CN102622503B - Urban land change analogy method - Google Patents

Abstract

A kind of urban land change analogy method.It incorporates the thought of Dynamic Neighborhood in cellular automaton and monte-Carlo model, according to the distribution of different cellular state and Monte Carlo method adjustment local neighborhood range size, solve the uncertain problem of urban land change on Spatial Dimension and the stochastic problems on time dimension, realize urban land change simulation.Its method comprises: data acquisition and pre-service; According to classification chart determination cellular size, cellular state and iteration time, Monte Carlo is used to determine Dynamic Neighborhood scope; Set up the simulation of land use changes system based on Monte Carlo and cellular Automation Model; The simulation of land use changes system of Monte Carlo and cellular automation method is used to carry out simulation and forecast; Based on the Analysis of Land Use Structure of analog result.Instant invention overcomes the uncertain problem run in urban land change simulation, improve effect and the precision of simulation.

Description

Urban land change analogy method

Technical field

The present invention relates to urban land change analogy method.The thought of main utilization Dynamic Neighborhood, combines Monte Carlo method with cellular automaton, realizes urban land change simulation, thus provides reference and the decision-making foundation of science for urban land resource planning management and sustainable development.Belong to field of earth sciences.

Background technology

Land change is the remarkable form of expression of global environmental change, and along with going deep into and development of global environmental change research, land change causes great attention and the concern of numerous scholar, has become the important science subject under discussion of 21 century at present.Along with China's population, the becoming increasingly conspicuous of resources and environment problem, the important research content that urban land change research has become in the sustainable development of current city.Strengthen the simulation of urban land change, be conducive to Optimizing City land use structure and pattern, promoting sustainable use and the development level of urban land resource, for deepening research and understanding urban environment change global behavior and operating mechanism, all there is important theory value and scientific meaning.

Domestic and international many experts and scholars have carried out a large amount of useful exploration to urban land change simulation.Finn have studied tolerance and the statistics that Markov chain is applied to land use change survey.De & Velloso etc. use fuzzy set theory and remote sensing technology to monitor land cover pattern change.Cellular Automation Model is applied to land use change survey and the growth simulation of urban slum in Brazil Bauru by DeAlmeida etc.Comber & Law etc. has inquired into utilization expertise and has automatically monitored land cover pattern change.Zhang Xinchang and Liang Jincheng have studied the Land use changes prediction of Markov Model.Li Xia and Liu Xiaoping propose the CA model of case-based reasioning (CBR).Li Yuechen and He spring sun-light proposes the Land_use change/covering change Spatio-temporal Evolution simulation based on artificial neural network and CA.In addition, the method such as intelligent agent technology (Yang Qingsheng, Li Xia etc.) and support vector machine (Xiong Hua, Liu Yaolin etc.) is also applied to simulation of land use changes research.

Comprehensive present Research both at home and abroad, at present two classes are mainly divided into for urban land change analogy method: (1) mathematical statistics method: as Markov model, regression analysis model, statistical model etc., land change process is considered as random transition process by these methods, lacks the consideration to urban land change spatial coherence.(2) intellectual evolution method: as cellular automaton, artificial neural network, genetic algorithm and support vector machine etc., this class model is based upon on the system emulation basis to process of land use change, but there is computation process complexity, be difficult to hold shortcoming.In order to overcome deficiency and the defect of the existence of above-mentioned model, need to introduce new technical method and thinking.

Summary of the invention

The object of the invention aims to provide a kind of urban land change analogy method, overcomes the deficiency of above-mentioned model existence and lacks

One of the present invention is towards urban land change analogy method, Monte Carlo method is combined with cellular automaton, according to the distribution of different cellular state and Monte Carlo method dynamic conditioning local neighborhood range size, solve the uncertain problem of urban land change on Spatial Dimension and the stochastic problems on time dimension, realize urban land change simulation.

One of the present invention is towards urban land change analogy method, and its step is followed successively by:

1) data acquisition and pre-service

Obtain the remote sensing image of survey region two different year, utilize ArcGIS, ERDAS software to Remote Sensing Image Matching, cutting and geometry correction process, finally carry out image interpretation classification, land use pattern is divided into waters, unused land, construction land and greenery patches four class, thus obtains the land use classes figure in two periods;

2). according to classification chart determination cellular size, cellular state and iteration time, use Monte Carlo method determination Dynamic Neighborhood scope;

(1) Dynamic Neighborhood is selected: for general cellular, uses the Moore neighborhood of 3 × 3, if the land use classes of cellular is tending towards average in neighborhood, then utilizes the Moore neighborhood of 5 × 5;

(2) the neighborhood type determined according to previous step calculates each land type number percent in cellular contiguous range;

(3) based on Monte Carlo random number method establishment cellular transformation rule

First the land type number percent obtained according to previous step obtains land type number percent vector

P＝[P ₁P ₂.....P _k](5)

It meets:

$\underset{i=0}{\overset{k}{\mathrm{\Σ}}}{P}_k=1,i=\mathrm{1,2},...,k---\left(6\right)$

Accumulation calculating is done to the element of land type number percent vector, obtains cumulative land type number percent vector

$P^{\left(s\right)}=\left[\begin{array}{ccc}{P}_1^{\left(s\right)}& P_{21}^{\left(s\right)}& ......P_k^{\left(s\right)}\end{array}\right]---\left(7\right)$

In formula:

${P_i}^{\left(s\right)}=\underset{n=0}{\overset{i}{\mathrm{\Σ}}}{P}_n,i=\mathrm{1,2},...,k---\left(8\right)$

Like this, P ^(s)in k element just define interval (0,1) one segmentation, k sub-range (0, P ₁ ^(s)], (P ₁ ^(s), P ₂ ^(s)] ..., (P _k-1 ^(s), P _k ^(s)] the corresponding result phase 1,2 of difference ..., k; Utilize uniform random number to produce function like this and carry out stochastic prediction, when namely the cellular land use pattern of subsequent time being simulated, just produce one at (0,1) upper equally distributed random number r, if r ∈ is (P _n-1 ^(s), P _n ^(s)), then judge that the state of subsequent time cellular is as n;

3). set up the simulation of land use changes system based on Monte Carlo and cellular Automation Model

Take Visualstudio2008 as development platform, under WindowsXP operating system environment, develop the simulation of land use changes system based on cellular automaton and Monte Carlo method, this system is mainly divided into two large modules: GIS basic function module and urban land change analog module;

A) GIS basic module mainly comprises some basic functions relevant with GIS, comprise data input and output, data query, data store, map amplifies, reduces, layer management, range finding, attribute query and hawk-eye function;

B) simulation of land use changes module, this module realizes simulation of land use changes forecast function based on Monte Carlo and cellular Automation Model, and in conjunction with the spatial analysis of GIS and thematic charting function, the quantity statistics and the thematic map that realize land use pattern are drawn;

4). use the simulation of land use changes system of Monte Carlo and cellular automation method to carry out simulation and forecast

Based on the land use pattern figure in the first period, simulate the state of each cellular subsequent time land use pattern, when analog result and next in period land use pattern the degree of fitting simulation precision that reaches setting be more than or equal to 90% time, the simulative iteration number of times of record cast, determine simulation iterations in the unit interval, on this basis, simulation obtains the land use change prediction figure in other periods;

5). based on the Analysis of Land Use Structure of analog result

The number percent that the quantity of each land use pattern in this period and each land use pattern account for sum is added up according to analog result, the further calculating land use pattern matrix of transition probabilities and conversion surface product matrix, thus draw conversion direction, all kinds of soil and conversion quantity, according to acquisition land use change survey information, relevant departments can take corresponding policy making steps, effectively plan and manage Land_use change.

The advantage of the inventive method and feature:

(1) apply in the transformation rule of cellular Automation Model by Monte Carlo method, the random number simulation ability utilizing Monte Carlo powerful solves the uncertain problem that cellular automaton runs in urban land change simulation.

(2) propose the thought of Dynamic Neighborhood, the cellular during urban land change is simulated develops and expansion is more reasonable and accurately, improve effect and the precision of simulation of land use changes.

Accompanying drawing explanation

An example of Fig. 1 cellular neighborhood

Fig. 2 uses CA-monte-Carlo model to simulate land use change survey process flow diagram

Embodiment

Urban land change analogy method flow process, as shown in Figure 2, comprises the following steps:

1. data acquisition and pre-service

Obtain the remote sensing image of survey region two different year, utilize the softwares such as ArcGIS, ERDAS to image procossing such as Remote Sensing Image Matching, cutting, geometry corrections, finally carry out image interpretation classification, in the present invention, land use pattern is divided into: waters (river, lake), unused land, construction land, greenery patches (farmland, meadow, forest land) four classes, thus obtains the land use classes figure in two periods.

2. based on the cellular Automation Model of Monte Carlo method

Cellular size, cellular state i (n=1 is determined according to classification chart (size of image and resolution), 2,3 ..., k. be the one in k kind land use pattern) and iteration time t, use Monte Carlo method determination Dynamic Neighborhood scope.

CA model can be expressed as form:

S ^t+！＝f(S ^t，N)(1)

In formula, S is state; F is contiguous function; N is nearby sphere; T is the interative computation time; The state when state of cellular depends on t during t+1 in nearby sphere.L is the state set of cellular: L={1,2 ..., k}.In an experiment, k=m.1,2 ..., m} represents m kind land use pattern respectively: waters, unused land, construction land, greenery patches ...

1) Dynamic Neighborhood is selected

In 3 × 3 neighborhood situations of calculating, in contiguous range, each land use pattern number accounts for number percent P (i) of sum.First the number g (L of the i-th class land use pattern in each cellular contiguous range is added up _i); G (L before calculating _i) initial value be all set to 0, whenever search the cellular of an i kind land type in contiguous range: g (L _i)=g (L _i)+1, i=1,2, ..., k, considers that the land use pattern of center cellular also has impact (if a certain cellular is construction land, so subsequent time has certain driving force and makes it still keep construction land constant) to subsequent time center cellular type, and this impact is greater than the impact (being set to the twice of neighborhood cellular here) of neighborhood cellular, so when center cellular is i kind land type: g (L _i)=g (L _i)+2, i=1,2 ..., k.Such as, suppose to there is certain cellular, as shown in Figure 1, its neighborhood is the Moore type of radius r=1.

Land use pattern number result in statistics contiguous range is:

g(L ₁)＝1+1＝2；g(L ₂)＝1+1＝2；g(L ₃)＝2+1＝3；g(L ₄)＝1+1+1＝3(2)

Then calculate each land type cellular number in cellular contiguous range and account for number percent p (i) of total cellular number.

$P\left(i\right)=\frac{g\left(L_i\right)}{\underset{n=0}{\overset{K}{\mathrm{\Σ}}}g\left(L_n\right)},n=\mathrm{1,2},...,k$ And meet $\underset{i=1}{\overset{k}{\mathrm{\Σ}}}p\left(i\right)=1,i=\mathrm{1,2},..,k.---\left(3\right)$

Often plant land type when the neighborhood cellular of 8 in 3 × 3 contiguous range and have two.The state of center cellular subsequent time is just random completely in this case, and transformation rule does not play obvious effect.Therefore, use the thought of Dynamic Neighborhood, expand the contiguous range of this cellular to 5 × 5 neighborhoods, allow more cellular participate in statistics and transfer process, make cellular evolution and expansion more reasonable and representative.Ultimate principle is: for general cellular, uses the Moore neighborhood of 3 × 3, if the land use classes of cellular is tending towards average in neighborhood, then utilizes the Moore neighborhood of 5 × 5.

$p\left(i\right)=\frac{g\left(L_i\right)}{\underset{n=0}{\overset{k}{\mathrm{\Σ}}}g\left(L_n\right)},n=\mathrm{1,2},...,k---\left(4\right)$

Analyze known p (1), p (2) ..., the value of the maximal value p (max) in p (k) is in the interval of (0.2,1).More close to 0.2, illustrate that land use pattern is overstepping the bounds of propriety loose, be more suitable for use 5 × 5 neighborhood; More close to 1, illustrate that land use pattern distribution is more concentrated, be more suitable for use 3 × 3 neighborhood.So the random number r in Monte Carlo method generation (0.2,1) interval can be used, if r is ∈ (0.2, p (max)), then use 3 × 3 neighborhoods; If r is ∈ (p (max), 1), then use 5 × 5 neighborhoods.

2) the neighborhood type determined according to previous step calculates each land type number percent in cellular contiguous range

Add up the number g (L of the i-th class land use pattern in each cellular contiguous range _i).G (L before calculating _i) initial value be all set to 0, whenever search the cellular of an i kind land type in contiguous range: g (L _i)=g (L _i)+1, i=1,2 ..., k, when center cellular is i kind land type: g (L _i)=g (L _i)+2, i=1,2 ..., k.Then calculate each land type cellular number in cellular contiguous range and account for number percent p (i) of total cellular number, $P\left(i\right)=\frac{g\left(L_i\right)}{\underset{n=0}{\overset{K}{\mathrm{\Σ}}}g\left(L_n\right)},n=\mathrm{1,2},...,k$ And $\underset{i=0}{\overset{k}{\mathrm{\Σ}}}p\left(i\right)=1,$ $i=\mathrm{1,2},..,k.$

3) based on Monte Carlo random number method establishment cellular transformation rule.

First the land type number percent obtained according to previous step obtains land type number percent vector

P＝[P ₁P ₂......P _k](5)

It meets:

$\underset{i=0}{\overset{k}{\mathrm{\Σ}}}{P}_k=1,i=\mathrm{1,2},...,k---\left(6\right)$

Accumulation calculating is done to the element of land type number percent vector, obtains cumulative land type number percent vector

$P^{\left(s\right)}=\left[\begin{array}{ccc}{P}_1^{\left(s\right)}& P_{21}^{\left(s\right)}& ......P_k^{\left(s\right)}\end{array}\right]---\left(7\right)$

In formula:

${P_i}^{\left(s\right)}=\underset{n=0}{\overset{i}{\mathrm{\Σ}}}{P}_n,i=\mathrm{1,2},...,k---\left(8\right)$

Like this, P ^(s)in k element just define interval (0,1) one segmentation, k sub-range (0, P ₁ ^(s)], (P ₁ ^(s), P ₂ ^(s)] ..., (P _k-1 ^(s), P _k ^(s)] the corresponding result phase 1,2 of difference ..., k.Utilize uniform random number to produce function like this and carry out stochastic prediction, when namely the cellular land use pattern of subsequent time being simulated, just produce one at (0,1) upper equally distributed random number r, if r ∈ is (P _n-1 ^(s), P _n ^(s)), then judge that the state of subsequent time cellular is as n.

Except rule defined above, binding district feature can also increase some restrictive conditions (such as: set other land use patterns in the present invention and waters can not be converted to, and other lands used can only be converted to by waters) in addition.

3. set up the simulation of land use changes system based on Monte Carlo and cellular Automation Model

Close C# programming language based on ARCGIS secondary development inclusion, take Visualstudio2008 as development platform, under WindowsXP operating system environment, develop the simulation of land use changes system based on cellular automaton and Monte Carlo method.This system is mainly divided into two large modules: GIS basic function module and urban land change analog module.

1) GIS basic module mainly comprises some basic functions relevant with GIS, comprises data input and output, the function such as data query, data store, map amplifies, reduces, layer management, range finding, attribute query, hawkeye.

2) simulation of land use changes module is the core of system, this module realizes simulation of land use changes forecast function based on Monte Carlo and cellular Automation Model, and in conjunction with the spatial analysis of GIS and thematic charting function, the quantity statistics and the thematic map that realize land use pattern are drawn.

4. use the simulation of land use changes system of Monte Carlo and cellular automation method to carry out simulation and forecast

Based on the land use pattern figure in the first period, simulate the state of each cellular subsequent time land use pattern, when analog result and next in period land use pattern degree of fitting reach simulation precision (as 95%) of setting time, the simulative iteration number of times of record cast, determines simulation iterations in the unit interval (year).On this basis, simulation obtains the land use change prediction figure in other periods.

5. based on the Analysis of Land Use Structure of analog result

The number percent that the quantity of each land use pattern in this period and each Land_use change class account for sum is added up according to analog result, the land use pattern matrix of transition probabilities and conversion surface product matrix can be calculated further, thus draw conversion direction, all kinds of soil and conversion quantity.According to acquisition land use change survey information, relevant departments can take corresponding policy making steps, effectively plan and manage Land_use change.Such as: when construction land rolls up; when exceeding planning construction land used; show that urbanization process is too fast; now should focus on the protection of ecologic environment; planning utilization has construction land area rationally and effectively; develop exposed wasteland as far as possible, account for less and even do not take agricultural land, greenery patches, forest land used, conscientiously protect farmland and ecological land.When farming land is more transfer exposed wasteland, meadow to time, the appearance of no longer cultivated phenomenon of being allowed to lie waste should be paid close attention to.Now relevant departments can play the effect of macro adjustments and controls, formulate and implement relevant adjustment policy and measure, and Reasonable adjustment, optimization land use structure and layout, improve land use intensively level, promotes sustainable use and the development of urban land resource.

Claims (1)

1. one kind towards urban land change analogy method, it is characterized in that, Monte Carlo method is combined with cellular automaton, according to the distribution of different cellular state and Monte Carlo method dynamic conditioning local neighborhood range size, solve the uncertain problem of urban land change on Spatial Dimension and the stochastic problems on time dimension, realize urban land change simulation;

Method step is:

1) data acquisition and pre-service

Obtain the remote sensing image of survey region two different year, utilize ArcGIS, ERDAS software to Remote Sensing Image Matching, cutting and geometry correction process, finally carry out image interpretation classification, land use pattern is divided into waters, unused land, construction land and greenery patches four class, thus obtains the land use classes figure in two periods;

2) according to classification chart determination cellular size, cellular state and iteration time, Monte Carlo method determination Dynamic Neighborhood scope is used;

(1) Dynamic Neighborhood is selected: for general cellular, uses the Moore neighborhood of 3 × 3, if the land use classes of cellular is tending towards average in neighborhood, then utilizes the Moore neighborhood of 5 × 5;

(2) the neighborhood type determined according to previous step calculates each land type number percent in cellular contiguous range;

(3) based on Monte Carlo random number method establishment cellular transformation rule;

First the land type number percent obtained according to previous step obtains land type number percent vector

P＝[P ₁P ₂......P _k](5)

, wherein k is land use pattern species number; It meets:

$\underset{i=0}{\overset{k}{\Σ}}{P}_k=1,i=1,2,...,k---\left(6\right)$

Accumulation calculating is done to the element of land type number percent vector, obtains cumulative land type number percent vector

$P^{\left(s\right)}=\left[\begin{array}{cccc}{P}_1^{\left(s\right)}& P_{21}^{\left(s\right)}& ......& P_k^{\left(s\right)}\end{array}\right]---\left(7\right)$

In formula (7):

${P_i}^{\left(s\right)}=\underset{n=0}{\overset{i}{\Σ}}{P}_n,i=1,2,...,k---\left(8\right)$

Like this, P ^(s)in k element just define interval (0,1) one segmentation, k sub-range (0, P ₁ ^(s)], (P ₁ ^(s), P ₂ ^(s)] ..., (P _k-1 ^(s), P _k ^(s)] the corresponding result phase 1,2 of difference ..., k; Utilize uniform random number to produce function like this and carry out stochastic prediction, when namely the cellular land use pattern of subsequent time being simulated, just produce one at (0,1) upper equally distributed random number r, if r ∈ is (P _n-1 ^(s), P _n ^(s)), then judge that the state of subsequent time cellular is as n;

3) the simulation of land use changes system based on Monte Carlo and cellular Automation Model is set up;

Take Visualstudio2008 as development platform, under WindowsXP operating system environment, develop the simulation of land use changes system based on cellular automaton and Monte Carlo method, this system is mainly divided into two large modules: GIS basic function module and urban land change analog module;

A) GIS basic function module mainly comprises the basic function relevant with GIS, comprise data input and output, data query, data store, map amplifies, reduces, layer management, range finding, attribute query and hawk-eye function;

B) simulation of land use changes module, this module realizes simulation of land use changes forecast function based on Monte Carlo and cellular Automation Model, and in conjunction with the spatial analysis of GIS and thematic charting function, the quantity statistics and the thematic map that realize land use pattern are drawn;

4) the simulation of land use changes system of Monte Carlo and cellular automation method is used to carry out simulation and forecast;

Based on the land use classes figure in the first period, simulate the state of each cellular subsequent time land use pattern, when analog result and next in period land use pattern degree of fitting reach the simulation precision of setting time, the simulation precision wherein set is more than or equal to 90%; The simulative iteration number of times of record cast, determine simulation iterations in the unit interval, on this basis, simulation obtains the land use change prediction figure in other periods;

5) based on the Analysis of Land Use Structure of analog result;

The number percent that the quantity of each land use pattern in this period and each land use pattern account for sum is added up according to analog result, the further calculating land use pattern matrix of transition probabilities and conversion surface product matrix, thus draw conversion direction, all kinds of soil and conversion quantity, according to acquisition land use change survey information, relevant departments can take corresponding policy making steps, effectively plan and manage Land_use change.