Land utilization data downscaling method based on generalized addition model and related equipment

Abstract

The invention discloses a land utilization data downscaling method based on a generalized addition model and related equipment, and the method comprises the steps: obtaining an LUH2 data set and a downscaling demand, the LUH2 data set comprising data of a plurality of land types; on the basis of the LUH2 data set, downscaling requirements and preset driving element data, fishing net construction is conducted on the preset area, and a thick fishing net and a thin fishing net are obtained; the thick fishing net and the thin fishing net are input into a preset initial generalized addition model for downscaling processing, and an intermediate land class area prediction result for the preset area is obtained; optimizing the initial generalized addition model based on an intermediate land class area prediction result to obtain a target generalized addition model; and outputting a target land class area prediction result for the preset region based on the target generalized addition model. The problem that the resolution ratio of the LUH2 data set is low is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a land use data downscaling method based on a generalized additive model and related equipment, and the related equipment includes a land use data downscaling system based on a generalized additive model, a computing device and a computer readable storage medium. BACKGROUND

[0002] Global-scale land use scenario data (taking the well-known LUH2 dataset as an example) provides an important data basis and support for many fields such as ecosystem models, climate models, biodiversity assessment models, etc. For example, the LUH2 (Land Use Harmonization 2) dataset is a standardized dataset of global land use and land cover change developed by international climate change research institutions and widely used. This dataset integrates historical land use change observation data and future land use scenario prediction data, covering a continuous time series from the historical period (AD 850) to the future scenario (2100), providing long-term data support for global-scale land use pattern evolution. However, since the LUH2 dataset is designed for global-scale research, its spatial resolution is low, generally 0.25°x0.25° (about 25 kilometers), which is sufficient in global-scale research, but there is a large spatial scale error in regional or local scale (such as city, township scale) research. This low spatial resolution land use data is difficult to directly apply to fine regional ecological environment management, land use planning, biodiversity protection and urban planning, etc. because land use changes at local scales often exhibit strong spatial heterogeneity, and this heterogeneity has an important impact on ecosystem services and land resource management. Therefore, there is an urgent need for a LUH2 data downscaling method to improve the resolution of land use data. SUMMARY

[0003] In order to overcome the problem of low resolution of the LUH2 dataset, the present application provides a land use data downscaling method based on a generalized additive model and related equipment, and the related equipment includes a land use data downscaling system based on a generalized additive model, a computing device and a computer readable storage medium.

[0004] In a first aspect, in order to solve the above technical problems, the present application provides a land use data downscaling method based on a generalized additive model, comprising: obtaining a LUH2 dataset and a downscaling requirement, the LUH2 dataset including data of multiple land class types; constructing a fishing net for a preset region based on the LUH2 dataset, the downscaling requirement and preset driving factor data to obtain a coarse fishing net and a fine fishing net; input the coarse fishnet and the fine fishnet into the preset initial generalized additive model for downscaling processing to obtain an intermediate land class area prediction result for the preset area; optimize the initial generalized additive model based on the intermediate land class area prediction result to obtain a target generalized additive model; output a target land class area prediction result for the preset area based on the target generalized additive model.

[0005] Further, the coarse fishnet and the fine fishnet are obtained by performing fishnet construction on the preset area based on the LUH2 data set, a downscaling requirement and preset driving factor data, including: performing fishnet construction on the preset area based on the LUH2 data set and the preset driving factor data to obtain the coarse fishnet; performing fishnet construction on the preset area based on the downscaling requirement and the driving factor data to obtain the fine fishnet.

[0006] Further, the coarse fishnet is obtained by performing fishnet construction on the preset area based on the LUH2 data set and preset driving factor data, including: cutting the LUH2 data set based on the preset area to obtain a low-resolution data set for the preset area; constructing a first vector fishnet matching the size of the low-resolution data set; correspondingly adding the low-resolution data set and the preset driving factor data to the first vector fishnet to form the coarse fishnet.

[0007] Further, the downscaling requirement is a requirement of downscaling the LUH2 data set to a target resolution; performing fishnet construction on the preset area based on the downscaling requirement and the driving factor data to obtain the fine fishnet, including: constructing a second vector fishnet corresponding to the preset area, the resolution of the second vector fishnet being the same as the target resolution; correspondingly adding the driving factor data to the second vector fishnet to form the fine fishnet.

[0008] Further, the initial generalized additive model includes a land class generalized additive model corresponding to each land class type, and the driving factor data includes a plurality of driving factors; input the coarse fishnet and the fine fishnet into the preset initial generalized additive model for downscaling processing to obtain an intermediate land class area prediction result for the preset area, including: input the coarse fishnet and the fine fishnet into the preset initial generalized additive model, and use the land class generalized additive model to predict based on the driving factors in the fine fishnet to obtain an original area prediction result of the corresponding land class type in the fine fishnet, the original area prediction result including an original land class area value and an original standard error; Based on the plurality of original area prediction results, an initial land class area prediction result for the preset area is formed; Based on the coarse fishing net and the fine fishing net, the initial land class area prediction result is corrected to obtain an intermediate land class area prediction result for the preset area.

[0009] Further, the initial generalized additive model includes a land class generalized additive model corresponding to each land class type, and the intermediate land class area prediction result includes an intermediate land class area value and an intermediate standard error of each land class generalized additive model; Based on the intermediate land class area prediction result, the initial generalized additive model is optimized to obtain a target generalized additive model, including: Using a preset constraint optimization function, based on the plurality of intermediate land class area values, the plurality of intermediate standard errors and the corresponding plurality of true land class area values, a constraint value is calculated; Based on the prediction value difference between each intermediate land class area value and the corresponding true land class area value, an average difference is calculated; Based on the constraint value and the average difference, the network parameters in the initial generalized additive model are adjusted to obtain the target generalized additive model.

[0010] Further, based on the constraint value and the average difference, the network parameters in the initial generalized additive model are adjusted to obtain the target generalized additive model, including: When the constraint value is greater than a preset value, and / or, the average difference is greater than a preset difference, the network parameters in the initial generalized additive model are adjusted to obtain an intermediate generalized additive model; The coarse fishing net and the fine fishing net are input into the intermediate generalized additive model for downscaling processing to obtain a transition land class area prediction result for the preset area; When the constraint value corresponding to the transition land class area prediction result is less than or equal to a preset value, and the corresponding average difference is less than or equal to a preset difference, the corresponding intermediate generalized additive model is determined as the target generalized additive model.

[0011] In a second aspect, the present application also provides a land use data downscaling system based on a generalized additive model, including: An acquisition module is configured to acquire a LUH2 data set and a downscaling requirement, and the LUH2 data set includes data of a plurality of land class types; A fishing net construction module is configured to construct a fishing net for a preset area based on the LUH2 data set, the downscaling requirement and preset driving factor data to obtain a coarse fishing net and a fine fishing net; A downscaling module is configured to input the coarse fishing net and the fine fishing net into a preset initial generalized additive model for downscaling processing to obtain an intermediate land class area prediction result for the preset area; a model optimization module, configured to optimize the initial generalized additive model based on the intermediate land class area prediction result, to obtain a target generalized additive model; a result output module, configured to output a target land class area prediction result for the preset area based on the target generalized additive model.

[0012] In a third aspect, the present application provides a computing device, comprising a memory, a processor, and a program stored in the memory and running on the processor, and the processor implements the steps of the land use data downscaling method based on a generalized additive model when executing the program.

[0013] In a fourth aspect, the present application provides a computer readable storage medium, which stores instructions, and the instructions, when running on a terminal device, cause the terminal device to execute the steps of the land use data downscaling method based on a generalized additive model.

[0014] The present application has the following beneficial effects. Firstly, the coarse mesh and the fine mesh of the preset area are constructed based on the LUH2 dataset, the downscaling requirement, and the preset driving factor data, the coarse mesh and the fine mesh are input into the preset initial generalized additive model for downscaling processing, and the initial generalized additive model is optimized by using the intermediate land class area prediction result obtained by the downscaling processing, so as to improve the downscaling precision of the target generalized additive model obtained by optimization, so that the target land class area prediction result output by the target generalized additive model for the preset area can meet the downscaling requirement, thereby improving the resolution of the land use scenario data corresponding to the preset area in the LUH2 dataset. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 a flowchart of a land use data downscaling method based on a generalized additive model according to an example embodiment of the present application; Figure 2 a flowchart of a land use data downscaling method based on a generalized additive model according to an example embodiment of the present application; Figure 3 an example diagram of an intermediate land class area prediction result according to an example embodiment of the present application; Figure 4 an example diagram of a target land class area prediction result according to an example embodiment of the present application; Figure 5 a structural diagram of a land use data downscaling system based on a generalized additive model according to an example embodiment of the present application. DETAILED DESCRIPTION

[0016] The following examples are further explanations and supplements of the present application and do not constitute any limitation on the present application.

[0017] Existing global or regional scale land use datasets (e.g. Land Use Scenario data LUH2) often have low spatial resolution, which cannot meet the needs of fine-scale land use change research in local areas. Specifically, such datasets (taking LUH2 as an example) are widely used in the fields of global-scale ecology, land use change, biodiversity assessment, and environmental policy research as an important data source. They are used to simulate and predict long-term trends in land use changes, support ecological environment management and land planning decisions. However, due to the current coarse spatial resolution of LUH2 data (about 0.25°, about 25 km), it is difficult to directly apply it to fine-scale land management, environmental risk assessment, and ecological protection planning. For analyses such as urban expansion, fine-scale management of farmland, and assessment of ecosystem services, which require high-precision spatial data support, it is difficult to meet the needs. Therefore, to more effectively utilize the value of such low-resolution data, there is an urgent need for a scientific, efficient, and accurate downscaling method to obtain higher spatial resolution land use data to meet the practical needs of regional-scale, especially urban and county-level, fine-scale land use change research and management planning.

[0018] Currently, there have been some attempts at downscaling methods in academia and industry, which mainly include the following categories: (1) Spatial interpolation method Commonly used spatial interpolation methods include inverse distance weighted interpolation (IDW) and Kriging interpolation (Kriging), which are based on the spatial continuity between adjacent grid data for downscaling. This method is usually simple to operate and fast to calculate, but it assumes that the spatial continuity of land use types is strong, which performs poorly in areas with obvious spatial boundaries and strong heterogeneity (such as urban boundaries and forest-farmland boundaries), and may produce unrealistic transition areas.

[0019] (2) Simple area allocation method (proportional area method) This method usually allocates low-resolution LUH2 land use types to fine-scale grids in proportion to the area based on existing high-resolution reference land cover data (such as MODIS, ESA WorldCover data). This method also has obvious defects: it does not consider the ecological and socio-economic driving forces of land use transition processes, resulting in a static area redistribution of the downscaling results, which cannot reflect the spatial dynamic characteristics of actual land use changes.

[0020] (3) Empirical downscaling method based on historical statistics or expert knowledge This method relies on historical land use statistics, regional planning data, or expert experience, and uses regression models or rule-driven methods for downscaling. For example, Giuliani et al. (2022) used expert knowledge combined with high-resolution remote sensing data to finely divide land use data. Although this method can reflect the actual land use characteristics at the regional scale to some extent, it requires high-quality data and expert experience, and is difficult to implement in areas where data is lacking or there is no historical record.

[0021] (4) Model coupling and machine learning downscaling methods In recent years, with the rapid development of geographic information technology and machine learning technology, models such as Random Forest and Convolutional Neural Network (CNN) have been gradually applied to the data downscaling process of LUH2 and other data. For example, Rashidi et al. (2023) used machine learning methods combined with land suitability models to downscale global LUH2 data to the national scale and achieved high accuracy. However, this method usually requires a large amount of sample data for model training, has high computational complexity, and may not have sufficient generalization ability in areas with limited data.

[0022] However, there are some obvious shortcomings in the application of various downscaling methods in the academic community at present: (1) Lack of spatial precision In existing downscaling methods, spatial interpolation and simple area allocation methods are effective in some cases, but they all have the problem of insufficient precision. Spatial interpolation methods (such as IDW and Kriging interpolation) assume that land use types are continuous in space, but this does not conform to reality, especially in areas with obvious spatial boundaries such as urban expansion and the boundary between farmland and forest, which are prone to unrealistic transition zones. Simple area allocation methods allocate low-resolution land use data to high-resolution grids in proportion, although they are simple to implement, they ignore the dynamic characteristics of land use change and cannot reflect the spatial heterogeneity of land use change.

[0023] (2) Lack of consideration of spatial heterogeneity Most existing downscaling methods fail to effectively consider the spatial heterogeneity of land use change. Land use types and ecological environments within a region often have significant differences, and existing methods usually use simplified assumptions that ignore the interactions between different land use types and the influence of environmental factors. For example, downscaling methods based on historical statistics or expert knowledge can simulate realistic land use type distributions in some cases, but they rely on a large amount of expert experience and historical data, lack consideration of current spatial heterogeneity, and are often difficult to apply in areas where data is lacking.

[0024] (3) High computational complexity Machine learning and model coupling methods (such as random forests, convolutional neural networks, etc.) have advantages in improving the accuracy of downscaling, but such methods often require a large amount of training data and high computing resources, and have high computational complexity. For example, when using a random forest model for downscaling, a large amount of sample data needs to be trained, and multiple hyperparameters need to be optimized, which can lead to low computational efficiency when processing large-scale data, especially when processing high-resolution data, the training and calculation time is significantly increased.

[0025] (4) Poor universality of the model Most existing downscaling methods are developed for specific regions or specific types of land use data, so they have poor universality. Many methods rely on high-quality auxiliary data (such as remote sensing images, climate data, etc.), and in some areas, especially in areas where data is scarce, the effectiveness of existing methods is difficult to guarantee. This results in great limitations in the application of these methods on a global scale, especially when auxiliary data is not available or incomplete, existing methods cannot provide reliable downscaling results.

[0026] To solve the above problems, the embodiments of the present application provide a land use data downscaling method based on a generalized additive model and related equipment, related equipment includes a land use data downscaling system based on a generalized additive model, a computing device and a computer readable storage medium, the following will be described in detail.

[0027] The land use data downscaling method based on a generalized additive model provided by the embodiments of the present application can be specifically executed by a server. It should be noted that the server can be an independent server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms, etc. Basic cloud computing services are not limited here. The automation control logic of the land use data downscaling method based on a generalized additive model of the present application can be realized by writing codes using R language or Python, etc. The downscaling of the present application refers to the process of converting the information of the low-resolution LUH2 dataset into higher-resolution, more local-scale data.

[0028] Please refer to Figure 1 , Figure 1 A land use data downscaling method based on a generalized additive model is shown in an exemplary embodiment of the present application, as shown in Figure 1 The present application provides a land use data downscaling method based on a generalized additive model, which comprises: S11, acquire a LUH2 data set and a downscaling requirement, the LUH2 data set including data of multiple land class types; S12, perform mesh construction on a preset region based on the LUH2 data set, the downscaling requirement, and preset driving factor data, to obtain a coarse mesh and a fine mesh; S13, input the coarse mesh and the fine mesh into a preset initial generalized additive model for downscaling processing, to obtain an intermediate land class area prediction result for the preset region; S14, optimize the initial generalized additive model based on the intermediate land class area prediction result, to obtain a target generalized additive model; S15, output a target land class area prediction result for the preset region based on the target generalized additive model.

[0029] The generalized additive model-based land use data downscaling method provided in this embodiment first constructs a coarse mesh and a fine mesh of a preset region based on a LUH2 data set, a downscaling requirement, and preset driving factor data, inputs the coarse mesh and the fine mesh into a preset initial generalized additive model for downscaling processing, and optimizes the initial generalized additive model by using an intermediate land class area prediction result obtained through the downscaling processing, to improve the downscaling precision of a target generalized additive model obtained through optimization, so that a target land class area prediction result output by the target generalized additive model for the preset region can meet the downscaling requirement, thereby improving the resolution of land use scenario data corresponding to the preset region in the LUH2 data set. The target land class area prediction result is land use scenario data after downscaling in the form of a planar mesh, and corresponding raster data is obtained after vector-to-raster processing.

[0030] In an example embodiment provided in this application, the LUH2 data set can include multiple land class types (such as 17 types), and theoretically, the data of these different land class types can be simultaneously subjected to downscaling processing, but in actual research, most cases do not require so many land class types, and therefore, the land class types of the LUH2 data set acquired in this application are classified and integrated into 5 types according to previous research, namely, PRI (primary forest land), SEC (secondary forest land), CROP (arable land), PAS (grassland), and URB (city).

[0031] Before downscaling, the corresponding driving factor data needs to be prepared, including but not limited to climate, soil, terrain, land use, human activity intensity, etc. The resolution of these driving factors needs to be consistent with the target resolution of the down-scaled target land class area prediction results to ensure the accuracy of the downscaling method, and the downscaling requirements are set based on the target resolution. At the same time, for the selected different layer driving factor data, preprocessing is needed, including logistic regression, normalization, logarithmic processing, etc. to ensure that the value range difference between different driving factor data is not too large, and the value distribution as much as possible follows the normal distribution.

[0032] The construction method of the initial generalized additive model (GAM model): The downscaling idea applied in this method is to convert the relationship between coarse-grained response data and fine-grained covariate data into fine-grained response prediction. Therefore, the generalized additive model is used to construct the regression relationship between the original low-resolution data set and the driving factor layer: the area value of each land class type is taken as the response variable, and each driving factor data is taken as the driving variable, to construct the generalized additive model of each land class type corresponding to the quasi-binomial and logistic connection function (see code for details). In the subsequent process, VIF collinearity detection, comparison of AIC values in the stepwise regression process, evaluation of whether random slope is needed, and simplification of the land class generalized additive model by backward stepwise can be used to determine the final optimal prediction model structure of each land class type, forming the target generalized additive model.

[0033] Optionally, based on the LUH2 data set, the downscaling requirement, and the preset driving factor data, a fishing net is constructed for the preset area to obtain a coarse fishing net and a fine fishing net, including: Based on the LUH2 data set and the preset driving factor data, a fishing net is constructed for the preset area to obtain a coarse fishing net; Based on the downscaling requirement and the driving factor data, a fishing net is constructed for the preset area to obtain a fine fishing net.

[0034] In the embodiment provided in the present application, the preset area is constructed based on the LUH2 data set and the preset driving factor data, a low-resolution coarse fishing net containing the LUH2 data set and the driving factor data is obtained, the preset area is constructed based on the downscaling requirement and the driving factor data, a fine fishing net is obtained, a fine fishing net containing the driving factor data and meeting the downscaling requirement and the target resolution is obtained, and the subsequent target land class area prediction result for the preset area obtained by performing downscaling processing on the coarse fishing net based on the fine fishing net can meet the downscaling requirement, so as to improve the resolution of the land use scenario data corresponding to the preset area in the LUH2 data set. The preset area can be a national territory area.

[0035] Optionally, the preset area is constructed based on the LUH2 data set and the preset driving factor data, and the coarse fishing net is obtained, including: The LUH2 data set is cropped based on the preset area, and a low-resolution data set for the preset area is obtained; A first vector fishing net matching the size of the low-resolution data set is constructed; The low-resolution data set and the preset driving factor data are correspondingly added to the first vector fishing net, and the coarse fishing net is formed.

[0036] In the embodiment provided in the present application, first, the LUH2 data set is cropped to obtain a low-resolution data set in the land range of the preset area, and a first vector fishing net matching the size of the low-resolution data set is constructed, so that the first vector fishing net can cover the pixels of the low-resolution data set. Second, the low-resolution data set and the preset driving factor data are correspondingly added to the first vector fishing net, and the coarse fishing net is formed, so that subsequent downscaling processing on the coarse fishing net based on the fishing net form using the fine fishing net can be performed, the downscaling of the land use scenario data corresponding to the preset area in the LUH2 data set is realized, and thus the resolution of the land use data is improved.

[0037] In an example embodiment provided by the present application, the scope of the low-resolution LUH2 dataset is generally global or large area, so it is necessary to crop the required part, that is, the low-resolution dataset within the land scope of the preset area, so that the dataset in different areas of the LUH2 dataset can be down-scaled in batches, the data processing amount of each down-scaling is reduced, and the down-scaling performance of the model can be improved. When dividing the preset area for each down-scaling, the preset area can be divided according to different conditions such as the economic zone, climate zone, and ecosystem type in which the study area is located. The corresponding dataset area of the required preset area in the LUH2 dataset is pre-cropped using QGIS 3.40.1 software (open source geographic information system GIS software), and a first vector mesh with a pixel size of 0.25° that can cover the preset area is established using the software. The information of various driving factors and the data area of each land class type in the low-resolution dataset are uniformly counted on the mesh to form a coarse mesh. After counting, the coarse mesh layer is exported as a CSV file (a pure text file).

[0038] Optionally, the down-scaling requirement is a requirement of down-scaling the LUH2 dataset to a target resolution; Based on the down-scaling requirement and the driving factor data, a mesh is constructed for the preset area to obtain a fine mesh, including: A second vector mesh corresponding to the preset area is constructed, and the resolution of the second vector mesh is the same as the target resolution corresponding to the down-scaling requirement; The driving factor data is added to the second vector mesh to form the fine mesh.

[0039] In this embodiment provided by the present application, the second vector mesh corresponding to the preset area is constructed so that the resolution of the second vector mesh is the same as the target resolution corresponding to the down-scaling requirement, and the driving factor data is added to the second vector mesh to form the fine mesh. This facilitates the subsequent down-scaling processing of the coarse mesh based on the mesh form to obtain a target land class area prediction result for the preset area, which can meet the down-scaling requirement, thereby improving the resolution of the land use scenario data corresponding to the preset area in the LUH2 dataset.

[0040] In an example embodiment provided by the present application, the final down-scaling target and the study range are determined. Here, the land range of a certain country with a resolution of 1 km is taken as an example of the preset area. A second vector mesh with a resolution of 1 km is generated in QGIS 3.40.1 software with the preset area as the range, and the position selection tool is used to remove the unnecessary mesh units. After successful generation, the various driving factor data with a resolution of 1 km is counted on the second vector mesh with a resolution of 1 km to form a fine mesh. After counting, the fine mesh with a resolution of 1 km is exported as a CSV file (a pure text file).

[0041] Optionally, the initial generalized additive model comprises a land class generalized additive model corresponding to each land class type, and the driving factor data comprises a plurality of driving factors; The coarse fishnet and the fine fishnet are input into the preset initial generalized additive model for downscaling processing to obtain an intermediate land class area prediction result for the preset area, comprising: The coarse fishnet and the fine fishnet are input into the preset initial generalized additive model, and the land class generalized additive model is used to make a prediction based on the driving factors in the fine fishnet to obtain an original area prediction result of the corresponding land class type in the fine fishnet, the original area prediction result comprising an original land class area value and an original standard error; Based on the plurality of original area prediction results, an initial land class area prediction result for the preset area is formed; The initial land class area prediction result is corrected based on the coarse fishnet and the fine fishnet to obtain the intermediate land class area prediction result for the preset area.

[0042] In this embodiment, the coarse fishnet and the fine fishnet are input into the preset initial generalized additive model, the land class generalized additive model is used to make a prediction based on the driving factors in the fine fishnet to obtain an original area prediction result of the corresponding land class type in the fine fishnet, based on the plurality of original area prediction results, an initial land class area prediction result for the preset area is formed, and the initial land class area prediction result is corrected based on the coarse fishnet and the fine fishnet to realize downscaling processing of the land use scenario data in the LUH2 dataset corresponding to the preset area in the coarse fishnet, to obtain the intermediate land class area prediction result for the preset area, which facilitates subsequent optimization of the initial generalized additive model based on the intermediate land class area prediction result, improves the downscaling precision of the target generalized additive model obtained through optimization, and enables the target generalized additive model to meet the downscaling requirement, so that the target land class area prediction result for the preset area output by the target generalized additive model can meet the downscaling requirement, and the resolution of the land use scenario data in the LUH2 dataset corresponding to the preset area is improved. The standard error refers to the standard deviation of the sampling distribution of a sample statistic (such as mean, regression coefficient, etc.) under repeated sampling, which measures the stability or reliability of the statistic.

[0043] In an example embodiment provided by the present application, the land class generalized additive model is used to predict the driving factors in the fine fishing net to obtain the original area prediction result of the corresponding land class type in the fine fishing net, and the specific steps are as follows: for each driving factor in the fine fishing net, the driving factor is input into the land class generalized additive model for prediction to obtain the corresponding original prediction area value; the average value of the plurality of original prediction area values is determined as the original land class area value of the corresponding land class type in the fine fishing net; the standard error is calculated based on the plurality of original prediction area values; and the original area prediction result of the corresponding land class type in the fine fishing net is formed based on the original land class area value and the standard error.

[0044] Using the land class generalized additive model of different land class types in the constructed initial generalized additive model, each driving factor in the fishing net unit is input respectively to predict the original area prediction result (including the original land class area value and the original standard error) of different land class types in the fishing net unit of the 1km fine fishing net, and based on the plurality of original area prediction results, the initial land class area prediction result for the preset area is formed. The initial land class area prediction result predicted for the first time is separately exported into a CSV file (a pure text file).

[0045] Based on the coarse fishing net and the fine fishing net, the initial land class area prediction result is corrected to obtain the intermediate land class area prediction result for the preset area, and the specific steps are as follows: First, after obtaining the initial land class area prediction result of different land class types of the land use scenario data corresponding to the preset area in the LUH2 data set of the 1km fine fishing net, the intersection function in the QGIS 3.40.1 software can be used to attach the coarse fishing net ID to the fine fishing net, and after success, the mixed fishing net will appear two fields, which are the fine fishing net ID field ID_2 and the coarse fishing net ID field ID_1 to which each pixel belongs.

[0046] Secondly, in the R language, the area average value of different land class areas in the fishing net unit of the 1km fine fishing net under the fishing net unit of the 0.25° coarse fishing net is calculated in the mixed fishing net, and the area average value is compared with the original LUH2 area value in the 0.25° coarse fishing net, and then the predicted 1km land use value is recalibrated according to the correction coefficient obtained after the comparison, and the specific formula is as follows: Among them, represents the intermediate land class area value included in the multiplication scaled intermediate land class area prediction result, represents the original land class area value included in the original area prediction result, represents the area value of the original LUH2 in the 0.25° coarse fishing net, The area of different land types in the fishing net unit of 1km fine fishing net is averaged in the area of the fishing net unit of 0.25° coarse fishing net.

[0047] Optionally, the initial generalized additive model comprises a land class generalized additive model corresponding to each land class type, and the intermediate land class area prediction result comprises an intermediate land class area value of each land class generalized additive model and an intermediate standard error; The initial generalized additive model is optimized based on the intermediate land class area prediction result to obtain a target generalized additive model, comprising: A constraint value is obtained by using a preset constraint optimization function and based on the plurality of intermediate land class area values, the plurality of intermediate standard errors and the plurality of corresponding real land class area values; An average difference is calculated based on the prediction value difference between each intermediate land class area value and the corresponding real land class area value; The network parameters in the initial generalized additive model are adjusted based on the constraint value and the average difference to obtain the target generalized additive model.

[0048] In the embodiment provided in the present application, a constraint value is obtained by using a preset constraint optimization function and based on the plurality of intermediate land class area values, the plurality of intermediate standard errors and the plurality of corresponding real land class area values, and an average difference is calculated based on the prediction value difference between each intermediate land class area value and the corresponding real land class area value, so as to adjust the network parameters in the initial generalized additive model based on the constraint value and the average difference, so that the initial generalized additive model converges to meet the downscaling accuracy requirement, thereby improving the accuracy of the target land class area prediction result for the preset area output by the target generalized additive model meeting the downscaling accuracy requirement obtained by adjustment, so that it can meet the downscaling requirement, thereby improving the resolution of the land use scenario data corresponding to the preset area in the LUH2 data set.

[0049] In the example embodiment provided in the present application, when the network parameters in the initial generalized additive model are adjusted based on the constraint value and the average difference, the network parameters of each land class generalized additive model are actually adjusted, so that the prediction value difference corresponding to each land class generalized additive model is less than a preset value, so that each land class generalized additive model converges, thereby completing the convergence of the initial generalized additive model and obtaining the target generalized additive model meeting the downscaling requirement.

[0050] The setting logic of the constraint optimization function is as follows: Since the downscaling prediction results for each land use type are based on a separate generalized additive model, and the area values ​​of different land use types within each fishing net cell must sum to 1, with each land use type's area value ranging from 0 to 1, using different generalized additive models for prediction may lead to the predicted area of ​​different land use types within the same cell violating the aforementioned two constraints. Therefore, a constraint optimization function needs to be set to ensure that the prediction results meet the actual conditions. The formula for the constraint optimization function is: in, Indicates the constraint value. Indicates the number of land use types. Indicates the first The actual land area value corresponding to the land type. Indicates the first The predicted area of ​​intermediate land categories corresponding to different land types includes the area values ​​of those intermediate land categories. Indicates the first The intermediate standard error is included in the intermediate land area prediction results output by the generalized additive model of land types corresponding to different land types.

[0051] After multiple rounds of iteration, ensuring that the constraints are optimized The smallest, and meets the general conditions for land area prediction ( (Less than or equal to the preset value), to obtain the target generalized additive model, and output the constrained optimized prediction result of the target land type area for the preset area.

[0052] Optionally, the network parameters in the initial generalized additive model are adjusted based on the constraint values ​​and average differences to obtain the target generalized additive model, including: When the constraint value is greater than the preset value, and / or the average difference is greater than the preset difference, the network parameters in the initial generalized additive model are adjusted to obtain the intermediate generalized additive model. The coarse and fine fishing nets are input into the intermediate generalized summation model for downscaling to obtain the predicted area of ​​transitional land types for the preset area. When the constraint value corresponding to the predicted area of ​​transitional land is less than or equal to the preset value, and the corresponding average difference is less than or equal to the preset difference, the corresponding intermediate generalized additive model is determined as the target generalized additive model.

[0053] In the embodiment provided in the present application, during the process of training and adjusting the initial generalized additive model, the convergence values (preset values and preset differences) of the constraint value and the average difference are set, and when the constraint value and the average difference both meet the corresponding convergence values, the intermediate generalized additive model obtained by this training is determined as the target generalized additive model, so that the target generalized additive model converges to a model that can meet the accuracy requirement of the downscaling, thereby improving the accuracy of the target land class area prediction result for the preset area output by the target generalized additive model obtained by the adjustment and meeting the downscaling requirement, so as to improve the resolution of the land use scenario data corresponding to the preset area in the LUH2 data set. The preset difference can be 0.001.

[0054] In an example embodiment provided in the present application, when the intermediate land class area prediction result is imported into the QGIS software, it is found that there are still some problems in the downscaling result, so when the network parameters in the initial generalized additive model are adjusted based on the constraint value and the average difference, multiple rounds of iterative adjustment are required to obtain the target generalized additive model. The specific process of the multiple rounds of iterative adjustment is as follows: The intermediate land class area prediction result is imported into the initial generalized additive model trained before as the response variable. This step may have the problem of too large data volume, so sampling can be performed according to the hardware conditions. For example, 20% random sampling at equal intervals is performed, all data pixels are divided into 10 intervals of equal intervals, 3000-5000 pixels are extracted from each interval, and a total of 50,000-100,000 constraint optimization results are imported as the response variable.

[0055] When the constraint value is greater than the preset value, and / or the average difference is greater than the preset difference, the network parameters in the initial generalized additive model are adjusted to retrain the initial generalized additive model and obtain an intermediate generalized additive model.

[0056] The intermediate land class area prediction result is imported into the intermediate generalized additive model for processing to obtain a transition land class area prediction result, which is compared with the intermediate land class area prediction result to judge the difference of the prediction values of different pixels and different land class types, i.e., the two are subtracted. If the average difference exceeds 0.001 (preset difference), and / or the constraint value of this training is greater than the preset value, the transition land class area prediction result of this round is imported into the intermediate generalized additive model for retraining, and the process is repeated until the average difference is less than or equal to 0.001 (preset difference) and the constraint value is less than or equal to the preset value. It is proved that the model has converged and the target generalized additive model is obtained.

[0057] The target land class area prediction result output by the target generalized additive model is the final downscaling result for the preset area.

[0058] The final downscaling result is saved in CSV format and imported into QGIS software, the result is connected with the fishing net according to ID_1 and visualized, and finally the downscaling result is converted into a visual 1km raster file using the vector to raster tool, thus the 1km downscaling result of LUH2 of five land class types can be obtained.

[0059] Please refer to Figure 2 , Figure 2 In an exemplary embodiment of the present application, the flowchart of the land use data downscaling method based on the generalized additive model provided by the application is shown in FIG. 1, the application flow steps are as follows: Figure 2 Input the LUH2 land use data (LUH2 data set) with a resolution of 0.25°, and the driving factor data with a resolution of 1km in the form of independent variable layer, form the coarse fishing net and the fine fishing net for the preset area; The coarse fishing net and the fine fishing net are input into the initial generalized additive model, the initial generalized additive model defines the explained and predicted variables and the explained variables, so that the coarse fishing net and the fine fishing net are downscaled in the initial generalized additive model, and the first prediction result (intermediate land class area prediction result) with a resolution of 1km can be obtained after the first iteration, as shown in FIG. 2; Figure 3 Based on the intermediate land class area prediction result, the initial generalized additive model is optimized to obtain the target generalized additive model, and the optimization adjustment process is as follows: During the model training process, the network parameters in each land class generalized additive model are adjusted, if there is a land class generalized additive model that has converged at a certain iteration, that is, the difference between the predicted values corresponding to the land class generalized additive model is less than or equal to 0.001, it means that the land class generalized additive model has converged, then the network parameters of the land class generalized additive model after convergence are fixed (only the network parameters in the land class generalized additive model that has not been fixed are adjusted in the next iteration), and the transition land class area prediction result output by each round of model is imported into the intermediate generalized additive model after adjusting the network parameters in the next round, until all land class types in the intermediate generalized additive model correspond to the land class generalized additive model that has converged, and the final target generalized additive model is output; Then, based on the target generalized additive model, the final iteration result (target land class area prediction result) for the preset area is output, as shown in FIG. 3. Figure 4

[0060] As can be seen from the above, the land use data downscaling method based on the generalized additive model can improve the downscaling precision, solve the spatial heterogeneity problem, improve the calculation efficiency, and enhance the universality of the model.

[0061] ​​​The improvement of downscaling precision is embodied in that: by introducing a generalized additive model (GAM) and other spatial statistical models, combined with multiple driving factors (such as climate, soil, terrain, etc.), the spatial distribution of land use types can be more accurately simulated, especially for regions with significant spatial heterogeneity. Through this method, accurate land use change prediction results can be provided while maintaining high computational efficiency, and global-scale LUH2 data (0.25° resolution) can be downscaled to higher resolutions (such as 1 km or finer). This downscaling process not only guarantees accuracy but also effectively improves computational efficiency, making it suitable for land use research in different regions.

[0062] The solution to spatial heterogeneity is embodied in that: by introducing multiple geographical and environmental driving factors (such as climate, soil, terrain, etc.), the spatial heterogeneity of land use change can be captured, taking into account the spatial heterogeneity of land use types and ecological environment characteristics. Using multiple source driving factors as input variables, the dynamic characteristics of land use change can be captured through regression analysis, effectively reflecting the spatial distribution and change rules of different land use types. Compared with traditional single variable methods, the method of the present application can more realistically reflect the spatial dynamics of different land use types and the impact of ecological systems.

[0063] The improvement of computational efficiency is embodied in that: R language and geographic information system (GIS) are used for data processing, efficient spatial analysis methods and optimized algorithms (such as stepwise regression and VIF collinearity detection) are applied, and through stepwise regression, VIF collinearity detection, and AIC value optimization, the model constructed can reflect the complex land use change rules, maintaining high generalization ability and precision in a variable environment, ensuring high accuracy and reduced computation time in the downscaling process, making it suitable for large-scale data processing and complex regional analysis. At the same time, R language and GIS technology are combined, and GIS software (such as QGIS) is used for data preprocessing, regional clipping, and grid division, combined with the powerful data processing capability of R language, simplifying the data processing process and significantly improving the efficiency and accuracy of the downscaling process.

[0064] The universality of the model is enhanced in that: it can be flexibly adjusted according to the characteristics of different research regions (such as different land use types, ecological environment background, etc.), and in the case of insufficient or incomplete data, it can provide more accurate downscaling results by selecting appropriate driving factors and optimizing model parameters. Whether in data-rich regions or data-scarce places, reliable downscaling results can be provided, with strong universality.

[0065] Please refer to Figure 5 , Figure 5A land use data downscaling system based on a generalized additive model is shown in an example embodiment of the present application, as shown in Figure 5 The present application provides a land use data downscaling system 500 based on a generalized additive model, comprising: The acquisition module 501 is configured to acquire a LUH2 dataset and downscaling requirements, wherein the LUH2 dataset comprises data of multiple land class types; The fishnet construction module 502 is configured to construct a coarse fishnet and a fine fishnet for a preset region based on the LUH2 dataset, the downscaling requirements, and preset driving factor data; The downscaling module 503 is configured to input the coarse fishnet and the fine fishnet into a preset initial generalized additive model for downscaling processing to obtain an intermediate land class area prediction result for the preset region; The model optimization module 504 is configured to optimize the initial generalized additive model based on the intermediate land class area prediction result to obtain a target generalized additive model; The result output module 505 is configured to output a target land class area prediction result for the preset region based on the target generalized additive model.

[0066] The land use data downscaling system 500 based on a generalized additive model provided in this embodiment first constructs a coarse fishnet and a fine fishnet for a preset region based on the LUH2 dataset, the downscaling requirements, and the preset driving factor data acquired by the acquisition module 501, inputs the coarse fishnet and the fine fishnet into a preset initial generalized additive model for downscaling processing by the downscaling module 503, and optimizes the initial generalized additive model using the intermediate land class area prediction result obtained by the downscaling processing by the model optimization module 504 to improve the downscaling precision of the target generalized additive model obtained by optimization, so that the target land class area prediction result for the preset region output by the target generalized additive model in the result output module 505 can meet the downscaling requirements, thereby improving the resolution of the land use scenario data corresponding to the preset region in the LUH2 dataset.

[0067] Optionally, the fishnet construction module 502 is specifically configured to: construct a coarse fishnet for the preset region based on the LUH2 dataset and the preset driving factor data; construct a fine fishnet for the preset region based on the downscaling requirements and the driving factor data.

[0068] Optionally, the fishnet construction module 502 is specifically configured to: crop the LUH2 dataset based on the preset region to obtain a low-resolution dataset for the preset region; construct a first vector fishnet matching the size of the low-resolution dataset; corresponding to the preset driving factor data is added into the first vector fishing net to form a coarse fishing net.

[0069] Optionally, the downsizing requirement is a requirement of downsizing the LUH2 dataset to a target resolution; The fishing net construction module 502 is specifically configured to: The second vector fishing net corresponding to the preset area is constructed, and the resolution of the second vector fishing net is the same as the target resolution; corresponding to the driving factor data is added into the second vector fishing net to form a fine fishing net.

[0070] Optionally, the initial generalized additive model includes a land class generalized additive model corresponding to each land class type, and the driving factor data includes a plurality of driving factors; The downsizing module 503 is specifically configured to: The coarse fishing net and the fine fishing net are input into a preset initial generalized additive model, and a land class type corresponding to the fine fishing net is predicted based on the driving factor in the fine fishing net by using the land class generalized additive model to obtain an original area prediction result, and the original area prediction result includes an original land class area value and an original standard error; Based on the plurality of original area prediction results, an initial land class area prediction result for the preset area is formed; The initial land class area prediction result is corrected based on the coarse fishing net and the fine fishing net to obtain an intermediate land class area prediction result for the preset area.

[0071] Optionally, the initial generalized additive model includes a land class generalized additive model corresponding to each land class type, and the intermediate land class area prediction result includes an intermediate land class area value and an intermediate standard error of each land class generalized additive model; The model optimization module 504 is specifically configured to: The constraint value is obtained by calculating, based on the plurality of intermediate land class area values, the plurality of intermediate standard errors and the plurality of true land class area values, by using a preset constraint optimization function; The average difference is calculated based on the difference between each intermediate land class area value and the corresponding true land class area value; The network parameters in the initial generalized additive model are adjusted based on the constraint value and the average difference to obtain a target generalized additive model.

[0072] Optionally, the model optimization module 504 is specifically configured to: When the constraint value is greater than a preset value, and / or, the average difference is greater than a preset difference, the network parameters in the initial generalized additive model are adjusted to obtain an intermediate generalized additive model; The coarse mesh and the fine mesh are input into the intermediate generalized additive model for downscaling processing to obtain a transition land class area prediction result for a preset area. When the constraint value corresponding to the transition land class area prediction result is less than or equal to a preset value, and the average difference corresponding to the constraint value is less than or equal to a preset difference, the corresponding intermediate generalized additive model is determined as a target generalized additive model.

[0073] It should be noted that the land use data downscaling system based on the generalized additive model provided in the above embodiments and the land use data downscaling method based on the generalized additive model provided in the above embodiments belong to the same concept, and the specific manner in which each module and unit performs operations has been described in detail in the method embodiments, which will not be repeated here. The land use data downscaling system based on the generalized additive model provided in the above embodiments can allocate the above functions to different functional modules according to the actual application, that is, divide the internal structure of the system into different functional modules to complete all or part of the functions described above, and this is not limited herein.

[0074] The computing device provided in the embodiments of the present application includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements some or all steps of the land use data downscaling method based on the generalized additive model.

[0075] The computing device can be a computer, and the program is computer software. The parameters and steps in the computing device are described above in the embodiments of the land use data downscaling method based on the generalized additive model, and will not be repeated here.

[0076] The computer readable storage medium in the embodiments of the present application stores instructions. When the instructions are executed, the steps of the land use data downscaling method based on the generalized additive model are executed.

[0077] The computer readable storage medium can be a transitory computer readable storage medium or a non-transitory computer readable storage medium.

[0078] The technical solutions of the embodiments of the present disclosure can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes one or more instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method of the embodiments of the present disclosure. The aforementioned computer readable storage medium can be a non-transitory computer readable storage medium, including: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes, or can be a transitory computer readable storage medium.

[0079] The flowcharts and block diagrams in the drawings illustrate the possible implementation architectures, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In the flowcharts or block diagrams, each block can represent a module, a program segment or a part of code, and the module, the program segment or the part of code include one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different orders from those noted in the drawings. For example, two blocks indicated in succession can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams or flowcharts, and the combination of blocks in the block diagrams or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0080] Those skilled in the art know that the present application can be implemented as a system, a method or a computer program product. Therefore, the present disclosure can be embodied in the form of a complete hardware, a complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, which is generally referred to as "module" or "system" herein. In addition, in some embodiments, the present application can also be implemented in the form of a computer program product in one or more computer readable media, which includes computer readable program codes. The computer readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device or component, or any combination of the above.

[0081] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are contained in at least one embodiment or example of the present application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0082] Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and are not to be construed as limiting the present application, and the person skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A land use data downscaling method based on generalized additive model, characterized in that, The method comprises the following steps: obtaining an LUH2 data set and a downscaling requirement, the LUH2 data set comprising data of multiple land class types; performing mesh construction on a preset region based on the LUH2 data set, the downscaling requirement and preset driving factor data to obtain a coarse mesh and a fine mesh; inputting the coarse mesh and the fine mesh into a preset initial generalized additive model for downscaling processing to obtain an intermediate land class area prediction result for the preset region; optimizing the initial generalized additive model based on the intermediate land class area prediction result to obtain a target generalized additive model; outputting a target land class area prediction result for the preset region based on the target generalized additive model.

2. The method of claim 1, wherein, The mesh construction on the preset region based on the LUH2 data set, the downscaling requirement and preset driving factor data to obtain the coarse mesh and the fine mesh comprises: performing mesh construction on the preset region based on the LUH2 data set and preset driving factor data to obtain a coarse mesh; performing mesh construction on the preset region based on the downscaling requirement and the driving factor data to obtain a fine mesh.

3. The method of claim 2, wherein, The mesh construction on the preset region based on the LUH2 data set and preset driving factor data to obtain the coarse mesh comprises: cropping the LUH2 data set based on the preset region to obtain a low-resolution data set for the preset region; constructing a first vector mesh matching the size of the low-resolution data set; correspondingly adding the low-resolution data set and preset driving factor data to the first vector mesh to form a coarse mesh.

4. The method of claim 2, wherein, The downscaling requirement is a requirement to downscale the LUH2 data set to a target resolution; The mesh construction on the preset region based on the downscaling requirement and the driving factor data to obtain the fine mesh comprises: constructing a second vector mesh corresponding to the preset region, the resolution of the vector mesh being the same as the target resolution; correspondingly adding the driving factor data to the second vector mesh to form a fine mesh.

5. The method according to any one of claims 1 to 4, characterized in that, The initial generalized additive model comprises a land class generalized additive model corresponding to each land class type, and the driving factor data comprises multiple driving factors; The inputting of the coarse mesh and the fine mesh into the preset initial generalized additive model for downscaling processing to obtain the intermediate land class area prediction result for the preset region comprises: inputting the coarse mesh and the fine mesh into the preset initial generalized additive model, and using the land class generalized additive model to predict based on the driving factors in the fine mesh to obtain an original area prediction result of a corresponding land class type in the fine mesh, the original area prediction result comprising an original land class area value and an original standard error; based on multiple original area prediction results, forming an initial land class area prediction result for the preset region; based on the coarse mesh and the fine mesh, correcting the initial land class area prediction result to obtain the intermediate land class area prediction result for the preset region.

6. The method according to any one of claims 1 to 4, characterized in that, The initial generalized additive model comprises a land class generalized additive model corresponding to each land class type, and the intermediate land class area prediction result comprises an intermediate land class area value of each land class generalized additive model and an intermediate standard error; The optimization of the initial generalized additive model based on the intermediate land class area prediction result comprises: A constraint value is obtained by using a preset constraint optimization function and based on the plurality of intermediate land class area values, the plurality of intermediate standard errors and a plurality of corresponding true land class area values; An average difference is calculated based on a difference between each intermediate land class area value and a corresponding true land class area value; The network parameters in the initial generalized additive model are adjusted based on the constraint value and the average difference, to obtain a target generalized additive model.

7. The method of claim 6, wherein, The optimization of the initial generalized additive model based on the intermediate land class area prediction result comprises: When the constraint value is greater than a preset value and / or the average difference is greater than a preset difference, the network parameters in the initial generalized additive model are adjusted to obtain an intermediate generalized additive model; The coarse fishing net and the fine fishing net are input into the intermediate generalized additive model for downscaling processing, to obtain a transition land class area prediction result for the preset area; When the constraint value corresponding to the transition land class area prediction result is less than or equal to the preset value and the average difference corresponding to the transition land class area prediction result is less than or equal to the preset difference, the corresponding intermediate generalized additive model is determined as a target generalized additive model. 8.A land use data downscaling system based on a generalized additive model, characterized in that, The method comprises: An acquisition module is configured to acquire an LUH2 dataset and a downscaling requirement, wherein the LUH2 dataset comprises data of a plurality of land class types; A fishing net construction module is configured to construct a fishing net for a preset area based on the LUH2 dataset, the downscaling requirement and preset driving factor data, to obtain a coarse fishing net and a fine fishing net; A downscaling module is configured to input the coarse fishing net and the fine fishing net into a preset initial generalized additive model for downscaling processing, to obtain an intermediate land class area prediction result for the preset area; A model optimization module is configured to optimize the initial generalized additive model based on the intermediate land class area prediction result, to obtain a target generalized additive model; A result output module is configured to output a target land class area prediction result for the preset area based on the target generalized additive model.

9. A computing device comprising a memory, a processor, and a program stored on the memory and running on the processor, wherein, The processor executes the program to implement the steps of the land use data downscaling method based on a generalized additive model according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores instructions, and when the instructions run on the terminal device, the terminal device executes the steps of the land use data downscaling method based on a generalized additive model according to any one of claims 1 to 7.

Images