sustainability
Article Spatial Patterns and the Regional Differences of Rural Settlements in Jilin Province, China
Xiaoyan Li 1,*, Huiying Li 1,2, Yingnan Zhang 3,4 and Limin Yang 1
1 College of Earth Sciences, Jilin University, Changchun 130012, China; [email protected] (H.L.); [email protected] (L.Y.) 2 Key Laboratory of Wetland Ecology and Environment, Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130102, China 3 Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100049, China; [email protected] 4 College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China * Correspondence: [email protected]
Received: 24 September 2017; Accepted: 21 November 2017; Published: 25 November 2017
Abstract: The spatial patterns of rural settlements are important for understanding the drivers of land use change and the relationship between human activity and environmental processes. It has been suggested that the clustering of houses decreases the negative effects on the environment and promotes the development of the countryside, but few empirical studies have quantified the spatial distribution patterns of houses. Our aim was to explore the regional differences in rural settlement patterns and expand our understanding of their geographic associations, and thus contribute to land use planning and the implementation of the policy of “building a new countryside”. We used spatial statistical methods and indices of landscape metrics to investigate different settlement patterns in three typical counties within different environments in Jilin Province, Northeast China. The results indicated that rural settlements in these three counties were all clustered, but to a varied degree. Settlement density maps and landscape metrics displayed uniformity of the settlement distributions within plain, hill, and mountainous areas. Influenced by the physical environment, the scale, form, and degree of aggregation varied. Accordingly, three types of rural settlements were summarized: a low-density, large-scale and sparse type; a mass-like and point-scattered type; and a low-density and high cluster-like type. The spatial patterns of rural settlements are the result of anthropogenic and complex physical processes, and provide an important insight for the layout and management of the countryside.
Keywords: Ripley’s K function; point pattern analysis; spatial statistics; geographic association
1. Introduction Rural development has been highlighted in the form of policy with the industrialization and urbanization of a country reaching a certain stage. For example, the “Rural White Papers” of England, Scotland, and Wales were comprehensive reports of the governments’ policies on issues relating to rural areas [1]. In the 1970s, the “New Countryside Campaign” was carried out by South Korea to accelerate the development of rural areas [2]. For China, since the opening up policy of 1978, and the economic development and urbanization that followed, a series of rural problems have come out, including: the shrinking of cultivated land, the inefficient and extensive utilization of resources, small-scale and disorderly distributions of rural settlements, the hollow village problem, etc. [3]. All of these widened the income gap between rural and urban residents, and led to a weak foundation of the agricultural sector for speeding up regional development [1]. Under the circumstances, the “building a new countryside” strategy was laid out by the central government of China, which expected to solve
Sustainability 2017, 9, 2170; doi:10.3390/su9122170 www.mdpi.com/journal/sustainability Sustainability 2017, 9, 2170 2 of 16 all of the above problems through coordinating the development of urban and rural areas. Taking the village as a place where local communities live and work, the specific solutions to this policy were realized through centralizing rural settlements, improving infrastructure, and directing developments according to local environmental conditions. Therefore, to realize the sustainable management of rural areas, a reasonable rural planning is necessary and thus can reduce the impact on the environment [4]. Rural settlements are places to inhabit, produce, and live for rural people, and are interactive results of the surrounding natural, economic, social, and cultural environment [5]. Residential settlements and ancillary infrastructure development have been recognized as primary causes of anthropogenic landscape changes in many countries [6]. The patterns of spatial dispersion and expansion in China and other countries have received enough attention at the urban fringe. However, all of these concentrated on “suburban sprawl” [6–8], or the importance of the urban–rural gradient that runs from the densely urbanized inner city to exurban environments [9–13], but neglected the profound impact brought by the settlements located in more remote rural areas. Many recent case studies have been conducted to analyze the area, shape, and density, as well as the evolution, of the landscape of rural settlements [2,14,15]. In China, due to the implementation of the policy to build a socialist countryside in recent years, more attention has been paid to the problems regarding rural urbanization, rural communities, the “hollowing out” of rural settlements, and the optimized regulation of rural settlements [16–19]. However, analysis of the spatial distribution and pattern of rural settlements is of great importance for understanding the influence of landscape changes and adopting policies to guide rural development. Therefore, the characteristics of typical patterns and the spatial reconstruction of rural settlements have become a hot issue in academic circles [20,21]. Several approaches have been developed to reveal the spatial characteristics of rural settlements with regular data [22–24]. In the past 20 years, due to the application of geographic information system (GIS), the spatial pattern of residential land has undergone a revolutionary change. Since Hodder and Orton suggested rather complicated and cumbersome models in the early 1970s, sophisticated built-in tests have been available in the ArcGIS software packages (Fletcher, 2008). However, the application of these tools is still in the initial stages [25,26]. Northeast China is a famous grain-producing region in China. The grain production capacity accounted for 19.3% of the country’s total in 2015 [27], and maize production accounted for approximately 30% of the nation’s total production [28]. The rural problem of northeast China has always been a matter of national concern, and rural development of northeast China has become a prerequisite for regional development. Jilin Province is a major agricultural province in northeast China, and arable land accounts for more than 30% of the total area. In addition, there are significant regional differences in the natural environment, economic development, and cultural context from west to east. Accordingly, the development patterns of different counties in Jilin Province should be discrepantly influenced by their particular geographical factors [18,29]. Understanding this spatial pattern and its geographical linking could provide valuable insights for planners and decision-makers to plan and renovate the rural settlements in different regions [19]. Our objectives are to: (1) compare spatial distribution patterns of rural settlements in different regions; (2) interpret the geographical factors affected by the spatial patterns of settlements; and (3) explore the development of modes suited to regional characteristics.
2. Materials and Methods
2.1. Study Area Jilin Province is located in northeast China (40◦520N to 46◦180N, 121◦380E to 131◦190E) (Figure1). The total land area is approximately 187,400 km2. The altitude drops gently from the southeast to the northwest, with the vast Songnen Plain in the mid-west of the province. The prevalent climate in the study area is a continental monsoon, and the annual average precipitation changes between 350 mm and 1500 mm gradually from east to west. Jilin Province is the largest commercial grain-producing Sustainability 2017, 9, 2170 3 of 16 Sustainability 2017, 9, 2170 3 of 16 baseof China. of China. A study A study on onits itsrural rural settlements settlements is isuseful useful for for regional regional land management andand economiceconomic development.development. InIn thisthis paper,paper, threethree typicaltypical counties,counties, Tongyu,Tongyu, Dehui,Dehui, andand Antu,Antu, werewere selectedselected fromfrom differentdifferent regions regions as as study study areas. areas. LocatedLocated in in western western Jilin Jilin Province, Province, Tongyu Tongyu extends extends from from 44 44°13◦13057′5700″NN to to 45 45°16◦160N′N inin latitude latitude and and fromfrom 122 122°02◦02013′1300″EE to to 123 123°30◦30057′5700″EE inin longitude,longitude, withwith a a territoryterritory of of 8459 8459 km km2.2. TheThe climateclimate is is characterized characterized byby a a temperate temperate continental continental monsoon monsoon with with four four distinct distinct seasons. seasons. The The average average annual annual temperature temperature is is 5.15.1◦ C,°C, and and extreme extreme temperatures temperatures range range from −from32 ◦−C32 to °C 38.9 to◦ C.38.9 The °C. mean The annual mean precipitation annual precipitation amounts toamounts 400–500 to mm, 400–500 70–80% mm, of 70–80% which isof concentratedwhich is concen in thetrated summer in the season.summer Accordingly, season. Accordingly, mean annual mean evaporationannual evaporation is up to 1500–1900is up to 1500–1900 mm. Its landformmm. Its landfo is generallyrm is generally similar to similar a dustpan, to a dustpan, because itsbecause eastern, its southern,eastern, southern, and western and parts western are comparativelyparts are comp higheraratively than higher its northern than its part.northern part.
FigureFigure 1. 1.Location Location of of the the study study area. area.
DehuiDehui isis situatedsituated inin the the middle middle of of Jilin Jilin province province (125 (125°45◦450E′E to to 126 126°23◦230E,′E, 43 43°32◦320N′N toto 4444°45◦450N),′N), 2 coveringcovering an an area area of of approximately approximately 3435 3435 km km2.. TheThe elevationelevation rangesranges fromfrom 149149 mm to to 241 241 m, m, dropping dropping slightlyslightly fromfrom thethe south to to the the no northrth and and fluctuating fluctuating from from east east to towest west across across the thestudy study area. area. The TheSonghua Songhua River, River, Yitong Yitong River, River, Ying Yingmama River, River, and and Wukai Wukai River River run run th throughrough the the middle middle of of Dehui. Dehui. It Ithas has a asemi-humid, semi-humid, continental, continental, monsoon-type monsoon-type climat climatee with with an annual mean temperaturetemperature ofof 4.44.4◦ °C,C, andand an an annual annual average average precipitation precipitation of of 530 530 mm. mm. Dehui Dehui is is one one of of the the best best agricultural agricultural demonstration demonstration regionsregions in in Jilin Jilin Province, Province, with with farmland farmland accounting accounting for for 70% 70% of of the the total total land land [29 [29].]. 2 AntuAntu isis located located in in the the east east of of Jilin Jilin Province Province and and covers covers an an area area of of 7438 7438 km km2.. The The topography topography fluctuatesfluctuates acrossacross thethestudy study area,area, withwith aarelative relativeelevation elevation ofof 23572357 m,m, andand dropsdrops fromfrom southeast southeast toto northwest,northwest, showing showing the the characteristics characteristics of of a a narrow narrow strip. strip. ThereThere areare convenientconvenient traffictraffic conditionsconditions here,here, andand thethe MinglongMinglong highwayhighway runsruns fromfrom northnorth to to south. south. AntuAntu isis richrich inin forests,forests, water,water,and and mineral mineral resources,resources, and and the the tourism tourism industry industry developed developed quickly quickly based based on on the the scenic scenic Changbai Changbai Mountains. Mountains.
2.2. Interpretation of Rural Settlements To establish the land use data sets, a total of six scenes of Landsat OLI images (30 m spatial resolution) were used for the land use classification in 2015. All of the images used in the classification Sustainability 2017, 9, 2170 4 of 16
2.2. Interpretation of Rural Settlements SustainabilityTo establish 2017, 9, 2170 the land use data sets, a total of six scenes of Landsat OLI images (30 m spatial4 of 16 resolution) were used for the land use classification in 2015. All of the images used in the classification were false colors composed of four, three, and two bands using the RGB method, and georeferenced were false colors composed of four, three, and two bands using the RGB method, and georeferenced to to 1:100,000 relief maps using at least 20 ground control points (GCPs) for each scene. The root mean 1:100,000 relief maps using at least 20 ground control points (GCPs) for each scene. The root mean squared error (RMS error) of the geometric reference was controlled within 1.5 pixels. squared error (RMS error) of the geometric reference was controlled within 1.5 pixels. Land use classified data of the study area were obtained by the object-oriented classification Land use classified data of the study area were obtained by the object-oriented classification (using eCognition 8.64 software [30]) combined with the nearest neighbor classifiers and visual (using eCognition 8.64 software [30]) combined with the nearest neighbor classifiers and visual interpretations (Figure 2). First, multi-resolution segmentation was used to process the Landsat OLI interpretations (Figure2). First, multi-resolution segmentation was used to process the Landsat OLI imageries, and then, the images were grouped into homogeneous pixels representing individual imageries, and then, the images were grouped into homogeneous pixels representing individual objects [31]. In this process, the scale, shape, and compactness were selected as image segmentation objects [31]. In this process, the scale, shape, and compactness were selected as image segmentation parameters. We classified image objects into different classes using the nearest neighbor classifiers parameters. We classified image objects into different classes using the nearest neighbor classifiers and visual interpretation after segmentation. In our study, there were eight aggregated classes of land and visual interpretation after segmentation. In our study, there were eight aggregated classes of land use types: rural settlement, urban built-upland, cropland, forestland, grassland, wetland, water body, use types: rural settlement, urban built-upland, cropland, forestland, grassland, wetland, water body, and barren land. The overall accuracy of the land use maps interpreted from remote sensing images and barren land. The overall accuracy of the land use maps interpreted from remote sensing images exceeded 90% (91.3% for Tongyu, 91.5% for Dehui, and 90.9% for Antu, respectively). The results exceeded 90% (91.3% for Tongyu, 91.5% for Dehui, and 90.9% for Antu, respectively). The results indicated that 99% of the polygon boundaries showed a shift of less than 1.5 pixels (45 m) from the indicated that 99% of the polygon boundaries showed a shift of less than 1.5 pixels (45 m) from the real boundary. real boundary.
FigureFigure 2. 2. LandLand use use maps maps of of To Tongyu,ngyu, Dehui, Dehui, and and Antu. Antu.
2.3.2.3. Spatial Spatial Statistic Statistic Method Method UsingUsing datadata ofof rural rural settlements settlements interpreted interpreted from from Landsat Landsat imagery,the imagery, central the points central of ruralpoints settlements of rural settlementswere extracted. were We extracted. used spatial We statistic used methods,spatial statistic such as methods, Ripley’s K ,such kernel as density Ripley’s estimation K, kernel (KDE), density and estimationGetis–Ord General(KDE), Gandmethods Getis–Ord to interpret General the G spatial methods distribution, to interpret spatial the variation, spatial spatialdistribution, agglomeration, spatial variation,and other heterogeneousspatial agglomeration, characteristics and other of the heteroge rural settlementsneous characteristics in the study area. of the These rural calculations settlements were in therealized study using area. the These GIS calculations software ArcGIS were 10.3. realized using the GIS software ArcGIS 10.3. Ripley’s K function is a tool used to characterize the strength of spatial dependence at multiple scales in a spatial pattern. The K function has been widely used to identify clustering, randomness, or regularity among events in a spatial point pattern [32]. The formula is as follows: Sustainability 2017, 9, 2170 5 of 16
Ripley’s K function is a tool used to characterize the strength of spatial dependence at multiple scales in a spatial pattern. The K function has been widely used to identify clustering, randomness, or regularitySustainability among 2017, 9, 2170 events in a spatial point pattern [32]. The formula is as follows: 5 of 16
n n δ (d) ij ( ) K(d) = A ∑ ∑ (i, j = 1, 2, 3, . . . , n; i 6= j, dij ≥ d) (1) ( ) = n2 ( , = 1,2,3, … , ; ≠ , ≥ ) (1) i=1 j=1 r K(d) L(d) = ( )− d (2) ( ) = π − (2) where A is the area of the study region, n is the number of settlements, d represents the spatial scale, δij(d)whereis a weight, A is thed ijarearefers of the to thestudy distance region, between n is the number settlement of settlements,i and settlement d representsj, and theδij =spatial 1 (dij scale,≤ d) or δij = 0δij ((d)dij is> ad ).weight, In order dij refers to achieve to the thedistance linearly-expected between settlement value i and and stabilitysettlement of j the, and standard δij = 1 (dij deviance, ≤ d) or Besagδijreplaced = 0 (dij > dK).( dIn) withorderL to(d ),achieve whose the expected linearly-expected value is zero value under and thestability assumption of the standard of a totally deviance, random spatialBesag distribution replaced K [(33d)]. with The L graph(d), whose of L expected(d) can be value used is tozero describe under the the assumption spatial pattern of a totally of a multi-scale random landscape.spatial distribution When L(d) [33]. > 0, ruralThe graph settlements of L(d) can show be used a clustered to describe distribution; the spatial whenpatternL of(d )a
Input layer Output layer
FigureFigure 3. Illustration3. Illustration of of the the kernel kerneldensity density method.
KDEKDE is described is described as following:as following: 1 ( , ) =1 n (d ) (3) f (x, y) = ℎ K( i ) (3) 2 ∑ nh i=1 n where f(x,y) is the estimated density at position (x,y), n is the number of observation wherepoints,f(x,y) his represents the estimated the bandwidth density at positionparameter, (x,y K), isn ais kernel the number function, of observation and di represents points, theh representsdistance the bandwidthfrom position parameter, (x,y) to observingK is a kernel position function, i. and di represents the distance from position (x,y) to observingGetis–Ord position iGeneral. G is used to measure the concentration of high or low values for a given studyGetis–Ord area. GeneralThe formulaG is is used given to as measure follows: the concentration of high or low values for a given study area. The formula is given as follows: ( ) = ( ) / (4) n n n n G(d) = ∑ ∑ Wij(d)XiXj/ ∑ ∑ XiXj (4) i=1 j=1 i=1 j=1 ( ) = ( ) − ( )/ ( ) (5) q where Xi and Xj are attribute valuesZ(G) for= Gsettlements(d) − E(G i) /andVar j, respectively;(G) Wij(d) refers to the spatial(5) weight between settlement i and j; and n is the total number of settlements in the study area. Z(G) is wheretheX resulti and Xofj theare normalization attribute values of G for(d); settlements E(G) and Vari (andG) denotej, respectively; the expectedWij (valued) refers and tovariance the spatial of weightG(d between), respectively. settlement If G(di) andis higherj; and thann is E the(G) total and numberZ(G) is significant, of settlements high-value in the studyin the area.study Zarea(G) is shows a clustering characteristic; if G(d) is lower than E(G) and Z(G) is significant, a low-value cluster Sustainability 2017, 9, 2170 6 of 16 the result of the normalization of G(d); E(G) and Var(G) denote the expected value and variance of G(d), respectively. If G(d) is higher than E(G) and Z(G) is significant, high-value in the study area shows a clustering characteristic; if G(d) is lower than E(G) and Z(G) is significant, a low-value cluster appears in the study area; and if G(d) approaches E(G), variables in the study area display a random distribution.
2.4. Indices of Landscape Metrics The landscape metrics were calculated using the Patch Analyst module of the ArcGIS system. The definitions of various patch metrics were given by FRAGSTATS [34]. The spatial distribution of rural settlements was described quantitatively by several indices, because no single metric can capture the complexity of the spatial arrangement of the settlements [35]. When working with landscape metrics, those indicators relevant for the problem under investigation with a small correlation can be considered [36]. In this analysis of the spatial characteristic of rural settlement metrics at the class level, six indices were selected: the average nearest-neighbor index (ANN), mean patch size (MPS), patch size standard deviation (PSSD), patch size coefficient of variation (PSCV), landscape shape index (LSI), and patch cohesion index (PCI). All of these metrics describe the heterogeneity and complexity of settlements within the landscape [32]. The vector data of rural settlements in study area were converted to grid data and all of the landscape metrics were realized in the patch analysis module of ArcGIS 10.3. The formula and description of the indices are as follows (Table1):
Table 1. The formula and description of the indices.
Landscape Metric Abbreviation Description Unit ∑n d 0.5 ANN = i i / √ . N Average Nearest- N/A ANN where d is the distance between two rural points; m Neighbor Index i A is the total area of the landscape, and N is the total number of patches. MPS = A/N. Mean Patch Size MPS where A is the total area of the landscape and ha N is the total number of patches. The standard deviation of patch size in the entire landscape. v u 2 u m n [ − A ] tch Size t ∑i=1 ∑j=1 aij N PSSD × 104 ha Standard Deviation N where A is the total area of the landscape and N is the total number of patch edges, aij is the area of path ij. The standard deviation of patch size divided by Patch Size Coefficient PSCV mean patch size for the entire landscape. % of Variation PSSD × 100 MPS A modified perimeter-area ration of the form: 0.25E LSI = √ , Landscape Shape Index LSI A unitless where E is the total length of patch edges and A is the total area of the landscape. " n # ∑j=1 pij 1 PCI = 1 − n √ × h i × 100, ∑j=1 pij aij 1 − √1 Patch Cohesion Index PCI A unitless where pij is perimeter of habitat patch ij, aij is area of patch ij, and A is total of the landscape. Number of Patch NP ≥1, Equals the number of patches in the landscape. unitless Pach Area PA ≥0, Equals the area of patches in the landscape. ha Sustainability 2017, 9, 2170 7 of 16
3.Sustainability Results 2017, 9, 2170 7 of 16 The landscape metrics of rural settlements in the study area are illustrated in Table2. 3. Results The landscapeTable metrics 2. Landscape of rural settlements metrics of in rural the settlements study area in are the illustrated studied area. in Table 2.
DistributionTable of 2. RuralLandscape Settlements metrics of rural Scale settlements of Rural Settlements in the studiedForm area. of Rural Settlements ANN MPS PSSD PSCV LSI PCI Distribution of Rural Settlements Scale of Rural Settlements Form of Rural Settlements Tongyu ANN 0.72 MPS 27.44 PSSD 0.42 PSCV 154.79 LSI 55.04 PCI 99.86 Dehui 0.84 16.55 0.25 153.71 78.43 99.82 Tongyu 0.72 27.44 0.42 154.79 55.04 99.86 Antu 0.51 10.08 0.16 161.20 39.69 99.78 Dehui 0.84 16.55 0.25 153.71 78.43 99.82 Antu 0.51 10.08 0.16 161.20 39.69 99.78 3.1. The Spatial Pattern of Rural Settlements 3.1.The The landscapeSpatial Pattern metric of Rural ANN, Settlements Ripley’s L(d) function, and kernel density estimation methods were used toThe describe landscape the characteristicsmetric ANN, Ripley’s of the spatialL(d) function, distribution and kernel of rural density settlements estimation (Table methods2). were usedANN to describe was used the to characteristics describe the spatialof the spatia distributionl distribution pattern of ofrural rural settlements settlements. (Table IfANN 2). is smaller than1,ANN rural settlementswas used to describe show an the agglomerated spatial distribution distribution; pattern on of the rural contrary, settlements. rural If settlements ANN is smaller display a randomthan1, rural distribution. settlements In thisshow study, an agglomerated the ANN values distribution; are 0.72, 0.84,on the and contrary, 0.51 for rural Tongyu, settlements Dehui and Antu,display respectively, a random showingdistribution. a significant In this study, agglomerated the ANN distributionvalues are 0.72, at the 0.84, 0.1 and level. 0.51 for Tongyu, DehuiThe and results Antu, of respectively, the analyses showing of Ripley’s a significantL(d) function agglomerated are shown distribution in Figure4 .at The the curve0.1 level. shows that the degreeThe results of the clusterof the analyses of rural of settlements Ripley’s L(d of) function Antu is are the shown highest, in Figure and the 4. valueThe curve of Tongyu shows isthat next. ThetheL (degreed) curves of the for cluster Antu andof rural Dehui settlements are all positive of Antu values, is the highest, suggesting and athe higher value agglomerated of Tongyu is next. degree, evenThe atL(d the) curves larger for scale. Antu However, and Dehui there are areall positive still changes values, in suggesting the agglomeration a higher levelagglomerated of rural settlements degree, even at the larger scale. However, there are still changes in the agglomeration level of rural at smaller scales. The curve of Antu county showed two peaks (appearing at approximately 20 km settlements at smaller scales. The curve of Antu county showed two peaks (appearing at and 80 km), indicating that the radiation effects of towns were obvious and rural settlements grew approximately 20 km and 80 km), indicating that the radiation effects of towns were obvious and around two important towns. The L(d) curves of Dehui county maintained a rising trend over 0–35 km, rural settlements grew around two important towns. The L(d) curves of Dehui county maintained a reached a maximum at 35 km, and changed slightly, indicating that the agglomeration level firstly rising trend over 0–35 km, reached a maximum at 35 km,and changed slightly, indicating that the increased,agglomeration and then level remained firstly increased, smooth asand the then distance remained increased. smooth The as theL(d distance) curve ofincreased. Tongyu countyThe L(d had) a similarcurve of trend Tongyu as thatcounty of Dehui,had a similar but showed trend clusteredas that of characteristicsDehui, but showed within clustered 105 km, characteristics and separated characteristicswithin 105 km, over and this separated distance. characteristics over this distance.
Figure 4. K Figure 4. Ripley’sRipley’s K curves of of rural rural settlements settlements in in the the three three counties. counties.
KDEKDE was was considered considered toto bebe anan effectiveeffective method by by which which density density is iscalculated calculated using using a kernel a kernel functionfunction superimposed superimposed over over eacheach location.location. The The application application of of KDE KDE req requiresuires the the choice choice of ofthe the type type of of kernelkernel function function and and the the definitiondefinition of three key para parameters:meters: bandwidth, bandwidth, cell cell size, size, and and intensity intensity [36]. [36 ]. BandwidthBandwidth is is a a very very import import parameterparameter influencing the the value value of of KDE. KDE. Increa Increasingsing the the bandwidth bandwidth will will notnot greatly greatly change change the the mean mean ofof thethe density values values calculated, calculated, but but greatly greatly affects affects standard standard deviations. deviations. TheThe standard standard deviations deviations of KDEof KDE for differentfor different bandwidths bandwidths in the in three the countiesthree counties were calculated were calculated (Figure 5), (Figure 5), and the curve changed slightly when the bandwidth was approximately 2000 m. Therefore, a radius of 2000 m was selected to estimate the kernel density, and the results were divided into four levels. Sustainability 2017, 9, 2170 8 of 16 and the curve changed slightly when the bandwidth was approximately 2000 m. Therefore, a radius of 2000Sustainability m was selected 2017, 9, 2170 to estimate the kernel density, and the results were divided into four levels.8 of 16
Sustainability 2017, 9, 2170 8 of 16
FigureFigure5. 5. Standard Standard deviations of of kernel kernel density density estimati estimationon (KDE) (KDE) for different for different bandwidths bandwidths in the three in the counties. threeFigure5. counties. Standard deviations of kernel density estimation (KDE) for different bandwidths in the three counties. The mean values of the kernel density for Tongyu, Dehui, and Antu were 0.22, 0.32, and 0.21, The mean values of the kernel density for Tongyu, Dehui, and Antu were 0.22, 0.32, and 0.21, and and the standard deviations of KDE were 0.07, 0.26, and 0.11, respectively, showing that the The mean values of the kernel density for Tongyu, Dehui, and Antu were 0.22, 0.32, and 0.21, thesettlement standard deviationsdensity of Dehui of KDE was were larger, 0.07, and 0.26,the rural and settlements 0.11, respectively, in Tongyu showing were relatively that the uniform. settlement densityand the of Dehuistandard was deviations larger, and of theKDE rural were settlements 0.07, 0.26, in and Tongyu 0.11, wererespectively, relatively showing uniform. that The the maps of settlementThe maps density of KDE of Dehuifor the was three larger, counties and the lookedrural settlements very different in Tongyu (Figure were 6). relatively Influenced uniform. by the KDE for the three counties looked very different (Figure6). Influenced by the distribution of sandy Thedistribution maps of ofKDE sandy for land the and three salinized counties land, looked settlements very differentdistributed (Figure in Tongyu 6). Influenced were sparse, by and the the landdistributiondensity and salinized value of sandy was land, lower, land settlements andespecially salinized in distributed theland, northwes settlements int and Tongyu distributed southern were parts in sparse, Tongyu (Figures and were 2 the and sparse, density 6). There and value the were was lower,densitymany especially clustervalue was centers in lower, the in northwest Dehui;especially therefore, in and the southernnorthwes the overallt partsand density southern (Figures value parts 2was and (Fig higher.6).ures There 2For and Antu, were 6). There the many physical were cluster centersmanyconditions incluster Dehui; in centers the therefore, mountainous in Dehui; the therefore, overallarea limited densitythe over the distributionall value density was value of higher. rural was residents, Forhigher. Antu, For and theAntu, the physical higherthe physical density conditions inconditions thevalues mountainous were in thepresented mountainous area as limited strips area along the limited distributionthe valleythe distribution and of main rural of roads, rural residents, leading residents, and to significantand the the higher higher heterogeneity density density values werevaluesof presentedthe were rural presented density. as strips as along strips thealong valley the valley and mainand main roads, roads, leading leading to to significant significant heterogeneity heterogeneity of the ruralof the density. rural density. Tongyu Dehui
Tongyu Dehui
Ant u
Ant u
FigureFigure 6. 6.Map Map ofof kernel density density in in the the three three counties. counties. Figure 6. Map of kernel density in the three counties. Sustainability 2017, 9, 2170 9 of 16
3.2. Scale Distribution Characteristics of Rural Settlements Three landscape indices, i.e., MPS, PSSD, and PSCV, and the Getis–Ord General G approach were employed to analyze the patterns of the scale distribution of rural settlements for these three counties (Table2). The MPS was 27.44 ha, 16.55 ha, and 10.08 ha for Tongyu, Dehui, and Antu, respectively, indicating the largest scale and lowest fragmentation of rural settlements in Tongyu. The PSSD of Antu was only 0.16, the smallest among the three counties, showing a largest landscape heterogeneity of rural settlements. We also calculated the PSCV for the three counties. We can see that the PSCV of Antu was larger than those of the other two counties, which indicates that the dispersion around the mean of the rural settlements in Antu was large. Taking the area of rural settlements as the evaluated field, the clustering condition of high or low value was calculated by the Getis–Ord General G approach. The calculation results are shown in Table3. We can see that the G(d) value of rural settlements in Tongyu and Antu was higher than the G(e) value, and their Z(G) values were all positive, indicating high value clustering for these two counties. The possibility of being influenced by a random process in these two counties was no more than 1%. For Dehui, the value of G(d) was equal to G(e), and the possibility that was influenced by a random process was 1.6%, showing there was no obvious phenomenon of gathering towers of large-scale settlements.
Table 3. Estimation of Getis–Ord General G for rural settlements.
County G(d) G(e) Z(G) p Tongyu 0.000009 0.000008 4.4177 0.000010 Dehui 0.000007 0.000007 1.3992 0.016176 Antu 0.000059 0.000033 19.1667 0.000008
3.3. Morphological Characteristics of Rural Settlements We used LSI and PCI as the qualitative metrics to compare the morphological difference of rural settlements (Table2). The LSI measured the irregularity or complexity of the morphology of rural settlements. Among the three counties, the LSI of rural settlements in Dehui was the highest (the value was 78.43), and that of Antu was the lowest (the value was 39.69), showing that the morphology of rural settlements in Antu was regular and simple. The PCI value reflected the level of connection or continuity. The PCI of rural settlements for Tongyu was slightly higher (the value was 99.86) than that of the other two counties. The relatively higher PCI value indicated there were good connections and integration for settlements in Tongyu. The fragmented patches and disconnection of rural settlements decreased the value of the patch cohesion index of Antu.
3.4. GeographicAssociation In terms of the spatial pattern of rural settlements, we learned that settlements in the three counties were all clustered. We analyzed the association between its pattern and geographic factors by landscape metrics. When the landscape metrics were used to characterize the rural settlement patterns, different features became apparent. The influence related to terrain was obvious (Figures7 and8). Due to the flat terrain and relatively small fluctuation, the influence of terrain was comparatively weak. All of the rural settlements were distributed within 200 m of elevation and a 5 degree slope, while the Tongyu settlements of Dehui were concentrated within 300 m elevation and 10 degree of slope, showing the opposite trend to terrain changes. Located in the mountainous area, the settlements of Antu represented obvious geographical differences. The number, area, and shape indices of settlements for Antu changed differently within and outside of the 800 m elevation range, showing a trend of rising first, and then decreasing. About 95% Sustainability 2017, 9, 2170 10 of 16 of the number and area were concentrated in the elevation range between 300 and 800 m. The number of patchSustainability (NP), 2017 patch, 9, 2170 area (PA), and MPS of Antu decreased when the slope increased. 10 of 16 Sustainability 2017, 9, 2170 10 of 16
1800 1800 300 1600 Tongyu Dehui Antu 300 Tongyu Dehui Antu 1600 Tongyu Dehui Antu Tongyu Dehui Antu 1400 250 1400 250 1200
ha) 200
1200 2 1000 ha) 200 2 1000 10 150
800 10 × NP 800 150 × NP 600 100 600 400 100 400 50 200 50 200 PA (Unit:
0 PA (Unit: 0 0 0 100-200 200-300 300-500 500-800 800-1000 >1000 100-200 200-300 300-500 500-800 800-1000 >1000 100-200 200-300 300-500 500-800 800-1000 >1000 100-200 200-300Elevation 300-500 (Unit: 500-800 m) 800-1000 >1000 Elevation (Unit: m) Elevation (Unit: m) Elevation (Unit: m) 80 35 80 Tongyu Dehui Antu 35 Tongyu Dehui Antu 70 Tongyu Dehui Antu 30 Tongyu Dehui Antu 70 30 60 25 60 25 50 20 50 20 40 15 40 15 30 10 LSI 30 LSI 20 10 20 MPS (Unit:MPS ha) 5 10 MPS (Unit:MPS ha) 5 10 0 0 0 0 100-200 200-300 300-500 500-800 800-1000 >1000 100-200 200-300 300-500 500-800 800-1000 >1000 100-200 200-300 300-500 500-800 800-1000 >1000 Elevation (Unit: m) 100-200 200-300 300-500Elevation 500-800 (Unit: m) 800-1000 >1000 Elevation (Unit: m) Elevation (Unit: m) Figure 7. Changes of the landscape indices with elevation. FigureFigure 7. 7.Changes Changes of the landscape landscape indices indices with with elevation. elevation.
2100 350 2100 Tongyu Dehui Antu Tongyu Dehui Antu 350 Tongyu Dehui Antu 1800 Tongyu Dehui Antu 300 1800 300 1500 250 ha)
1500 2 250 ha)
2 200 1200 10 200 1200 10 NP ×
NP 150 900 × 900 150 600 100 600 100
300 PA (Unit: 50
300 PA (Unit: 50 0 0 0 0 － 5-10 10-15 15-25 >25 0－5 5-10 10-15 15-25 >25 0 5 0－5 5-10 10-15 15-25 >25 0－5 Slope5-10 (Unit: 10-15 degree) 15-25 >25 Slope (Unit: degree) Slope (Unit: degree) Slope (Unit: degree) 35 35 Tongyu Dehui Antu 70 30 Tongyu Dehui Antu 70 Tongyu Dehui Antu 30 Tongyu Dehui Antu 25 65 25 65 20 20 60 15 LSI 60 15 LSI
MPS (Unit: ha) (Unit: MPS 10 MPS (Unit: ha) (Unit: MPS 10 55 5 55 5 0 50 0 50 0－5 5-10 10-15 15-25 >25 0－5 5-10 10-15 15-25 >25 － 5-10 10-15 15-25 >25 0－5 Slope5-10 (Unit: 10-15 degree) 15-25 >25 0 5 Slope (Unit: degree) Slope (Unit: degree) Slope (Unit: degree)
Figure 8. Changes of the landscape indices with gradient. FigureFigure 8. 8.Changes Changes of the landscape landscape indices indices with with gradient. gradient. The influence brought by rivers is very complex (Figure 9). Theoretically, a water source is an The influence brought by rivers is very complex (Figure 9). Theoretically, a water source is an important factor affecting the distribution of rural settlements in Tongyu because of the dry climate, importantThe influence factor affecting brought the by distri riversbution is very of rural complex settlements (Figure in9). To Theoretically,ngyu because of a waterthe dry source climate, is an but the number, area, and mean size of settlements increased when far away from the river. The importantbut the number, factor affecting area, and the mean distribution size of settlements of rural settlements increased when in Tongyu far away because from of the the river. dry climate,The number and area of settlements 6 km away from the river accounted for 79% and 77% of the total, butnumber the number, and area area, of and settlements mean size 6 km of settlements away from increasedthe river accounted when far awayfor 79% from and the 77% river. of the The total, number respectively. For Dehui, the peak value for settlements distribution appeared in the 6–10km distance andrespectively. area of settlements For Dehui, 6km the awaypeak value from for the settlements river accounted distribution for 79% appeared and 77% in ofthe the 6–10km total, respectively.distance range from the river, with the number and area decreasing within, or outside, of this distance. Forrange Dehui, from the the peak river, value with for the settlements number and distribution area decreasing appeared within, in the or 6–10 outside, km distanceof this distance. range from Compared with the settlement patterns of Tongyu, that of Antu showed a reverse trend, that is, the theCompared river, with with the the number settlement and areapatterns decreasing of Tongyu, within, that of or Antu outside, showed of this a reve distance.rse trend, Compared that is, the with closer to the river, the more residents. The percentages of the number and area of settlements within closer to the river, the more residents. The percentages of the number and area of settlements within 2 km of rivers were 50.16% and 69.58%, respectively. 2 km of rivers were 50.16% and 69.58%, respectively. Sustainability 2017, 9, 2170 11 of 16 the settlementSustainability patterns 2017, 9, 2170 of Tongyu, that of Antu showed a reverse trend, that is, the closer to11 the of 16 river, the more residents. The percentages of the number and area of settlements within 2 km of rivers were 140 Tongyu Dehui AnTu 50.16%Sustainability and 69.58%, 2017600 , 9, 2170 respectively. Tongyu Dehui AnTu 11 of 16 120 500 140100 Tongyu Dehui AnTu Tongyu Dehui AnTu 600400 ha) 2 12080 500300 10 × 10060 NP 400200 ha) 40 2 80 10
300100 PA (Unit: 20 × 60
NP 200 0 40 0 0－22－4 4-6 6－10 >10 0－22－4 4-6 6－10 >10
100 PA (Unit: Distance (Unit: km) 20 Distance (Unit: km) 35 65 0 0 Tongyu Dehui AnTu Tongyu Dehui AnTu 30 0－22－4 4-6 6－10 >10 0－22－4 4-6 6－10 >10 Distance (Unit: km) Distance (Unit: km) 25 35 6560 Tongyu Dehui AnTu 20 Tongyu Dehui AnTu 30 15 25
60LSI 55 10 20 MPS (Unit:MPS ha) 5 15 50 0 LSI 55 10 0－22－4 4-6 6－10 >10 0－22－4 4-6 6－10 >10 MPS (Unit:MPS ha) 5 Distance (Unit: km) Distance (Unit: km) 0 50 0－22Figure－4 9.4-6 Changes6－10 of landscape>10 indices with0－ distance22－ 4from4-6 rivers.6 －10 >10 Distance (Unit: km) Distance (Unit: km) When the landscape index was examined by road transects, we find that the curves of NP and Figure 9. Changes of landscape indices with distance from rivers. PA for the three Figurecounties 9. haveChanges similar of landscape downward indices trends with when distance far away from from rivers. the roads, while that of the LSI showed the opposite (Figure 10). The MPS of Tongyu was the highest among the three When the landscape index was examined by road transects, we find that the curves of NP and counties, and that of Dehui fluctuated slightly across the whole scale of the research. PAWhen for the three landscape counties index have was similar examined downward by road trends transects, when far we away find from that the the roads, curves while of NP that and PA The landscape shape index was quite variable across the three counties, with an opposite trend for theof the three LSI counties showed havethe opposite similar (Figure downward 10). The trends MPS when of Tongyu far away was fromthe highest the roads, among while the three that of the compared with NP, PA, and MPS analyzed by slope, proximity to rivers, and roads. A regional counties, and that of Dehui fluctuated slightly across the whole scale of the research. LSI showeduniversal the characteristic opposite (Figure is that the10). greater The MPS the slop of Tongyue, the greater was the index, highest and among the farther the threeaway counties,from The landscape shape index was quite variable across the three counties, with an opposite trend and thatthe ofroad, Dehui the greater fluctuated the index. slightly across the whole scale of the research. compared with NP, PA, and MPS analyzed by slope, proximity to rivers, and roads. A regional universal characteristic is that the greater the slope, the greater the index, and the farther away from 1000 200 the road, the greater theTongyu index. Dehui AnTu Tongyu Dehui AnTu 800 160 1000 200 ha) Tongyu Dehui AnTu 600 Tongyu Dehui AnTu 2 120 800 16010 400 × 80 NP ha)
600 2 120 200 40 10 PA (Unit: 400 × 80 NP 0 0 － － 4-6 － >10 200 0－22－4 4-6 6－10 >10 40 0 224 6 10 Distance (Unit: km) PA (Unit: Distance (Unit: km) 0 0 40 65 0－22Tongyu－4 Dehui4-6 6－AnTu10 >10 0－22－4Tongyu4-6 Dehui6－10 >10AnTu Distance (Unit: km) Distance (Unit: km) 30 60 40 65 Tongyu Dehui AnTu Tongyu Dehui AnTu 20 55 60
30 LSI
10 50 MPS (Unit:MPS ha) 20 55
0 LSI 45 10 － － 4-6 － >10 50 MPS (Unit:MPS ha) 0 224 6 10 0－22－4 4-6 6－10 >10 Distance (Unit: km) Distance (Unite: km) 0 45 0－22Figure－4 10. 4-6Changes6－10 of landscape>10 indices with0－ distance22－4 from4-6 roads.6－10 >10 Distance (Unit: km) Distance (Unite: km) Sustainability 2017, 9, 2170 12 of 16
The landscape shape index was quite variable across the three counties, with an opposite trend compared with NP, PA, and MPS analyzed by slope, proximity to rivers, and roads. A regional universal characteristic is that the greater the slope, the greater the index, and the farther away from the road, the greater the index. In order to further analyzethe association between rural settlements and geographic factors, we conducted some correlation analyses between the settlement patterns and influencing factors at the test sites. In Tongyu, only the relationship between the settlement density and slope was statistically significant (Table4; the total estimated settlement points in Tongyu was 666). The density of settlements in Dehui covaried positively with the elevation, slope, and proximity to water but covaried negatively with proximity to roads (the statistical number was 1793). For Antu (the statistical number was 581), the density of rural settlements was negatively correlated with DEM and the proximity to rivers and roads. The correlation coefficients were −0.411, −0.27, and −0.135 respectively, which was significant at the 0.01 level.
Table 4. Correlations between settlements density and influencing factors.
DEM Slope Proximity to Rivers Proximity to Roads Tongyu 0.028 0.087 * 0.03 0.013 Dehui 0.139 ** 0.071 ** 0.206 ** −0.104 ** Antu −0.411 ** 0.009 −0.270 ** −0.135 ** ** Means correlation was significant at the 0.01 level (bilateral). * Means correlation was significant at the 0.05 level (bilateral).
4. Discussion This paper compared the spatial distribution patterns of rural settlements in different regions, and studied the relationships between the rural settlement distribution and DEM, slope, distance to river, and road. These aspects of the paper provide a valuable contribution to our understanding of geographic dependence on the human landscape, which count a great deal towards formulating rural community planning according to local conditions.
4.1. Spatial Pattern of Rural Settlements We found that the density value in the middle part of Jilin province, typified by Dehui, was higher than the west and east parts of Jilin province because of substantially intensive agricultural use in this area. The west of Jilin province was one of the most important agricultural and livestock areas, but soil in this area was poor, and sandy land and salinized land were serious issues. The physical environment made a high density of rural settlements here impossible. For the eastern part, the physical conditions of the mountainous area limited the distribution of rural settlements, which brought a lower density and higher PSCV. The MPS of Tongyu (the value was 27.44 ha) was the highest among the three studied counties, even higher than the MPS of northeast China (the value was 21.46 ha [20]), indicating that the distribution of rural settlements tended to have a larger scale, but the larger PSSV showed a relatively large scale gap. In terms of the spatial pattern of rural settlements, we found that settlements in the three counties were all clustered. This is good news from habitat conservation and countryside planning standpoints, because it limits habitat loss [37]. Still, settlement clusters were not randomly located, and occurred largely within close proximity to better amenities and good environments. For example, in Antu, more settlements were distributed close to rivers than farther away (Figure9). This is troublesome, because there is intensive human activity in important ecosystems for many species that are easily damaged. Previous research studies also demonstrated that clustered settlements in artificial landscapes led to lower levels of habitat loss [37,38]. Therefore, our results suggested that not only the pattern of rural settlements, but also the location of the clustering, should be considered during conservation planning and countryside management. Thus, clusters must be located away from Sustainability 2017, 9, 2170 13 of 16 particularly sensitive and important habitats. In reality, the settlement patterns are very complex, and the measures seem to be more sensitive to the scale of analysis than what appears to be a centralized pattern to one better described as dispersed in other contexts [12]. Previous studies showed the rural settlements were agglomerated at a smaller scale (within 63 km [21]) or finer scale [37] when using Ripley’s K function. However, in our study, at the county scale, rural settlements showed clustering characteristics at a larger scale (over 100 km). We thought the results reflected the regional differences, to a certain extent.
4.2. Geographic Association The clustering of settlements may result from many factors. The localized distribution of resources and the emergence of political or regional centers have often been highlighted [12]. In reality, the determinants varied because of the analytical resolution and the size and shape of the study area [12]. All of the factors can be divided into two types: physical factors and socioeconomic factors. The rural settlement patterns were the result of combined physical and socioeconomic factors [39]. Natural geographic conditions were the direct driving factor of the location and layout of rural settlements, while socioeconomic factors were considered decisive driving factors that determine the degree and speed of rural settlement development [40]. Among all of these factors, the terrain factors and river were taken as the leading physical factors [40]. Socioeconomic conditions, including traffic location, policies, the level of economic development, and the size and structure of the village population all have great influences on the layout and development of rural residential areas. As demonstrated by previous studies, the regional location of the rural settlements was determined by the influences exerted spatially outside of that site, e.g., the latitude, annual mean precipitation, annual mean temperature at a large scale, topography, and the distance to water and roads at a small scale [20]. Tian found that DEM, annual mean temperature, annual mean precipitation, and latitude were all factors influencing the distribution of rural settlements in northeast China [20]. On a county scale, the spatial pattern of rural settlements was more likely associated with terrain, proximity to water and roads, and agriculture suitability [41]. After referring to previous research results and the regional conditions, we selected DEM, slope, and distance to water and roads to capture the geographic associations with the distribution of rural settlements. We found that there was obvious heterogeneity for geographic dependence on the human landscape. In our study, for example, settlement patterns of Tongyu were not determined predominantly by the proximity of neighbors, while those of Antu and Dehui were more closely correlated with physical factors. For Dehui, settlements were regularly distributed near the rivers, but were clustered far from them. This is very different from the river inclination of rural settlements in previous studies [37,41]. We speculate that due to the risk of flooding, settlements near the river were more sparse. However, for Antu, a mountainous county, more rural settlements were concentrated in the valley, within close proximity to the river. Dehui and Antu showed a closer relationship with physical factors, but represented different associations. All of these reflected the basis of the regional gradient of plain–hill–mountain. Roadswere taken to be information landscape corridors between rural settlement landscapes, as well as between rural settlement landscape and other landscapes. Previous studies showed that there was significant positive correlation between transportation land areas and rural settlement areas [39,42]. On the one hand, with the economic developments, the distribution pattern of residential settlements directly affects the direction and flow of traffic lines. On the other hand, the distribution of traffic lines will change the spatial distribution of settlement structures. More and more farmers prefer to dwell in the places that have convenient transportation [39]. In our study, the landscape indices of rural settlements in different buffer rangesnear roads were counted, and we found similar trends across the three counties: the closer to the road, the larger the number and area of rural settlements. However in Tongyu, correlation analysis showed that the correlation between rural patterns and road distance was not significant. According to the study results Sustainability 2017, 9, 2170 14 of 16 of Li [42], rural settlements and arable land showed an obvious symbiotic relationship. In Tongyu, land desertification and salinization changed the distribution pattern of arable land, and thus influenced settlement patterns, which represents the impact of agricultural suitability and the environmental adaptability of rural settlement patterns.
4.3. Distribution Pattern of Rural Settlements and Implications for Land Use Planning According to the characteristics of the density,scale, and landform of rural settlements, we generalized three distribution patterns for the west, middle and east of Jilin province. Pattern 1, represented by Tongyu, located in the west of Jilin province, is one of the most important agricultural and livestock areas. The physical environment for this region is characterized by flat alluvial plain, a semi-arid continental climate, lower precipitation, intensive evaporation, and poor soil conditions, resulting in serious desertification and salinization, and relatively low land productivity. The rural settlements in this region were relatively homogeneous and sparse, with obvious high-value clustering, indicting a larger scale. We labeled it the low-density, large-scale, and sparse type. For this type, we must fully consider the radiation of the regional center and ecological security, adjust the rural structure, and optimize the spatial layout. These goals can be realized by demolishing abandoned rural settlements and consolidating the scattered settlements. Pattern 2, represented by Dehui, which is one of the major grain production bases in China, is located in the black soil zone, and has a relatively flat alluvial and diluvial platform, fertile soil, a better environment, and higher land productivity. Many people live here, and the density of rural settlements is higher. We labeled it the mass-like and point-scattered type. For this type, we should form a development pattern with multiple centers, adjust the scattered and extensive land use, tap the internal potential of rural settlements, utilize the land in an economical and intensive way, reclaim the abandoned homestead, and consolidate the rural settlements in an orderly fashion. Pattern 3, represented by Antu, is called the Changbai mountain region, and is constituted by mountains and hills. Most of the scenery of this region is beautiful, so tourism has been developed. Limited by terrain, rural settlements in this region are centralized upon roads lying in the valley. The rural settlements were labeled as the low-density and high cluster-like type. For this type, we should improve the traffic conditions, strengthen the structure of public service facilities, make full use of the radiation effect of towns, and make reasonable planning and layouts for rural settlements.
5. Conclusions In this research, we compared the regional differences of rural settlements by several methods, and explored the geographic associations of settlement patterns. Our study shows that the rural settlements in three counties were all clustered, but there are different characteristics in the aspects of concentration scale, distribution density, landscape morphology, and so on. Different geographic associations were the basis of the regional gradient of plain–hill–mountain. Three patterns of rural settlements representing characteristics of different regions were summarized, and measures were put forward according to the patterns. It is expected that the studied results will be useful for land use planning and the implementation of the regional policy of “building a new countryside”. We should note that there are limitations of the study presented in this paper. For example, the results we obtained were restricted to the spatio-temporal scale of the data; the geographical determinants associated with spatial patterns we considered were only at a regional scale, and over one year. In addition, the study is static, and the settlement development over time was not taken into account. We used only morphological data—the distance from road and water—to quantitatively analyze the spatial characteristics of rural settlements in the studied counties. Other determinants, including physical and social variables, such as geology, land use, land cover, climate variables, the distance to natural amenities, and hazards, should be considered in the future. Furthermore, it is important to understand how to couple social and economic factors to build optimal regulation models. Sustainability 2017, 9, 2170 15 of 16
Acknowledgments: This work is supported by the Natural Science Foundation of China (No. 41671219) and the Regional Project: Comparative study on land use and geochemistry in global black soil zones (No. DD201601). Author Contributions: Xiaoyan Li conceived and wrote the paper; Yingnan Zhang and Huiying Li analyzed the data; and Limin Yang analyzed and summed up the literatures concerned. Conflicts of Interest: The authors declare no conflict of interest.
References
1. Long, H.L.; Liu, Y.S.; Wu, X.Q. Spatio-temporal dynamic patterns of farmland and rural settlements in Su-Xi-Chang region: Implications for building a new countryside in coastal China. Land Use Policy 2009, 26, 322–333. [CrossRef] 2. Woods, M. Rural Geography: Processes, Responses and Experiences in Rural Restructuring; Sage: London, UK, 2005. 3. Wang, C.; Wu, H.; Xu, H.C.; Zheng, J.; Zhou, H. Analysis on characteristics and distribution pattern of settlement in river valley: A case study in Fuping County of Hebei Province. Sci. Geogr. Sin. 2010, 21, 170–176. (In Chinese) 4. Huber, N.L. The role of land use planning in sustainable rural systems. Landsc. Urban Plan. 1998, 41, 83–91. 5. Shoshany, M.; Goldshelger, N. Land-use and population density changes in Israel—1950 to 1990: Analysis of regional and local trends. Land Use Policy 2002, 2, 123–133. [CrossRef] 6. Hammer, R.B.; Stewart, S.I.; Winkler, R.L.; Radeloff, V.C.; Voss, P.R. Characterizing dynamic spatial and temporal residential density patterns from 1940–1990 across the North Central United States. Landsc. Urban Plan. 2004, 69, 183–199. [CrossRef] 7. Radeloff, V.C.; Hammer, R.B.; Stewart, S.I. Rural and suburban sprawl in the U.S. midwest from 1940 to 2000 and its relation to forest fragmentation. Conserv. Biol. 2005, 3, 793–805. [CrossRef] 8. Manonmani, R.; Mary, G.; Suganya, D. Remote sensing and GIS application in change detection study in urban zone using multi temporal satellite. Int. J. Geomat. Geosci. 2010, 4, 339–348. 9. Modica, G.; Vizzari, M.; Pollino, M.; Fichera, C.R.; Zoccali, P.; Di Fazio, S. Spatio-temporal analysis of the urban–rural gradient structure: An application in a Mediterranean mountainous landscape (Serra San Bruno, Italy). Earth Syst. Dyn. 2012, 3, 263–279. [CrossRef] 10. Amy, K.; Hahs, M.; McDonnell, J. Selecting independent measures to quantify Melbourne’s urban–rural gradient. Landsc. Urban Plan. 2006, 78, 435–448. 11. Theohald, D.M. Land use dynamics beyond the urban fringe. Geogr. Rev. 2001, 91, 544–564. [CrossRef] 12. Andrew, B.; James, C. Multiscalar approaches to settlement pattern analysis. In Confronting Scale in Archaeology; Springer: Boston, MA, USA, 2007. 13. Kong, F.H.; Nakagoshi, N. Spatial-temporal gradient analysis of urban green spaces in Jinan, China. Landsc. Urban Plan. 2006, 78, 147–164. [CrossRef] 14. Andrén, H.; Delin, A.; Seiler, A. Population response to landscape changes depends on specialization to different landscape elements. Oikos 1997, 1, 193–196. [CrossRef] 15. Guo, X.D.; Zhang, Q.Y.; Ma, L.B. Analysis of the spatial distribution character and its influence factors of rural settlement in transition-region between mountain and hilly. Econ. Geogr. 2012, 32, 114–120. (In Chinese) 16. Long, H.L.; Li, Y.R.; Liu, Y.S. Analysis of evolutive characteristics and their driving mechanism of hollowing village in China. Acta Geogr. Sin. 2009, 64, 1203–1213. (In Chinese) 17. Wang, C.X.; Yao, S.M.; Chen, C.H. Empirical study on ‘village-hollowing0 in China. Sci. Geogr. Sin. 2005, 21, 257–262. (In Chinese) 18. Wu, Y.J.; Guo, F.; Zhang, S.W.; Zhang, Y.Z. The study on spatial pattern of urban and rural residential landuse in Jilin Province. J. Arid Land Resour. Environ. 2006, 1, 108–122. (In Chinese) 19. Zhang, Z.H.; Xiao, R.; Ashton, S.; Wu, J.P. Spatial point pattern analysis of human settlements and geographical associations in eastern coastal China—A case study. Int. J. Environ. Res. Public Health 2014, 11, 2818–2833. [CrossRef][PubMed] 20. Tian, G.; Qiao, Z.; Zhang, Y. The investigation of relationship between rural settlement density, size, spatial distribution and its geophysical parameters of China using Landsat TM images. Ecol. Model. 2012, 231, 25–36. [CrossRef] 21. Ma, X.D.; Qiu, F.D.; Li, Q.L.; Shan, Y.B.; Cao, Y. Spatial pattern and regional types of rural settlements in Xuzou city, Jiangsu Province, China. Chin. Geogr. Sci. 2013, 4, 482–491. [CrossRef] Sustainability 2017, 9, 2170 16 of 16
22. Robert, H. Describing and modeling rural settlements maps. Ann. Assoc. Am. Geogr. 1982, 2, 221–223. 23. Zomeni, M.; Tzanopoulos, J.; Pantis, J.D. Historical analysis of landscape change using remote sensing techniques: An explanatory tool for agricultural transformation in Greek rural areas. Landsc. Urban Plan. 2008, 86, 38–46. [CrossRef] 24. Chakraborty, J.; Armstrong, M.P. Exploring the use of buffer analysis for the identification of impacted areas in environmental equity assessment. Cartogr. Geogr. Inf. Syst. 2013, 3, 145–157. [CrossRef] 25. Fletcher, R. Some spatial analyses of Chalcolithic settlement in Southern Israel. J. Archaeol. Sci. 2008, 35, 2048–2058. [CrossRef] 26. Anderson, T.K. Kernel density estimation and K-means clustering to profile road accident hotspots. Accid. Anal. Prev. 2009, 41, 359–364. [CrossRef][PubMed] 27. Liu, X. Study on the Security of Grain Reserves in Northeast China. Ph.D. Thesis, Jilin University, Changchun, China, 2013. (In Chinese) 28. Wang, C.Y.; Tang, J.; Li, Z.Y.; Mao, Z.; Wang, X. Driving force analysis of land use and land cover temporal-spatial change in the west of Jilin province. Ecol. Environ. 2008, 17, 1914–1920. 29. Li, X.Y.; Zhang, S.W.; Wang, Z.M. Spatial variability and pattern analysis of soil properties in Dehui city, Jilin province. J. Geogr. Sci. 2004, 14, 503–511. 30. Definiens, A.G. Definiens Professional 8.6 User Guide; Definiens A.G: Munchen, Germany, 2011. 31. Lu, C.Y.; Wang, Z.M.; Li, L.; Wu, P.Z. Assessing the conservation effectiveness of wetland protect areas in Northeast China. Wetl. Ecol. Manag. 2016, 24, 381–398. [CrossRef] 32. Ripley, B.D. The second-order analysis of stationary point processes. J. Appl. Probab. 1976, 13, 255–266. [CrossRef] 33. Besag, J.E. Commentson Ripley’s paper. J. R. Stat. Soc. Ser. B 1977, 40, 193–195. 34. McGarigal, K.; Marks, B. Fragstats—Spatial Pattern Analysis Program for Quantifying Landscape Structure; Forest Science Department, Oregon State University: Corvallis, OR, USA, 1994. 35. Riitters, K.H.; O’Neill, V.; Hunsaker, C.T.; Wickham, J.D.; Yankee, D.H.; Timmins, S.P.; Jones, B.K.; Jackson, B.L. A factor analysis of landscape pattern and structure metrics. Landsc. Ecol. 1995, 10, 23–39. [CrossRef] 36. Uuemaa, E.; Mander, Ü.; Marja, R. Trends in the use of landscape spatial metrics as landscape indicators: A review. Ecol. Indic. 2013, 28, 100–106. [CrossRef] 37. Gonzalez-Abraham, C.E.; Radeloff, V.C.; Hawbarker, T.J.; Hamer, R.B.; Stewart, S.I.; Clayton, M.K. Patterns of houses and habitat loss from 1937 to 1999 in Northern Wisconsin, USA. Ecol. Appl. 2007, 17, 2011–2023. [CrossRef][PubMed] 38. Odell, E.A.; Theobald, D.M.; Knight, R.L. Incorporating ecology into land use planning-the songbirds’ case for clustered development. J. Am. Plan. Assoc. 2003, 69, 72–82. [CrossRef] 39. Jiang, G.H.; Zhang, F.R.; Qin, J. Relationship between distribution changes of rural residential land and environment in mountainous areas of Beijing. Trans. CSAE 2006, 22, 85–92. 40. Gude, P.H.; Hansen, A.J.; Rasker, R.; Maxwell, B. Rates and drivers of rural residential development in the Greater Yellowstone. Landsc. Urban Plan. 2006, 77, 131–151. [CrossRef] 41. Min, J.; Yang, Q.Y.; Weng, C.Y. The analysis of spatial distribution and optimization on rural settlement based on villageregion. Chin. Agric. Sci. Bull. 2012, 28, 283–290. 42. Li, D.M.; Wang, D.Y.; Li, H. The effects of rural settlement evolution on the surrounding land ecosystem service values: A Case Study in the Eco-Fragile Areas, China. ISPRS Int. J. Geo-Inf. 2017, 6, 49. [CrossRef]
© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).