A credibility assessment method for land use change simulation based on spatial dislocation

Abstract

The present invention relates to a credibility assessment method for land use change simulation based on spatial dislocation, comprising the following steps: collecting land use classification maps and driving factor maps for constructing a CA model; obtaining a land conversion probability map based on an intelligent optimization algorithm and establishing a CA model; simulating urban patterns using the CA model and outputting and saving the simulation results; performing individual numerical evaluations of the CA model simulation results using a pixel-by-pixel comparison method; generating spatial distribution maps of the simulation result state-stationary index and the result state-change index using a spatial dislocation-moving window analysis method; and outputting and saving the simulation results and spatial distribution map as credibility assessment results. Compared with existing technologies, the present invention has the advantages of generating assessment maps reporting the simulation accuracy / error for each pixel, displaying the spatial distribution of simulation accuracy and error, and capturing spatial heterogeneity in accuracy, thus facilitating the credibility assessment of CA models and remote sensing classifiers.

Description

Technical Field

[0001] The present invention relates to the field of land use change simulation, and in particular to a land use change simulation credibility assessment method based on spatial dislocation. Background Art

[0002] Cellular automaton (CA) models are the most commonly used spatial simulation methods for simulating and predicting dynamic urban growth. The reliability of CA models underpins their ability to simulate historical patterns and predict future scenarios. In CA-based urban growth simulations, model performance evaluation is often influenced by the accuracy of data sources and driving factors, the choice of CA parameters, and the effectiveness of evaluation methods. Therefore, assessing simulation accuracy is crucial for determining whether CA models are suitable for urban planning and decision-making. Methods for evaluating model simulation results typically include visual inspection, pixel-by-pixel comparison, spatial structure assessment, and landscape index calculation.

[0003] For each simulation result map, when state-static and state-change metrics are generated using pixel-by-pixel comparisons, each evaluation metric is assigned a numerical value. The values ​​of these metrics are closely related to the modeling method and the study area, and often cannot accurately estimate the simulation accuracy. Because simulation results often exhibit spatial heterogeneity, the resulting accuracy results will also exhibit similar spatial heterogeneity. For CA-based simulations, a globally high evaluation metric value may result in low simulation accuracy in a local area, while a globally low evaluation metric value may result in high simulation accuracy in a local area. Therefore, a single numerical evaluation cannot reflect simulation accuracy across space, making it impossible to assess simulation performance in a local area. Therefore, the key issue that needs to be addressed is how to evaluate simulation results in a local area. Summary of the Invention

[0004] The purpose of the present invention is to provide a credibility assessment method for land use change simulation based on spatial dislocation, which can not only report the quantitative evaluation indicators of each pixel, but also capture the spatial heterogeneity of accuracy, which is helpful to detect the credibility of CA models and remote sensing classifiers, thereby supporting land resource management and decision-making.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A credibility assessment method for land use change simulation based on spatial dislocation includes the following steps:

[0007] Step 1) Collect land use classification maps and driving factor maps for building CA models;

[0008] Step 2) Obtaining a land conversion probability map based on an intelligent optimization algorithm and establishing a CA model;

[0009] Step 3) Use the CA model to simulate the urban pattern and output and save the simulation results;

[0010] Step 4) Perform a single numerical evaluation of the CA model simulation results using a pixel-by-pixel comparison method;

[0011] Step 5) Generate a spatial distribution map of the state-stationary index of the simulation results using a spatial dislocation-moving window analysis method;

[0012] Step 6) Generate a spatial distribution map of the state-change index of the simulation results using a spatial dislocation-moving window analysis method;

[0013] Step 7) Output and save the simulation results and spatial distribution map as the credibility assessment results.

[0014] The step 1) comprises the following steps:

[0015] Step 1-1) Using land satellite remote sensing images, a supervised classification method is used to obtain a land use classification map;

[0016] Steps 1-2) use vector datasets and raster images to extract the driving factors affecting urban expansion.

[0017] The step 2) comprises the following steps:

[0018] Step 2-1) Systematically sample the urban land use classification map and driving factor map to provide training samples for constructing CA conversion rules. The function of the conversion rule is expressed as:

[0019] P·g~UrbanTranRule(State cu ,P tr ,NE,Res,Sto)

[0020] Where Pg is the global transition probability, State cu is the current cell state, P tr is the transition probability defined by the driving factor, NE is the neighborhood effect, Res is the global and local restrictions, and Sto is a random factor;

[0021] Step 2-2) Based on the cross entropy optimization algorithm CEO search CA parameters and establish a CA model.

[0022] The step 2-2) comprises the following steps:

[0023] Step 2-2-1) Construct an objective function that links the CA model with actual urban growth;

[0024] Step 2-2-2) construct a set of random sequence samples according to the probability distribution;

[0025] Step 2-2-3) Find a better solution by updating the probability distribution parameters and the objective function, where in order to find the most appropriate probability P tr (a), automatically updated probability parameters P = {P1, P2, ..., P k} is represented as:

[0026]

[0027] Where, is the indicator function, γ is a predefined small residual, a ij is the jth parameter of the i-th driving factor, and k is the population size of CEOs;

[0028] Step 2-2-4) When the CEO iteration is completed, the optimal solution a, that is, the CA parameter, is obtained, and the CA model is determined by the CA parameter.

[0029] The step 3) comprises the following steps:

[0030] Step 3-1) Calibrate the constructed CA model: In a GIS modeling and simulation environment, select the urban distribution pattern of a certain year as the initial state of the calibration period, run the constructed model A times, obtain the simulation results of the urban distribution pattern, and output the simulation results, where A is the difference between the beginning and end years of the model calibration period;

[0031] Step 3-2) Verify the constructed CA model: In the GIS modeling and simulation environment, select the urban distribution pattern of a certain year as the initial state of the verification period, run the constructed model B times, obtain the simulation results of the urban distribution pattern, and output the simulation results, where B is the difference between the beginning and end years of the model verification period.

[0032] The step 4) comprises the following steps:

[0033] Step 4-1) Using the contingency matrix in the pixel-by-pixel comparison method to evaluate the state-stationary index of the simulation results;

[0034] Step 4-2) uses the real-simulation overlay in the pixel-by-pixel comparison method to generate four indicators: hit, missed, false alarm, and correct rejection, and calculates a state-change indicator for evaluating the dynamic growth of the city based on the four indicators.

[0035] The state-stationary indicators include overall accuracy, user accuracy, and producer accuracy.

[0036] The state-change indicators include quality factor, accuracy, and recall rate.

[0037] The step 5) comprises the following steps:

[0038] Step 5-1) Generate basic evaluation indicator layer:

[0039] Layer a. The real model and simulation results are both for the city category Cor_Ur;

[0040] Layer b. The real model is the urban category, and the simulation result is the non-urban category Fal_Ur;

[0041] Layer c. The real model is the non-urban category, and the simulation result is the urban category Fal_NUr;

[0042] Layer d. The real model and simulation results are both for the non-urban category Cor_NUr;

[0043] Step 5-2) Use spatial dislocation-moving window analysis: Use the focal statistics tool in ArcGIS to calculate the focal level index;

[0044] Step 5-3) Generate accuracy and error graph: Generate an evaluation graph of the state-rest indicator focus level using the following formula:

[0045]

[0046]

[0047]

[0048] Among them, Map mw () represents a layer, SPOA represents the spatialization result of overall accuracy, SPPA represents the spatialization result of producer accuracy, and SPUA represents the spatialization result of user accuracy.

[0049] The step 6) comprises the following steps:

[0050] Step 6-1) Generate basic evaluation indicator layer:

[0051] Layer a. Simulation results and actual models are both urban URHit;

[0052] Layer b. The simulation result is urban, and the actual mode is non-urban URFalse;

[0053] Layer c. The simulation result is non-urban, and the actual mode is urban URMiss;

[0054] Layer d. Simulation results and actual patterns are both non-urban CR;

[0055] Step 6-2) Use spatial dislocation-moving window analysis: Use the focal statistics tool in ArcGIS to calculate focal level indicators;

[0056] Step 6-3) Generate accuracy and error graphs: Generate an evaluation graph of the state-change indicator focus level using the following formula:

[0057]

[0058]

[0059]

[0060] Among them, Map mw () represents the layer, SPFOM represents the spatialized result of quality factor, SPPRE represents the spatialized result of precision, and SPRE represents the spatialized result of recall rate.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] The proposed method can identify the quantitative accuracy of each pixel while taking into account the influence of its neighborhood, rather than simply providing a single numerical value to summarize simulation quality and model performance. This method better reflects the effectiveness of the CA model in simulating urban growth in different regions. Its advantages include: 1) generating an evaluation map reporting the simulation accuracy / error for each pixel, displaying the spatial distribution of simulation accuracy and error; and 2) quantifying the relationship between driving factors and evaluation indicators. This spatial evaluation method can be used not only to evaluate CA model simulation results, but also to evaluate remote sensing image classifiers. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 Schematic diagram of the method flow of the present invention;

[0064] Figure 2 A historical land use distribution map of the case area in the embodiment of the present invention;

[0065] Figure 3 This is a map of driving factors of land use change from 1995 to 2015;

[0066] Figure 4 The simulation results and evaluation diagrams for 2005 and 2015 are shown;

[0067] Figure 5 This is the spatial pattern diagram of the state-stationary index of the simulation results in 2005;

[0068] Figure 6 This is the spatial pattern diagram of the state-change indicators of the simulation results in 2005;

[0069] Figure 7 This is the spatial pattern diagram of the state-stationary index of the simulation results in 2015;

[0070] Figure 8This is the spatial pattern diagram of the state-change indicators of the simulation results in 2015. DETAILED DESCRIPTION

[0071] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0072] Different from the global level and the local level, moving window analysis is an effective method to evaluate simulation results at the focal level. Some GIS scholars have applied this method to detect local differences or similarities between simulation results and actual patterns, and pointed out that moving window analysis can be used to evaluate the similarity of land use categories in composition and spatial structure. The characteristic of this method is the fuzzy evaluation using a single numerical evaluation indicator, such as Kappa. Compared with the precise comparison of pixel-by-pixel comparison, the credibility assessment of land use change simulation based on spatial dislocation overcomes the shortcomings of previous evaluation methods that may underestimate the simulation accuracy. Therefore, the present invention provides a credibility assessment method for land use change simulation based on spatial dislocation, such as Figure 1 As shown, the following steps are included:

[0073] Step 1) Collect land use classification maps and maps of driving factors affecting urban expansion covering the initial and final years of the study for constructing the CA model.

[0074] Step 1-1) Using Landsat remote sensing images, supervised classification methods were used in ENVI software to obtain the land use classification maps for the initial and final years required for CA model calibration and validation.

[0075] Steps 1-2) use vector datasets and raster images to extract the driving factors affecting urban expansion.

[0076] Specifically, by integrating topographic maps, population distribution maps, administrative division maps, road traffic maps, POI and other data, the Euclidean distance method is used in a GIS environment to obtain the distances to key elements such as city centers, district and county centers, main roads, and infrastructure, thereby extracting the driving factors affecting urban expansion.

[0077] Step 2) Obtain the land conversion probability map based on the intelligent optimization algorithm and establish the CA model.

[0078] Step 2-1) Systematically sample the urban land use classification map and driving factor map to provide training samples for constructing CA conversion rules.

[0079] Cellular Automata (CA) is a general term for a class of self-organizing models that use a bottom-up approach to automatically evolve and are composed of rules constructed by the model.cu ), the conversion probability based on the driving factor (P tr ), neighborhood effect (NE), global and local restrictions (Res), and random factors (Sto), the global transition probability P·g and the transition rule UrbanTranRule can be expressed as:

[0080] P·g~UrbanTranRule(State cu ,P tr ,NE,Res,Sto) (1)

[0081] Where Pg is the global transition probability, State cu is the current cell state, P tr is the transition probability defined by the driving factor, NE is the neighborhood effect, Res is the global and local restrictions, and Sto is a random factor.

[0082] The construction of CA model is realized in UrbanCA software, and the global conversion probability P all It can be calculated by the following formula:

[0083]

[0084] Where TIP is the scaling parameter for adjusting the probability attenuation effect, ranging from 0.0 to 0.1. A larger value indicates a stronger scaling effect. LAP is the scaling parameter for adjusting the neighborhood effect, ranging from 0.5 to 1.0. A smaller value indicates a weaker scaling effect.

[0085] The conversion probability P determined by the driving factor tr It is the core part of the CA transformation rule, which represents the impact of driving factors on land use change and urban development and affects the cell state at the next moment in a probabilistic way.

[0086] Step 2-2) Based on the cross entropy optimization algorithm CEO search CA parameters and establish a CA model.

[0087] Cross Entropy Optimization (CEO) is a heuristic algorithm that is currently widely used to search for the difference between the target value and its predicted value. In CA modeling, the objective function can be used to guide the model to achieve automatic parameterization. Specifically, it includes the following steps:

[0088] Step 2-2-1) Construct an objective function that links the CA model with actual urban growth;

[0089] Step 2-2-2) construct a set of random sequence samples according to the probability distribution;

[0090] Step 2-2-3) Find a better solution (i.e. CA parameters) by updating the probability distribution parameters and the objective function. In order to find the most appropriate probability P tr (a), automatically updated probability parameters P = {P1, P2, ..., P k} is represented as:

[0091]

[0092] Where, is the indicator function, γ is a predefined small residual, a ij is the jth parameter of the i-th driving factor, and k is the population size of CEOs;

[0093] Step 2-2-4) When the CEO iteration is completed, the optimal solution a, that is, the CA parameter, is obtained, and the CA model is determined by the CA parameter.

[0094] Step 3) Use the CA model to simulate the urban pattern and output and save the simulation results.

[0095] Step 3-1) Calibrate the constructed CA model: In the GIS modeling and simulation environment, select the urban distribution pattern of a certain year as the initial state of the calibration period, run the constructed model A times, obtain the simulation results of the urban distribution pattern, and output the simulation results, where A is the difference between the beginning and end years of the model calibration period.

[0096] Step 3-2) Verify the constructed CA model: In the GIS modeling and simulation environment, select the urban distribution pattern of a certain year as the initial state of the verification period, run the constructed model B times, obtain the simulation results of the urban distribution pattern, and output the simulation results, where B is the difference between the beginning and end years of the model verification period.

[0097] Step 4) Use pixel-by-pixel comparison method to perform individual numerical evaluation on the simulation results of the CA model.

[0098] Step 4-1) uses the contingency matrix in the pixel-by-pixel comparison method to evaluate the state-stationary index of the simulation results. In this embodiment, the state-stationary index includes overall accuracy (OA), producer accuracy (PA), and user accuracy (UA).

[0099] In step 4-2, the real-simulation overlay in the pixel-by-pixel comparison method is used to generate four metrics: hit (URHit), miss (URMiss), false positive (URFalse), and correct rejection (CR). Based on these four metrics, a state-change index for assessing the dynamic growth of the city is calculated. In this embodiment, the state-change index includes the figure of merit (FOM), precision (PRE), and recall (RE).

[0100] Step 4) is as follows: a pixel-by-pixel comparison method is used to measure whether the state of each pixel in the simulation results matches the actual urban pattern. OA is selected as an indicator for quantitative evaluation of the CA model. OA is a single numerical probability representing the consistency between the simulation results and the actual urban pattern, which can be expressed as:

[0101]

[0102] Where p nn is the number of correctly simulated pixels, SUM is the number of all pixels, t cn is the number of urban pixels in the actual result.

[0103] The spatial superposition of the actual map of the starting year, the actual map of the ending year, and the simulated map of the ending year can reveal the differences in structural patterns. The real-simulated superposition generates four indicators, including hits (URHit), missed reports (URMiss), false positives (URFlase), and correct rejections (CR). Among them, URHit means that a cell is in an urban state in both the actual model and the simulation results, URMiss means that a cell is in an urban state in the actual model but in a non-urban state in the simulation results, URFlase means that a cell is in a non-urban state in the actual model but in an urban state in the simulation results, and CR means that a cell is in a non-urban state in both the actual model and the simulation results. The indicators generated by the above spatial superposition can be used to calculate state-change indicators for evaluating the dynamic growth of cities, such as FOM:

[0104]

[0105] Where FOM represents the ability of the CA model to capture new urbanized pixels.

[0106] Step 5) Use the spatial dislocation-moving window analysis method to generate the spatial distribution map of the state-stationary index of the simulation results.

[0107] Moving window analysis typically uses a fixed-size and -shaped window centered on a pixel. Statistical calculations are performed within the window, and the results are returned to the pixel at the center of the window. The calculation process is repeated for each pixel, generating a new raster surface layer. Using moving window analysis for model evaluation can yield spatial distributions of accuracy and error.

[0108] Based on the traditional pixel-by-pixel comparison method, the spatial visualization of the state-stationary indicator mainly includes the following steps:

[0109] Step 5-1) Based on the actual land use pattern and simulated land use pattern at time T2, obtain the four basic evaluation index TIF layers:

[0110] Layer a. The real model and simulation results are both for the city category Cor_Ur;

[0111] Layer b. The real model is the urban category, and the simulation result is the non-urban category Fal_Ur;

[0112] Layer c. The real model is the non-urban category, and the simulation result is the urban category Fal_NUr;

[0113] Layer d. Both the real model and the simulation results are for the non-urban category Cor_Nur.

[0114] Step 5-2) Use spatial dislocation-moving window analysis: Use the focus statistics tool in ArcGIS, input the indicator layers extracted in step 5-1), set the shape and size of the moving window, and determine the numerical statistics method to calculate the focus level indicators. The generated results are used to calculate the spatial visualization results of each state-stationary indicator.

[0115] Step 5-3) Generate accuracy and error graph: Generate an evaluation graph of the state-rest indicator focus level using the following formula:

[0116]

[0117]

[0118]

[0119] Among them, Map mw () represents a layer, SPOA represents the spatialization result of overall accuracy, SPPA represents the spatialization result of producer accuracy, and SPUA represents the spatialization result of user accuracy.

[0120] Step 6) Use the spatial dislocation-moving window analysis method to generate a spatial distribution map of the state-change index of the simulation results.

[0121] Based on the traditional pixel-by-pixel comparison method, the spatial visualization of state-change indicators mainly includes the following steps:

[0122] Step 6-1) Based on the actual land use pattern at time T1, the actual land use pattern at time T2, and the simulated land use pattern, a basic evaluation index layer is generated to obtain spatial visualization results of the state-change index:

[0123] Layer a. Simulation results and actual models are both urban URHit;

[0124] Layer b. The simulation result is urban, and the actual mode is non-urban URFalse;

[0125] Layer c. The simulation result is non-urban, and the actual mode is urban URMiss;

[0126] Layer d. Both the simulation results and the actual model are non-urban CR.

[0127] Step 6-2) Use spatial dislocation-moving window analysis: Use the focal statistics tool in ArcGIS, input the indicator layers extracted in step 6-1), set the shape and size of the moving window, and determine the numerical statistical method. The generated results are used to calculate the spatial visualization results of each state-change indicator.

[0128] Step 6-3) Generate accuracy and error graphs: Generate an evaluation graph of the state-change indicator focus level using the following formula:

[0129]

[0130]

[0131]

[0132] Among them, Map mw () represents the layer, SPFOM represents the spatialized result of quality factor, SPPRE represents the spatialized result of precision, and SPRE represents the spatialized result of recall rate.

[0133] Step 7) Output and save the simulation results and spatial distribution map as the credibility assessment results.

[0134] This example uses land satellite remote sensing imagery and vector datasets to determine the urban land pattern and driving factors of urban expansion in Ningbo in 1995, 2005, and 2015. Using the CEO method, CA conversion rules are constructed in the UrbanCA software to generate a land conversion probability map for Ningbo. A CEO-CA model is then constructed to simulate urban expansion from 1995 to 2015. After obtaining a single numerical evaluation result using a pixel-by-pixel comparison method, the proposed spatial evaluation method is used to evaluate the simulation results from the calibration and validation phases, obtaining the SPOA, SPPA, SPUA, SPFOM, SPPRE, and SPRE for the 2005 and 2015 simulations.

[0135] In step 1), Ningbo City was used as the study area, and land satellite remote sensing images from 1995, 2005, and 2015, as well as administrative division maps, traffic network maps, population maps, topographic maps, and POI data were collected as basic data for constructing the CA model in this study.

[0136] The supervised classification method was used in ENVI software to classify the land use of the Landsat remote sensing images of Ningbo City, thereby generating the real urban spatial pattern maps of Ningbo City in 1995, 2005 and 2015, as shown in the figure. Figure 2 shown.

[0137] Based on the administrative division layer, road traffic layer, and infrastructure layer, the Euclidean distance in the spatial analysis tool was used in the GIS environment to calculate the distance of each cell to the city center, district and county center, main roads, educational institutions, and medical facilities. The raster maps of terrain and population were also obtained to form a map of driving factors affecting the urban spatial expansion of Ningbo, as shown in Table 1 and Figure 3 shown.

[0138] Table 1. Driving factors affecting Ningbo's urban development

[0139] name category source explain Distance to city center center National basic geographic data The impact of Ningbo's administrative center on urban development Distance to town center center National basic geographic data The impact of Ningbo's urban center on urban development Distance to main roads network OpenStreetMap road data The impact of Ningbo's main roads on urban development population density WorldPOP Population density per pixel terrain altitude Topographic remote sensing products The elevation of each pixel Distance to hospital POI Baidu Maps The impact of Ningbo City Hospital on urban development Distance to school POI Baidu Maps The impact of Ningbo schools on urban development

[0140] In steps 2) to 4), the CEO-CA model was used to simulate the urban pattern of Ningbo in 2005, taking the actual urban pattern of Ningbo in 1995 as the initial state; and the CEO-CA model was used to simulate the urban pattern of Ningbo in 2015, taking the actual urban pattern of Ningbo in 2005 as the initial state. The simulation results were evaluated by single numerical values, as shown in Tables 2 and Figure 4 The simulation results and accuracy.

[0141] Table 2. Accuracy evaluation of simulation results during calibration and validation periods

[0142]

[0143] In steps 5) and 6), a moving window analysis method was used to visualize the accuracy of the focal level and analyze its statistics (range, mean, and SD) for the CA model simulation results. Tables 3 and 4 show the statistical information of the evaluation indicators in the state-stationary and state-changing spaces during the calibration and validation periods of the CA model, respectively.

[0144] Table 3. Statistics of evaluation indicators for the state-stationary and state-changing spaces during the correction period

[0145]

[0146] Table 4. Statistics of evaluation indicators for state-stationary and state-changing spaces during the verification period

[0147]

[0148] Figure 5 and Figure 6 They are the spatial evaluation maps of the simulation results during the CA model calibration period, where Figure 5Shown are SPOA, SPPA, and SPUA for the simulation results in 2005. Figure 6 Shown are the SPFOM, SPPRE, and SPRE for the 2005 simulation results.

[0149] Figure 7 and Figure 8 They are the spatial evaluation diagrams of the simulation results during the CA model verification period, where Figure 7 Shown are SPOA, SPPA, and SPUA for the 2015 simulation results. Figure 8 Shown are the SPFOM, SPPRE, and SPRE for the 2015 simulation results.

[0150] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A credibility assessment method for land use change simulation based on spatial dislocation, characterized in that: The following steps are involved: Step 1) Collect land use classification maps and driving factor maps for building CA models; Step 2) Obtain the land conversion probability map based on the intelligent optimization algorithm and establish the CA model; Step 3) Use the CA model to simulate the urban pattern and output and save the simulation results; Step 4) Perform a single numerical evaluation of the CA model simulation results using a pixel-by-pixel comparison method; The step 4) includes the following steps: Step 4-1) Using the contingency matrix in the pixel-by-pixel comparison method to evaluate the state-stationary index of the simulation results; the state-stationary index includes overall accuracy, user accuracy, and producer accuracy; Step 4-2) Using the real-simulation overlay in the pixel-by-pixel comparison method to generate four metrics: hits, missed alerts, false positives, and correct rejections, and calculating state-change metrics for assessing the city's dynamic growth based on these four metrics; the state-change metrics include quality factor, precision, and recall; Step 5) Generate a spatial distribution map of the state-stationary index of the simulation results using a spatial dislocation-moving window analysis method; The step 5) includes the following steps: Step 5-1) Generate basic evaluation indicator layer: Layer a. The real model and simulation results are both for the city category Cor_Ur; Layer b. The real model is the urban category, and the simulation result is the non-urban category Fal_Ur; Layer c. The real model is the non-urban category, and the simulation result is the urban category Fal_NUr; Layer d. The real model and simulation results are both for the non-urban category Cor_NUr; Step 5-2) Use spatial dislocation-moving window analysis: Use the focal statistics tool in ArcGIS to calculate focal level indicators; Step 5-3) Generate accuracy and error graphs: Generate an evaluation graph of the state-rest indicator focus level using the following formula: in, Represents a layer, SPOA represents the spatialization result of overall accuracy, SPPA represents the spatialization result of producer accuracy, and SPUA represents the spatialization result of user accuracy; Step 6) Generate a spatial distribution map of the state-change index of the simulation results using the spatial dislocation-moving window analysis method; The step 6) includes the following steps: Step 6-1) Generate basic evaluation indicator layer: Layer a. Simulation results and actual models are both urban URHit; Layer b. The simulation result is urban, and the actual mode is non-urban URFalse; Layer c. The simulation result is non-urban, and the actual mode is urban URMiss; Layer d. Simulation results and actual patterns are both non-urban CR; Step 6-2) Use spatial dislocation-moving window analysis: Use the focal statistics tool in ArcGIS to calculate focal level indicators; Step 6-3) Generate accuracy and error graphs: Generate an evaluation graph of the state-change indicator focus level using the following formula: in, Represents the layer, SPFOM represents the spatialized result of quality factor, SPPRE represents the spatialized result of precision, and SPRE represents the spatialized result of recall rate; Step 7) Output and save the simulation results and spatial distribution map as the credibility assessment results.

2. The method for credibility assessment of land use change simulation based on spatial dislocation according to claim 1, characterized in that: The step 1) includes the following steps: Step 1-1) Using Landsat remote sensing images, a supervised classification method is used to obtain a land use classification map; Steps 1-2) Use vector datasets and raster images to extract a map of driving factors affecting urban expansion.

3. The method for credibility assessment of land use change simulation based on spatial dislocation according to claim 1, characterized in that: The step 2) includes the following steps: Step 2-1) Systematically sample the urban land use classification map and driving factor map to provide training samples for constructing CA conversion rules. The function of the conversion rule is expressed as: Where, is the global transition probability, is the current cell state, is the transition probability defined by the driver, is the neighborhood effect, are global and local restrictions, is a random factor; Step 2-2) Based on the cross entropy optimization algorithm CEO search CA parameters and establish a CA model.

4. The method for credibility assessment of land use change simulation based on spatial dislocation according to claim 3 is characterized in that: The step 2-2) includes the following steps: Step 2-2-1) Construct an objective function that links the CA model with actual urban growth; Step 2-2-2) Construct a set of random sequence samples according to the probability distribution; Step 2-2-3) Find a better solution by updating the probability distribution parameters and the objective function, in which, in order to find the most appropriate probability , automatically updated probability parameters Expressed as: Where, is the indicator function, is a predefined small residual, For the i The first driving factor j parameters, is the CEO population size; Step 2-2-4) When the CEO iteration ends, the optimal solution is obtained a , namely CA parameters, the CA model is determined by the CA parameters.

5. The method for credibility assessment of land use change simulation based on spatial dislocation according to claim 1, characterized in that: The step 3) includes the following steps: Step 3-1) Calibrate the constructed CA model: In the GIS modeling and simulation environment, select the urban distribution pattern of a certain year as the initial state of the calibration period, run the constructed model A times, obtain the simulation results of the urban distribution pattern, and output the simulation results, where A is the difference between the beginning and end years of the model calibration period; Step 3-2) Verify the constructed CA model: In the GIS modeling and simulation environment, select the urban distribution pattern of a certain year as the initial state of the verification period, run the constructed model B times, obtain the simulation results of the urban distribution pattern, and output the simulation results, where B is the difference between the beginning and end years of the model verification period.

Images