The Influence of Natural Factors on the Spatio-Temporal Distribution Of

Acta Tropica 164 (2016) 194–207

Contents lists available at ScienceDirect

Acta Tropica

jo urnal homepage: www.elsevier.com/locate/actatropica

The inﬂuence of natural factors on the spatio-temporal distribution of

Oncomelania hupensis

a a b b,∗

Gong Cheng , Dan Li , Dafang Zhuang , Yong Wang

School of Geosciences and Info-Physics, Central South University, Changsha, China

State Key Laboratory of Resources and Environmental Information Systems, Institute of Geographical Sciences and Natural Resources Research, Chinese

Academy of Sciences, Beijing, China

a r t i c l e i n f o a b s t r a c t

Article history: Background: We analyzed the inﬂuence of natural factors, such as temperature, rainfall, vegetation and

Received 4 August 2016

hydrology, on the spatio-temporal distribution of Oncomelania hupensis and explored the leading factors

Received in revised form 7 September 2016

inﬂuencing these parameters. The results will provide reference methods and theoretical a basis for the

Accepted 17 September 2016

schistosomiasis control.

Available online 19 September 2016

Methods: GIS (Geographic Information System) spatial display and analysis were used to describe the

spatio-temporal distribution of Oncomelania hupensis in the study area (Dongting Lake in Hunan Province)

Keywords:

from 2004 to 2011. Correlation analysis was used to detect the natural factors associated with the spatio-

Schistosomiasis japonica

temporal distribution of O. hupensis. Spatial regression analysis was used to quantitatively analyze the

Oncomelania hupensis

effects of related natural factors on the spatio-temporal distribution of snails and explore the dominant

Natural factors

Spatial regression analysis factors inﬂuencing this parameter.

Snail control Results: (1) Overall, the spatio-temporal distribution of O. hupensis was governed by the comprehensive

effects of natural factors. In the study area, the average density of living snails showed a downward trend,

with the exception of a slight rebound in 2009. The density of living snails showed signiﬁcant spatial

clustering, and the degree of aggregation was initially weak but enhanced later. Regions with high snail

density and towns with an HH distribution pattern were mostly distributed in the plain areas in the

northwestern and inlet and outlet of the lake. (2) There were space-time differences in the inﬂuence

of natural factors on the spatio-temporal distribution of O. hupensis. Temporally, the comprehensive

inﬂuence of natural factors on snail distribution increased ﬁrst and then decreased. Natural factors played

an important role in snail distribution in 2005, 2006, 2010 and 2011. Spatially, it decreased from the

northeast to the southwest. Snail distributions in more than 20 towns located along the Yuanshui River

and on the west side of the Lishui River were less affected by natural factors, whereas relatively larger

in areas around the outlet of the lake (Chenglingji) were more affected. (3) The effects of natural factors

on the spatio-temporal distribution of O. hupensis were spatio-temporally heterogeneous. Rainfall, land

surface temperature, NDVI, and distance from water sources all played an important role in the spatio-

temporal distribution of O. hupensis. In addition, due to the effects of the local geographical environment,

the direction of the inﬂuences the average annual rainfall, land surface temperature, and NDVI had on

the spatio-temporal distribution of O. hupensis were all spatio-temporally heterogeneous, and both the

distance from water sources and the history of snail distribution always had positive effects on the

distribution O. hupensis, but the direction of the inﬂuence was spatio-temporally heterogeneous. (4) Of

all the natural factors, the leading factors inﬂuencing the spatio-temporal distribution of O. hupensis were

rainfall and vegetation (NDVI), and the primary factor alternated between these two. The leading role of

rainfall decreased year by year, while that of vegetation (NDVI) increased from 2004 to 2011.

Conclusions: The spatio-temporal distribution of O. hupensis was signiﬁcantly inﬂuenced by natural fac-

tors, and the inﬂuences were heterogeneous across space and time. Additionally, the variation in the

spatial-temporal distribution of O. hupensis was mainly affected by rainfall and vegetation.

license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

∗

Corresponding author at: 11A Datun Road, Chaoyang District, Beijing 100101, Schistosomiasis is a parasitic infectious disease that is harmful

China. to human health and restricts the development of the economy and

E-mail address: [email protected] (Y. Wang).

http://dx.doi.org/10.1016/j.actatropica.2016.09.017

G. Cheng et al. / Acta Tropica 164 (2016) 194–207 195

society. It is prevalent mainly in the 73 countries of Asia, Africa and 2. Materials and methodology

Latin America, and the number of patients affected is approximately

2.1. Overview of the study area

200 million (Mcmanus et al., 2011). Six parasite species have been

reported to infect humans, including Schistosomiasis haematobium,

Dongting Lake is located in the north of Hunan province,

Schistosomiasis japonicum, Schistosomiasis mansoni, Schistosomiasis

the south of Hubei province, and the southern bank of the Jin-

intercalatum, Schistosomiasis mekongi and Schistosomiasis malayen- ◦ ◦

jiang River (east longitude 111 19 –113 34 , northern longitude

sis. They are parasitic in speciﬁc types of intermediate hosts of ◦ ◦

27 39 –30 25 ). It ranks as the second largest freshwater lake in

freshwater snails (Wang et al., 2013). Schistosomiasis japonicais is

China. It is mainly divided into three lakes, West, South and East

mainly prevalent in China, the Philippines, and parts of Indonesia.

Dongting Lake. It connects to the Yangtze River in the north and

It has existed in China for more than 2100 years, and its virulence

to the Xiangjiang River, Zi River, Yuan River, and Feng River in

has aroused the attention of all previous governments in our coun-

the south. Vast amounts of water from the Yangtze River ﬂow

try (Zou and Ruan, 2015; Madsen and Hung, 2015). In recent years,

into the lake through Songzi, Taiping and Ouchi estuaries (Tiaox-

repeated infections or other factors have led to the reemergence

ian estuary was earthed up in 1954). Small rivers also ﬂow in,

of schistosomiasis, which is becoming one of the most important

such as the Luojiang and Xinqiang River. Finally, there is water

infectious diseases currently affecting public health in China (Zhou,

ﬂow into the Yangtze River through Chenglingji. Dongting Lake is

2012; Yang et al., 2015; Hu et al., 2016). Oncomelania hupensis is the

located in an area with a humid subtropical monsoon climate. It

only intermediate host of Schistosoma japonicum, and its distribu-

is warm and humid, with rain and heat over the same period. The

tion is consistent with that of S. japonicum. Therefore, investigating ◦ ◦

regional average annual temperature is 16.4 C–17.0 C, the aver-

the number of Oncomelania hupensis is the key chain in the control

age daily temperature is stable, and the coldest month is January,

of schistosomiasis (Wu et al., 2014). Many studies have shown that ◦ ◦

with an average temperature of 3.8 C–4.5 C. From late July to

the distribution of O. hupensis is closely related to natural factors,

early August, mostly, the average temperature is approximately

especially climate and environmental factors, such as tempera- ◦

29.6 C. The annual rainfall is 1328.8 mm, ranging from 800 mm

ture, rainfall, soil, hydrology, and vegetation (Wang and Zhuang,

to 2000–2300 mm. Precipitation mainly occurs from May- July and

2015; Zhu et al., 2015; Zhang et al., 2008, 2013). Therefore, study-

decreases from west to east. The annual evaporation is 1270 mm

ing the spatio-temporal distribution of O. hupensis and determining

and mainly occurs from May to September (64.4% of the total evap-

the factors that inﬂuence this distribution will be of great impor-

oration). The maximum evaporation is 233 mm in July. Because

tance in predicting snail status and controlling endemic areas of

schistosomiasis. the area is graced with rich rivers, ditches and wetland resources,

Dongting Lake is known as the land of ﬁsh and rice and plays an

Studies of the inﬂuence of natural factors on the spatio-temporal

important role in Hunan economic development.

distribution of snails have mainly focused on statistical analysis,

However, as a major lake in the middle of the Yangtze River, the

correlation analysis and normal regression analysis, generalized

complicated water regime and suitable natural environment pro-

linear models, generalized additive models, and Bayesian models

vide good living conditions for the breeding and reproduction of

(Wang et al., 2014; Sun et al., 2011; Liu et al., 2016; Hu et al., 2013;

Oncomelania hupensis, which is the host of S. japonicum. Therefore,

Yang et al., 2009a,b). These studies all ignored some attributions

the Dongting Lake district has become the most serious schistoso-

(spatial autocorrelation, spatial heterogeneity, scale-dependent

miasis endemic area in our country (Zhou et al., 2012; Gray et al.,

and so on) of schistosomiasis and snail (its host) data, and the

2014).

spatio-temporal heterogeneity of the effects of natural factors on

In particular, in recent years, water level, climate and environ-

snails, which might lead to some biased research conclusions and

mental factors have changed with the effect of climatic variation.

a limited understanding of these effects. A geographical weighted

The Three Gorges Project, the Returning Arable Land to Lake Project,

regression model (GWR), a local spatial regression analysis method,

and ﬂoods and droughts all affect snail breeding and distribution.

can successfully compensate for this deﬁciency (O’Sullivan, 2010).

For example, according to the report on national epidemiology

Using GWR, not only can we quantitatively analyze the relation-

of S. japonica in 2014, the 2 cases of acute schistosomiasis were

ship between snail distribution and its inﬂuencing factors, but we

both from Lixian in Hunan. Besides, the snail area increased by

can also deeply reveal the complex spatial variation of relation-

531.13 hm2 and Hunan province accounts for 81.7%. Therefore,

ships between snail distribution and its inﬂuencing factors with

schistosomiasis control is still grim in this region. Therefore, study-

respect to spatial-temporal heterogeneity (Lee et al., 2015). Rec-

ing the distribution of snails and related inﬂuencing factors plays a

ognizing these advantages, researchers have applied this method

very important role in controlling schistosomiasis, improving the

to schistosomiasis studies, but these studies were only preliminary

health of residents, and promoting regional socio-economic devel-

investigations taking place over a relatively short time period and

opment.

were more concerned with the superiority of the model ﬁt. Besides,

According to the distributions of S. japonica and O. hupensis (the

there is a lack of thorough studies on the factors inﬂuencing snail

host) in Hunan province, we selected the following 15 counties

distribution and a lack of comprehensive comparisons based on

as the study area: Yueyanglou, Junshan, Yunxi, Yueyang, Xiangyin,

multiple sections of time (Wu et al., 2014).

Huarong, Miluo, Anxiang, Dingcheng, Jinshi, Lixian, Hanshou, Nanx-

This study aims at exploring the possible impacts of natural

ian, Yuanjiang, and Ziyang. They are mostly located in Yueyang City,

factors on the spatio-temporal distribution of snails through a spa-

Changde City and Yiyang City, in the northeast of Hunan province.

tial autocorrelation analysis, by using living snail density in the

See Fig. 1.

Dongting Lake in Hunan Province during 2004–2011 as well as

certain relevant environmental factors such as average annual tem-

perature, rainfall, land surface temperature, normalized difference 2.2. Data sources and processing methods

vegetation index, and distance from water sources. Correlation and

regression analyses of the snail distribution and the related natu- 2.2.1. Data sources

ral environmental elements were carried out so that the effects Class A included basic geographic data (administrative division

of natural elements on snail distribution and the primary factors map of Hunan Province, vector map of the study area), climate

contributing to these effects could be identiﬁed. The results will and vegetation data (annual average temperature in the study area,

provide methodological references and a scientiﬁc basis for achiev- precipitation, normalized difference vegetation index (NDVI) and

ing schistosomiasis control. 1 km resolution raster data from 2003 to 2010). Class B included

196 G. Cheng et al. / Acta Tropica 164 (2016) 194–207

Fig. 1. Location of the study area.

n n

land surface temperature data (1 km resolution MODIS monthly 1 1

S2 = W X − X¯ , X¯ = X

(2)

LST product of the study area from 2003 to 2010). Class C included n ij i n i

snail density data (living snail density data of the study area from j=1 i=1

2004 to 2011). Class A and Class C were provided by the Resources

Where Xi and Xj are the observed values of the variable X at the

and Environmental Science Data Center of the Chinese Academy

X¯ X W

of Sciences. Class B was derived from the Geospatial Data Cloud ith and jth regions, respectively; is the average of ; ij is a

(http://www.gscloud.cn/). binary weighting matrix for the adjacent spaces; i = 1,2,3,. . .,n, and

j . . . i j W

Research has shown that the inﬂuence of natural elements on = 1,2,3, ,n. When the regions and are adjacent, ij = 1; other-

W −

snails has hysteresis. Given the data features of this study, we wise, ij = 0. The range of Moran’s I value is [ 1,1].

selected one year as the lag period (Hu et al., 2013). Therefore, snail Hypothesis testing was used to further verify the coefﬁcient, and

density data from 2004 to 2011 correspond to natural factor data the Z-value, expected value and Moran’s variation coefﬁcient were

from 2003 to 2010. obtained. The formula is as follows:

I −

E (I) Z (I) = (3)

VAR I

2.2.2. Data prepossessing ( )

Based on the vector map at town-scale, zonal statistics

(Pistocchi, 2014) on the average temperature, rainfall, NDVI and

E (I) = − (4)

n − 1

LST raster data, the township’s annual average temperature, rain-

fall, NDVI and LST data were obtained. Adjacent analysis was used

The variance is:

to obtain the value of the distance from water sources. After testing,

2 2

these data met the accuracy requirements of this study. n n − 3n + 3 W − nW + 3W 1 2 0 2

VAR (I) = − E(I) (5)

2 n n n

W − − −

n ( 1) ( 2) ( 3)

2.3. Methodology

When the z-score or p-value indicates statistical signiﬁcance, a

positive Moran’s I index value indicates a snail distribution with

2.3.1. Spatial autocorrelation

a tendency toward clustering, whereas a negative Moran’s I index

(1) Global spatial autocorrelation

value indicates a tendency toward dispersion. Zero indicates a ten-

dency toward randomness (Ii et al., 2016).

Global spatial autocorrelation was used to study the distri-

butional characteristics of snail density. This section uses the (2) Local spatial auto-correlation

Moran index (Moran’s I) to conduct spatial autocorrelation analy-

sis. The spatial autocorrelation tool evaluates whether the pattern

Moran’s I of local statistics can judge whether the spatial clus-

expressed is clustered, dispersed, or random. The formula is as fol-

tering pattern of the snail density is high or low. The formula is as

lows (Ii et al., 2016):

follows (Anselin, 1995): n n W X − X¯ X − X¯ x − x i= j= ij i j i ¯

1 1 Ii = wij xj − x¯ (6) Moran sI = n n (1) S2 S2 W i=1 j=1 ij j

G. Cheng et al. / Acta Tropica 164 (2016) 194–207 197

Table 1

The z-score formula is (Wu et al., 2014):

Results of general auto-correlation analysis of snail density.

Ii − E (Ii)

Z (Ii) = (7) Year Moran’s I Expected I Variance z-score p-value Result

VAR (Ii)

2004 0.3402 −0.0066 0.0028 6.5459 0.0000 clustered

2005 0.2520 −0.0066 0.0029 4.7712 0.0000 clustered

Where x¯ is the average of x, and Wij is a binary weighting matrix for

− n 2006 0.1509 0.0066 0.0029 2.9178 0.0035 clustered

−

j= ,j =/ j 2007 0.0975 0.0066 0.0028 1.9495 0.0412 clustered

2 1 i 2

adjacent spaces,S = − x¯ . A positive Ii indicates spatial

n−1 2008 0.1023 −0.0066 0.0028 2.0468 0.0407 clustered

clustering of similar values surrounding the unit space , and a nega- 2009 0.3166 −0.0066 0.0028 6.0226 0.0000 clustered

2010 0.2640 −0.0066 0.0028 5.0823 0.0000 clustered

tive Ii indicates spatial clustering of non-similar values surrounding

2011 0.3403 −0.0066 0.0028 6.5459 0.0000 clustered

the unit i.

A high positive z-score for a feature indicates that the surround-

ing features have similar values (either high or low). The feature

density of living snails showed a downward trend, with the excep-

class is HH for a statistically signiﬁcant cluster of high values and

tion of a slight rebound in 2009, especially in the south Dongting

LL for a statistically signiﬁcant cluster of low values. A low negative

Lake region.

z-score for a feature indicates a statistically signiﬁcant spatial data

Spatially, the living snail density in towns along lakes and water

outlier. The feature class indicates if the feature has a high value

systems was relatively high, and towns with high levels of liv-

and is surrounded by features with low values (HL) or if the feature

ing snail density were mostly distributed in the Lishui River Plain,

has a low value and is surrounded by features with high values (LH)

Yuanshui River Estuary, and Xinqiang River Estuary of East Dongt-

(Wu et al., 2014). This process was implemented in Arcgis 10.1.

ing Lake and along the tributaries of the Yangtze River Basin and

nearby Chenglingji port.

2.3.2. Simple correlation analysis In the temporal domain, areas with a low density of living snails

Simple correlation analysis (Pearson correlation analysis) was increased year by year; however, high density areas decreased sig-

used to explore natural factors that may inﬂuence the spatial dis- niﬁcantly. Statistical analysis showed that towns without snails

tribution of snails. The formula is as follows (Rujirakul and So-In, accounted for 11.2% of the total number of towns in 2004 and

2015): increased to 30.9% by 2011. Towns where snail density was higher

than 3.0/0.11 m decreased from 15.8% in 2004 to 2.6% in 2011. The n x − x y − y

i= ( i ¯) ( i ¯)

1 rxy = (8) snail density of Lishui Plain slowly decreased from 2004 to 2007 and

n n

x − x 2 y − y 2 changed little after 2007, and the snail density in West Dongting

i=1( i ¯) i=1( i ¯)

Lake Region changed little from 2004 to 2006 before decreasing

signiﬁcantly in 2007 and then continuing decline thereafter. In

Where rxy is the simple correlation coefﬁcient of x and y, x,¯ y¯ are the

the regions around South and East Dongting Lake, snail density

average values of x and y, respectively, and xi, yi are the ith observed

decreased signiﬁcantly, especially at South Dongting Lake; snail

values of x and y, respectively. This process was implemented in

density rebounded in these four regions in 2009.

SPSS 19.0.

3.1.2. Auto-correlation spatial analysis of snail density

2.3.3. Spatial regression analysis

(1) Global spatial auto-correlation analysis

Because the independent variable and dependent variable all

have spatial heterogeneity, there was a difference between the two

Snail density from 2004 to 2011 showed a signiﬁcantly spatially

numbers of different spatial units.

aggregated distribution, which ﬁrst weakened and then enhanced.

Using geographically weighted regression (GWR), we could

In this paper, global spatial auto-correlation analysis was used to

study the spatial heterogeneity of the spatial data, and the formula

study the snail density data across 8 years. The results are shown in

is as follows (O’Sullivan, 2010):

Table 1. The Moran’s I values were 0.3402, 0.2520, 0.1509, 0.0975,

i = ˇk (ui, vi) + εkˇk (ui, vi) xik + εi (9) 0.1023, 0.3166, 0.2640, and 0.3403. All of them were larger than

their corresponding expectation indices, p < 0.05 and Z > 0, which

Where, y is the dependent variable of point i; x is the kth inde-

i ik indicates that the snail density distributions were signiﬁcantly spa-

pendent variable at point i; k is the numeration of the independent

tially auto-correlated across these years. Thus, there was a spatial

variable, i is the numeration of sample points,ε is the residual,

i aggregation phenomenon of HH and LL aggregation. Meanwhile,

ˇ u , are the space coordinates of the ith point, and ˇ u ,

k ( i vi) k ( i vi) Moran’s I decreased from 2004 to 2007 and then showed an increas-

is the value of the continuous function at point i.

ing trend after 2007, indicating that the spatial aggregation of snail

GWR is based on the local weighted least squares method. The

density ﬁrst weakened and then enhanced from 2004 to 2011.

weight value is the distance function of the geographical space posi-

tion between points i and j. The regression coefﬁcient is no longer

(2) Local spatial auto-correlation analysis

a ﬁxed value in the global model but changes with spatial position

variation (Lin and Wen, 2011). This process was implemented in

High-value clustering was the main distribution pattern of snail

ArcGIS 10.1.

density, and areas with high-value clusters were mainly distributed

in the estuary of the Xinqiang River, the Lishui Plain and the estu-

3. Results and analysis ary of the Xiang River. Based on the snail density clustering and

outlier analysis at the town level, the spatial distribution patterns

3.1. Analysis of the spatio-temporal distribution of snail density of snail density were signiﬁcantly different in different years. The

signiﬁcance levels of each year’s snail density distribution index

3.1.1. Spatio-temporal distribution of snail density were above 0.05 (see Table 2). The spatial distribution patterns

As shown in Figs. 2 and 3, in the Dongting Lake area, Hunan were HH, HL, and LH in the years 2004, 2005, 2006, 2007, 2009

Province, snails were mainly distributed along the river and near and 2010. The patterns were HH and HL in 2008 and HH and LH in

the inlet and outlet of the lake, and a few lived in the region far 2011. The HH pattern was the most common. Because snail breeds

from the water source area from 2004 to 2011. Overall, the average are closely related to geographic environmental factors, it could

198 G. Cheng et al. / Acta Tropica 164 (2016) 194–207

Fig. 2. Distribution of snail density from 2004 to 2011.

Fig. 3. Variation in living snail density from 2004 to 2011.

G. Cheng et al. / Acta Tropica 164 (2016) 194–207 199

Table 2

Rates of snail density distribution patterns (␣ = 0.05).

Year HH HL LH

Number Percentage Number Percentage Number Percentage

2004 11 7.24% 2 1.32% 4 2.63%

2005 18 11.84% 2 1.32% 2 1.32%

2006 7 4.60% 3 1.97% 1 0.65%

2007 1 0.65% 2 1.32% 1 0.65%

2008 3 1.97% 2 1.32% 0 0.00%

2009 6 3.94% 2 1.32% 1 0.65%

2010 6 3.94% 1 0.65% 1 0.65%

2011 6 3.94% 0 0.00% 3 1.97%

Table 3

Results of the correlation analysis on snail density and natural factors.

Year TEMP RAIN NDVI LST DIST PRE ON

** ** ** ** **

2004 −0.535 0.399 0.236 −0.257 −0.294 − ** ** ** * ** ** 2005 0.488 −0.333 0.375 0.012 −0.244 0.683

** ** ** ** **

2006 −0.307 −0.284 0.196 0.287 −0.062 0.632

** * ** ** **

2007 −0.237 −0.107 0.188 −0.212 0.026 0.704

* ** **

2008 −0.019 −0.148 0.122 −0.106 −0.208 0.249

** ** ** **

2009 −0.314 0.077 0.337 −0.273 0.038 0.433

− ** ** ** ** ** − − 2010 0.327 0.312 0.38 0.224 0.098 0.872 ** ** ** * ** **

2011 −0.439 −0.513 0.468 −0.168 0.244 0.783

Signiﬁcant correlation at the 0.05 level (bilateral).

Signiﬁcant correlation at the 0.01 level (bilateral).

be inferred that the environmental characteristics of these towns relations featured the same trend of an initial decrease followed

are similar. Outliers appeared in a few towns. It could be inferred by an increase from 2004 to 2011. Annual average temperature,

that there were signiﬁcant differences between the snail breeding land surface temperature and annual rainfall had a strong negative

environments of these towns and those of surrounding towns. correlation with snail density and annual average NDVI, and histor-

According to Fig. 4, from the standpoint of time, the distribution ical snail information had a strong positive correlation with snail

patterns of living snails were signiﬁcantly different across 8 years. density. From the standpoint of inter-annual variability, the cor-

Around the estuary of the Xinqiang River, the snail density distri- relation between snail density and the distance from water sources

bution pattern was HH from 2004 to 2008. In all years except 2007, was very unstable. The two variables had a negative correlation in

there were towns with an HH pattern in the Lishui Plain. In partic- 2004, 2005 and 2008, a signiﬁcant positive correlation in 2011, and

ular, in areas with a dense river network in 2005, approximately no statistically signiﬁcant relation in other years. By comparing

1/10 towns had an HH pattern. In 2004, several towns around the the correlation coefﬁcients of snail density and inﬂuencing factors

estuary of the Xiang River (such as Jinghe Village and Zhangshu each year, it was found that historical snail information had the

Town in Xiangyin County) had HH patterns. In addition, the spatial greatest inﬂuence on these correlations.

position patterns of snail density were signiﬁcantly different across The above research results indicate that snail density was closely

8 years. The patterns changed profoundly from 2004 to 2007 and related to natural factors, such as climate and vegetation. However,

remained unchanged from 2008 to 2009. in different years, signiﬁcant variations existed in the magnitude,

direction, and signiﬁcance of these correlations. There may be two

reasons for these results. First, these factors are linked, inﬂuence

3.2. Correlation analysis of snail density and natural elements each other, and, together, constitute the breeding environment of

the snails (Remais et al., 2009; Qiu et al., 2014), which means that a

Snail density was closely related to natural factors, such as correlation analysis based on a single factor cannot fully reﬂect their

climate and vegetation. However, in different years, signiﬁcant inﬂuence on snail distribution. Second, snail breeding and living

variations existed in the magnitude, direction, and signiﬁcance of environments have relatively strict requirements with respect to

the correlation because changes in natural environmental factors, environmental factors, such as temperature, humidity, and light

such as temperature, humidity, rainfall, height, and vegetation, intensity, as extremes of these parameters are not suitable for snail

inﬂuenced snail distribution (Mcmanus et al., 2011; Wu et al., survival. Therefore, only natural elements, such as annual average

2015). To further explore this inﬂuence, annual rainfall, annual temperature, rainfall, and vegetation, in the appropriate range can

average temperature, surface temperature, normalized difference promote the growth of snails.

vegetation index, distance from water sources, and historical snail

information (the density of living snails the previous year) were

subjected to a correlation analysis along with snail density. The 3.3. Regression analysis of snail density and natural elements

main reason for selecting historical factors was to cover the com-

prehensive inﬂuence of other environmental factors (Wu et al., This paper used spatial regression analysis (GWR model) to

2014). quantitatively study the impact of climate, vegetation, rainfall, and

The results (Table 3) showed that the annual average tem- historical snail information on the spatio-temporal distributions

perature (TEMP), land surface temperature (LST), annual rainfall and differential characteristics of snails.

(RAIN), normalized difference vegetation index (NDVI) and histori- Square root transformation and normalization were applied

cal snail situation (PRE ON) signiﬁcantly affected snail density; the over the data regarding snail density and natural factors to create

results were mostly statistically signiﬁcant. Additionally, their cor- normal distributions of snail density and eliminate the differ-

200 G. Cheng et al. / Acta Tropica 164 (2016) 194–207

Fig. 4. Spatial distribution pattern of snail density from 2004 to 2011.

ences in magnitude and dimension among various natural factors. where the ﬁt was higher than 0.5. Fig. 5 and Table 5 show the spa-

Then, based on the “AIC minimization” principle (Akaike, 1974), tial differentiation of the combined effects of natural elements on

we selected ﬁxed kernels and AICc bandwidth to make a local snails. Table 6 shows the range of explanatory variable regression

estimate of the factors inﬂuencing snail density and tested the cor- coefﬁcients and the maximum and minimum values of the regres-

rectness of the model and the coefficient significance (Bowman, sion coefficients of the 154 towns for each factor. The signs and

1984). In the collinearity test of the regression model (Wheeler value sizes of the regression coefﬁcients represent the direction

and Tiefelsdorf, 2005), the condition index showed that the annual and extent of the impact of natural factors on snails, respectively.

average temperature had signiﬁcant collinearity with the annual A positive regression coefﬁcient indicates that the relevant natural

rainfall and surface temperature so that it could not be entered element has a positive impact on snails. The bigger the absolute

into the local spatial regression equation. The explanatory vari- value of the regression coefﬁcient, the greater the impact of the

ables of the regression model were annual rainfall (RAIN), annual factor.

average land surface temperature (LST), annual average NDVI, dis-

tance from water sources (DIST) and historical snail information 3.3.1. Spatio-temporal analysis of the impact of natural factors

(PRE ON). Their regression coefﬁcients passed the test at the 0.05 on snails

signiﬁcance level (Kelejian and Prucha, 2010). In terms of time, the impact of natural factors on snails initially

Results are shown in Tables 4–6, Fig. 5. Table 4 shows the out- increased and then decreased. As shown in Table 5, the result of

puts of the diagnosis models. Fig. 5 shows the demonstration of the GWR model based on land surface temperature (LST), annual

the spatial distribution of the ﬁt of the local model. Table 5 shows rainfall (RAIN), normalized difference vegetation index (NDVI), dis-

␣

the local R ( = 0. 05) and the statistics on the number of towns tance from water sources and historical snail information (PRE ON)

G. Cheng et al. / Acta Tropica 164 (2016) 194–207 201

Table 4

Diagnosis results of the GWR models.

Year 2004 2005 2006 2007 2008 2009 2010 2011

R2 0.56 0.75 0.63 0.61 0.50 0.76 0.82 0.77

AIC 155.1 109.5 121.1 67.1 137.7 90.5 −23.7 −4.5

Table 5

Local R statistics (␣ = 0. 05).

Year 2004 2005 2006 2007 2008 2009 2010 2011

Local MIN 0.03 0.235 0.317 0.17 0.046 0.06 0.229 0.443

R MAX 0.643 0.907 0.751 0.869 0.643 0.969 0.99 0.791

Town number(Local R > 0.5) 15 91 91 35 19 45 141 146

Table 6

Estimated result descriptions of local regression coefﬁcients (␣ = 0. 05).

Variable RAIN NDVI LST DIST PRE − − 2004 (−0.12,1.6) ( 0.18,0.15) ( 0.26,0.05) (0,0.46) 151:1 66:86 27:125 152:0 − − − 2005 ( 0.52,0.60) ( 0.13,0.27) ( 0.387,0.14) (0,1.0) (0.04,0.52) 32:120 123:29 75:77 152:0 152:0

2006 (−2.12,2.12) (−0.51,0.15) (−0.78,0.21) (0,1.2) (−0.34,0.66) 35:117 32:120 106:46 152:0 145:7

2007 (−0.31,0.27) (−0.05,0.06) (0.11,0.1) (0,0.15) (0.14,0.51) 89:63 76:76 26:126 152:0 152:0

2008 (−0.4,0.57) (−0.44,0.21) (−0.1,0.1) (0,0.23) (−0.01,0.32) 81:71 90:62 70:85 152:0 151:1

2009 (−0.81,0.41) (−0.32,0.28) (−0.75,0.13) (0,0.41) (−0.01,1.7) 110:42 119:33 51:101 152:0 151:1

2010 (−0.21,0.24) (−0.22,0.1) (−0.12,0.04) (0,0.14) (0.19,0.56) 84:68 77:75 48:104 152:0 152:0 − − − 2011 ( 0.55, 0.07) (0.02,0.46) ( 0.01,0.16) (0,0.46) (0.35,1.93)

0:152 152:0 129:23 152:0 152:0

Note: The numbers in brackets are the regression coefﬁcient ranges of the explanatory variables (natural elements); the ratio below shows the ratio of the number of positive

to negative towns. For example, (66:86) indicates that the regression coefﬁcient was positive in 66 towns and negative in the 86 towns.

showed that they could all indicate snail density in 2007 and from West Dongting Lake and west of the Lishui River, such as Yazigang

2009 to 2011. The variations in snail density, respectively, were and Yougang, had lower local R values than other towns in the

3%–64.3%, 23.5%–90.7%, 31.7%–75.1%, 17.0%–86.9%, 4.6%–64.3%, same year. This result indicates that in these areas, natural factors

6%–96.9%, 22.9%–99%, and 44.3%–79.1% in 2004–2011. Fit greater were weak indicators of snail density.

than 0.5 indicates a larger impact of natural elements on snails. The

ratio of towns where ﬁt was higher than 0.5 was less than 30% in

3.3.2. Impact analysis of natural elements on snail

2004 and 2008, approximately 60% in 2007 and 2009 and more than

spatio-temporal distribution

90% in 2010 and 2011. Obviously, the impact of natural factors on

The results show that the annual rainfall, average annual land

snails initially increased and then decreased.

surface temperature, normalized difference vegetation index, dis-

There was little effect in 2004 and 2008. Based on this informa-

tance from water sources, and historical snail information had

tion, we found that the regression model in 2004 was missing the

signiﬁcant effects on the spatio-temporal distribution of snails.

factor of historical snail information, and the results in Table 3 show

There was space-time heterogeneity in the impact direction of

that historical snail information had a greater impact on snails than

annual rainfall, land surface temperature and normalized differ-

other factors. In addition, there was a weak correlation between

ence vegetation index. Distance from water sources and historical

snail density and RAIN, NDVI, and LST in 2008, and the correla-

snail information had stable positive effects on snails. The local

tion between snail density and historical snail information was

regression coefﬁcients of the natural factors (Table 6) indicate that

weaker than that of other years. These results were consistent with

the larger the absolute value, the greater the impact of natural

previous research.

factors on snails. Additionally, if the regression coefﬁcient was pos-

Spatially, the combined impact decreased from northeast to

itive, the natural factor had a positive effect on snails. To perform

southwest. The results of the GWR model (Fig. 5) show that the

comparative analysis, we constructed distribution graphs of aver-

natural factors in different regions had very different inﬂuences on

age annual temperature, rainfall, NDVI, and LST (Fig. 6).

snail density. Thus, in different regions, the combined inﬂuences

The effect of rainfall on snails is shown in Table 6. In 2004, almost

of RAIN, NDVI, LST, NDVI and PRE ON on the growth and devel-

all town rainfall regression coefﬁcients were positive, and almost

opment of snails were very different. The ﬁt was higher near the

all were negative in 2011. The ratios of the number of positive to

water source. There was higher ﬁt between 0.6 and 0.95 for the

negative towns were different in 2005–2010, and the values were

20 towns around the outlet (Chenglingji) in the northeast of the

between the ratios of 2004 and 2011. By analyzing the distribu-

lake. This result indicates that in these areas, the natural factors

tional characteristics of rainfall (Fig. 6), we found that the annual

were strong predictors of snail density. The ﬁt was very low in

rainfall in 2011 was signiﬁcantly higher than in previous years,

the towns around the southwest lake. In particular, towns south of

ranging from 1156.1 mm to 1482.4 mm. Annual rainfall in 2004

202 G. Cheng et al. / Acta Tropica 164 (2016) 194–207

Fig. 5. Distribution of local R of the regression model.

ranged from 1068.8 mm to 1192.8 mm. The other six years of rain- where vegetation positively affected snail density occurred in 2011,

fall were generally lower than those two years. Based on these and the maximum number of villages where vegetation negatively

results, we can infer that the snail has a suitable range of rain- affected snail density occurred in 2006. However, the distribution

fall that is not too high or too low because that would interfere of NDVI (Fig. 6) shows that although there is large spatial dif-

with snail breeding. In this paper, based on the methodology of Wu ference between each region, the inter-annual variations of the

et al. (2014), we further analyzed the range of annual rainfall in spatial distributional characteristics are much smaller. Therefore,

the towns with positive regression coefﬁcients and the towns with the relationship between NDVI and snail density should be compre-

negative regression coefﬁcients, and the results produced ranges hensively inﬂuenced by other factors, such as temperature, rainfall

of 802.9 mm–1192.2 mm and 802.9 mm–1482.4 mm, respectively. and topography. For example, the NDVI regression coefﬁcients in

Thus, in our research area, rainfall was intermediate between approximately 80% of the towns were negative in 2006, and the

promoting and inhibiting the growth of snails. However, when NDVI regression coefﬁcients in all towns were positive in 2011.

rainfall was higher than 1192.2 mm, there was a stable inhibitory This difference was clearly not caused by the NDVI itself (Fig. 6).

effect. Observing the distributional characteristics of the other natural

The relationship between NDVI and snail density had large factors in these two years, we found that there were large differ-

temporal and spatial variation, but the annual difference was not ences in annual average temperature, land surface temperature and

caused by changes in NDVI itself. The maximum number of towns annual rainfall between the two years (Fig. 6). As a result, the above

G. Cheng et al. / Acta Tropica 164 (2016) 194–207 203

Fig. 6. Distribution of annual average temperature, rainfall, NDVI, and LST.

changes may have been driven by the combined effects of the other degree of inﬂuence may be affected by the regional environment

factors. In addition, some vegetation types that may have had an or other factors that have spatial-temporal heterogeneity.

impact on snail density were not introduced in this paper. The huge The historical snail situation (PRE ON) had an impact on snails.

impacts of abnormal climate change and human activity should also In almost every year and in all towns, the regression coefﬁcients

not be ignored and need further discussion. of PRE ON were positive. This result indicates that PRE ON was a

Land surface temperature (LST) had an impact on snail distri- strong predictor of the spatio-temporal distribution of snails, which

bution. The change in the trend of the impact of LST on snails was consistent with the results of the previous correlation analysis.

was roughly “M” shaped. In 2006 and 2011, 70% and 84% of the

towns had a positive regression coefﬁcient to LST (Table 4), respec-

4. Discussion

tively, and approximately half had a positive regression coefﬁcient

in 2005 and 2008. In 2004, 2007, 2009 and 2010, the LST regres-

4.1. Analysis of the leading factors in the spatio-temporal

sion coefﬁcients of most of the towns were negative. Furthermore,

distribution of Oncomelania hupensis

the distributional characteristics of LST (Fig. 6) can be broadly sep-

arated into the following cases by average levels: the minimum

The results indicate that the distribution of O. hupensis was

levels occurred in 2006, 2004–2005, 2007, and 2009 were between

formed by the comprehensive inﬂuence of various natural factors.

2006 and 2011, and the maximum occurred in 2008 and 2010. With

Even in the same area, rainfall, land surface temperature, veg-

the variation in the two series, the trend of positive effects of LST on

etation (NDVI) and distance from water sources (DIST) still had

snails was roughly “M” shaped. For further exploration, we counted

different effects on O. hupensis. Of all of them, rainfall and vegeta-

the ranges of LST in areas where the LST regression coefﬁcient was

tion (NDVI) were the primary factors exerting effects. Additionally,

◦ ◦

positive and negative, respectively. The results were 16.32 C–19 C

spatio-temporal variations showed that the dominant effect of rain-

◦ ◦

and 16.74 C–19.93 C, respectively. When the LST was in the range

fall weakened annually, whereas vegetation (NDVI) followed the

◦ ◦

of 16.32 C–16.74 C, it had a positive impact on snails. If the LST was

opposite trend from 2004 to 2011, especially in regions in the north

◦ ◦

16.74 C–19 C, it had both a stimulatory effect and an inhibitory

of the Lishui River plain. The land surface temperature and the dis-

effect on snails at different times and in different regions. How-

tance from the water source had a relatively weak effect on the

◦

ever, when LST was higher than 19 C, it would inhibit the growth

spatio-temporal distribution of O. hupensis.

of snails. This conclusion is consistent with the results of Zhang

To investigate which natural factor governed the snail distri-

et al. (2008).

bution of each time node and region, this study, based on the

There was an impact of distance from the water sources (DIST).

comparison of the absolute values of their local regression coef-

The DIST local regression coefﬁcients were all positive. However,

ﬁcients, explored the speciﬁc factors with the greatest impact on

the magnitude of the regression coefﬁcient was different across snails.

space and time. This variation shows that study areas closer to

The results are shown in Fig. 7. Except in 2007, 80%–90%

water sources were more suitable for snail living. However, the

of snail distributions were dominated by variation in rainfall and

NDVI. Speciﬁcally, more than half of the towns featured snail dis-

204 G. Cheng et al. / Acta Tropica 164 (2016) 194–207

Fig. 7. Leading factors in the spatio-temporal distribution of Oncomelania hupensis.

tributions that were mainly affected by rainfall. In particular, from 4.2. Cause analysis of the variation in the inﬂuence of natural

2004 to 2005, the proportions were all approximately 80%. The vari- factors on O. hupensis

ation in time indicated that towns that were dominated by rainfall

gradually decreased, especially in the Lishui River Plain and near Dongting Lake is the second largest freshwater lake in our coun-

west Dongting Lake, whereas NDVI increased in all years, with try, and 17 counties along the lake are the most serious S. japonica

exception of for 2007. In addition, snails in towns distributed in endemic areas (Cao et al., 2015). The water level of Dongting Lake

the north of the Lishui River plain were more affected by NDVI fol- shows seasonal variation, and vegetation there is lush, making it

lowing the year 2007. The effects of land surface temperature and very suitable for breeds of O. hupensis. There are many factors inﬂu-

distance from the water source on the distribution of O. hupensis encing the distribution and survival of O. hupensis, which can be

were relatively weak, especially the land surface temperature. The divided into natural and human factors. Natural factors include

towns that were mainly inﬂuenced by distance from water sources rainfall, vegetation, temperature, distance from water sources and

were all far from lakes and rivers. Land surface temperature had others, and the human factors consist of artiﬁcial snail control, envi-

less inﬂuence on snail distribution than other natural factors, which ronmental pollution, water conservancy and protection projects.

may be because the spatial disparity and inter-annual variability in To assure integrated and effective research, our study furthered

land surface temperature were not signiﬁcant and suitable for O. our understanding of natural elements, including climate and the

hupensis survival (Fig. 6). environment.

G. Cheng et al. / Acta Tropica 164 (2016) 194–207 205

(1) The inﬂuence of annual average temperature on the distri- on snails, both positive and negative, when rainfall is between

bution of O. hupensis was more sensitive to the local environment. 8029.5 mm and 1192.2 mm in different times and regions. There-

Annual average temperature and snail density showed a signiﬁcant fore, the conclusion that annual rainfall is positively associated with

negative correlation, which was different from previous research snail density, as indicated by some researchers, has many limita-

results (Yang et al., 2007; Wu et al., 2014) and incompatible with tions.

the laws of growth and development of O. hupensis. This result may In fact, the effects that rainfall, distance from water sources and

be explained by the fact that annual average temperature exhib- other related hydrological factors have on snails are mainly realized

ited little temporal or spatial change (with respect to the effect of by changing the humidity conditions of the snail breeding envi-

snail growth) and that the negative correlations between the two ronment. There are strict humidity requirements for snail growth.

appeared to be only statistical results. Temperature affected snails Humidity that is too high or too low will hinder snail growth. Thus,

in all stages of their life cycle, including snail egg hatching, young we could infer that the effects of variability in distance from water

snail growth, snail breeding and mating and snail infection with S. sources on snails (Table 3 and Fig. 5.) may be attributed to the

japonicum miracidia, which inﬂuenced the distributions of snails humidity of the micro-environment, which is also affected by other

and infected snails (Stensgaard et al., 2013; Yang et al., 2010; Zhou hydrological elements (Lai et al., 2013; Lu et al., 2014). In this study,

et al., 2008). Liang et al. (1996) believed that snails are mainly dis- when annual rainfall was higher than 1192.2 mm, it had a negative

◦

tributed in areas where the average annual temperature is >14 C inﬂuence on O. hupensis, regardless of changes in other factors.

◦

and the average temperature in January is >0 C. The temperature (3) The relationship between NDVI and snail density cannot

conditions of every town in the study area around Dongting Lake fully reﬂect the effect of vegetation on O. hupensis. Land vegetation

meet these requirements. Additionally, relevant research results can adjust temperature, preserve the soil moisture of the environ-

(Hong et al., 2002, 2003) show that the half lethal low temperatures ment and protect against heat and sun in midsummer and chill and

◦ ◦

for snails in dry and wet environments are −2.34 C and −2.72 C, blizzards in the winter, providing a suitable environment for snail

◦

respectively, and that the half lethal high temperatures are 40.01 C survival. In schistosomiasis endemic areas, if there is no human

◦

and 4.212 C, respectively. In this study, the extreme minimum intervention, the relationship between snail distribution and the

temperature and maximum temperature in Dongting Lake met the related vegetation index will be highly stable (McCracken, 2014).

conditions for the half lethal temperatures. Therefore, the effects Some research results show that snail density achieves a maximum

of extreme temperature on O. hupensis may be far greater than when NDVI reaches a certain value. Too large of an NDVI index

the annual average temperature, which masked the actual effect may shield them from the sun and restrict snail growth (Robins

of the annual average temperature. In addition, the effects of other and Wang, 2000). However, in this study, the space-time differ-

non-climatic factors on snails should not be ignored. ences in the effect that NDVI had on snails were not caused by

To verify the above speculation, this paper performed a spatial spatio-temporal variability itself. Based on this situation, we should

regression analysis on data of 2005, which showed larger nega- consider the inﬂuence of not only other natural factors but also

tive correlations between annual average temperature and snail changes in the local environment related to snail survival caused

density. The results showed that the towns where the regression by human activities (Liang et al., 2012; Xie et al., 2014). For exam-

coefﬁcient was negative were all distributed east of the study area, ple, the impoundment of the Three Gorges project and the South to

along the lake (Junshan county, Yueyanglou county, etc.). Com- North Water Diversion Project have changed the wetland pattern

pared with the eastern region relatively far from the lake, the lake of Dongting Lake in recent years, thereby affecting the distributions

region is crisscrossed by irrigation canals and ditches, and resident of snails there (Hu et al., 2015a,b; Liu et al., 2016). Artiﬁcial planting

production and living activities are frequent, making the schis- reeds and afforestation on beach lands have changed the ecologi-

tosomiasis prevalence in the region more serious, which is why cal environment of wetlands in Dongting Lake to a certain extent,

it has attracted increasing attention from relevant departments. leading to variations in the breeding environment of O. hupensis

Therefore, snail density changes may objectively be caused by (Cui et al., 2012). Vegetation succession on beach land may have

some local environmental differences, such as channel type; weed an insigniﬁcant impact on NDVI, but it has an obvious effect on

density and man-made factors, including agricultural and pastoral the development and reproduction of O. hupensis (Sun et al., 2011).

activities; snail control projects; and so on (Qiu et al., 2014; Shan Thus, with the increasing frequency of human activities, it is not

et al., 2014). Of course, the speciﬁc mechanisms underlying these suitable to study the effect of vegetation on O. hupensis based only

changes require more research and validation. on NDVI. We should perform a speciﬁc analysis combined with the

(2) Rainfall inﬂuenced the growth and development of O. hupen- micro-environment of speciﬁc vegetation in various regions.

sis mainly by changing the humidity of the natural environment. (4) In addition, global warming, which results in increases in

Water is one of the necessary conditions for snails to breed and extreme weather/climate, such as ﬂooding rain and snow, also

calve. Young snails often live in water surroundings, and adult snails has a great impact on snails, which has attracted the attention

living in areas of soil with plenty of water and food (Wang et al., of researchers (Wu et al., 2008). Remais et al. (2009) found that

◦

2014). Usually, the annual rainfall in areas with snail distribution the number of heat waves (>20 C) or rainstorms (>15 mm) will

is greater than 750 mm (Yang et al., 2009a, 2009b). Additionally, also inﬂuence the survival and proliferation of snails. Wang et al.

annual rainfall in the Dongting Lake district, which is the key (2010) observed the snail densities and the mortality distributions

schistosomiasis endemic area, was much higher than the lower of the original ecological marshes without a snail control project in

value. However, rainfall is a macro indicator, and variation in rain- areas of Junshan county near Dongting Lake and found that snail

fall changed the micro-environment for snail breeding, including mortality was up to 88.8% in February 2008. The snail area was

vegetation and humidity. As a result, the snail distribution was decreased by 2053.7 km2 from 2007 to 2008, which was directly

inﬂuenced. For example, rainfall may increase water levels, thereby related to the once-in-a-century sustained rain-snow and freez-

affecting the distribution of snails. Flooding can inﬂuence the dis- ing weather Hunan province suffered in early 2008. This ﬁnding

tribution of snails in two directions. It can inhibit the reproductive may also explain why there was no signiﬁcant correlation between

function of screw like spawning, leading to snail death. However, it snail density and related factors in 2008 (Table 1) and why ﬁt of the

also provides favorable conditions for the growth of juvenile snails model of snail density was low in the regression model for 2008.

(Wu et al., 2015). Therefore, in different environments, the effects (5) This paper did not discuss human factors in detail. How-

of rainfall on snails may contrast with each other. This research ever, the space-time distributions of the local ﬁt of the GWR model

result also showed that annual rainfall may have varying impacts demonstrated that the inﬂuence of human activity was a matter

206 G. Cheng et al. / Acta Tropica 164 (2016) 194–207

of concern, especially in areas with a worse ﬁt. In the analy- would have had a negative impact throughout the entire study

sis of the relationship between current snail density and history area. Average annual surface temperature may not only pro-

of snail density, the results showed that history of snail density mote but also inhibit snail breeds at different time points and

◦ ◦

played an important role in predicting current snail density. This in different regions when between 16.32 C and 16.74 C and

relation may be because the snail density was relatively stable, would play a negative role in the growth and development of

◦

and the strict natural, geographical, environmental requirements O. hupensis when higher than 19 C. Analysis of the effects of the

and the distribution of schistosomiasis are strictly endemic (Yin normalized difference vegetation index (NDVI) on the spatio-

et al., 2013). Additionally, under the backdrop that the perfect snail temporal distribution of Oncomelania hupensis still needs to be

investigation, management and analysis system would include the connected with speciﬁc vegetation types.

establishment of regional monitoring points of schistosomiasis (4) Of all the natural factors, the leading factors affecting the

control stations, if there was a strong negative correlation between spatio-temporal distribution of O. hupensis were rainfall and

snail density and historical snail information, in addition to the vegetation (NDVI), and the primary factor alternated between

inﬂuences of extreme natural environments, it could be concluded the two. The effect of rainfall decreased year by year, while

that results would be best achieved by promptly mastering snail vegetation (NDVI) increased from 2004 to 2011.

information and taking relevant prevention and control measures.

Overall, this paper was based on full consideration of the spatial

effect of snail density and of the spatial correlation between nat-

5. Conclusions

ural factors and snail distribution in the view of space and time.

The result showed the complex spatial variation in the relation-

This paper examined 15 counties around Dongting Lake, com-

ship between snails and the factors inﬂuencing their distribution.

bined the data on snails and natural factors, such as climate,

Additionally, the main factors affecting the spatio-temporal distri-

vegetation, hydrology and others, and used correlation analysis and

bution of O. hupensis were also explored. All of these results provide

spatial regression analysis to explore the inﬂuence of natural fac-

speciﬁc referential methods and a strongly theoretical basis for

tors on the spatio-temporal distribution of O. hupensis. The results

schistosomiasis control in this region.

were as follows.

Of course, there were some problems with and deﬁciencies of

this study. For example, the effects of climatic factors on snail den-

(1) The spatio-temporal distribution of O. hupensis was governed sity are macroscopic and are thus relatively appropriate to study

by the comprehensive effects of natural factors. Speciﬁcally, in at the town level based on the results of this paper. However,

this study, the average density of living snails showed a down- the relationship between snail distribution and environmental fac-

ward trend, with the exception of a slight rebound in 2009. The tors, such as normalized difference vegetation index, land surface

density of living snails showed signiﬁcant spatial clustering dis- temperature and humidity, was more obvious on a smaller scale.

tribution, and the degree of aggregation was initially weak but Therefore, in a subsequent study, we will use the appropriate

later enhanced. Regions with high snail density and towns with research scale to consider the above factors and perform a com-

an HH distribution pattern were mostly distributed in the Lishui prehensive analysis.

River Plain; the estuary of the Yuanshui River; XinQiang River,

Ethics statement

the tributary of the Yangtze River; and near Chenglingji port.

(2) There were space-time differences in the inﬂuence of natural

The study protocol was approved by the Resources and Envi-

factors on the spatio-temporal distribution of O. hupensis. Tem-

ronmental Science Data Center of the Chinese Academy of Sciences

porally, the comprehensive inﬂuence of natural factors on snails

and the Schistosomiasis Control Center of Hunan Province. Written

ﬁrst increased and then decreased. Natural factors played an

informed consent had also been obtained. No speciﬁc permits were

important role in predicting snail distribution in 2005, 2006,

required for the ﬁeld studies focusing on the O. hupensis as they did

2010 and 2011. Spatially, snail distribution decreased from the

not involve endangered or protected species.

northeast to the southwest. Snail information of more than 20

towns, such as YaZigang and YouGang, which are located along

Author summary

the Yuanshui River and on the west side of Lishui River which

are in the south of West Dongting Lake, was less affected by the Oncomelania hupensis is the intermediate only host of Schisto-

natural factors that were relatively more important in areas soma japonicum, which infect approximately 600 million people

around the outlet of the lake (Chenglingji). living environment worldwide, and its distribution is consistent

(3) The effects of natural factors on the spatio-temporal distri- with that of S. japonicum. The natural elements especially envi-

bution of O. hupensis had spatio-temporal heterogeneity. The ronment and climate elements such as water distribution, rainfall,

single most important factor is that snails require speciﬁc sur- land surface temperature LST, vegetation index are closely related

vival conditions, and extremes of these conditions will inhibit to the distribution of O. hupensis. Using correlation analysis and

the growth of snails. However, in the speciﬁc geographical envi- spatial regression analysis, we quantitatively analyzed the rela-

ronment of this study, snail distribution was the result of the tionship between distribution of O. hupensis and its inﬂuencing

comprehensive effects of multiple factors. factors, deeply revealed the complex spatial variation of relation-

Average annual rainfall, land surface temperature, NDVI, dis- ships between distribution of O. hupensis and its inﬂuencing factors

tance from water sources and history of snail distribution all with respect to spatial-temporal heterogeneity and explored the

had a signiﬁcant impact on the spatio-temporal distribution of leading factors.

O. hupensis. In particular, with the impact of the local geograph- The results will provide reference methods and theoretical a

ical environment, the effects of average annual rainfall, land basis for the schistosomiasis control.

surface temperature, and NDVI all had spatio-temporal hetero-

geneity. Both distance from water sources and historical snail Acknowledgements

information had stable, positive effects on O. hupensis. When

annual rainfall ranged from 802.9 mm–1192.2 mm, it may have This study was partially funded by the High Resolution Earth

had stimulatory or inhibitory effects on O. hupensis under differ- Observation Systems of National Science and Technology Major

ent local environments. When higher than 1192.2 mm, rainfall Projects (10-Y30B11-9001-14/16). The funders had no role in study

G. Cheng et al. / Acta Tropica 164 (2016) 194–207 207

design, data analysis, decision to publish, or preparation of the Qiu, J., Li, R.D., Xu, X.J., et al., 2014. Identifying determinants of Oncomelania

hupensis habitats and assessing the effects of environmental control strategies

manuscript, which exclude data collection and laboratory test. We

in the plain regions with the waterway network of china at the microscale. Int.

like to thank the editors and anonymous reviewers for their helpful

J. Environ. Res. Public Health 11, 6571–6585.

remarks. Remais, J., Bo, Z., Carlton, E.J., et al., 2009. Model approaches for estimating the

inﬂuence of time-varying socio-environmental factors on macroparasite

transmission in two endemic regions. Epidemics 1 (4), 213–220.

References

Robins, J.M., Wang, N., 2000. Inference for imputation estimators. Biometrika 87

(1), 87–91.

Akaike, H., 1974. A new look at the statistical model identiﬁcation. IEEE Trans. Rujirakul, K., So-In, C., 2015. P-PCC: parallel pearson correlation condition for

Autom. Control 19 (6), 716–723. robust cosmetic makeup face recognitions. Lect. Notes Electr. Eng. 339,

Anselin, L., 1995. Local indicators of spatial association—LISA. Geogr. Anal. 27 (2), 259–266.

93–115. Shan, C., Zhou, X., Zhu, H., 2014. The dynamics of growing islets and transmission of

Bowman, A.W., 1984. An alternative method of cross-validation for the smoothing schistosomiasis japonica in the yangtze river. Bull. Math. Biol. 76, 1194–1217.

of density estimate. Biometrika 71 (2), 353–360. Stensgaard, A.S., Utzinger, J., Vounatsou, P., et al., 2013. Large-scale determinants

Cao, C.L., Bao, Z.P., Li, S.Z., et al., 2015. Schistosomiasis in a migrating population in of intestinal schistosomiasis and intermediate host snail distribution across

the lake region of china and its potential impact on control operation. Acta africa: does climate matter? Acta Trop. 128 (2), 378–390.

Trop. 145, 88–92. Sun, Q., Peng, Z., Zhang, J., et al., 2011. Response of Oncomelania snail distribution

Cui, M., Zhou, J.X., Huang, B., 2012. Beneﬁt evaluation of wetlands resource with on land use in Sichuan, China. Afr. J. Biotechnol. 10 (63), 13835–13840.

different modes of protection and utilization in the Dongting Lake region. Wang, Y., Zhuang, D.F., 2015. A rapid monitoring and evaluation method of

Procedia Environ. Sci. 13 (10), 2–17. schistosomiasis based on spatial information technology. Int. J. Environ. Res.

Gray, D.J., Li, Y.S., Williams, G.M., et al., 2014. A multi-component integrated Public Health 12 (12), 15843–15859.

approach for the elimination of schistosomiasis in the People’s Republic of Wang, H.Y., Zong, H.E., Zhou, Y.B., et al., 2010. Observation on the living status of

China: design and baseline results of a 4-year cluster-randomised intervention Oncomelania hupensis in dongting lake area. Fudan Univ. J. Med. Sci. 37 (4),

trial. Int. J. Parasitol. 44, 659–668. 430–433.

Hong, Q., Zhou, X., Sun, L., et al., 2002. Impact of global warming on the Wang, W., Liang, Y.S., Hong, Q.B., et al., 2013. African schistosomiasis in mainland

transmission of schistosomiasis in China I. The hibernation and lethal china: risk of transmission and countermeasures to tackle the risk. Parasites

temperature of oncomelania hupensis in laboratory. Chin. J. Schistosomiasis Vectors 6 (1), 261–265.

Control 14, 192–195. Wang, Z., Lu, D.B., Guo, C.X., et al., 2014. Compatibility of schistosoma japonicum

Hong, Q., Zhou, X., Sun, L., et al., 2003. Impact of global warming on the from the hilly region and Oncomelania hupensis from the marshland region

transmission of schistosomiasis in china ii: the aestivation and lethal within anhui: china. Parasitol. Res. 113 (12), 4477–4484.

hyperthermy temperature of oncomelania hupensis in laboratory. Chin. J. Wheeler, D., Tiefelsdorf, M., 2005. Multicollinearity and correlation local

Schistosomiasis Control 15, 24–26. regression coefﬁcients in geographically weighted regression. J. Geogr. Syst. 7

Hu, Y., Zhang, Z., Chen, Y., et al., 2013. Spatial pattern of schistosomiasis in Xingzi, (2), 161–187.

Jiangxi province, China: the effects of environmental factors. Parasites Vectors Wu, X.H., Zhang, S.Q., Xu, X.J., et al., 2008. Effect of ﬂoods on the transmission of

6 (1), 1–8. schistosomiasis in the yangtze river valley: People’s Republic of China.

Hu, Y.X., Huang, J.L., Du, Y., et al., 2015a. Monitoring wetland vegetation pattern Parasitol. Int. 57 (3), 271–276.

response to water-level change resulting from the Three Gorges project in the Wu, J.Y., Zhou, Y.B., Li, L.H., et al., 2014. Identiﬁcation of optimum scopes of

two largest freshwater lakes of China. Ecol. Eng. 74, 274–285. environmental factors for snails using spatial analysis techniques in dongting

Hu, Y., Huang, J., Du, Y., et al., 2015b. Monitoring spatial and temporal dynamics of lake region: china. Parasites Vectors 7 (1), 1–12.

ﬂood regimes and their relation to wetland landscape patterns in dongting Wu, J.Y., Zhou, Y.B., Chen, Y., et al., 2015. Three gorges dam: impact of water level

lake from MODIS time-series imagery. Remote Sens. 7 (6), 7494–7520. changes on the density of schistosome-transmitting snail Oncomelania

Hu, Y., Ward, M.P., Xia, C.C., et al., 2016. Monitoring schistosomiasis risk in East hupensis in dongting lake area, China. PLoS Negl. Trop. Dis. 9 (6).

China over space and time using a Bayesian hierarchical modeling approach. Xie, Y.H., Yue, T., Chen, X.S., et al., 2014. The impact of Three Gorges dam on the

Sci. Rep. 6, 24173, http://dx.doi.org/10.1038/srep24173. downstream eco-hydrological environment and vegetation distribution of East

Ii, A.C.S., Vidt, M.E., Tuohy, C.J., et al., 2016. Quantitative analysis of Dongting Lake. Ecohydrology 8 (4), 738–746.

three-dimensional distribution and clustering of intramuscular fat in muscles Yang, G.J., Utzinger, J., Sun, L.P., et al., 2007. Effect of temperature on the

of the rotator cuff. Ann. Biomed. Eng. 44 (7), 1–10. development of schistosoma japonicum within Oncomelania hupensis: and

Kelejian, H.H., Prucha, I.R., 2010. Speciﬁcation and estimation of spatial hibernation of o. hupensis. Parasitol. Res. 100 (100), 695–700.

autoregressive models with autoregressive and heteroskedastic disturbances. Yang, G.J., Sun, L.P., Hong, Q.B., et al., 2009a. Application and perspective of climatic

J. Econom. 157 (1), 53–67. forecasting models related to schistosomiasis transmission. Chin. J.

Lai, X., Jiang, J., Huang, Q., 2013. Effects of the normal operation of the Three Gorges Schistosomiasis Control, 432–436.

reservoir on wetland inundation in Dongting Lake, China: a modelling study. Yang, K., Zhou, X.N., Wu, X.H., et al., 2009b. Landscape pattern analysis and

Hydrol. Sci. J. 58 (7), 1467–1477 (11). Bayesian modeling for predicting Oncomelania hupensis distribution in Eryuan

Lee, Y.H., Oh, Y.T., Ahn, H.C., et al., 2015. The prognostic value of the grey-to-white County, People’s Republic of China. Am. J. Trop. Med. Hyg. 81 (3), 416–423.

matter ratio in cardiac arrest patients treated with extracorporeal membrane Yang, G.J., Utzinger, J., Lv, S., et al., 2010. The regional network for asian

oxygenation. Resuscitation 99, 50–55. schistosomiasis and other helminth zoonoses (rnas(+)) target diseases in face

Liang, Y.S., Xiao, R.W., Dai, J.R., et al., 1996. Field observations on the survival and of climate change. Adv. Parasitol. 73, 101–135.

multiplication of oncomelania snails in the different latitude regions in china. Yang, K., Zhou, X.N., Jia, T.W., et al., 2015. Eco-social determinants of schistosoma

Chin. J. Schistosomiasis Control 5, 259–262. japonicum infection supported by multi-level modelling in eryuan county,

Liang, Y.S., Wei, W., Li, H.J., et al., 2012. The south-to-north water diversion project: People’s Republic of China. Acta Trop. 141 (Pt B), 391–398.

effect of the water diversion pattern on transmission of Oncomelania hupensis, Yin, H., Cai, Y., Duan, H., et al., 2013. Use of DGT and conventional methods to

the intermediate host of schistosoma japonicum in china. Parasites Vectors 5, predict sediment metal bioavailability to a ﬁeld inhabitant freshwater snail

398. (bellamya aeruginosa) from chinese eutrophic lakes. J. Hazard. Mater. 264 (2),

Liu, K., Xu, X., Chen, Y., et al., 2016. Analysis on inﬂuence of meteorological factors 184–194.

on Oncomelania density. Zhonghua Liu Xing Bing Xue Za Zhi 36 (11), Zhang, Z., Ong, S.H., Peng, W., et al., 2008. A model for the prediction of

1274–1278. Oncomelania hupensis in the lake and marshland regions, china. Parasitol. Int.

Lu, L., Yang, L., Liao, J., et al., 2014. Wetlands explain most in the genetic divergence 57 (2), 121–131.

pattern of Oncomelania hupensis. Infect. Genet. Evol. 27, 436–444. Zhang, Z., Bergquist, R., Chen, D., et al., 2013. Identiﬁcation of parasite-host

Madsen, H., Hung, N.M., 2015. Reprint of an overview of freshwater snails in asia habitats in anxiang county: hunan province, china based on multi-temporal

with main focus on vietnam. Acta Trop. 141, 372–384. china-brazil earth resources satellite (cbers) images. PLoS One 8 (7), e69447.

McCracken Stephanie L., Utilizing Landsat TM and OLI in Predicting Oncomelania Zhou, X., Yang, G.K., Wang, X., et al., 2008. Potential impact of climate change on

hupensis Habitats Around Poyang Lake Before and After Three Gorges Dam schistosomiasis transmission in china. Am. J. Trop. Med. Hyg. 78 (2), 188–194.

Completion. Honors Theses. 2014, Paper 268. http://aquila.usm.edu/honors Zhou, Y.B., Song, L., Jiang, Q.W., 2012. Factors impacting on progress towards

theses/268. elimination of transmission of schistosomiasis japonica in china. Parasites

Mcmanus, D.P., Gray, D.J., Ross, A.G., et al., 2011. Schistosomiasis research in the Vectors 5 (6), 1–7.

dongting lake region and its impact on local and national treatment and Zhou, X.N., 2012. Prioritizing research for one health—one world. Infect. Dis.

control in china. PLoS Negl. Trop. Dis. 5, 639. Poverty 1 (1), 1–5.

O’Sullivan, D., 2010. Geographically weighted regression: the analysis of spatially Zhu, H.R., Liu, L., Zhou, X.N., et al., 2015. Ecological model to predict potential

varying relationships, by a S. Fotheringham, C. Brunsdon, and M. Charlton. habitats of Oncomelania hupensis, the intermediate host of schistosoma

Geogr. Anal. 35 (3), 272–275. japonicum in the mountainous regions, China. PLoS Negl. Trop. Dis. 9 (8).

Pistocchi, Alberto, 2014. 6. Spatial statistics and neighborhood modeling in GIS. In: Zou, L., Ruan, S., 2015. Schistosomiasis transmission and control in china. Acta

GIS Based Chemical Fate Modeling: Principles and Applications. John Wiley & Trop. 143, 51–57.

Sons, Inc., pp. 149–170.