Probing the Spatial Cluster of Meriones Unguiculatus Using the Nest Flea Index Based on GIS Technology

Home , Dingbian County, Erenhot, Hanggin Banner, Huade County, Kangbao County, Otog Banner

Accepted Manuscript

Title: Probing the spatial cluster of Meriones unguiculatus using the nest ﬂea index based on GIS Technology

Author: Dafang Zhuang Haiwen Du Yong Wang Xiaosan Jiang Xianming Shi Dong Yan

PII: S0001-706X(16)30182-6 DOI: http://dx.doi.org/doi:10.1016/j.actatropica.2016.08.007 Reference: ACTROP 4009

To appear in: Acta Tropica

Received date: 14-4-2016 Revised date: 3-8-2016 Accepted date: 6-8-2016

Please cite this article as: Zhuang, Dafang, Du, Haiwen, Wang, Yong, Jiang, Xiaosan, Shi, Xianming, Yan, Dong, Probing the spatial cluster of Meriones unguiculatus using the nest ﬂea index based on GIS Technology.Acta Tropica http://dx.doi.org/10.1016/j.actatropica.2016.08.007

This is a PDF ﬁle of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its ﬁnal form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Probing the spatial cluster of Meriones unguiculatus using the nest flea index based on GIS Technology

Dafang Zhuang1, Haiwen Du2, Yong Wang1*, Xiaosan Jiang2, Xianming Shi3, Dong Yan3

1 State Key Laboratory of Resources and Environmental Information Systems, Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China. 2 College of Resources and Environmental Science, Nanjing Agricultural University, Nanjing, China. 3 Institute of Plague Prevention and Control of Hebei Province, Zhangjiakou, China.

*Correspondence: Yong Wang, [email protected]

Abstract: The nest flea index of Meriones unguiculatus is a critical indicator for the prevention and control of plague, which can be used not only to detect the spatial and temporal distributions of Meriones unguiculatus, but also to reveal its cluster rule. This study used global spatial autocorrelation and spatial hot spot detection methods to describe the relationship between different years and the autocorrelation coefficient of nest flea indexes; it also used a spatial detection method and GIS technology to detect the spatial gathered hot spot of Meriones unguiculatus in the epidemic areas. The results of this study showed that (1) there were statistically significant spatial autocorrelations in the nest flea indexes in 2006, 2012, 2013 and 2014. (2) Most of the distribution patterns of Meriones unguiculatus were statistically significant clusters of high values. (3) There were some typical hot spot regions of plague distributed along the Inner Mongolia plateau, north of China. (4) The hot spot regions of plague were gradually stabilized after increasing and decreasing repeatedly. Generally speaking, the number of hot spot regions showed an accelerated increase from 2005 to 2007, decreased slowly from 2007 to 2008, rapidly increased again after decreasing slowly from 2008 to 2010, showed an accelerated decrease from 2010 to 2011, and ultimately were stabilized after rapidly increasing again from 2011 to 2014. (5) The migration period of the hot spot regions was 2-3 years. The epidemic area of plague moved from southwest to east during 2005, 2007, 2008 and 2010, from east to southwest during 2007 and 2008, from east to west during 2010 and 2011, and from Midwest to east during 2011 and 2014. (6) Effective factors, such as temperature, rainfall, DEM, host density, and NDVI, can affect the spatial cluster of Meriones unguiculatus. The results of this study have important implications for exploring the temporal and spatial distribution law and distribution of the hot spot regions of plague, which can reduce the risk of plague, help support the decision making process for the control and prevention of plague, and form a valuable application for plague research.

Keywords：Spatial autocorrelation; GIS; Nest fleas; Meriones unguiculatus

1、 Introduction

Plague is a natural focal disease and is the first notifiable disease. New epidemic areas have continually been discovered, and the intensity of plague has constantly been enhanced around the world since the 1990s. Plague is a zoonotic, insect-borne, infectious disease with a natural focal infectious endemic that is caused by Yersinia pestis and is spread through the bite of fleas that are naturally infected with plague

(Davis et al., 2006). The main mode of glandular plague of transmission is from rodents to fleas to people. The flea is one of the most important groups of medical insects; fleas bite people and livestock to feed on their blood, causing direct damage and the spread of plague and endemic typhus (Nakazawa et al., 2007). Several fleas spread plague bacillus widely between hosts; fleas of higher richness can suppress the immune reactions in hosts, thus the susceptibility of hosts is increased in favour of plague bacillus, which breed to a high density in the hosts, and then the probability of infecting new hosts rises again. The natural epidemic focus of Meriones unguiculatus was discovered in 1954 (Fan et al., 2010; Williams 1974.). Previous research has mainly focused on the general situation of plague epidemic in each epidemic focus (Parmenter et al., 1999), the relationship between plague and environment factors (Zhang et al., 2007), the popularity of plague vector fleas in epidemiology, the relationship between the quantity of vector fleas and meteorological factors (Pham et al., 2009; Li 1999; Zhong 2000) etc. There was a lack of research on the association of the plague epidemic and the quantity of vector fleas with the spatial distribution characteristics of the natural epidemic focus, as well as geographical factors that are associated with the spread of plague, which are not independently distributed in space, and have a complex spatial-temporal topology. The mathematical statistics, ecological and epidemiological methods are unable to analyse the complex spatial-temporal relationships between these geographical factors; thus, it is imperative to solve this problem using a new technology. Geographical Information Systems (GIS) is based on the time-space relationship between surface features using a spatial-temporal analysis of a geographic model with

3 the support of computer. It also analyses multiple dynamic spatial-temporal geographic data, and the new technical system is established for research and decision services (Yang et al., 2010). GIS is a new scientific study method because it has powerful spatial data management and analytical capabilities. Its application to the plague indicates that GIS has extended to more fields than ever before. Yang et al. (2000) analysed the spatial-temporal distribution law and characteristics in the natural foci of China since 1840. Qian et al. (2014) established the plague prevention and control GIS to predict the tendency of plague development using a spatial-temporal model. Eisen et al. (2007a; 2007b) determined the location of plague relative to the human population using the GIS model of a geographical landscape and created the risk map of plague and the established Habitat-Suitability model. Lieshout et al. (2004) performed a spatial clustering analysis on ecological types of plague bacillus in Arizona grassland. Previously, research on the spatial distribution of the plague vector flea, the regularity of the plague vector flea distribution and the spatial aggregation of the plague vector flea by GIS was uncommon. This study combines spatial statistical analysis research methods using spatial analysis tools to investigate the Meriones unguiculatus nest flea indexes and explore the regularity and aggregation of the nest flea index distribution, which provided a new method and technology for analysing plague and assisted in decisions regarding plague prevention and control.

2、 Materials and methods

2.1 The general situation of the survey region The Meriones unguiculatus natural epidemic focus is mainly distributed in the Inner Mongolian Plateau, also known as the Inner Mongolian Plateau Meriones unguiculatus epidemic focus (Thomas et al., 2004). The Meriones unguiculatus natural epidemic focus is located in the southwest edge of the ordos plateau, and the boundary is the Huanghe River, which divided the natural epidemic focus into two separate natural epidemic foci: the north focus is the Wulanchabu Plateau medium temperature desert steppe and the south focus is ordos Warm-temperate desert steppe.

The natural epidemic foci is located from latitude 37°34′N-45°00′N and longitude

106°30′E-114°50′E, as illustrated by Fig. 1.(Eisen et al., 2007c), with an elevation of

1,000-1,600 metres, a stratiform upland plain topography, and a semi-desert landscape. The Chahar hills dry steppe is mainly low mountains and hills, and dotted with marsh land, valley floor and plain in the hills; the area of the natural epidemic focus is

13,4803 km2. It accounts for 10.2 percent of twelve types of natural epidemic foci in

China. This is one of the more active natural epidemic foci. There are a total of 29 counties (District, Banner and City). An analysis of the variation in the nest flea indexes has an important value for monitoring and controlling flea epidemics and preventing plague outbreak. 2.2 Data sources and processing methods The data of the epidemic focus of the Meriones unguiculatus administrative map extracted by the Chinese Academy of Sciences Resources and the Environment Scientific Information Center provided a 1:50,000 scale Chinese county electronic map, climate and vegetation data (annual average temperature in the study area, digital elevation model (DEM), rainfall, normalized difference vegetation index (NDVI) and 1 km resolution raster data from 2005 to 2010).The quantity of Meriones unguiculatus nest fleas in the sentinel surveillance sites from 2005 to 2014 and host density data of the study area from 2005-2010 were provided by the China Center for Disease Control. Using Moran’s I, Anselin Local Moran’s I, Getis-Ord and spatial regression which are based on the spatial statistical analysis module of ArcGIS10.2, the aggregated spatial-temporal distribution characteristics of the nest flea indexes were analysed and the statistical significance was determined. In the Meriones unguiculatus natural epidemic focus, the quantity of Meriones unguiculatus nest fleas was investigated from April to July and October to November from 2005 to 2014. The data from April to July and October to November of every year were merged into one set, and the flea indexes of the nest fleas were quantified. The nest flea indexes were investigated with an insect vector monitoring method used in the pilot programmes of every major national monitoring point for plague, and calculated using the following formula: 5

VNFI = TNNF(1) TNM where VNFI is value of the nest flea index, TNNF is the total number of nest fleas, and TNM is the total number of Meriones unguiculatus. 2.3 research method The spatial autocorrelation means that the statistical correlation between the attribute values of a geographical region is distributed in a different spatial position, such that the closer the distance, the stronger the correlation between the two values. The spatial autocorrelation is measured through the spatial autocorrelation coefficient, which examines a property of spatial matter, regardless of whether it cross-distributes with “High-High” (“Low-Low”) or “High-Low” (“Low-High”). Elements of the geographical region between the spatial relationships are based on the premise of the spatial autocorrelation and give a quantitative expression of the spatial weight matrix (Seya et al., 2013), which is automatically generated from the Inverse Distance in this study. The spatial autocorrelation includes a global autocorrelation and local autocorrelation; the local spatial autocorrelation analysis is the necessary complement to the global spatial autocorrelation analysis, which includes the Cluster and Outlier Analysis and Hot Spot Analysis. The global spatial autocorrelation cannot differentiate the spatial clustering regions of either high or low values, cannot measure the local aggregation between the observed value and the data surrounding the observed value in each spatial unit position, and cannot determine the location of the studied phenomenon in the clustering regions. Therefore, Moran’s I indexes of spatial autocorrelation were used to analyse whether the distribution of nest fleas indexes in the Meriones unguiculatus natural epidemic focus is aggregated, explore the extent of aggregation and the patterns of spatial distribution at the County (District, Banner, City) level from 2005 to 2014. Anselin local Moran’s I of the Cluster and Outlier Analysis of the local spatial autocorrelation analysis was used to obtain nest flea indexes that were higher or lower than the spatial clustering model of the regions,

∗ while the Hot Spot Analysis and Getis-Ord G 푖 local statistics can identify hot spot regions with high nest flea indexes (Table 2).

Trend surface is created by applying polynomial regression which is used to fit the least square surface and the input points. Dependent variable and independent variable are individually the nest index and the latitude, longitude on the spatial location. In order to build the trends, two degree polynomial fitting interpolation is applied to achieve the fitting of the least square surface for the nest flea index of Meriones unguiculatus. Spatial data has the effect correlation among the locations, in case of a significant spatial autocorrelation; Anselin (1999) proposes Ordinary Least Square estimation (OLS), Spatial Lag Model (SLM) and Spatial Error Models (SEM), depending on the source of spatial autocorrelation. Dependent variable was the nest index, many of the independent variable: DEM, annual average temperature, rainfall, NDVI and host density. The formal test for the significance of and is conducted by means of the Lagrange Multiplier test (Anselin, 1999).

3、 Results and analysis

3.1 Global Spatial Autocorrelation Analysis The spatial autocorrelation analysis module is used to test the random distribution hypothesis by determining the Global Moran’s I statistic of the Meriones unguiculatus nest flea indexes from 2005 to 2014 with ArcGIS10.2. The results (Table 3) show that the spatial autocorrelation coefficients of the Meriones unguiculatus nest flea indexes from 2005 to 2014 were -0.0338, 0.2303, -0.0097, -0.0288, -0.0205, 0.1174, -0.0396, 0.4801, 0.2701 and 0.2985, respectively. The Moran’s I statistics were greater than 0 in 2006, 2010, 2012, 2013 and 2014, and the z-scores were greater than 1.96 in 2006, 2012, 2013 and 2014. The results showed that the nest flea indexes exhibited spatial clustering. The z-scores were between -1.96 and 1.96 in the other years and show a random distribution, suggesting that there is no correlation; the p-values were less than 0.05 in 2006, 2012, 2013 and 2014. There were obvious spatial autocorrelations of the nest flea indexes in the Meriones unguiculatus natural epidemic focus in 2006, 2012, 2013 and 2014. The spatial autocorrelations in the other years were randomly and sporadically distributed throughout the region and were not statistically

7 significant. Moran’s I value of the nest flea indexes was sharply reduced after 2012; the spatial clustering of the nest flea index in 2012 was stronger than those in the subsequent years. 3.2 Local Spatial Autocorrelation Analysis 3.2.1 Cluster and outlier analysis Cluster analysis and outlier analysis were used to analyse the Meriones unguiculatus nest flea indexes from 2005 to 2014, and the result shows that the distribution patterns of the Meriones unguiculatus nest flea indexes and the significance levels for each County (District, Banner, City) were greater than 0.05, except for 2007. Moreover, the cluster and outlier maps showed the distribution pattern of the Meriones unguiculatus nest flea indexes (Fig 1). Spatial distribution exists in three phenomena: High-High (HH), High-Low (HL) and Low-High (LH). The HH regions appeared in Dingbian County in 2005 and 2008; in Kangbao County, Taibus Banner and Zhengxiangbai Banner in 2006; in Huade County in 2010; in Shangdu County, Huade County, Boarder Yellow Banner, Zhengxiangbai Banner and Taibus Banner in 2012; in Huade County and Taibus Banner in 2013; and in Boarder Yellow Banner, Huade County and Taibus Banner in 2014. The HL regions appeared in 2008 in Kangbao County, in 2009 in Sonid Right Banner and in 2011 in Otog Banner. The LH regions appeared in 2012 in Kangbao County. There were two cases of outliers that only appeared in some regions. One was that the nest flea indexes were larger than the surrounding regions, and the other was that the nest flea indexes were smaller than the surrounding regions. We can infer that the characteristics of the geographical environment in these regions were different than the adjacent regions. These regions with specific distribution patterns require additional studies of the growth and decline of the environment of the nest flea indexes. The differences in the other regions with nest flea indexes were not statistically significant and exhibited a random distribution. The HH region of nest flea indexes was distributed in Dingbian County in 2005 and 2008 and in Boarder Yellow Banner, Huade County and Taibus Banner in 2012 and 2014; thus it can be observed that the important time nodes of the nest flea 8 indexes have changed in 2005, 2006, 2010, 2012, 2013 and 2014. Therefore, we can focus on analysing the growth and decline of the environment of the nest flea indexes in these years in the study area, particularly in Dingbian County, Boarder Yellow Banner, Huade County, Taibus Banner and the regions in which the nest flea index patterns have changed. 3.2.2 Hot spot analysis The hot spot regions of the nest flea indexes were relatively concentrated from

* 2005 to 2014 (Fig 2). The significance levels of each Gi index were greater than

0.05. Yanchi County and Dingbian County were hot spot regions of the nest flea index distributions in 2005, with 7.41% of the total number of regions. Zhengxiangbai Banner, Taibus Banner, Kangbao County, Boarder Yellow Banner and Huade County were hot spot regions of the nest flea index distributions in 2006, with 18.52% of the total number of regions. Zhengxiangbai Banner, Taibus Banner and Kangbao County were hot spot regions of the nest flea index distributions in 2007, with 11.11% of the total number of regions. Yanchi County, Dingbian County and Otog Front Banner were hot spot regions of the nest flea index distributions in 2008, with 11.11% of the total number of regions. Sonid Right Banner was hot spot region of nest flea index distributions in 2009, with 3.70% of the total number of regions. Kangbao County, Boarder Yellow Banner, Huade County, Shangdu County and Chahar Right Back Banner were hot spot regions of the nest flea index distributions in 2010, with 18.52% of the total number of regions. Wushen County, Otog Front Banner, Otog Banner and Hanggin Banner were hot spot regions of the nest flea index distributions in 2011, with 14.81% of the total number of regions. Zhengxiangbai Banner, Taibus Banner, Kangbao County, Boarder Yellow Banner, Huade County, Shangdu County and Chahar Right Back Banner were hot spot regions of the nest flea index distributions in 2012, with 25.93% of the total number of regions. Zhengxiangbai Banner, Taibus Banner, Kangbao County, Boarder Yellow Banner, Huade County and Chahar Right Back Banner were hot spot regions of the nest flea index distributions in 2013, with 22.22% of the total number of regions. Zhengxiangbai Banner, Taibus Banner,

Kangbao County, Boarder Yellow Banner, Huade County, Shangdu County and Chahar Right Back Banner were hot spot regions of the nest flea index distributions in 2014, with 25.00% of the total number of regions. Urad Front Banner was cold spot region of the nest flea index distributions in 2012 and 2013, with 3.70% of the total number of regions. The nest flea indexes were aggregated in Urad Front Banner and formed low regions. The percentage of hot spot regions of the nest flea indexes decreased after 2006 and increased again in 2009; the percentage of hot spot regions of the nest flea indexes decreased, but with less fluctuation after 2012. Hanggin Banner in Erdos city of the Inner Mongolia Autonomous Region became a new natural epidemic focus in which Yersinia pestis was isolated in 2014; the nest flea index of the Hanggin Banner in 2011 was a hot spot region, as shown in Fig 2. Otog Banner, Wushen County, Urad Middle Banner and Jiuyuen District around Hanggin Banner had relatively high nest flea indexes and formed a spatial cluster distribution; therefore, we can focus our attention on these hot spot regions and their surrounding regions. Kangbao County was a hot spot region of the nest flea index distributions in 2006, 2007, 2010, 2012, 2013 and 2014. Zhengxiangbai Banner and Taibus Banner were hot spot regions of the nest flea index distributions in 2006, 2007, 2012, 2013, 2014. Boarder Yellow Banner and Huade County were hot spot regions of the nest flea index distributions in 2006, 2010, 2012, 2013, 2014; Chahar Right Back Banner was a hot spot region of the nest flea index distributions in 2010, 2012, 2013, 2014. All of these regions are priority regions for plague control and prevention. The regional positions of the hot spot regions were basically the same in 2012 and 2013 and included Chahar Right Back Banner, Boarder Yellow Banner, Huade County, Zhengxiangbai Banner, Taibus Banner, Kangbao County. Only Urad Front Banner

* was a hot spot region of the nest flea index distributions in 2012 and 2013. The Gi value of the nest flea indexes was relatively high in the northeast corner from 2005 to 2014. The important time nodes of the nest flea indexes have changed in 2005, 2007, 2008, 2009, 2010, 2011, and 2014. We can focus on analysing the growth and decline

10 of the environment of the nest flea indexes in these years in the study area, particularly in Kangbao County, Zhengxiangbai Banner, Taibus Banner, Boarder Yellow Banner, Huade County, and Chahar Right Back Banner. 3.3 Trend Analysis

By examining the global trends through trend analysis, the tendency chart shows that every long string represents the value and position of the average nest flea indexes for each County (District, Banner and City) over 10 years in the Meriones unguiculatus natural epidemic focus (Fig 4). As shown in Fig 4, the green trend line projected onto the east-west direction gradually declined from the east to the west; in the north-south direction, the blue trend line is U-shaped. The nest flea indexes of Meriones unguiculatus natural epidemic focus appeared to show a decreasing trend from east to west; in the north south direction, there is a low-side among the high phenomenon. The influence of the trend is gradually strengthened from the centre of the epidemic focus to the various boundaries; the minimum value appeared at the centre of the epidemic focus, and the maximum value appeared on the north of the epidemic focus.

3.4 Spatial Regression Analysis In modelling spatial data, which consists of spatial dependency, the dependencies caused spatial autocorrelation. Testing the spatial dependency in regression models is very important because ignored it will lead to inefficient estimates and less valid conclusions in Table 4. There are two regression models for modeling the spatial dependency of data which is ordinary linear regression model (OLS) and spatial lag model (SLM) in Table 5. One of the statistical tests used to determine two types of model are the Lagrange Multiplier (LM). However, the LM test often experience uncertainty in summing the two types of models. Therefore, this study examined Robust LM to solve these problems and used to modeling relationships between possible impacting factors and the distribution of nest flea index, the result showed that testing with the LM test statistic, both LM lag and error are equally significant results so that is not yet clear which model will be used. Furthermore, the use of

Robust LM test statistics to identify the SLM is more appropriate to be used in modeling the relationships between impacting factors and the distribution of nest flea index except 2007; the OLS is appropriate to be used in modeling in 2007. Host density, temperature, DEM, NDVI and rainfall has a significant influence on the distribution of nest flea index from 2005 to 2010. The results of relationships between impacting factors and the distribution of nest flea index between 2005 and 2010 in the natural epidemic focus of Meriones unguiculatus were summarized in Table 6 and 7. The results showed that the nest flea index exist the positive spatial autocorrelation between one region and other regions.

The nest flea index increased with host density (β=3.0362,P=0.0352), temperature (β=0.5953,P=0.0072) and DEM (β=0.592,P=0.0324), the nest flea index decreased with increasing NDVI (β= -1.4333,P= 0.0001); No statistically significant dissimilarity (P>0.05) was observed between the nest flea index and rainfall in 2005.The nest flea index decreased with increasing temperature (β= -0.7878,P= 0.001) and DEM (β= -0.6547,P= 0.0113), the nest flea index increased with NDVI (β= 0.8342,P=0.0335) and rainfall (β=1.1528,P= 0.0001); No statistically significant dissimilarity was observed between the nest flea index and host density in 2006. The nest flea index decreased with increasing host density (β=-3.9906,P=0.0434) and temperature (β= -0.9167,P= 0.0001), the nest flea index increased with NDVI (β= 1.4958, P= 0.0001) and rainfall (β= 0.7544,P= 0.0029); No statistically significant dissimilarity was observed between the nest flea index and DEM in 2007. The nest flea index decreased with increasing temperature (β= -0.4398,P= 0.0481)，increased with rainfall (β= 0.68,P= 0.0069), host density; No statistically significant dissimilarity was observed between the nest flea index and host density, DEM and NDVI in 2009.No statistically significant dissimilarity was observed between the nest flea index and Host density, temperature, DEM, NDVI and Rainfall in 2008 and 2010.

4、 Discussion

4.1 Regularities in the distribution and hot spot regions Kangbao County was a hot spot region of the nest flea indexes from 2006 to 2007, in

2010, and from 2012 to 2014; Zhengxiangbai Banner and Taibus Banner were hot spot regions from 2006 to 2007 and from 2012 to 2014. Taibus Banner was an HH region in 2006 and from 2012 to 2014. Chahar Right Back Banner was a hot spot region in 2010 and from 2012 to 2014. Boarder Yellow Banner and Huade County were hot spot regions in 2006, 2010 and from 2012 to 2014. Boarder Yellow Banner was an HH region in 2012 and 2014, and Huade County was an HH region in 2010 and from 2012 to 2014. The outbreak and prevalence of plague in animals increased on three occasions from 2002 to 2003 and in 2005 in Kangbao County; the density of the main host Meriones unguiculatus increased yearly, and there was a plague outbreak from 2001 to 2005. The density decreased for several years from 2006 to 2010, and gradually increased in 2011, with an apparent increase in nest flea indexes (Ning et al., 2012). Vegetation in Kangbao County has further increased with climate warming and an increase in rainfall in recent years; the changes in these conditions cause rodents to multiply rapidly, and nest flea reproduction is also accelerated with climate warming and increased rainfall. Due to global warming, the temperature of

Kangbao County rose gradually, particularly in the winter. Climate warming is convenient for rodents, which are the epidemic host of Meriones unguiculatus. Rodents do not hibernate and exhibit frequent activity, thus their range of activity expanded, reproduction was accelerated, the viability of young rodents increased, and their population has increased (Dong et al., 2009). Most scholars agree that the plague epidemic was spread into Kangbao pasture by Zhengxiangbai Banner and Huade County of the Inner Mongolia Autonomous Region; Erenhot City, Sonid Right Banner and Siziwang Banner are located in the northeast corner of the Meriones unguiculatus natural epidemic focus and were High-incidence regions. The plague epidemic occurred in Erenhot City for 7 years, Sonid Right Banner and Siziwang Banner for 5 years, Boarder Yellow Banner and Huade County for 3 years, Zhengxiangbai Banner, Shangdu County, Chahar Right Back Banner and Darhan Muminggan Joint Banner for 2 years, and Kangbao County, Sonid Left Banner, Taibus Banner for 1 year. The plague epidemic of 2004 in Chahar Right Back Banner and Shangdu County and the plague epidemic of Ulanqab (Huade County, Shangdu County, Siziwang Banner and 13

Chahar Right Back Banner) are more serious (Fan et al., 2012). Zhengxiangbai Banner, Taibus Banner and Boarder Yellow Banner are linked with Kangbao County and Huade County. Some of the landscape transitions from a typical steppe to desert steppe, some of the hill steppe is cultivated, and the soil trends towards desertification; thus, large numbers of Meriones unguiculatus were invading the habitat at a high density. Suxi and Erlian are located along the Ji-Er Railway and have the appropriate environmental conditions for Meriones unguiculatus, which included each side of railroad bed and some azonal habitat. The continuous plague epidemic of Erenhot City, Sonid Right Banner, Siziwang Banner, Darhan Muminggan Joint Banner and Huade

County may be caused by several factors. First, those regions have a large area with a suitable habitat for Meriones unguiculatus and provided good conditions for a long-term plague epidemic. Second, this is a wide area of epidemic focus, and there are plenty of rodents for the plague epidemic to spread widely. Third, the diversity of host animals and vectors can produce a plague epidemic cycle in various animals. In my opinion, the rodent plague will endure for a long time in these regions that have not done much for the ecological landscape, and they will also maintain a continuous

Meriones unguiculatus epidemic (Liu et al., 1986). 4.2 Related influencing factors

1) The season and spatial distribution produce spatial clustering of the nest flea indexes. Climatic variation can directly affect the reproduction of the vector and change the habitat of the vectors or hosts. The hot spot regions of the nest flea indexes were concentrated in the northeast corner of the Meriones unguiculatus natural epidemic focus. The Inner Mongolian Plateau area belongs to a mainland monsoon-type climate with the principal characteristics of a dry and windy spring, a warm and rainy summer, a pleasantly cool autumn with a large temperature difference, and a long, dry and cold winter. Kangbao County has typical continental climate, with large temperature differences between day and night (Liu et al., 2005). The quantity of nest fleas peaks in the cold season because hosts spend more time in the nest and reduce their time outside the nest, which favours blood feeding and reproduction of the nest flea. The large numbers of nest fleas are sectioned such that rodents can 14 survive. Meriones unguiculatus often centralized their habitat in autumn and winter, where the high density of rodents provided a favourable habitat for fleas during the low temperatures and increased nest time in the winter, and some principal species can carry Yersinia pestis throughout the winter. Nest fleas are not influenced by host capture, the state of the host or other human factors (Liu et al., 1986). The nests of most of infected plague fleas were removed during a rodent plague epidemic, but a minority of infected plague fleas were dug up from abandoned nests after the plague epidemic, thus the nests play an important role in infection and preserving the infected fleas. Although the density of Meriones unguiculatus was tentatively increased in the local area and some infected plague fleas were preserved in nests in regions that Meriones unguiculatus does not inhabit, this phenomenon can cause a new plague epidemic. Therefore, fleas play an important role in conserving Yersinia pestis and causing plague epidemic.

2) The number of Meriones unguiculatus, rainfall capacity and temperature influence the spatial clustering of nest fleas. Plague is natural focus infectious disease, and climatic variations and the ecological environment changes have deep impacts on the hosts and vector fleas. A moist environment is likely to increase seed yield and vegetation growth, and the rodents that are hosts of parasitic fleas have plentiful food resources (Li et al., 2013). The annual rainfall of the previous and current years and warm climate are beneficial to the wide spread in the Inner Mongolia Autonomous Region. Increased rainfall will provide abundant food for the hosts, and warm conditions are good for hibernating hosts to awaken and look for food earlier and hibernate later. The scope of the hosts’ activities expanded in each epidemic focus, as fleas breed and Yersinia pestis lives through the winter (Li et al., 2005). Rainfall played an important role in the primary food sources of rodents, which include seeds in the monsoon climate and tropical grassland, and the temperature directly influences the quantity of fleas (Zhang et al., 2003). An annual mean temperature of 1-6 degrees centigrade, an annual rainfall of 250-450 millimetres and a progressive decrease in rainfall were observed in the border regions of the Inner Mongolian plateau. Plants were growing poorly in the most suitable habitat, such that the fecundity of Meriones 15 unguiculatus was depressed; it even appeared that migration and density of the major host were reduced, producing quiescent conditions of the plague epidemic over a short period of time. However, the host density quickly increased as the rainfall increased and vegetation was reconstructed. When the quantity of Meriones unguiculatus surges, most of the nests were new and diluted, and the internal and external environment were convenient for breeding fleas. The quantitative variations in rodents and nests had a significant negative correlation (Ning et al., 2012).

3) Natural environmental factors influence the spatial clustering of nest fleas. The border regions of the Inner Mongolian plateau can be divided into typical steppe and desert steppe (Liu et al., 2005Error! Bookmark not defined.). Sparse and undersized

Chinese wild rye, Cold pole, Cleistogenes squarrosa, Stipa krylovii and Caragana microphylla are the dominant vegetation species in Kangbao County, which belongs to the temperate steppe vegetational form. The soil of Kangbao County is a typical chestnut soil and carbonate chestnut soil; the soil is thin, and its structure is soft. Some regions have a large occupancy, as some of the soil is not ploughed and some of the stub land is not planted (abandoned field) annually. However, some regions had planted rapeseed, flax, wheat, etc., in previous years, which are Meriones unguiculatus’ favourite living environment. This caused the host density to remain high and allowed the vector fleas to rapidly multiply as the host density increased. From the perspective of organizational ecology, the strength of the Meriones unguiculatus plague epidemic is the result of multiple factors (biotic and abiotic factors).

4) The spatial clustering of body fleas compared with the spatial clustering of nest fleas. The number of body fleas peaks in the warm season, and the number of nest fleas peaks in the cold season. The quantity of body fleas decreased as the quantity of nest fleas was increasing, such that the quantity of nest fleas was well above that of the body fleas. Although the variations in body fleas and nest fleas reflected the quantity and dynamics of parasitic fleas from two different perspectives, body fleas are influenced by the capture of their hosts, the state of the host, and other human factors, whereas nest fleas were not affected. Different seasons not only influence the 16 spatial distribution and quantity of Meriones unguiculatus but also influence the spatial distribution and quantity of fleas (Liu et al., 1986).

5、 Conclusions

The Meriones unguiculatus natural epidemic focus was studied by analysing the regularity of the distributions of the nest flea indexes from 2005 to 2014 using a new GIS method, quantitatively describing the spatial distribution of the nest flea indexes, and probing the hot spot regions of the nest flea indexes. This method compensated for the lack of ecological, epidemiological and mathematical statistics method. 1) In the Meriones unguiculatus natural epidemic focus, the nest flea indexes have shown a decreasing trend from east to west, and the influence of the trend gradually increased from the centre of the natural epidemic focus to the northeast and southwest. The maximum value appeared in the northeast of the natural epidemic focus, and the minimum value appeared in the centre of the natural epidemic focus. 2) There were significant spatial autocorrelation results for the nest flea indexes in 2006, 2012, 2013 and 2014; the distribution patterns of the nest flea index were mainly in the high value cluster. 3) The high-occurrence hot spot regions of the nest flea indexes focus on the northeast region of the natural epidemic focus, and aggregation is increasingly apparent over time. 4) Temperature, rainfall, NDVI, host density and DEM affect the spatial clustering of the nest flea indexes. The results of this study provide clues for additional spatial modelling analyses and lay a theoretical foundation for a spatial regression analysis of the epidemic situation, environment and other factors. The results of this study have important implications for reducing the risk of flea outbreaks in the natural epidemic focus, predicting plague epidemics, and promptly handling the epidemic area.

Ethics statement

The study protocol was approved by the Ethics Review Committee of the Inner

Mongolia Autonomous Regional Center for Disease Control and Prevention, Hohhot, China, the Ningxia Hui Autonomous Regional Center for Disease Control and Prevention, Yinchuan, China, Shaanxi Provincial Center for Disease Control and Prevention, Xi’an, China and Hebei Provincial Center for Disease Control and Prevention, Shijiazhuang, China. Written informed consent had also been obtained. No specific permits were required for the field studies focusing on the Meriones unguiculatus’ nest fleas as they did not involve endangered or protected species.

Acknowledgements

This study was partially funded by the High Resolution Earth Observation Systems of National Science and Technology Major Projects (10-Y30B11-9001-14/16). The funders had no role in study design, data analysis, decision to publish, or preparation of the manuscript, which exclude data collection and laboratory test. We like to thank the editors and anonymous reviewers for their helpful remarks.

References

Anselin L., 1999. Short Course on “Spatial Econometrics”, The World Bank, Washington,

DC.

Barbosa, D.S., Belo, V.S., Rangel, M.E.S., Werneck, G.L., 2014. Spatial analysis for identification of priority areas for surveillance and control in a visceral leishmaniasis endemic area in Brazil. Acta Tropica, 131(1), 56-62.

Davis,S., Makundi, R.R., Leirs, H., 2006. Demographic and spatio-temporal variation in human plague at a persistent focus in Tanzania. Acta Tropica, 100(1-2), 133-141.

Dong, G.R., Guo-Yi, D.U., 2009. Analysis of epidemic characteristics of mice plague at

Kangbao county of Hebei province. Chinese Journal of Vector Biology & Control.

Eisen, R.J., Enscore, R.E., Biggerstaff, B.J., Reynolds, P.J., Paul, E., Ted, B., et al., 2007a.

Human plague in the Southwestern United States, 1957-2004: spatial models of elevated risk of human exposure to Yersinia pestis. Journal of Medical Entomology, 44(3), 530-7.

Eisen, R.J., Glass, G.E., Lars, E., James, C., Enscore, R.E., Paul, E., et al., 2007b. A spatial model of shared risk for plague and Hantavirus pulmonary syndrome in the Southwestern United

States. American Journal of Tropical Medicine & Hygiene, 77(6), 999-1004.

Eisen, R.J., Reynolds, P.J., Ettestad, P., Brown, T., Enscore, R.E., Biggerstaff, B.J., et al.,

2007c. Residence-linked human plague in New Mexico: a habitat-suitability model. American

Journal of Tropical Medicine & Hygiene, 77(1), 121-5.

Fan M.G., Zhang Z.B., Liu J., 2010. Analysis of epidemic characteristics of Inner Mongolia

Meriones unguiculatus Plague in recent years. Chinese Journal of Control of Endemic Diseases, 4,

274-275.

Fan, M.G., Na, T.A., 2012. Analysis of the test results of animal plague in Inner Mongolia from 2000 to 2010. Chinese Journal of Vector Biology & Control, 23(4), 349-351.

Hu, Y., Zhou, G., Ruan, Y., Lee, M. C., Xu, X., Deng, S., et al., 2016. Seasonal dynamics and microgeographical spatial heterogeneity of malaria along the China-Myanmar border. Acta

Tropica, 157, 12-19.

Li H.R., Wang, W.Y., Yang, L.S., Tan, J.A., 2005. Coupling analysis of climate change with

19 human plague prevalence. Chinese Journal of Zoonoses, 21(10), 887-891.

Li, G.D., Zhang, J., Jiao, G., Zhao, Z., 2013. Advances in impacts of climate change on infectious diseases outbreak. Acta Ecologica Sinica, 33(21), 6762-6773.

Li, Z., Chen, D., 1999. Studies on relationships among parasitic flea index of Meriones unguiculatus and meteorological factors. Acta Entomologica Sinica, 42.

Lieshout, M.V., Kovats, R.S., Livermore, M.T.J., Martens, P., 2004. Climate change and malaria: analysis of the SRES climate and socio-economic scenarios. Global Environmental

Change, 14(1), 87-99.

Liu, H.Z., Liu, M.F., 2005. Study on naturally infected plague fleas in the plague natural focus of Hebei province. Chinese Journal of Vector Biology & Control, 74(3), 362-370.

Liu, J.Y., 1986. Seasonal fluctuation of fleas parasitizing gerbilsin the northern desert-steppe area of Neimongol Autonomous Region. Acta Entomologica Sinica, 29(2), 167-173.

Nakazawa, Y., Williams, R., Peterson, A.T., Mead, P., Staples, E., Gage, K.L., 2007. Climate change effects on plague and tularemia in the United States. Vector Borne & Zoonotic Diseases,

7(4), 529-40.

Ning, Z.B., Shao J.L., Yan, D., Wang, Z.Y., Gao, W.L., Wang, X.H., et al., 2012. Plague control in Kangbao County in 2001-2010. Chinese Journal of Hygienic Insecticides & Equipments,

127(20).

Parmenter, R.R., Yadav, E.P., Parmenter, C.A., Ettestad, P., Gage, K.L., 1999. Incidence of plague associated with increased winter-spring precipitation in New Mexico. American Journal of

Tropical Medicine & Hygiene, 61(5), 814-821.

Pham, H.V., Dang, D.T., Tran Minh, N.N., Nguyen, N.D., Nguyen, T.V., 2009. Correlates of environmental factors and human plague: an ecological study in Vietnam. International Journal of

Epidemiology 2009 Vol.38 No.6 pp. 1634-1641, volume 38(6), 1634-1641(8).

Qian, Q., Zhao, J., Fang, L., Zhou, H., Zhang, W., Wei, L., et al., 2014. Mapping risk of plague in Qinghai-Tibetan plateau, China. Bmc Infectious Diseases, 14(2), 1-8.

Seya, H., Yamagata, Y., Tsutsumi, M., 2013. Automatic selection of a spatial weight matrix in spatial econometrics: application to a spatial hedonic approach. Regional Science & Urban

Economics, 43(3), 429-444.

Thomas, C.J., Gemma, D., Dunn, C.E., 2004. Mixed picture for changes in stable malaria 20 distribution with future climate in Africa. Trends in Parasitology, 20(5), 216-220.

Williams, W.M., 1974. The anatomy of the Mongolian gerbil (Meriones unguiculatus).

Anatomy of the Mongolian Gerbil.

Yang, C., Kafatos, M., Wong, D.W., Wolf, H.D., Yang, R., 2010. Geographic Information

System. US, US7725529.

Yang, L.S., Chen, R.G., Wang, W.Y., Tan, J.A., 2000. The temporal and spatial distribution of the plague foci since 1840 in China. Geographical Research, 19(3), 243-248.

Zhang, N., Song, Z.Z., 2003. Impact of climate change on plague. Climate change monitoring report. Japan Meteorological Agency, 61-73.

Zhang, Z., Li, Z., Tao, Y., Chen, M., Wen, X., 2007. Relationship between increase rate of human plague in china and global climate index as revealed by cross-spectral and cross-wavelet analyses. Integrative Zoology, 2(3), 144-153.

Zhong, L.I., Zhang, W., & Yan, W. (2000). Studies on dynamics of body and burrow nest fleas of Meriones unguiculatus. Acta Entomologica Sinica.

Fig 1. Location of the study region

Fig 2. The spatial distribution patterns maps of nest flea indexes

Fig 3. Maps of the hot spot regions of the nest flea indexes

Fig 4. Analysis of the spatial three-dimensional trend of the nest flea index. Where the X-axis represents longitude, the Y-axis represents latitude, and the Z-axis represents the average annual nest flea indexes in the Meriones unguiculatus natural epidemic focus from 2005 to 2014.

Table 1. Nest flea index used in the study Time The data area Zhengxiangbai Banner, Erenhot City, Sonid Right Banner, Wushen County；Dingbian 2005 County；Kangbao County Zhengxiangbai Banner, Erenhot City, Sonid Left Banner, Sonid Right Banner, Taibus 2006 Banner, Wuchuan County, Wushen County; Dingbian County; Lingwu City, Pingluo County, Yanchi County; Kangbao County Erenhot City, Sonid Left Banner, Sonid Right Banner, Taibus Banner, Hondlon District, 2007 Wushen County, Otog Front Banner;Dingbian County;Lingwu City, Pingluo County;Kangbao County Erenhot City, Sonid Left Banner, Sonid Right Banner, Jiuyuen District, Hondlon 2008 District, Otog Front Banner; Dingbian County; Pingluo County, Yanchi County; Kangbao County Erenhot City, Sonid Right Banner, Siziwang Banner, Huade County, Shangdu County, 2009 Jiuyuen District, Wushen County, Otog Front Banner; Dingbian County; Lingwu City, Pingluo County, Yanchi County; Kangbao County Erenhot City, Sonid Right Banner, Taibus Banner, Siziwang Banner, Huade County, Shangdu County, Chahar Right Back Banner, Jiuyuen District, Otog Front Banner; 2010 Dingbian County; Xingqing District, Lingwu City, Pingluo County, Yanchi County; Kangbao County Zhengxiangbai Banner, Boarder Yellow Banner, Sonid Right Banner, Siziwang Banner, Huade County, Shangdu County, Chahar Right Back Banner, Jiuyuen District, Otog 2011 Banner, Wushen County, Otog Front Banner; Dingbian County; Xingqing District, Lingwu City, Pingluo County, Yanchi County; Kangbao County Zhengxiangbai Banner, Boarder Yellow Banner, Sonid Left Banner, Sonid Right Banner, Taibus Banner, Siziwang Banner, Huade County, Shangdu County, Chahar Right Back 2012 Banner, Jiuyuen District, Darhan Muminggan Joint Banner, Otog Front Banner;Dingbian County;Xingqing District, Lingwu City, Pingluo County, Yanchi County;Kangbao County Zhengxiangbai Banner, Boarder Yellow Banner, Sonid Left Banner, Sonid Right Banner, Taibus Banner, Siziwang Banner, Huade County, Shangdu County, Chahar Right Back 2013 Banner, Jiuyuen District, Darhan Muminggan Joint Banner, Otog Banner, Otog Front Banner; Dingbian County; Xingqing District, Lingwu City, Pingluo County, Yanchi County; Kangbao County Zhengxiangbai Banner, Boarder Yellow Banner, Sonid Right Banner, Taibus Banner, Siziwang Banner, Huade County, Shangdu County, Chahar Right Back Banner, Urad 2014 Middle Banner, Jiuyuen District, Darhan Muminggan Joint Banner, Otog Front Banner, Hanggin Banner; Dingbian County; Xingqing District, Lingwu City, Pingluo County, Yanchi County;Kangbao County

Table 2. Analytical methods Indicator Calculating formula Reference

n n n n WCWXij ij  ij (X i  ) I i1 j  1 i  1 j  1 n n n n1 n WSWX22 (X  ) ij  ijn  i i1 j  1 i  1 j  1 i  1 I E (I) Z-score Z(I)  Global VAR(I) Moran's I value of 1 E(I)  expectation n 1 (Barbosa et al., 2014) n[(n22 3n  3) W  nW  3W ] variance VAR(I)1 2 0 E (I)2 W2 (n 1)(n  2)(n  3) n xx Anselin Ii w() x x iS 2  ij j Local j

Moran's I IEii (I ) Z-score Z(Ii )  VAR() Ii

* wij x j- i x Getis-Ord ∗  ∗ G 푖 statistic * j (Hu et al., 2016) G Gi = 푖 sw(ns*2 ) / (n 1) 1ii Ordinary Least Square 푦 = 휌푊푦 + 푋훽 + 휀 estimation Spatial Lag Model 푦 = (퐼 − 휌푊)−1푋훽 + (퐼 − 휌푊)−1휀 Anselin, 1999

Spatial Error Models 푦 = 휆푊푦 + 푋훽 − 휆푊푋훽 + 휀

Table 3. Result of the Global Autocorrelation Analysis of the nest flea indexes Time Moran’s I Expected I Variance z-score p-value Result 2005 -0.03382 -0.03448 0.006864 0.007971 0.99364 random 2006 0.230311 -0.03448 0.008565 2.861229 0.00422 clustered 2007 -0.00965 -0.03448 0.002858 0.464585 0.642229 random 2008 -0.02876 -0.03448 0.012821 0.05054 0.959692 random 2009 -0.02051 -0.03448 0.013477 0.120396 0.90417 random 2010 0.11744 -0.03448 0.009111 1.591622 0.11147 random 2011 -0.03955 -0.03448 0.006707 -0.06187 0.950668 random 2012 0.480096 -0.03448 0.014078 4.33691 0.000014 clustered 2013 0.270083 -0.03448 0.014279 2.548757 0.010811 clustered 2014 0.298491 -0.03448 0.009078 3.494728 0.000475 clustered

Table 4 The spatial dependence test results of the nest flea index and impacting factors Lagrange Multiplier MI/DF t-value P-value Lagrange Multiplier (lag)

2005 1 18.4458 0.0000 2006 1 10.7829 0.0010 2007 1 0.0124 0.9112 2008 1 13.9147 0.0002 2009 1 7.2863 0.0069 2010 1 17.6809 0.0000 Robust LM (lag)

2005 1 19.1307 0.0000 2006 1 9.2934 0.0023 2007 1 1.9944 0.1579 2008 1 9.6655 0.0019 2009 1 14.1932 0.0002 2010 1 14.4201 0.0001 Lagrange Multiplier (error)

2005 1 5.8391 0.0157 2006 1 2.8263 0.0927 2007 1 2.7220 0.0990 2008 1 6.6408 0.0100 2009 1 0.1624 0.6869 2010 1 7.7676 0.0053 Robust LM (error)

2005 1 6.5240 0.0106 2006 1 1.3368 0.2476 2007 1 4.7040 0.0301 2008 1 2.3916 0.1220 2009 1 7.0694 0.0078 2010 1 4.5069 0.0338

Table 5 The OLS and SLM results of the nest flea index and impacting factors Model Year R2 Log Likelihood Akaike Info Criterion Schwarz Criterion OLS 2007 0.8564 19.7775 -27.5549 -19.1477 2005 0.7925 13.9354 -13.8708 -4.0625 2006 0.8287 15.3565 -16.713 -6.9046 SLM 2008 0.8565 19.7848 -25.5696 -15.7612 2009 0.7291 12.4102 -10.8204 -1.012

Table 6 The OLS results of the nest flea index and impacting factors in 2007 Coefficient of Standard Variable t value P value Regression deviation Host density -3.9906 1.8710 -2.1329 0.0434 Temperature -0.9167 0.2032 -4.5107 0.0001 DEM -0.2215 0.2792 -0.7933 0.4354 NDVI 1.4958 0.3225 4.6382 0.0001 Rainfall 0.7544 0.2278 3.3124 0.0029

Table 7 The SLM results of the nest flea index and impacting factors Coefficient of Standard Variable t value P value Regression (β) deviation Host density

2005 3.0362 1.4413 2.1065 0.0352 2006 -1.0688 1.5212 -0.7026 0.4823 2008 0.8216 1.6300 0.5041 0.6142 2009 -0.9526 2.2227 -0.4286 0.6682 2010 1.8249 1.5536 1.1746 0.2402 Temperature

2005 0.5953 0.2217 2.6855 0.0072 2006 -0.7878 0.2386 -3.3021 0.0010 2008 -0.5039 0.2561 -1.9673 0.0491 2009 -0.4398 0.2226 -1.9763 0.0481 2010 -0.0754 0.2298 -0.3281 0.7429 DEM

2005 0.5920 0.2767 2.1393 0.0324 2006 -0.6547 0.2585 -2.5323 0.0113 2008 0.1919 0.2451 0.7830 0.4336 2009 0.0346 0.3044 0.1137 0.9095 2010 0.4239 0.2496 1.6985 0.0894 NDVI

2005 -1.4333 0.3728 -3.8444 0.0001 2006 0.8342 0.3925 2.1256 0.0335 2008 0.0379 0.3137 0.1207 0.9040 2009 0.4517 0.4649 0.9716 0.3312 2010 -0.2955 0.2717 -1.0877 0.2767 Rainfall

2005 0.0344 0.2625 0.1311 0.8957 2006 1.1528 0.2867 4.0215 0.0001 2008 0.4995 0.2641 1.8912 0.0586 2009 0.6800 0.2518 2.7003 0.0069 2010 0.2159 0.2115 1.0208 0.3073