Measurement of Land Use Pattern__ Based on

0 downloads 0 Views 109KB Size Report
that mostly Cultivated land and Forest land converted to Grass land, which could be concluded that there is a trend that. Cultivated land and Forest land was ...
Measurement of Land Use Pattern Based on Spatial Adjacency 1

Lina Lv1, Xinqi Zheng1,2*, Jing Deng1, Dept. Land Science and Technology in China University of Geosciences, Beijing, China 2 Institute of Land Information technology and Application, Beijing, China * Corresponding author: [email protected]

Abstract—The land use structure can be understood as a concept of land use types’ composition, contrast relations and spatial patterns, which including the numeral structure and spatial structure. Spatial distribution characteristics and spatial relationships of land use are important contents in land use planning. The precondition of traditional statistical methods and interpretation is that the data must be statistically independent and uniformly distributed. However, spatial data often have dependencies on themselves which were named spatial correlation, and the value of a variable is measured different as the distance unit changed. In current research, traditional qualitative analysis method is usually adopted, but quantitative analysis method is rarely used. However, the attribute of land use class is nominal variable, which is the simplest property type, namely: the measurement of geographical entities, the classification of geographical entities. The function is only to distinguish particular entities. Therefore, it is meaningless for arithmetic calculations. Based on land use types classification, this paper proposed a spatial correlation index named adjacency index to quantitatively analyze spatial correlation of nominal variables, and used Visual Studio platform to realize the algorithm’s implementation. In the end, this research set Changping District of Beijing as an example. Through adjacency, it could be concluded that cultivated land and forest land distributed more concentrated, while transportation land had large adjacency with unused land, which was basically accord with actual conditions. Based on the spatial distribution location of polygons, and their auto-adjacency index changes, it showed that mostly Cultivated land and Forest land converted to Grass land, which could be concluded that there is a trend that Cultivated land and Forest land was changing to Grass land. Results showed that the method could not only quantitatively measure spatial autocorrelation, and characterize spatial distribution and spatial relations of land use, but also predict the trend of land use changing pattern. Keywords- spatial correlation; nominal variables; adjacency index; land use

I.

INTRODUCTION

Land use planning is an activity, through which to arrange various different functional areas to achieve economic and social development. Direct object of land use planning is land parcel; therefore, the main content of land use planning is reasonable arrangements of land use. Currently, land use has a regular feature of the spatial distribution, which plays a very important role in urban planning and land use planning. The land use structure can be understood as a concept of land use types’ composition, contrast relations and spatial patterns,

which including the numeral structure and spatial structure [1] .The current research mainly focuses on macro-regional urban areas, while few studies on micro-plot land use types. This may lead to lacking of pertinence and direction deviating when relevant administrative departments make land use planning and policy. However, as the inherent complexity, multiplicity existing in land use process, how to use quantitative model to simplify and abstract spatial pattern of land use as well as development process may become research tendency in the future. The precondition of traditional statistical methods and interpretation is that the data must be statistically independent and uniformly distributed. However, spatial data often have dependencies on themselves which were named spatial autocorrelation, and the value of a variable is measured different as the distance unit changed. Conventional statistic method is the most commonly used quantitative method with which to analysis restriction factors of land use spatial distribution [2]. A problem occurs while using conventional statistical methods, because these methods assume the data to be statistically independent while spatial land use data have the tendency to be dependent, which is known as spatial autocorrelation [3]. Land utilization spatial data tend to have a certain spatial autocorrelation and contain some useful information which need appropriate methods to solve. Since Gould first put forward in 1970 [4], some scholars had proposed spatial autocorrelation analysis on Moran 'I index [5] and spatial autoregressive model [6]. Related research has a long history and obtains a series of achievements. In 2003, Overmars K P [7] proposed that land use might have spatial autocorrelation and need to make spatial autocorrelation analysis and space regression analysis for research. Gao Kai analyzed the land use quantitative structure and its spatial autocorrelation characteristic in the Yangtze River Basin based on WESTDC_ Land_ Cover_ Products 2.0dataset which was reclassified as cultivated land, forestland, grassland, water, urban and rural construction land and other unused land in terms of CAS resource and environment classification system[8]. XIE Zhengfeng studied land use structure change by measuring spatial autocorrelation change based on Moran ’s I of land use degree of Guangzhou in 1990, 1995, 2000 and 2004 [9]. In QIU Bingwen’s research, Moran’s I is used to describe spatial autocorrelation of dependent and independent variables; spatial autoregressive models which incorporate both regression and spatial autocorrelation were constructed basing on six land use types and their 27 candidate driving force

variables [10]. XIE Hualin constructed Moran's I and spatial lag models to describe the spatial autocorrelation of Inner Mongolia [11]. Usually, according to analytical methods, spatial data can be divided into nominal, ordinal, interval, and ratio variable [12]. An attribute in a data set can be classified as either being numerical or categorical, depending on whether its values can be described with standard numeric. The categorical attribute can be further divided as ordinal and nominal, depending on whether the sequence of the values could be defined [13]. In this paper, the attribute of land use class is nominal variable, which is the simplest property type, namely: the measurement of geographical entities, the classification of geographical entities. The function is only to distinguish particular entities. Therefore, it is meaningless for arithmetic calculations. This paper attempts to explore a method based on nominal variables (land type code) to measure spatial autocorrelation, in order to analyze land use spatial distribution in quantitative way. II.

Spatial autocorrelation analysis can do correlation analysis based on spatial relationship, it can take account of values and two-dimensional space which traditional statistical model cannot [16].Spatial autocorrelation is an important content of exploratory spatial data analysis and it refers to the relation of attribute values. Researchers can inspect certain geographical phenomena or a certain attribute value distribution through spatial autocorrelation analysis, they can also find out gathered characteristics in space between phenomenon or attribute value exist, including global spatial autocorrelation and local spatial autocorrelation [17-18]. The most commonly used autocorrelation is Moran's index, n

I =

n

i =1

indicators

w ij ( x i − x )( x

j

of

spatial

− x)

∑ ( xi − x) 2

n

n

i=1

j≠ i

∑ ∑

x =

n

1 n

j

In this formula: I - Moran index;

xi



i=1

xi

- the observation value

wij

of the region i; - matrix of spatial weights; values of Moran index I are generally in [-1, 1], the indicates less than 0 means a negative correlation, equal to 0 indicates not relevant, greater than 0 shows a positive correlation. B. Spatial Adjacency Index The current land use spatial autocorrelation analysis reflects autocorrelation by margin. There is no significance to calculate the difference of nominal variables aforementioned, and the current use of spatial autocorrelation index measuring method is established on general cross-product statistics [19-20] which was based on matrix operations, such as formula (2) shown.

Γ=

∑∑ i

W ij C ij

j

(2)

In formula (2), W represents space degree in statistical unit,C means similarity of property in statistical unit. It is impossible to compare nominal variables directly, so we need to re-construct autocorrelation index of nominal variables. For the simplicity of nominal variables properties, the construction of the similarity matrix is relatively simple, such as formula (3) shown [5]. C =

1 0

A =A A ≠A

(3)

Therefore, this paper drew on a measure index - Adjacency Index, which is used to describe spatial correlation. This index is usually presented in the concept of landscape ecology, and reflecting the heterogeneity of spatial pattern. It is the matrix that was expressed by boundary representation of space adjacent relationship in essence, and it includes relatively rich information compared to the general adjacency matrix, which is shown in formula (4).

Pij

Wij Pij

=

Pi

(4)

∑P

ij

=

j

(5)

In the formula above: P means the public arc length between the geographical entities of i and j, P represents the circumference of entity i. To sum up, nominal variables’ adjacency index can be derived as follows:

j =1 S 2

1 n

METHODS

A. Spatial autocorrelation Waldo Tobler’s First Law of Geography proposed: Everything is related to everything else, but closer things are more closely related" [14]. Spatial autocorrelation is a correlation among different spatial locations of a single variable. This is a common phenomenon for geochemical variables, as their spatial distribution patterns are controlled by several factors (such as rock type, soil type, etc.) and thus unlikely to be randomly distributed (except in rather small areas where the controlling factors are uniformly distributed) [15]. The identification of spatial autocorrelation is crucial for geochemical research. Spatial autocorrelation analysis is a method which can statistically analyze the degree of spatial autocorrelation about a certain characteristic between regions and its surroundings, in order to analyze the spatial distribution of these features in the area.

∑ ∑

S2 =

w ij

(1)

L=

∑∑ Pij A = A , i ≠ j i j i j ∑ Pi

Create SHP File

i

Since L is evolved from a neighboring index concept, and it represents self-similarity of geographical entities, so the reflected relationship is called spatial adjacency. Here, L is a reflection of global spatial autocorrelation index, and therefore is called adjacency index. In formula (6), value of L is (0, 1), which indicates overall stronger autocorrelation with a greater value. Also that we named C. Realization of Adjacency Index Calculating This paper implemented algorithms under the Visual Studio platform though C # programming. The specific process is as follows: Firstly, check the around adjacent block polygon, and calculate public arc length of adjacent polygons. And then, calculate ratios of the perimeter of the polygon, so that adjacency index between adjacent polygons can be obtained. Finally, distinguish the properties and store their adjacency in the corresponding adjacent index column.

III.

Calculate Local Spatial Autocorrelation

(6)

CASE ANALYSIS

A. Data Acquisition and Processing This research used land use data of Changping District in 2001, 2005, 2009, which are obtained from Beijing Land Change Survey, see Figure 1. The conversion of land type codes and calculation of adjacency are achieved through the C # language program on the Visual Studio platform, Figuer 2 shows the data processing.

Classify Adjacency Index

Convert Land Types

Classify Land Types

Calculate Adjacency Index

Figure 2. Data Processing Procedure

B. Results Discussion In the procedure of data processing, this research divided land use types into eight classes, namely: cultivated land, garden land, forest land, grass land, residential construction land, transportation land, water, and unused land. Then this research validated algorithm through Changping District land use data based on the aforementioned calculation method. Spatial adjacency data of land use of Changping District in2001, 2005, 2009 are calculated as follows. TABLE I.

Statistics of land use Adjacency of Changping District in 2001

TYPE

L1

L2

L3

L4

L5

L6

L7

L8

1

0.188

0.035

0.049

0.000

0.134

0.008

0.067

0.029

2

0.049

0.166

0.150

0.000

0.126

0.005

0.037

0.051

3

0.058

0.062

0.219

0.000

0.104

0.022

0.055

0.043

4

0.000

0.251

0.749

0.000

0.000

0.000

0.000

0.000

5

0.102

0.087

0.111

0.000

0.169

0.025

0.053

0.049

6

0.049

0.012

0.083

0.000

0.164

0.132

0.016

0.026

7

0.111

0.038

0.098

0.000

0.140

0.003

0.129

0.041

8

0.069

0.085

0.186

0.000

0.143

0.005

0.048

0.105

1- Cultivated land, 2- Garden land, 3- Forest land, 4-Grassland, 5- Residential and Industrial land 6-Transportation land, 7- Waters, 8-Unused land. TABLE II.

Statistics of land use Adjacency of Changping District in 2005

Figure 1. Land Use Pattern of Changping District.

TYPE

L1

L2

L3

L4

L5

L6

L7

L8

1- Cultivated land, 2- Garden land, 3- Forest land, 4-Grassland, 5- Residential and Industrial land 6-Transportation land, 7- Waters, 8-Unused land.

1

0.179

0.035

0.060

0.001

0.125

0.016

0.054

0.025

2

0.043

0.158

0.134

0.000

0.120

0.005

0.034

0.043

3

0.058

0.051

0.185

0.002

0.099

0.038

0.053

0.034

4

0.092

0.032

0.100

0.080

0.135

0.025

0.083

0.006

5

0.081

0.068

0.091

0.001

0.172

0.019

0.046

0.038

6

0.056

0.014

0.061

0.001

0.109

0.144

0.027

0.016

7

0.086

0.035

0.095

0.003

0.129

0.013

0.126

0.035

8

0.057

0.073

0.143

0.000

0.127

0.008

0.042

0.111

1- Cultivated land, 2- Garden land, 3- Forest land, 4-Grassland, 5- Residential and Industrial land 6-Transportation land, 7- Waters, 8-Unused land.

TABLE III.

Statistics of land use Adjacency of Changping District in 2009

TYPE

L1

L2

L3

L4

L5

L6

L7

L8

1

0.203

0.029

0.045

0.014

0.127

0.006

0.061

0.023

2

0.046

0.182

0.130

0.003

0.124

0.005

0.034

0.045

3

0.054

0.058

0.207

0.006

0.102

0.022

0.054

0.039

4

0.130

0.024

0.055

0.206

0.109

0.011

0.040

0.021

5

0.092

0.075

0.091

0.007

0.174

0.021

0.051

0.046

6

0.045

0.011

0.083

0.006

0.156

0.120

0.017

0.024

7

0.102

0.034

0.080

0.007

0.133

0.004

0.146

0.039

8

0.060

0.079

0.133

0.005

0.138

0.005

0.048

0.126

1- Cultivated land, 2- Garden land, 3- Forest land, 4-Grassland, 5- Residential and Industrial land 6-Transportation land, 7- Waters, 8-Unused land.

Results of adjacency index in Table I,II,III reasonably reflected a spatial accumulation state. For the research area, they directly indicated the status of land use intensive. The results showed that local adjacency index of various land use types were discrepant. In accordance with Adjacency Index, the results were divided into three categories: (1) Cultivated land had significant autocorrelation, which indicated that cultivated land distribution is relatively concentrated. Forest land also had significant autocorrelation, which indicated that distribution of forest land was concentrated and it formed a group pattern. (2) The mean of transportation land is less than 0.15 which presented that distribution of transportation land was linear from north to south in Changping District. On the one hand the mean was less significant; on the other hand, the mean changes not much, and the maximum value was small. This is associated with its shape, which belongs to linear geographical factors. Transportation land had large association with other geographic elements, so it had the weakest autocorrelation and the adjacency index was not obvious. (3) Other types of land use had concentration index mean. But unused land distributed more dispersedly, mainly in the northern region, and unused land needed to be integrated. TABLE IV.

Statistics of land use Auto-adjacency of Changping District in 2001,2005,2009

TIME

TYPE 1

TYPE 2

TYPE 3

TYPE 4

TYPE 5

TYPE 6

TYPE 7

TYPE 8

2001

0.188

0.166

0.219

0.000

0.169

0.132

0.129

0.105

2005

0.179

0.158

0.185

0.080

0.172

0.144

0.126

0.111

2009

0.203

0.182

0.207

0.206

0.174

0.120

0.146

0.126

1- Cultivated land, 2- Garden land, 3- Forest land, 4-Grassland, 5- Residential and Industrial land 6-Transportation land, 7- Waters, 8-Unused land

Through table IV, it can be seen that the auto-adjacency index had shown an increasing trend in the nine years since 2001. The reason of this phenomenon is due to contiguous pattern of development of land use, especially based on the related land use overall planning and policies, which focused on improving land utilization. Also we could see that the auto-adjacency index of Grass land had the biggest changes, its auto-adjacent index was only the 0 and it had no adjacent to other land use types. While the number of Grass land polygons in 2005 had reached to 169 and the auto-adjacency index had increased from 0 to 0.206 since 2001. This increasing of Grass land is not only on the number of polygons, but also on area of contiguous

regional scope. Based on the spatial distribution location of polygons, and their auto-adjacency index changes, it showed that mostly Cultivated land and Forest land converted to Grass land, which could be concluded that there is a trend that Cultivated land and Forest land was changing to Grass land.

IV.

CONCLUSION

In this paper, the research defined adjacency index, implemented its algorithm, and calculated spatial correlation indices based on nominal variables, which provided better visualization. Through the analysis of the result, it can be found that adjacency reflect spatial association characteristics of land use. Through adjacency, it could be concluded that cultivated land and forest land distributed more concentrated, while transportation land had large adjacency with unused land, which was basically accord with actual conditions. It could further verify adjacency’s rationality of definition and realization. However, the current relative definitions of the research are not perfect, such as, definition of adjacency index based on adjacent area, and measurement methods; what ’s more, the theoretical support is not well developed. ACKNOWLEDGMENT (HEADING 5) This research was supported by the National Public Benefit (Land) Research Foundation of China (No.20101018) and the Science and Technology Innovation Program for Graduate Student in China University of Geosciences, Beijing, China (No.2009121105). REFERENCES [1]

Chen Qichun , LüChengwen , Li Bicheng,Quan Bin. Quantitative analysis of land use structure characteristics at county scale [J]. Transactions of the Chinese Society of Agricultural Engineering, 2009,25(1):223-231. [2] Veldkamp A,Fresco L O.Reconstructing land use drivers and their spatial scale dependence for Costa Rica(1973 and 1984)[J].Agricultural System,1997,55:19-43. [3] K.P. Overmars a,b, G.H.J. de Koning c, A. Veldkamp b. Spatial autocorrelation in multi-scale land use models[J]. Ecological Modelling , 2003,164: 257–270. [4] Gould P R.Is. Statistics inference the geographical name for a wild goose[J]. Economic Geography,1970,46:439-448. [5] Cliff A D,Ord J K. Spatial Processes: Models and Applications[M]. London: Pion,1981. [6] Anselin L. Spatial Econometrics:Methods and Models[M]. Dordrecht : Kluwer Academic Publishers,1988.284. [7] Overmars K P,De Koning G H J,Veldkamp A.Spatial autocorrelation in multi-scale land use models[J].Ecological Modelling,2003,164:257-270 [8] GAO Kai1, ZHOU Zhi-xiang, YANG Yu-ping. Land use structure and its spatial autocorrelation analysis in the yangtze river basin [J]. 长 Resources and Environment in the Yangtze Basin,2010,(S1):13-20. [9] XIE Zhengfeng, WANG Qian. Spatial Autocorrelation Analysis of Land Use Intensity in Guangzhou City[J]. Tropical Geography,2009,(2):129-133. [10] QIU Bing-wen, WANG Qin-min, CHEN Chong-cheng, CHI Tian-he. Spatial Autocorrelation Analysis of Multi-scale Land Use in Fujian Province[J]. Journal of Natural Resources,2007,(2). 311-320. [11] XIE Hualin, LIU Liming, LI Bo, ZHANG Xinshi. Spatial Autocorrelation Analysis of Multi-scale Land-use Changes:A Case Study in Ongniud Banner, Inner Mongolia [J]. Acta Geographica Sinica,2006,(4):389-400.

[12] Wang Yuanfei. Spatial Data AnalysisM]. Beijing: Science Press, 2007,6. [13] Josep Domingo-Ferrer, Agusti Solanas. A measure of variance for hierarchical nominal attributes[J] (vol 178, pg 4644, 2008)[J]. INFORMATION SCIENCES, 2009, 20(179): 3732-3732. [14] Tobler W. A computer movie simulating urban growth in the Detroit region. Economic Geography, 1970, 46(2): 234-240. [15] Chaosheng Zhanga and Olle Selinus. Spatial analyses for copper, lead and zinc contents in sediments of the Yangtze River basin[J]. Science of The Total Environment ,1997,3(204):251-262. [16] LIU Deqin, LIU Yu, XUE Xinyu. Spatial Distribution and Autocorrelation Analysis of Population in China [J]. Science of Surveying and Mapping,2002(6):1-6.

[17] DALE M R T , FORITIN M J. Spatial autocorrelation and staticalcal tests:some solutions [J].Journal Of Agricultural Biological and Environmental Statistics, 2009,14(2):188-206. [18] DORMANN C F, MCPHERSON J M, ARAUJO M Bl. Methods to account for spatial autocorrelation in the analysis of species distributional data: A review[J]. Ecography, 2007,30(5):609-628. [19] Hubert L J, Golledge R G, Costanzo C M. Generalized Procedures for Evaluating Spatial Autocorrelation[J]. Geographical Analysis, 1981,13(3): 224-233. [20] LIU Yongxue; ZHANG Renshun; LI Machun. Automatic Extracting Method of Land Cover in Jiangsu Tidal Flat [J]. Remote Sensing Information,2004,(1):23-26.