Markov-FLUS model-based feedback land utilization change simulation method and system

Abstract

The invention belongs to the technical field of geographic information science, and particularly relates to a Markov-FLUS model-based feedback land utilization change simulation method and system, and the method comprises the steps: firstly predicting an initial area target vector based on historical data and a Markov chain, and inputting the initial area target vector into a FLUS model to carry out the first space distribution simulation; by calculating the achievement degree of the simulation result and the area target, the space conflict type which cannot be achieved due to the space constraint is identified, the area target vector is dynamically corrected according to the space conflict type, and the corrected target is fed back to the FLUS model for iterative simulation until the result is converged. According to the method, the technical problem of disjunction of area prediction and spatial layout in a traditional coupling model is effectively solved, and the scientificity and accuracy of land utilization change simulation are remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of geographic information science, and particularly relates to a Markov-FLUS model-based feedback land use change simulation method and system. BACKGROUND

[0002] Accurate simulation of future land use change is crucial for national spatial planning, ecosystem protection, and sustainable social and economic development. In the prior art, the Markov chain model can quantitatively predict the area demand of each land use type in the future based on historical land use data, but lacks spatial distribution information. The FLUS model can well simulate the spatial distribution of land use types, which calculates the land use conversion probability based on artificial neural networks and combines cellular automata with an adaptive inertia competition mechanism for spatial allocation.

[0003] The invention patent with publication number CN114723142A introduces a scheme based on the non-dominated sorting genetic algorithm (NSGA-II) and the FLUS model. In this scheme, first, the NSGA-II algorithm is used for multi-objective optimization to obtain a set of Pareto optimal area structure solutions that meet economic, ecological, and social benefits; then, this area solution set is used as the input target of the FLUS model, and its neural network and cellular automaton mechanism are used to complete the allocation in space, thereby realizing complete simulation from quantity structure to spatial structure. This method optimizes the initial setting of land demand through intelligent algorithms and improves the diversity of simulation scenarios.

[0004] However, the above-mentioned prior art still has a significant defect: its coupling process is one-way and open-loop. Specifically, the area target optimized by the NSGA-II algorithm is directly and unconditionally passed to the FLUS model to perform spatial allocation, but the FLUS model may encounter spatial constraints (such as ecological protection red lines and basic farmland) that cannot be overcome during allocation, resulting in that part of the area target cannot be fully realized in the actual space. The conflict between this spatial allocation result and the initial area target cannot be fed back to the area prediction module, which may cause the area target to deviate from the actual space, thereby affecting the rationality and accuracy of the final simulation result. Therefore, the prior art lacks a feedback mechanism that can dynamically adjust the area demand according to the spatial feasibility, making it difficult to ensure the coordinated optimization of area target and spatial layout. SUMMARY

[0005] The purpose of the present application is to provide a Markov-FLUS model-based feedback land use change simulation method and system to solve the problem of disconnection between area prediction and spatial simulation in existing land use change simulation methods.

[0006] The present application achieves the above-mentioned purpose through the following technical solutions: The first aspect, the application provides a feedback land use change simulation method based on a Markov-FLUS model, and the method comprises the following steps: S1. Obtain land use data sets of a target region at different historical periods, and multi-source driving factor data sets related to land use change; S2. Construct a Markov transition probability matrix based on the land use data sets, predict the area demand of each land use type in a future target year, and mark the initial area target vector; S3. Combine the multi-source driving factor data sets, input the initial area target vector into the FLUS model, and perform first land use spatial distribution simulation to obtain an initial simulation result; S4. Determine a corrected area target vector based on the difference between the initial simulation result and the initial area target vector; S5. Input the corrected area target vector into the FLUS model for iterative simulation, and determine whether convergence is achieved according to the simulation result; S6. If convergence is achieved, output a final land use change simulation result; if convergence is not achieved, return to step S4, generate a corrected area target vector again based on the current simulation result and perform iterative simulation until convergence is achieved or the maximum number of iterations is reached.

[0007] Further, the multi-source driving factor data sets comprise natural geographical factors, location traffic factors, social and economic factors, and constraint factors; wherein, The natural geographical factors comprise one or more of elevation, slope, slope direction, precipitation, and soil type; The location traffic factors comprise one or more of distances to highways, railways, ports, airports, and city centers; The social and economic factors comprise one or more of population density, spatial distribution of GDP, and nighttime light index; The constraint factors are spatially characterized by one or more of ecological protection red lines, permanent basic farmland ranges, urban development boundaries, and natural protection land ranges.

[0008] Further, step S3 comprises the following steps: S31. Utilize the land use data sets at the historical periods and the multi-source driving factor data sets, train a artificial neural network model to obtain a suitability probability distribution map of each land use type; S32. Combine the suitability probability distribution map and the initial area target vector, and perform first land use spatial distribution simulation by the FLUS model; Wherein, the calculation formula of the total suitability probability of land use type i on cell (x, y) at the kth iteration in the FLUS model is: ​ ; wherein, is a suitability probability obtained by the artificial neural network, is a neighborhood influence factor, is a conversion cost from the current land use type to type i.

[0009] Further, in step S31, the training by the artificial neural network model to obtain the suitability probability distribution map comprises: standardizing each factor data in the multi-source driving factor dataset to form a standardized driving factor layer; spatially registering and superimposing the land use dataset of the historical period and the standardized driving factor layer; training the artificial neural network with any period of land use data in the historical period as a training label and the driving factor layer corresponding to the land use data of the previous period of the any period as an input feature; using the trained model to predict the driving factor data of the target year to generate a suitability probability distribution map of each land use type in the target year.

[0010] Further, the neighborhood influence factor is determined in the following manner: In a preset n x n neighborhood window, the ratio Q of the number of cells belonging to land use type i around the center cell (x, y) to the total number of cells in the neighborhood window is calculated as follows: ; wherein NumTypeI is the number of cells belonging to type i in the neighborhood, and (n x n - 1) is the total number of cells in the neighborhood excluding the center cell; The ratio Q is weighted calculated by a distance decay function to obtain a neighborhood influence factor representing different influence degrees of adjacent cells on the center cell, as follows: ; wherein d traverses each cell in the neighborhood window excluding the center cell; is the land use type of the dth cell in the neighborhood; is a conditional function, which is 1 when , and 0 otherwise; is a weight calculated based on the distance d between cell d and the center cell (x, y), and the value of decreases with the increase of the distance d.

[0011] Further, step S4 comprises: S41. Calculate the ratio of the simulated area of each land use type in the initial simulation result to the target area of the corresponding type in the initial area target vector as an area achievement degree Di, as follows: ; wherein, is the simulated area of type i, is the initial area target of type i; S42. Based on the area achievement degree Di, identify the land use type whose area does not reach a preset threshold as a spatial conflict type; S43. Modify the initial area target vector according to the area difference of the spatial conflict type to generate the modified area target vector.

[0012] Further, in step S43, the modification according to the area difference of the spatial conflict type includes: Determine an unachieved area difference amount ΔAi based on the identified spatial conflict type, as follows: ; Distribute the area difference amount ΔAi to one or more other land use types as allocation weights with the probability of transition from type i in the Markov transition probability matrix M; Further distribute the area difference amount ΔAi to one or more other land use types as allocation weights with the probability of transition from type i in the Markov transition probability matrix M; For each allocated non-conflict type j, the modified area target is obtained by the following formula: ; wherein k is the traversal of all allocated land use types; is the probability of transition from conflict type i to type j, is the sum of the probabilities of all other land use types except the current conflict type i; The modified area targets of all land use types collectively constitute the modified area target vector.

[0013] Further, the specific criterion for determining convergence in step S5 is: Calculate the overall deviation δ of the area achievement degrees Di of all land use types in the current iteration simulation result from 1, as follows: ; wherein N is the number of land use types; When the overall deviation δ is less than a preset tolerance threshold ε, it is determined to be converged.

[0014] In a second aspect, the application provides a Markov-FLUS model feedback land use change simulation system for implementing the steps of the method described above, the system comprising: a data acquisition module for acquiring land use data sets at different historical periods and multi-source driving factor data sets related to land use change; an area prediction module for constructing a Markov transition probability matrix based on the land use data sets and predicting the area demand of each land use type in the target year as an initial area target vector; an initial simulation module for inputting the initial area target vector into the FLUS model in combination with the multi-source driving factor data sets to perform a first land use spatial allocation simulation and obtain an initial simulation result; a feedback correction module for determining a corrected area target vector based on the difference between the initial simulation result and the initial area target vector; a judgment module for inputting the corrected area target vector into the FLUS model again for iterative simulation and determining whether to converge according to the simulation result; if converging, outputting the final land use change simulation result; if not converging, triggering the feedback correction module to work again based on the current simulation result to generate a new corrected area target vector and perform the next iteration simulation until converging or reaching the maximum number of iterations.

[0015] The application has the following beneficial effects: The application effectively solves the technical problem of disconnection between area prediction and spatial layout in traditional simulation models by introducing a dynamic feedback mechanism. The method first predicts the initial area target based on historical land use data through Markov chain, then performs spatial allocation simulation using the FLUS model, automatically identifies the land use types that cannot achieve the target due to spatial constraints such as ecological red line and basic farmland by calculating the area achievement degree, and reallocates the unachieved area to other types according to the Markov transition probability matrix based on the historical conversion rule, and iteratively simulates until the result converges, so that the simulation result not only maintains the rationality of the macro area demand, but also ensures the feasibility of the micro spatial layout, effectively solving the coordination problem between area prediction and spatial allocation. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A flowchart of the Markov-FLUS model feedback land use change simulation method proposed in the application; Figure 2 Another flowchart of the Markov-FLUS model feedback land use change simulation method proposed in the application; Figure 3A flowchart for obtaining the corrected area target vector in the Markov-FLUS model-based feedback land use change simulation method proposed in the application is shown. DETAILED DESCRIPTION

[0017] The application will be further described in detail below with reference to the accompanying drawings. It is necessary to point out here that the following detailed description is only used to further illustrate the application and cannot be understood as limiting the scope of protection of the application. Those skilled in the art can make some non-essential improvements and adjustments to the application according to the above application content.

[0018] Embodiment 1 Please refer to Figure 1 and Figure 2 An embodiment of the application proposes a Markov-FLUS model-based feedback land use change simulation method. The method solves the problem of disconnection between area prediction and spatial simulation in traditional coupling models by introducing a dynamic feedback loop. The method comprises the following steps: S1. Obtain historical land use raster data of at least two periods (for example, 2010 and 2020) of a target area. At the same time, collect a multi-source driving factor data set closely related to land use change. The data set should comprehensively cover natural geographical, location traffic, social economic and constraint factors.

[0019] In a specific implementation, the multi-source driving factor data set includes natural geographical factors, location traffic factors, social economic factors and constraint factors. The natural geographical factors include one or more of elevation, slope, slope direction, precipitation and soil type. The location traffic factors include one or more of distances to highways, railways, ports, airports and urban centers. The social economic factors include one or more of population density, spatial distribution of GDP and night light index. The constraint factors are spatialized by one or more of ecological protection red lines, permanent basic farmland ranges, urban development boundaries and natural protection land ranges.

[0020] All data need to be unified to the same spatial coordinate system and raster pixel size and preprocessed, such as normalization, to eliminate the influence of dimension and form standardized data that can be used for model calculation.

[0021] S2. Based on the historical land use dataset obtained in step S1, calculate the area transfer amount of each land type between two periods of data, and construct a Markov transition probability matrix. Based on the existing Markov chain theory, the possibility of mutual conversion between different land use types is quantitatively described. Then, apply this matrix to the latest historical data as the starting point to deduce the area demand of each land use type in the future target year (for example, 2035), and record this prediction result as the initial area target vector. This vector represents the quantity structure of future land use in an ideal situation without considering spatial constraints.

[0022] As an example, input two periods of historical land use raster data (for example, T1 time (2010) and T2 time (2020)), spatially superimpose the two periods of land use maps, and count the number of pixels of each land type converted into all other types (including itself) from T1 time to T2 time; organize the above statistical results into an n x n matrix (n is the number of land use types), and each element S ij in the matrix represents the area converted from type i to type j within the time interval.

[0023] Calculate each row of the transition area matrix respectively, divide each element S ij by the total area of the original land class i at T1 time corresponding to the row, as follows: ; represents the probability of transferring from type i to type j.

[0024] Finally, an n x n Markov transition probability matrix P is obtained.

[0025] After obtaining the transition probability matrix P, the area in the future year is predicted. Assuming that the area state vector at T2 time is S(t), then the calculation formula of the area state vector S(t+1) at T3 time (for example, 2035) is: ; S3. Combine the multi-source driving factor dataset, input the initial area target vector obtained in step S2 into the FLUS model, and perform the first land use spatial allocation simulation.

[0026] In this step, the FLUS model first uses historical land use data and a multi-source driving factor dataset to train its built-in artificial neural network (ANN) to generate a suitability probability distribution map reflecting the development suitability of each land class at different spatial locations. Subsequently, the model combines this suitability probability map with the initial area target vector and performs spatial iterative allocation through its adaptive inertia competition mechanism-based cellular automaton (CA) to ultimately obtain the initial simulation result, i.e., the first version of the future land use spatial distribution map.

[0027] S4. Calculate the actual simulation area of each land use type in the initial simulation result and compare it with the initial area target vector in S2.

[0028] Specifically, step S4 calculates the area achievement degree of each land class (the ratio of simulation area to target area). Based on the area achievement degree, identify those spatial conflict types that cannot achieve the original area target due to spatial constraints (such as ecological protection red line, basic farmland, etc.). Then, according to the area deficit (or excess) of these conflict types, modify the initial area target vector. It can be understood that the core of the modification logic is to redistribute the unachieved spatial area to other land classes according to reasonable rules (e.g., based on historical transfer rules or preset planning priorities), thereby generating a modified area target vector that is more consistent with spatial feasibility.

[0029] S5. Input the modified area target vector obtained in S4 into the FLUS model again. The FLUS model will perform a new round of land use spatial allocation simulation based on the same suitability probability distribution map and spatial constraints, guided by the new area target. After the simulation is completed, repeat the evaluation process in S4 to calculate the area achievement degree of the new round and determine whether the model converges.

[0030] S6. Convergence judgment is based on the overall deviation of the area achievement degree of all land use types from the ideal value. If the deviation is less than the preset tolerance threshold, it is considered that the simulation result has stabilized and the model has converged, and the current land use simulation map is output as the final land use change simulation result. If it does not converge, it automatically returns to step S4 to modify and simulate the area target again based on the latest simulation result. This process will continue until the convergence condition is met or the maximum number of iterations is reached.

[0031] As a preferred way, the land use spatial allocation simulation in step S3 is implemented through the FLUS model, specifically including: S31. Generation of land use suitability probability distribution map First, the multi-source driving factor dataset (such as elevation, slope, distance from traffic line, etc.) obtained in S1 is standardized to eliminate the dimensional influence and form a set of spatially registered and numerically consistent standardized driving factor layers. Subsequently, a training sample set is constructed: from the historical land use data set, two consecutive periods of data are selected, and the later period of land use data is taken as the training label, and the standardized driving factor layer corresponding to the former period is taken as the input feature. With this sample set, the artificial neural network model is supervised trained to make the model learn the complex nonlinear mapping relationship from the driving factor to the land use pattern. After training, the driving factor data of the target year is input into the trained model, and the suitability probability distribution map of each land use type in the target year can be generated. The value of each pixel (0-1) of the layer represents the inherent possibility of the spatial location suitable for the development of a certain land class under the premise of not being affected by the total amount and the neighborhood.

[0032] S32. Spatial allocation based on adaptive inertial competition mechanism The model integrates the suitability probability distribution map generated in S31, the user's preset conversion cost matrix (representing the difficulty of conversion between land classes), and the self-calculated neighborhood influence factor (representing the spatial aggregation and diffusion effect).

[0033] The specific spatial allocation is completed through an adaptive inertial cellular automaton mechanism. In each iteration, for each spatial cell (x, y), the total suitability probability of the kth iteration in the FLUS model from the current type to the target type i is calculated by the following formula: ; wherein, is the suitability probability obtained by the artificial neural network; is the neighborhood influence factor, whose value is determined by the density and distance of the existing land class i in the neighborhood window around the cell (x, y); is the conversion cost from the current land use type to type i, which is taken from the preset conversion cost matrix and defines the permissibility of conversion from the current land class to land class i (which can be set to 0 or 1, or a probability value between 0 and 1).

[0034] The model determines the most likely type of conversion for each cell through the roulette selection mechanism according to the total suitability probability, and balances the global area target and the local conversion potential by combining the adaptive inertia coefficient. After multiple iterations, the simulated land use quantity reaches the initial area target vector set in S2, and the initial simulation result is finally output.

[0035] More specifically, in step S32, the neighborhood influence factor The determination manner comprises: In a preset n*n neighborhood window, a ratio Q of a number of cells belonging to a land use type i around a center cell (x, y) to a total number of cells in the neighborhood window is calculated, as follows: ; Where NumTypeI is the number of cells belonging to type i in the neighborhood, and (n*n - 1) is the total number of cells in the neighborhood except the center cell; The ratio Q is calculated by a distance attenuation function to obtain a neighborhood influence factor , which is used to represent different influence degrees of neighboring cells on the center cell, as follows: ; Where d traverses each cell in the neighborhood window except the center cell; is a land use type of the dth cell in the neighborhood; is a conditional function, which is 1 when , and 0 otherwise; is a weight calculated based on a distance d between the cell d and the center cell (x, y), and The value of decreases with the increase of the distance d.

[0036] The neighborhood influence factor calculated by the formula is a value between 0 and 1. The closer the value is to 1, the more conducive the surrounding environment of the center cell (x, y) is to the development or maintenance of type i, so that a higher conversion probability is obtained in subsequent cellular automaton competition.

[0037] As a preferred manner, referring to Figure 3 , step S4 comprises: S41. Calculating an area achievement degree A ratio of a simulated area of each land use type in the initial simulation result to a target area of the corresponding type in the initial area target vector is calculated as an area achievement degree Di, as follows: ; Where is the simulated area of type i, is the initial area target of type i.

[0038] The ratio accurately reflects the implementation of the area of each type of land. When is close to 1, it indicates that the area target of the land type has been basically achieved; when is significantly less than 1, it indicates that the land type is insufficiently expanded due to spatial constraints; and when is significantly greater than 1, it indicates that the land type is passively expanded beyond the expectation.

[0039] S42. Spatial conflict type identification Based on the area achievement degree Di, the land use type whose area does not reach the preset threshold is identified as a spatial conflict type; for example, if the Di of a certain construction land type is 0.85, which is lower than the preset lower limit of 0.95, it indicates that the original expansion target cannot be realized under the given spatial constraints, and this construction land is identified as a spatial conflict type that needs to be modified.

[0040] S43. Area target vector modification According to the area difference of the spatial conflict type, the initial area target vector is modified, that is, the absolute difference between the target and the actual simulation area, to generate the modified area target vector, as follows: ; The area difference amount of the conflict type i is redistributed to one or more other land use types according to the transition probability in the Markov transition probability matrix M. ; For each allocated non-conflict type j, the modified area target is obtained by the following formula: ; where k is the traversal of all allocated land use types. is the probability of transition from conflict type i to type j, is the sum of the probabilities of all other land use types except the current conflict type i; is a normalization factor representing the sum of the probabilities of transition from conflict type i to all currently included land use types k.

[0041] Finally, the modified area targets of all land use types (including unmodified types and all modified types) together constitute the modified area target vector for the next iteration.

[0042] As a preferred way, in the iterative simulation process of step S5, it is necessary to objectively and quantitatively judge whether the simulation process has reached a stable state, so as to decide whether to terminate the iteration and output the result, or to continue the feedback and modification. The convergence judgment standard in the past decade is as follows: After each iteration simulation is completed, the overall deviation δ of the area achievement degree Di of all land use types in the current iteration simulation result from 1 is calculated. The calculation of this index is based on the deviation of the area achievement degree Di of each type from the ideal achievement degree 1 in the current iteration simulation result as follows: ; Wherein, N is the number of land use types; Σ represents the sum of the squared deviations of all N land use types.

[0043] When the overall deviation δ is less than a preset tolerance threshold ε, it is determined to be converged; the threshold ε is a decimal greater than 0 (for example, 0.01, 0.05, etc.), which is preset by a user according to the accuracy requirement required by simulation.

[0044] According to the above embodiment, the present application firstly predicts the future area demand based on the historical land use data through the Markov chain as the initial target; then the FLUS model is used to combine the multi-source driving factors to perform the first spatial allocation simulation. By calculating the achievement degree of the simulation result and the area target, the conflict type that cannot achieve the target due to spatial constraints is automatically identified, and the unachieved area is redistributed to other types according to the historical conversion rule based on the Markov transition probability to generate a corrected area target. The corrected target is input into the model again for iterative simulation, and the overall deviation of the achievement degree of all land class areas is compared with the preset threshold to automatically judge the convergence.

[0045] Embodiment 2 Another embodiment of the present application proposes a Markov-FLUS model-based feedback land use change simulation system for realizing the steps of the method in embodiment 1, which comprises a processor and a memory, the memory stores a computer program, and the processor is configured to execute the computer program to realize the functions of the following modules; the system comprises a data acquisition module, an area prediction module, an initial simulation module, a feedback correction module and a judgment module.

[0046] The data acquisition module is used to acquire land use data sets at different historical periods and multi-source driving factor data sets related to land use change.

[0047] The land use data set can be acquired from a geographic information database, a remote sensing image library and a statistical database; and the original data is preprocessed through a spatial registration unit, a coordinate unification unit and a data standardization unit to form a spatially consistent standardized data set The area prediction module is used to construct a Markov transition probability matrix based on the land use data set, and to use the preinstalled area prediction unit to deduce the area demand of each land use type in the future target year according to the matrix to generate an initial area target vector.

[0048] The initial simulation module is configured to input an initial area target vector into the FLUS model in combination with the multi-source driving factor dataset, and the module includes a suitability probability calculation unit and a spatial allocation unit, wherein the suitability probability calculation unit processes the multi-source driving factor data by an artificial neural network algorithm to generate a suitability probability distribution map of each land use type, and the spatial allocation unit converts the area target into a spatial distribution pattern based on an adaptive inertia competition mechanism to output an initial simulation result.

[0049] The feedback correction module includes an area achievement calculation unit, a conflict identification unit and a target correction unit, which identifies the spatial conflict type by calculating the ratio of the simulation area to the target area, and dynamically corrects the area target vector based on a Markov transition probability matrix.

[0050] The judgment module includes a convergence judgment unit and an iteration control unit, which judges whether the system reaches a convergence state by calculating the overall deviation of the area achievement of all land use types, and controls the continuation of iteration or the output of the result of the system according to the judgment result.

[0051] For specific limitations of the Markov-FLUS model-based feedback land use change simulation system, refer to the limitations of the Markov-FLUS model-based feedback land use change simulation method described above, which will not be repeated here. It should be noted that the modules in the above prediction system correspond to steps S1 to S6 in the implementation of the above prediction method, and the instances and application scenarios realized by the modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment 1.

[0052] In the above embodiments, all or part of them can be realized by software, hardware, firmware or any combination thereof. When realized by software, all or part of them can be realized in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated.

[0053] The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in or transmitted from a computer-readable storage medium, such as from one website, computer, server, or data center to another website, computer, server, or data center, via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) medium. The computer-readable storage medium can be any available medium or data storage device that is accessible by a computer, or a data storage device, such as a server, data center, etc., that includes one or more of the available media integrated therein. The available medium can be a magnetic medium (e.g., a floppy diskette, a hard disk drive, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a Solid State Disk (SSD)), etc.

[0054] Those skilled in the art can understand that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0055] In addition, each functional module in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0056] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent ones; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A simulation method of land use change based on Markov-FLUS model feedback, characterized in that, The method comprises: S1. Obtain land use data sets of different historical periods of a target area, and multi-source driving factor data sets related to land use change; S2. Construct a Markov transition probability matrix based on the land use data sets, predict the area demand of each land use type in the future target year, and record it as an initial area target vector; S3. Combine the multi-source driving factor data sets, input the initial area target vector into the FLUS model, and perform a first land use spatial allocation simulation to obtain an initial simulation result; S4. Based on the difference between the initial simulation result and the initial area target vector, determine a corrected area target vector; S5. Input the corrected area target vector into the FLUS model for iterative simulation, and judge whether to converge according to the simulation result; S6. If it converges, output the final land use change simulation result; if it does not converge, return to step S4, generate a corrected area target vector again based on the current simulation result and perform iterative simulation until it converges or reaches the maximum number of iterations. 2.The method of claim 1, wherein, The multi-source driving factor data set includes natural geographical factors, location traffic factors, social and economic factors, and constraint factors; wherein, The natural geographical factors include one or more of elevation, slope, slope direction, precipitation, and soil type; The location traffic factors include one or more of distances to highways, railways, ports, airports, and urban centers; The social and economic factors include one or more of population density, spatial distribution of GDP, and nighttime light index; The constraint factors are spatially characterized by one or more of ecological protection red lines, permanent basic farmland ranges, urban development boundaries, and natural conservation areas. 3.The method of claim 1, wherein, Step S3 comprises: S31. Use the land use data sets of the historical periods and the multi-source driving factor data sets to train an artificial neural network model to obtain suitability probability distribution maps of each land use type; S32. Combine the suitability probability distribution maps and the initial area target vector to perform a first land use spatial allocation simulation by the FLUS model; Wherein, the total adaptive probability of land use type i on the cell (x, y) in the kth iteration of the FLUS model is The calculation formula is: ； wherein, is the suitability probability obtained by the artificial neural network, is the neighborhood influence factor, is the conversion cost from the current land use type to type i. 4.The method of claim 3, wherein, In step S31, the training by the artificial neural network model to obtain the suitability probability distribution maps comprises: Standardize each factor data in the multi-source driving factor data set to form a standardized driving factor layer; Spatially register and superimpose the land use data sets of the historical periods and the standardized driving factor layer; Train the artificial neural network using any period of land use data in the historical periods as a training label and the driving factor layer corresponding to the land use data of the previous period of the any period as an input feature; Use the trained model to predict the driving factor data of the target year to generate suitability probability distribution maps of each land use type in the target year. 5.The method of claim 3, wherein the Markov-FLUS model-based feedback land use change simulation method is characterized by, The neighborhood influence factor The determination manner comprises: In a preset n*n neighborhood window, calculate the ratio Q of the number of cells belonging to land use type i around the center cell (x, y) to the total number of cells in the neighborhood window, as follows: ； Wherein, NumTypeI is the number of cells belonging to type i in the neighborhood, (n×n - 1) is the total number of cells in the neighborhood except the center cell; The ratio Q is weighted by a distance attenuation function to obtain a neighborhood influence factor for representing different influence degrees of the neighboring cells on the center cell As follows: ； where d is each cell in the neighborhood window except the center cell; is the land use type of the dth cell in the neighborhood; is a conditional function that takes the value 1 when and 0 otherwise; is a weight calculated based on the distance d between cell d and the center cell (x, y), and the value of decreases as the distance d increases. 6.The method of simulating feedback of land use change based on Markov-FLUS model according to claim 1, wherein, Step S4 comprises: S41. Calculate the ratio of the simulated area of each land use type in the initial simulation result to the target area of the corresponding type in the initial area target vector as the area achievement degree Di, as follows: ； wherein, is the simulated area of type i, is the initial area target of type i; S42. Based on the area achievement degree Di, identify the land use type whose area does not reach the preset threshold as the spatial conflict type; S43. According to the area difference of the spatial conflict type, modify the initial area target vector to generate the modified area target vector. 7.The method of claim 6, wherein the Markov-FLUS model is based on a Markov chain model and a fuzzy logic-based unsupervised soft clustering (FLUS) model. In step S43, the modification according to the area difference of the spatial conflict type comprises: Determine the unachieved area difference amount ΔAi based on the identified spatial conflict type, as follows: ； allocating weights to the area difference amounts with probabilities of transition from the type i in the Markov transition probability matrix M to other one or more land use types; for each assigned non-conflicting type j, its revised area target is obtained by ； where k is the number of all assigned land use types; Pij is the probability of shifting from conflict type i to type j, Pj is the sum of probabilities of all other land use types except the current conflict type i. The modified area targets of all land use types together constitute the modified area target vector. 8.The method of claim 6, wherein the Markov-FLUS model-based feedback land use change simulation method is characterized by, The specific criterion for determining whether to converge in step S5 is: Calculate the overall deviation degree δ of the area achievement degrees Di of all land use types in the current iteration simulation result and 1, as follows: ； Wherein, N is the number of land use types; When the overall deviation δ is less than the preset tolerance threshold ε, it is determined to be converged.

9. A feedback land use change simulation system based on Markov-FLUS model for implementing the steps of the method according to any one of claims 1 to 8, characterized in that, The system comprises: A data acquisition module configured to acquire a land use data set of different historical periods and a multi-source driving factor data set related to land use change; An area prediction module configured to construct a Markov transition probability matrix based on the land use data set, and predict an area demand of each land use type in a target year, denoted as an initial area target vector; An initial simulation module configured to combine the multi-source driving factor data set, input the initial area target vector into a FLUS model, and perform a first land use spatial allocation simulation to obtain an initial simulation result; A feedback correction module configured to determine a modified area target vector based on a difference between the initial simulation result and the initial area target vector; A judgment module configured to input the modified area target vector into the FLUS model again for iteration simulation, and determine whether to converge according to a simulation result; if converged, output a final land use change simulation result; if not converged, trigger the feedback correction module to work again based on the current simulation result, generate a new modified area target vector, and perform next iteration simulation until converged or a maximum iteration number is reached.

Images