: The Influence of Herbivory, Pollination, and Establishment site. American Journal of Botany 81:307-313.

(23)

22

Daehler C.C, Strong D.R. 1996. Status, prediction and prevention of introduced cordgrass Spartina spp. invasions in pacific estuaries, USA. Biological Conversation 78: 51-58.

Dai T, Wiegert R.G. 1997. A field study of photosynthetic capacity and its response to nitrogen fertilization in SA. Estuarine, Coastal and Shelf Science 45:273-283

Deng Z.F, An S.Q, Zhi Y.B, Zhou C.F, Chen L, Zhao C.J, Fang S.B, Li H.L. 2006. Preliminary studies on invasive model and outbreak mechanism of exotic species, Spartina alterniflora Loisel. Acta Ecological Sinica 26(8):2678-2686.

Drake J.A, Mooney H.A, Castri F, Groves R.H, Kuger F.J, Rejmanek M, Williamson M.1989.

Biological invasions: a global perspective. John Wiley and Sons, New York.

Elith J, Graham C H, Anderson P R, Dudik M, Ferrier S, Guisan A, Hijmans R J, Huettmann F, Leathwick J R, Lehmann A, Li J, Lohmann L G, Loiselle B A, Manion G, Moritz C, Nakamura M, Nakazawa Y, Overton J McC, Peterson A T, Phillips S J, Richardson K, Scachetti-Pereira R, Schapire R E, Soberon J, Williams S, Wisz M S, Zimmermann N E. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29(2):129-151.

Elith J., Phillips S.J, Hastic T, Dudik M, Chee YE and Yates CJ. 2011. A statistical explanation of MAXENT for eologists. Diversity and Distributions17:43-57.

Farber O. and Kadmon R.2003. Assessment of alternative approaches for bioclimatic modeling with special emphasis on the Mahalanobis distance. Ecological Modeling 160: 115-130.

Fideling A H, Bell JF. 1990. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environ Consery 24: 38-49.

Friedman J.H. 1991. Multivariate Adaptive Regression Splines. Annals of Statistics 19:1-141.

Friedman J.H, Hastie T, Tibshirani R. 2000. Additive logistic regression: a statistical view of boosting. Annals of Statistics 28: 337-407.

Gao JH, Yang GS, Ou WX. 2007. The influence after introduction of Spartina Alterniflora on the distribution of TOC, TN and TP in the national Yanchen rare birds’ nature reserve, Jiangsu Province, China. Geographical research 27(4): 799-808.

Gewin V. 2005. Industry lured by the gains of going green. Nature 436:173.

Guan YJ & An SQ. 2003. Ecological benefits and ecological consequences of exotic Spartina in China. Journal of Nanjing Forestry University (special issue of ecological studies), 95-101.

Guo YW, Chen LL, Lu BL, Zeng YQ, Meng DM, Lei J. 2007. Research advances of Spartina alterniflora in China. Forestry and Animal farming 142:1-5.

(24)

Hastie, T.J. and Tibshirani, R. 1990. Generalized Additive Models. Chapman and Hall, London.

Hausser A.H, Chessel J, Perrin D. 2002. Ecological-niche factor analysis: How to compute habitat-suitability maps without absence data? Ecology 83:2027-2036.

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

Hirzel AH, Hasser J, Chessel D, Perrin N .2002. Ecological-niche factor analysis: how to compute habitat-suitability maps without absence data. Ecology 83, 2027-2036.

Hirzel AH., Hausser J and Perrin N.2006. Biomapper 3.2. Laboratory for Conservation Biology, Department of Ecology and Evolution, University of Lausanne, Switzerland. URL:

http://www.unil.ch/biomapper

Jenks, George F. 1967. The Data Model Concept in Statistical Mapping. International Yearbook of Cartography 7: 186–190.

Li FR, Chen JQ, Chen MR, Yu YN, Chen LY, Li J, Peng SL. 2007. Research process in the control of smooth cordgrass. Ecology and Environment 16(6): 1795-1800.

Liu CY, Zhaung SQ, Jiang HX, Wang H, Sun Y. 2009. Spatio-temporal dynamicas of coastal wetland landscape patterns in Yancheng, Jiangsu Province. Remote sensing for land & resouces 3 (81): 78-83.

Loiselle BA, Howell CA, Graham CH, Goerck JM, Brools T, Sminth KG, Williams PH.2003.

Avoiding pitfalls of using species distribution models in conservation planning. Conservation Biological 17: 1591-1600.

McCullagh, P. and Nelder, J.A.1989. Generalized linear models, Second Edition. Monographs on Statistics and Applied Probability. Chapman and Hall, London, UK. pp. 511 .

Mckee K.L, Patrick, W.H JR. 1988. The relationship of Smooth cordgrass (Spartina alterniflora) to tidal datums: a review. Estuaries 11(3):143-151.

Metcalfe W S, Ellison A M, Bertness M D. 1986. Survivorship and spatial development of Spartina alterniflora Loisel.(Gramineae) seedlings in a new England salt marsh. Annals of Botany 58: 249-258.

Mooney H, Cropper A, Reid A .2005. Confronting the human dilemma. Nature 434: 561-562.

Nestler, J. 1977. Interstitial salinity as a cause of ecophenic variation in Spartina alterniflora.

Estuarine Coastal Mar. Sci. 16: 103-109.

Partridge T. R. 1987. Spartina in New Zealand. New Zealand Journal of Botany 25: 567-575.

(25)

24

Pearce J. and Ferrier S. 2000. An evaluation of alternative algorithms for fitting species distribution models developed using logistic regression. Ecological Modelling128:127-147.

Phillips S.J, Dudik M, Schapire RE .2004. A maximum entropy approach to species distribution modeling. In: Proceedings of 21st International Conference on Machine Learning(eds Brodley C, Cavazos J, Chan P, Jacob N, Kak A, Lippman R, Moss E, Pavlopoulou C, Precup D, Scheeff D, Stefanovic D, Utgoff P, Yurcik B), pp. 655-662. ACM Press, New York.

Phillips S.J, Anderson R.P, Schapire R.E. 2006. Maximum entropy modeling of species geographic distributions. Ecological modeling. 190: 231-259.

Phillips S.J and Dudik M. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31:161-175.

Riggs S R. 1992. Distribution of Spartina alterniflora in Padilla Bay Washington, in 1991.

Padilla Bay National Estuarine Research Reserve Technical Report No.3. Washington:

Washington State Department of Ecology 63.

Salla V, Hardaway C.J, Sneddon J. 2011. Preliminary investigation of Spartina alterniflora for phytoextraction of selected heavy metals in soils from Southwest Louisiana. Microchemical Journal 97(2), 207-212.

Simenstad C A, Thom R M. 1995. Spartina alterniflora (smooth cordgrass) as an invasive halophyte in Pacific Northwest estuaries. Hortus Northwest 6:9-12, 38-40.

Stockwell D, and Peters D. 1999. The GARP modeling system: problems and solutions to automated spatial prediction. International Journal of Geographical Information Science 13:

143-158.

Stribling J M. 1997. The relative importance of sulfate availability in the growth of Spartina alterniflora and Spartina cynosuroides. Aquatic Botany 56(2):131-143.

Suarez AV, Tsutsui ND. 2004. The value of museum collections for research and society.

Bioscience 54:66-74.

Sutherst RW, Maywald GF, Skarratt DB.1995. Predicting insect distributions in a changed climate. In: Insects in Changing Environment(eds Harrington R, Stork NE), pp.59-91.

Academic Press, London.

Tang TG, Zhang WJ. 2003. A discussion of ecological engineering benefits of Spartina spp and its ecological invasion. Engineering Science 5:15-20.

Thuiller W, Albert C, Araujo M.B, Berry P.M, Cabeza M, Guisan A, Hickler T, Midgley G.F,

(26)

Paterson J, Schurr F.M, Sykes M.T, Zimmermann N.E. 2008. Predicting global change impacts on plant species’ distributions: Future challenges. Perspectives in Plant Ecology, Evolution and Systematics 9:137-152.

Wang Q, An SQ, Ma ZJ, Zhao B, Chen JK, Li B. 2006. Invasive Spartina alterniflora: biology, ecology and management. Acta Phytotaxonomica Sinica 44(5): 559-588.

Wang Q. 2007. The dynamics of plant community distribution of the salt marshes in the Yangtze River estuary as influenced by Spartina alterniflora invasions. Doctor project. Fudan University.

Wang YS, Xie BY, Wan FH, Xiao QM, Dai LY. 2007a. Application of ROC curve analysis in evaluating the performance of alien species’ potential distribution models. Biodiversity Science 15(4): 365-372.

Wang YS, Xie BY ,Wan FH, Xiao QM, Dai LY.2007b. Potential Geographic Distribution of Radopholus similis in China. Scientia Agricultura Sinica 40（11）：2502-2506.

Ward D.F.2007. Modeling the potential geographic distribution of invasive ant species in New Zealand. Biological Invasions 9:723-735.

Wu WH, Li MY.2009. Methods for Predicting Potential Habitat of Bursaphelenchus xylophilus based on Ecological Niche Model. Forest Inventory and Planning 10:33-38.

Xu GW, Zhuo RZ. 1985. Preliminary studies of introduced Spartina alterniflora Loisel. in China(I). Journal of Nanjing University (Natural Science Edition) 40(2):212-225.

Yee TW, Mitchell ND.1991. Generalized additive models in plant ecology. Journal of Vegetation Science 2: 587-602.

Yuan L, Zhang LQ, Xiao DR. 2011. The application of cutting plus waterlogging to control Spartina alterniflora on saltmarshes in the Yangtze Estuary, China. Estuarine, Coastal and Shelf Science 92: 103-110.

Yuan ZN, Shi FC. 2008. Effect of salt stress on seed germination and embryo growth of Spartina alterniflora (Gramineae). Acta Botanica Yunnanica 30(2): 227-231.

Zuo P, Liu CA, Zhao SH, Wang CH, Liang YB. 2009. Distribution of Spartina plantations along the China’s coast. Acta Oceanologica Sinica 5:101-111.

References used for collecting Spartina Alterniflora locations

An SQ, Gu BH, Zhou CF, Wang ZS, Deng ZF, Zhi YB, Li HL,Chen L, Yu DH, Liu

YH.2007.Spartina invasion in China: implications for invasive species management and future

(27)

26

research. European Weed Research Society 47: 183-191.

Chen HL, Li YJ, Li B, Chen JK, Wu JH. 2005. Impacts of exotic plant invasions on soil biodiversity and ecosystem processes. Biodiversity Science 13(6): 555-565.

Chen L, Deng ZF, An SQ, Zhao CJ, Zhou CF and Zhi YB. 2007. Alternate irrigation of fresh and salt water restrains clonal growth and reproduction of Spartina Alterniflora. Journal of Plant Ecology 31(4): 645-651.

Chung CH. 1993. Thirty years of ecological engineering with Spartina plantations in China.

Ecological Engineering 2:261-289.

Deng ZF, An SQ, Zhi YB,Zhou CF, Chen L, Zhao CJ, Fang SB, Li HL. 2006. Preliminary studies on invasive model and outbreak mechanism of exotic species, Spartina alterniflora Loisel. Acta Ecological Sinica 26 (8):2678-2686.

Deng ZF, Xie XL, Wang ZS, An SQ. 2010. Effects of substrate and water level on the growth of invasive species Spartina Alterniflora. Chinese Journal of Ecology 29(2): 256-260.

Ding WX, Zhang YH, Cai ZC. 2010. Impact of permanent inundation on methane emissions from a Spartina alterniflora coastal salt marsh. Atmospheric Environment 44:3894-3900.

Gan L, Zhao H, Qing H, Xiao Y, Li M, Hu QX, An SQ. 2011. Interactive effects of nitrogen and sulfur on the reproduction, biomass accumulation and allocation of the clonal plant Spartina Alterniflora. Acta Ecologica Sinica 31(7): 1794-1802.

Gao JH, Yang GS, Ou WX. The influence after introduction of Spartina Alterniflora on the distribution of TOC, TN and TP in the national Yanchen rare birds nature reserve, Jiangsu Province, China. 2007. Geographical research 27(4): 799-808.

Gao ZQ. 1996. The problem on the development and utilization of Spartina Resource on Fujian Seashore Shoaly Land. Journal of Fujian Agricultural University 25(1): 72-77.

Li FR, Chen JQ, Chen MR, Yu YN, Chen LY, Li J, Peng SL. 2007. Research process in the control of smooth cordgrass. Ecology and Environment 16(6): 1795-1800.

Li HP, Zhang LQ, Wang DH. 2005. Distribution of an exotic plant Spartina alterniflora in Shanghai. Biodiversity Science 14(2): 114-120.

Li JL. 2005. Granularity analyses of superficial sediment of Spartina Alterniflora flat on south coast of Hangzhou Bay. Project work, Coastal Resources and Environment Research Center, Ningbo University.

Li WZ. 2008. Distribution investigation and evaluation of Spartina alterniflora in Shankou

(28)

Mangrove reserve. Science Column 7: 39-41.

Li YL, Wang L, Zhang WQ, Zhang SP, Wang HL, Fu XH, Le YQ. 2010. Variability of soil carbon sequestration capability and microbial activity of different types of salt marsh soils at Chongming Dongtan. Ecological Engineering 36: 1754-1760.

Lin Y, Zhang LQ, Xiao DR. 2011. The application of cutting plus waterlogging to control Spartina alterniflora on saltmarshes in the Yangtze Estuary, China. Estuarine, Coastal and Shelf Science 92: 103-110.

Mo ZC, Fan HQ, Liu L. 2010. Investigation on Smooth Cordgrass (Spartina Alterniflora ) along Guangxi coastal tidal zone. Guangxi Sciences 17(2): 170-174.

Ping Y, Zhang LQ. 2010. Study on long-term effects of physical control measures on Spartina alterniflora. Marine environmental science 29(1): 32-35.

Shen YM, Yang JS, Zeng H, Liu YM, Chen ZY. 2008. The Spatial Distributing Characters of the Soil Humus in Spartina Alterniflora Loisel Salt Marsh. Journal of Agro-Environment Science 27(6): 2279-2284.

Shen YM, Liu YM, Chen QZ. 2002. Analysis of the expanding process of the Spartina Alterniflora Loisel salt marsh on Jiangsu Province coast by remote sensing. Journal of Plant Resources and Environment 11(2): 33-38.

Shen YM. 2002. The distribution of the manpower salt mash of Spartina Alterniflora and its effect on Jiangsu coast in ecology and economy. Territory & Natural resources study 2: 45-47.

Wang AJ, Chen J, Li DY. 2008. Impact of Spartina Alterniflora on sedimentary environment of coastal wetlands of the Quanzhou Bay. The ocean engineering 26(4): 60-69.

Wang BX, Zeng CS, Chen D, Wang WQ, Zhang LH. 2010. Effects of Spartina Alterniflora invasion on soil organic carbon in the Phragmites australis estuary wetland of Minjiang River.

Science of Soil and Water Conservation 8(2): 114-118.

Wang Q, An SQ, Ma ZJ, Zhao B, Chen JK, Li B. 2006. Invasive Spartina alterniflora: biology, ecology and management. Acta Phytotaxonomica Sinica 44(5): 559-588.

Wang Q. 2007. The dynamics of plant community distribution of the salt marshes in the Yangtze River estuary as influenced by Spartina alterniflora invasions. Doctor project. Fudan University.

Wang RZ, Yuan L, Zhang LQ. 2010. Impacts of Spartina alterniflora invasion on the benthic communities of salt marshes in the Yangtze Estuary, China. Ecological Engineering 36:

799-806.

(29)

28

Xiao DR, Zhang LQ, Zhu ZC. 2010. The range expansion patterns of Spartina alterniflora on salt marshes in the Yangtze Estuary, China. Estuarine, Coastal and Shelf Science 88: 99-104.

Xiao Y, Tang JB, An SQ. 2011. Responses of growth and sexual reproduction of Phragmites australis and Spartina Alterniflora to salinity stress. Chinese Journal of Ecology 30(2): 267-272.

Xiao Y, Tang JB, Qing H, Gan L, Tang BB, Zhou CF, An SQ. 2010. Effects of substrate, storage way and salinity on the seed germination of three coastal plants. Acta Ecologica Sinica

30:1675-1683.

Yu X, Tian JY, Li JQ, Sun JK. 2009. Distribution area and expanding speed of Alien invasive species Spartina spp. in Yellow River Delta. Marine Environmental Science 28(6): 684-686.

Yuan HW, Li SZ, Zheng HZ, Fang ZY. 2009. Evaluation of the influences of foreign Spartina alterniflora on ecosystem of Chinese coastal wetland and its countermeasures. Marine Science Bulletin 28(6): 122-128.

Yuan ZN, Shi FC, Li JJ, Guo SY. 2008. Sexual reproductive characteristics of Spartina alterniflora Loisel. in Tianjin coastal wetland. Chinese Journal of Ecology 27(9): 1537-1542.

Yuan ZN, Shi FC. 2008. Effect of salt stress on seed germination and embryo growth of Spartina alterniflora (Gramineae). Acta Botanica Yunnanica 30(2): 227-231.

Zeng Y, Tian GH ,Chen LY, Li J, An D, Lei ZS, Tang H, Peng SL. 2011. Influence of Spartina Alterniflora invasion on soil ecosystem: A review. Chinese Journal of Ecology 30(9):

2080-2087.

Zhang XL,Shi SL, Pan GX, Li LQ Zhang XH, Li ZP. 2008. Changes in Eco-chemical properties of a Mangrouve Wetland under Spartina Invasion from Zhang Jiangkou, Fujian, China.

Advances in earth science 23(9): 974-981.

Zhang Y, Lu JB. 2010. Progress on monitoring of two invasive species Smooth Cordgrass (Spartina alterniflora) and Water Hyacinth (Eichhornica crassipes) by remote sensing. Bulletin of Science and Technology 26(1): 130-137.

Zhang YH, Zhang FC, Zhou XD, Xie XJ, Wang XW, Li Q, Lei J. 2011. Effects of plant invasion along a Spartina Alterniflora chronosequence on organic carbon dynamic in coastal wetland in north Jiangsu. China Environmental Science 31(2): 271-276.

Zhao GQ, Zhang LQ, Liang X. 2005. A comparison of photosynthetic characteristics between an invasive plant Spartina Alterniflora and an indigenous plant Phragmites australis. Acta Ecologica Sinica 25(7): 1604-1611.

Prediction of potential survival areas of Smooth cordgrass (Spartina alterniflora) in China