Vegetation prediction model construction method, vegetation prediction method, device and equipment

Abstract

The invention provides a vegetation prediction model construction method, a vegetation prediction method, a vegetation prediction device and vegetation prediction equipment, and relates to the technical field of data processing. The vegetation prediction model construction method comprises the following steps: taking ecological environment data of a target area as an independent variable, taking vegetation spatial distribution data corresponding to the target area as a dependent variable, and training a first random forest model to obtain a vegetation spatial distribution prediction model; taking the output of the vegetation parameter spatial distribution prediction model and the period-by-period meteorological data as independent variables, taking the period-by-period vegetation feature change data as dependent variables, training a second random forest model, and obtaining a vegetation feature time change prediction model; and splicing the vegetation spatial distribution prediction model and the vegetation feature time change prediction model to obtain a vegetation prediction model. According to the technical scheme, the data utilization efficiency can be improved, the model training time is shortened, and meanwhile the adaptability of the vegetation prediction model and the accuracy of vegetation prediction are improved.

Description

Background Art

[0002] As an important carrier of carbon absorption, accurate prediction of the temporal and spatial distribution of vegetation coverage has become a key link in carbon assessment. At present, most commonly used vegetation coverage assessment methods are based on remote sensing data, such as the normalized vegetation index, leaf area index and vegetation coverage obtained by inversion of remote sensing images. However, in historical periods and future climate scenarios, especially in areas lacking remote sensing observation data or in obscured areas (such as photovoltaic fields and agricultural greenhouses), remote sensing methods cannot be directly used for vegetation assessment, which poses a challenge to vegetation prediction.

[0003] Traditional vegetation prediction models, including vegetation models based on statistical relationships, process-based static vegetation models, and global dynamic vegetation models, often have problems such as complex models, numerous parameters, and long parameter adjustment time. In practical applications, these models need to rely on a large amount of observation data for parameter correction. If the model is directly applied without parameter adjustment, its accuracy is usually low, which limits its universality and ease of use in engineering. In addition, the architecture of traditional models varies greatly, and some models have high requirements for operating environment and data conditions, which increases the difficulty of use. In addition, the accuracy of vegetation prediction of the above models is also low. Therefore, there is still room for improvement in the adaptability and accuracy of vegetation prediction models.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute relevant technology known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of the embodiments of the present disclosure is to provide a method for constructing a vegetation prediction model, a vegetation prediction method, a device for constructing a vegetation prediction model, a vegetation prediction device, an electronic device and a computer-readable storage medium, thereby improving data utilization efficiency and reducing model training time to at least a certain extent, while improving the adaptability of the vegetation prediction model and the accuracy of vegetation prediction.

[0006] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0007] According to a first aspect of an embodiment of the present disclosure, a method for constructing a vegetation prediction model is provided, the method comprising: taking ecological environmental data of a target area as an independent variable, and vegetation spatial distribution data corresponding to the target area as a dependent variable, training a first random forest model to obtain a vegetation spatial distribution prediction model; taking the output of the vegetation parameter spatial distribution prediction model and time period meteorological data as independent variables, and time period vegetation feature change data as a dependent variable, training a second random forest model to obtain a vegetation feature time change prediction model; splicing the vegetation spatial distribution prediction model with the vegetation feature time change prediction model to obtain a vegetation prediction model.

[0008] In some example embodiments of the present disclosure, based on the aforementioned scheme, before taking the ecological environment data of the target area as the independent variable and the vegetation spatial distribution data corresponding to the target area as the dependent variable, it also includes: obtaining the ecological environment data composed of annual cumulative precipitation data, annual accumulated temperature data, annual average direct radiation data, annual average scattered radiation data, altitude data, annual average soil moisture data, soil texture data, soil organic matter data, soil cation exchange capacity data and land use type data of the target area; taking the vegetation coverage data corresponding to the target area as the vegetation spatial distribution data; and unifying the temporal and spatial resolutions of the ecological environment data and the vegetation spatial distribution data.

[0009] In some example embodiments of the present disclosure, based on the aforementioned scheme, the output of the vegetation parameter spatial distribution prediction model and the meteorological data from time period are used as independent variables, and the vegetation characteristic change data from time period is used as the dependent variable, including: taking the output of the vegetation parameter spatial distribution prediction model and the cumulative precipitation data from time period, the soil accumulated temperature data from time period, the average surface temperature data from time period, the average daily scattered radiation data from time period, and the average soil moisture data from time period as independent variables, and the vegetation coverage change data from time period as the dependent variable.

[0010] In some example embodiments of the present disclosure, based on the aforementioned scheme, the training of the first random forest model includes: constructing first sample data based on the ecological environment data and the vegetation spatial distribution data; in the node splitting process of each decision tree in the first random forest model, randomly sampling the first sample data to obtain a feature subset; based on the feature subset, determining the splitting threshold of each node through the Gini coefficient; at each node, recursively constructing a decision tree through multiple splits until a preset termination condition is reached; and weighted averaging the prediction results of each decision tree for the input sample to obtain the prediction result of the vegetation spatial distribution prediction model.

[0011] In some example embodiments of the present disclosure, based on the above solution, the randomly sampling the first sample data to obtain a feature subset includes: according to: The first sample data is randomly sampled to obtain a feature subset, where: represents the feature subset matrix, Indicates i The node sample data matrix of a decision tree, represents the number of features, represents a random sampling process; The step of determining the splitting threshold of each node based on the feature subset by using the Gini coefficient includes: according to: Determine the splitting threshold of each of the nodes, where represents the target splitting threshold, represents the candidate split threshold, and Respectively represent the split threshold The two sample subsets after segmentation, Represents the Gini coefficient calculation function; The Gini coefficient calculation function is based on: To determine, Indicates the sample category ratio of the node.

[0012] In some example embodiments of the present disclosure, based on the above scheme, weighted averaging the prediction results of each decision tree on the input sample to obtain the prediction result of the vegetation spatial distribution prediction model includes: according to: The prediction result of the vegetation spatial distribution prediction model is obtained, wherein: represents the prediction result of the vegetation spatial distribution prediction model, represents the number of decision trees in the first random forest model, Indicates i The prediction output of a decision tree, represents a column vector of all ones, A column vector representing the predictions of all decision trees.

[0013] In some example embodiments of the present disclosure, based on the above scheme, the training of the second random forest model includes: according to: The second random forest model is trained, wherein: represents the predicted value of the vegetation coverage change data for each period, represents the number of decision trees in the second random forest model, Indicates j The prediction function of a decision tree, represents the output of the vegetation parameter spatial distribution prediction model, represents the cumulative precipitation data for each period, represents the soil accumulated temperature data for each period, represents the average surface temperature data for each period, represents the average daily diffuse radiation data for each period, It represents the average soil moisture data for each time period.

[0014] According to the second aspect of the embodiments of the present disclosure, a vegetation prediction method is provided, which includes: obtaining ecological environmental data and time period meteorological data of the area to be predicted; inputting the ecological environmental data and time period meteorological data into a vegetation prediction model to obtain time period vegetation characteristic change data of the area to be predicted; wherein the vegetation prediction model is obtained by the vegetation prediction model construction method described in any one of the above embodiments.

[0015] According to a third aspect of an embodiment of the present disclosure, a device for constructing a vegetation prediction model is provided, the device comprising: a spatial distribution prediction module, used to take the ecological environment data of a target area as an independent variable, and the vegetation spatial distribution data corresponding to the target area as a dependent variable, to train a first random forest model to obtain a vegetation spatial distribution prediction model; a time change prediction module, used to take the output of the vegetation parameter spatial distribution prediction model and the meteorological data at each time period as independent variables, and the vegetation feature change data at each time period as a dependent variable, to train a second random forest model to obtain a vegetation feature time change prediction model; a prediction model construction module, used to splice the vegetation spatial distribution prediction model with the vegetation feature time change prediction model to obtain the vegetation prediction model.

[0016] According to the fourth aspect of the embodiments of the present disclosure, a vegetation prediction device is provided, which includes: a data acquisition module for acquiring ecological environmental data and time period meteorological data of an area to be predicted; a vegetation prediction module for inputting the ecological environmental data and time period meteorological data into a vegetation prediction model to obtain time period vegetation characteristic change data of the area to be predicted; wherein the vegetation prediction model is obtained by the vegetation prediction model construction method described in any one of the above embodiments.

[0017] According to the fifth aspect of an embodiment of the present disclosure, there is provided an electronic device, comprising: a processor; and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the above-mentioned vegetation prediction model construction method or vegetation prediction method is implemented.

[0018] According to a sixth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for constructing a vegetation prediction model or the vegetation prediction method described above is implemented.

[0019] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects: The construction method of the vegetation prediction model in the exemplary embodiment of the present disclosure, on the one hand, through the first layer of random forest model, the ecological environment data of the target area is associated with the vegetation spatial distribution data, and a high-precision vegetation spatial distribution prediction model is trained. The model can capture the key ecological factors that affect the distribution of vegetation, and in the absence of remote sensing observation data, it can still provide accurate spatial vegetation coverage prediction, thereby improving the adaptability of the model; on the other hand, in the second layer model, the spatial distribution prediction results of the first layer model are combined with the meteorological data of each period, and the temporal changes of vegetation coverage are dynamically predicted by the ecological data of each period. This process enhances the dynamic response ability of the model, so that it can accurately reflect the changes in vegetation characteristics in different time periods; on the other hand, a two-layer random forest model structure is adopted, which is divided into two parts: spatial prediction and temporal change prediction. By using the output of the first layer model as the input of the second layer model, the complex prediction process is avoided, the data utilization efficiency is improved, the model training time is reduced, and the structural stability of the model is also improved. The hierarchical design effectively reduces the redundancy between data inputs and enhances the adaptability of the vegetation prediction model under different environmental conditions. In addition, the vegetation spatial distribution prediction model and the vegetation characteristic temporal change prediction model are combined to obtain a complete vegetation prediction model, which realizes a comprehensive prediction of the spatial and temporal distribution of vegetation coverage in the target area. This model can not only provide the vegetation status under the current ecological conditions, but also infer the future vegetation change trend based on meteorological data.

[0020] Therefore, the vegetation prediction model in the present invention significantly improves the accuracy of vegetation prediction, especially in areas with complex ecological environment or lack of continuous remote sensing observation. It can still predict the spatial and temporal changes of vegetation coverage with high accuracy, providing scientific support for ecological environment assessment and management.

[0021] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0023] Figure 1 A flowchart of a method for constructing a vegetation prediction model according to some embodiments of the present disclosure is schematically shown.

[0024] Figure 2 A schematic diagram schematically illustrates the annual average vegetation coverage of a target area observed by remote sensing according to some embodiments of the present disclosure.

[0025] Figure 3 A schematic diagram schematically shows the annual average vegetation coverage of a target area predicted by a vegetation spatial distribution prediction model according to some embodiments of the present disclosure.

[0026] Figure 4 A schematic diagram schematically illustrates the difference between the annual average vegetation coverage of a target area predicted by a remote sensing observation and a vegetation spatial distribution prediction model according to some embodiments of the present disclosure.

[0027] Figure 5 A scatter plot of remote sensing observation values ​​and simulation values ​​of a vegetation feature temporal variation prediction model according to some embodiments of the present disclosure is schematically shown.

[0028] Figure 6 A diagram schematically shows the average annual change in vegetation coverage according to some embodiments of the present disclosure.

[0029] Figure 7 A flowchart of a vegetation prediction method according to some embodiments of the present disclosure is schematically shown.

[0030] Figure 8 A block diagram of a device for constructing a vegetation prediction model according to some embodiments of the present disclosure is schematically shown.

[0031] Fig. 9 A block diagram of a vegetation prediction device according to some embodiments of the present disclosure is schematically shown.

[0032] Fig.10 A schematic diagram of the structure of a computer system of an electronic device according to some embodiments of the present disclosure is schematically shown.

[0033] Fig.11 A schematic diagram of a computer-readable storage medium according to some embodiments of the present disclosure is schematically shown.

[0034] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0035] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this specification. Instead, they are merely examples of devices and methods consistent with some aspects of this specification as detailed in the appended claims.

[0036] The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit this specification. The singular forms "a", "the" and "the" used in this specification and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0037] It should be understood that although the terms first, second, third, etc. may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this specification, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0038] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be more comprehensive and complete and will fully convey the concept of the example embodiments to those skilled in the art.

[0039] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the present disclosure.

[0040] Furthermore, the drawings are only schematic illustrations and are not necessarily drawn to scale. The block diagrams shown in the drawings are only functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0041] In an exemplary embodiment of the present disclosure, a method for constructing a vegetation prediction model is first proposed. Figure 1 The following is a schematic diagram of a process flow of a method for constructing a vegetation prediction model according to some embodiments of the present disclosure. Figure 1 As shown, the method for constructing the vegetation prediction model may include the following steps: Step S110, taking the ecological environment data of the target area as an independent variable and the vegetation spatial distribution data corresponding to the target area as a dependent variable, training the first random forest model to obtain a vegetation spatial distribution prediction model; Step S120, using the output of the vegetation parameter spatial distribution prediction model and the meteorological data at each time period as independent variables, and the vegetation characteristic change data at each time period as the dependent variable, training the second random forest model to obtain a vegetation characteristic time change prediction model; Step S130, combining the vegetation spatial distribution prediction model with the vegetation feature temporal change prediction model to obtain a vegetation prediction model.

[0042] Specifically, first, the ecological environment data of the target area, such as precipitation, temperature, soil moisture, etc., are used as independent variables, and the vegetation spatial distribution data of the area, such as vegetation coverage, leaf area index and normalized vegetation index, etc., are used as dependent variables, and input into the first layer of random forest model for training. After the training is completed, a model for predicting the spatial distribution characteristics of vegetation is obtained, that is, the vegetation spatial distribution prediction model.

[0043] Then, the output results of the first-layer model, the predicted values ​​of vegetation spatial distribution data and meteorological data at each time period, such as precipitation and surface temperature at each time period, are used as independent variables, and the vegetation characteristic change data at each time period is used as the dependent variable. They are input into the second-layer random forest model for training to obtain a vegetation characteristic temporal change prediction model. Finally, the vegetation spatial distribution prediction model and the vegetation feature temporal change prediction model are spliced ​​together, and the output results of the two models are integrated to obtain a complete vegetation prediction model. This model can be used to predict the spatial distribution of vegetation in the target area and its changes over time, and to achieve accurate prediction of vegetation data.

[0044] The following further describes the method for constructing the vegetation prediction model in the above exemplary embodiment.

[0045] In step S110, the ecological environment data of the target area is used as an independent variable, and the vegetation spatial distribution data corresponding to the target area is used as a dependent variable, and the first random forest model is trained to obtain a vegetation spatial distribution prediction model.

[0046] Among them, ecological and environmental data can represent external environmental variables that affect vegetation growth and distribution, usually including meteorological, geographical and soil information data of the target area. Vegetation spatial distribution data can represent variables that describe the vegetation coverage characteristics in the target area, such as fractional vegetation cover (FVC), leaf area index (LAI) and normalized difference vegetation index (NDVI), etc. These data can be inverted from remote sensing images or observation data. Vegetation spatial distribution data can describe the coverage of vegetation at different locations in the target area, reflecting the differences in vegetation in the spatial dimension. The random forest model is an integrated learning method that makes predictions based on a combination of multiple decision trees. It generates multiple sub-sample sets through random sampling and builds a decision tree model on each sub-sample set, thereby reducing the bias and variance of a single model.

[0047] In the actual training process, first, the ecological environment data and vegetation spatial distribution data of the target area are collected. In the process of training set construction, the data set is divided into training set and test set, of which the training set accounts for 80% of the total data and is used for model training; the test set accounts for 20% and is used to evaluate the generalization performance of the model.

[0048] Then, the parameters that are likely to have a significant impact on the prediction accuracy in the first random forest model are set, including tree depth, number of decision trees, minimum number of leaf node samples, and minimum number of split samples. Exemplarily, the tree depth is set to 11 to ensure that each decision tree can be fully split to capture the complex relationship between ecological environment data and vegetation distribution. The number of decision trees is set to 60 to balance the model training speed while ensuring the stability of the prediction. Although increasing the number of decision trees helps to reduce the model variance, too many decision trees will increase the calculation time, so the number of trees is moderately reduced to improve the calculation efficiency. The minimum number of leaf node samples is set to 2 to prevent overfitting, while ensuring that the leaf node contains at least two samples to improve the robustness of the model. At the same time, the minimum number of split samples is set to 3 to ensure that there are at least three samples each time the node splits to simplify the tree structure and prevent overfitting.

[0049] Next, the first random forest model is trained on the assigned training set. The model randomly samples from the training data and generates a decision tree based on each subsample set. Each decision tree is built on a subsample by randomly selecting features and node splitting strategies to reduce the model's dependence on specific data patterns and improve generalization performance. During the training process, the tree depth, number of decision trees, and splitting parameters are gradually adjusted to optimize the accuracy and efficiency of the model. After training, the model performance is evaluated on the test set, mainly using the determination coefficient ( ) value, mean square error (MSE) and root mean square error (RMSE) were used to evaluate the prediction accuracy of the model for vegetation spatial distribution. The low MSE and RMSE values ​​indicate that the model can accurately predict the vegetation distribution in the target area. The first random forest model finally obtained is the vegetation spatial distribution prediction model, which can be used to predict vegetation coverage characteristics in the target area based on ecological and environmental data, and is suitable for areas that lack direct observation data in practice. Figure 3 Schematic diagram showing the annual average vegetation coverage of the target area predicted by the vegetation spatial distribution prediction model, and Figure 2 Compared with the annual average vegetation coverage of the target area observed by remote sensing, it can be seen that the vegetation spatial distribution prediction model has a higher accuracy. Figure 4 A schematic diagram schematically illustrates the difference between the annual average vegetation coverage of a target area predicted by a remote sensing observation and a vegetation spatial distribution prediction model according to some embodiments of the present disclosure.

[0050] In some embodiments, before taking the ecological environment data of the target area as the independent variable and the vegetation spatial distribution data corresponding to the target area as the dependent variable, it also includes: obtaining the ecological environment data composed of annual cumulative precipitation data, annual accumulated temperature data, annual average direct radiation data, annual average scattered radiation data, altitude data, annual average soil moisture data, soil texture data, soil organic matter data, soil cation exchange capacity data and land use type data of the target area; taking the vegetation coverage data corresponding to the target area as the vegetation spatial distribution data; and unifying the temporal and spatial resolutions of the ecological environment data and the vegetation spatial distribution data.

[0051] Specifically, first, the ecological environment data of the target area are collected through a variety of data sources. These data provide the model with external environmental information closely related to vegetation growth and distribution. The ecological environment data may include annual cumulative precipitation data, annual accumulated temperature data, annual average direct radiation data, annual average diffuse radiation data, altitude data, annual average soil moisture data, soil texture data, soil organic matter data, soil cation exchange capacity data, and land use type data. Among them, the annual cumulative precipitation data can represent the total precipitation in the target area in one year, which is used to measure the regional water supply and has a direct impact on the growth and distribution of plants. The annual accumulated temperature data can represent the cumulative amount of the average daily temperature in the target area in one year (only the part above 0°C is accumulated), which is a measure of the heat required for plant growth. The annual average direct radiation data can represent the amount of solar radiation that directly reaches the surface in one year. The annual average diffuse radiation data can represent the amount of solar radiation that reaches the surface through atmospheric scattering in one year. The altitude data can represent the vertical height of the surface above sea level in the target area. The annual average soil moisture data can represent the average moisture content in the soil in one year. Soil texture data can indicate the composition structure of soil particles, usually including the ratio of sand, silt and clay. Soil organic matter data can indicate the content of organic matter in the soil, reflecting the soil fertility and nutrient supply capacity. Soil cation exchange capacity data can indicate the soil's adsorption capacity for exchangeable cations, usually based on the average cation exchange capacity at different depths.

[0052] Then, the vegetation coverage data of the target area is obtained and used as the dependent variable. Vegetation coverage is a quantitative indicator of the vegetation distribution characteristics in the target area. It can be expressed in the form of annual average or annual maximum value to reflect the spatial coverage of vegetation and provide key output reference data for the model.

[0053] Next, all ecological environment data and vegetation coverage data are unified to the same spatial resolution. In this example, a horizontal grid resolution of 0.1 degrees is used, which is the lowest resolution standard among all types of data sets. At this resolution, the study area is divided into 235×520 grid points, and the resolution of each grid point is consistent with the location of the model input, ensuring the precise matching of all data in spatial distribution. At the same time, soil texture, organic matter and cation exchange capacity data are usually divided into observations of different depths, namely the shallow surface layer (10 cm), the middle layer (30 cm) and the deep layer (60 cm). Therefore, for the uniformity of the model input variables, the average values ​​of these three layers of data need to be calculated; for soil moisture, the 30cm depth data that contributes the most to the modeling is selected. Other independent variable data use surface observations to maintain the consistency of the modeling data in the vertical direction.

[0054] Secondly, in the time dimension, all ecological environment data and vegetation coverage data also need to be processed consistently. The time span of the training data set is 12 years from 2001 to 2012. Therefore, all independent and dependent variable data are processed in the form of annual average values ​​or annual cumulative values ​​to ensure the uniformity of temporal resolution. During this 12-year period, a yearly data input sequence is generated, and the ecological environment data and vegetation coverage characteristics updated each year are provided to the model to construct input features with temporal consistency, so that the model can accurately reflect the spatial distribution change characteristics of different years.

[0055] In some embodiments, the training of the first random forest model includes: constructing first sample data based on the ecological environment data and the vegetation spatial distribution data; in the node splitting process of each decision tree in the first random forest model, randomly sampling the first sample data to obtain a feature subset; based on the feature subset, determining the splitting threshold of each node through the Gini coefficient; at each node, recursively constructing a decision tree through multiple splits until a preset termination condition is reached; and weighted averaging the prediction results of each decision tree for the input sample to obtain the prediction result of the vegetation spatial distribution prediction model.

[0056] Specifically, first, a first sample data set is constructed based on the ecological environment data of the target area and the corresponding vegetation spatial distribution data. The ecological environment data is used as the input feature of the model, and the vegetation spatial distribution data is used as the output label to provide a training basis for the random forest model. Through this sample data set, the model can learn the correlation characteristics between ecological environment factors and vegetation spatial distribution.

[0057] Then, during the training of the first random forest model, random features are sampled from the first sample dataset when each decision tree node splits. By randomly selecting a subset of features, the model generates different feature combinations when splitting each node, thereby increasing the diversity and stability of the model. This process is designed to prevent overfitting of the model and improve the generalization ability of the model.

[0058] Next, the Gini coefficient is used as a measure of node purity, and the optimal node splitting threshold is determined by calculating the Gini coefficients of different split points. On the feature subset of the current node, the lower the Gini coefficient, the higher the purity of the data after segmentation. Therefore, the split point with the lowest Gini coefficient is selected as the splitting threshold of the node, thereby effectively improving the prediction effect of the model after segmentation.

[0059] In addition, at each node, node splitting is performed recursively to construct child nodes of the decision tree in sequence until a preset termination condition is met. The termination condition may include reaching a maximum tree depth or the number of node samples being lower than a set threshold. Through the recursive splitting process, a complete decision tree structure is gradually generated, so that the decision tree can capture the relationship between different ecological environment characteristics and vegetation distribution characteristics.

[0060] Finally, when all decision trees are constructed, the first random forest model performs a weighted average of the prediction results of each decision tree for the input sample to form the final vegetation spatial distribution prediction value. By integrating the output results of multiple decision trees, the model can combine the predictions of different trees to provide more stable and accurate vegetation spatial distribution prediction results.

[0061] In some embodiments, the randomly sampling the first sample data to obtain a feature subset includes: according to: The first sample data is randomly sampled to obtain a feature subset, where: represents the feature subset matrix, Indicates i The node sample data matrix of a decision tree, represents the number of features, represents a random sampling process; In some embodiments, the determining the splitting threshold of each node based on the feature subset by using the Gini coefficient includes: according to: Determine the splitting threshold of each of the nodes, where represents the target splitting threshold, represents the candidate split threshold, and Respectively represent the split threshold The two sample subsets after segmentation, Represents the Gini coefficient calculation function; The Gini coefficient calculation function is based on: To determine, Indicates the sample category ratio of the node.

[0062] In some embodiments, performing weighted averaging on the prediction results of each decision tree for the input sample to obtain the prediction results of the vegetation spatial distribution prediction model includes: according to: The prediction result of the vegetation spatial distribution prediction model is obtained, wherein: represents the prediction result of the vegetation spatial distribution prediction model, represents the number of decision trees in the first random forest model, Indicates i The prediction output of a decision tree, represents a column vector of all ones, A column vector representing the predictions of all decision trees.

[0063] In step S120, the output of the vegetation parameter spatial distribution prediction model and the meteorological data at each time period are used as independent variables, and the vegetation characteristic change data at each time period is used as the dependent variable, and the second random forest model is trained to obtain a vegetation characteristic temporal change prediction model.

[0064] Among them, the time-period meteorological data refers to the ecological and meteorological characteristic data recorded at specific time intervals (such as daily, 8-day or monthly) in the target area. The vegetation characteristic temporal change prediction model is constructed through the random forest model, which is a model used to predict the temporal changes of vegetation characteristics (such as vegetation coverage, leaf area index, etc.) in the target area. The vegetation characteristic change data by time period can represent variables that describe the vegetation coverage characteristics in the target area at specific time intervals, such as vegetation coverage change data by time period, leaf area index change data by time period, and normalized vegetation index change data by time period. The model uses the time-period meteorological data (such as precipitation, temperature, diffuse radiation, etc. by time period) and the output of the previously generated vegetation spatial distribution prediction model to dynamically reflect the changing trend of vegetation in different time periods.

[0065] In the actual training process, first, the output results, i.e. the predicted vegetation parameters (such as the annual average vegetation coverage), are extracted from the first-layer vegetation spatial distribution prediction model. These prediction results will be used as part of the independent variables of the second-layer model, providing the spatial background information of vegetation distribution in the target area. Since the second-layer model is mainly based on the time period (8-day) resolution, the annual average value of the first-layer model can be assigned to each 8-day time period to ensure that the spatial distribution information is consistent with the time resolution of the meteorological data by time period. The independent variables of the second-layer model include the output of the first-layer model and the meteorological data by time period. The dependent variable is the vegetation characteristic change data by time period (such as the vegetation coverage change by 8 days). To ensure sufficient data volume and temporal continuity, this example uses 12 years of data from 2001 to 2012, with 46 8-day time periods each year, and a total of 132×12×46 data samples.

[0066] Then, in order to improve the efficiency and quality of model training, based on the 132 observation points, the selection criteria included quality control of meteorological observation data, and the elimination of sites with a large number of missing data and sites with zero vegetation coverage to ensure the accuracy and effectiveness of the training data.

[0067] Next, since the vegetation parameters predicted by the first-layer model are annual averages, and the independent variables of the second-layer model are mainly based on 8-day resolution, in order to ensure the consistency of temporal resolution, the same first-layer model prediction values ​​are assigned to the 46 8-day time periods each year. This makes the temporal resolution match uniform, and at the same time, the model contains the spatial background information of the annual average vegetation cover characteristics on the 8-day time scale.

[0068] Finally, the sorted independent variables and dependent variables are input into the second-layer random forest model for model training to learn the dynamic relationship of vegetation characteristics in time period. The second-layer model uses a longer 12-year time series data in terms of time length, and uses the random forest algorithm to capture the impact of meteorological data and annual average vegetation parameters on vegetation characteristics changes in time period, thereby constructing a vegetation characteristics time change prediction model with time dynamic prediction capabilities.

[0069] In some embodiments, the output of the vegetation parameter spatial distribution prediction model and the meteorological data for each time period are used as independent variables, and the vegetation characteristic change data for each time period are used as dependent variables, including: taking the output of the vegetation parameter spatial distribution prediction model and the cumulative precipitation data for each time period, the soil accumulated temperature data for each time period, the average surface temperature data for each time period, the average daily scattered radiation data for each time period, and the average soil moisture data for each time period as independent variables, and the vegetation coverage change data for each time period as dependent variables.

[0070] Specifically, the meteorological data by time period may include cumulative precipitation data by time period, accumulated soil temperature data by time period, average surface temperature data by time period, average daily scattered radiation data by time period, and average soil moisture data by time period. The time period may be other suitable time periods such as 8 days, 10 days, and 15 days. Among them, the cumulative precipitation data by time period may represent the cumulative precipitation in a specific time period, the accumulated soil temperature data by time period may represent the cumulative value of the soil temperature in a specific time period, the average surface temperature data by time period may represent the average daily surface temperature value in a specific time period, the average daily scattered radiation data by time period may represent the average scattered solar radiation in a specific time period, and the average soil moisture data by time period may represent the average soil moisture value in a specific time period.

[0071] In some embodiments, the training of the second random forest model comprises: according to: The second random forest model is trained, wherein: represents the predicted value of the vegetation coverage change data for each period, represents the number of decision trees in the second random forest model, Indicates j The prediction function of a decision tree, represents the output of the vegetation parameter spatial distribution prediction model, represents the cumulative precipitation data for each period, represents the soil accumulated temperature data for each period, represents the average surface temperature data for each period, represents the average daily diffuse radiation data for each period, It represents the average soil moisture data for each period.

[0072] In the actual training process, the model parameters need to be optimized and memorized to prevent overfitting. For example, when training the second-layer random forest model, when the tree depth is set to 13 or greater, the model shows higher accuracy on the training set, but the accuracy on the test set is not significantly improved. Instead, the accuracy gap increases, indicating that the model has a tendency to overfit. Therefore, the tree depth should be set below 13 to prevent the model from overfitting the training data and improve the generalization performance of the model on the test data.

[0073] In step S130, the vegetation spatial distribution prediction model and the vegetation feature temporal change prediction model are spliced ​​to obtain a vegetation prediction model.

[0074] Specifically, the first-layer model, namely the vegetation spatial distribution prediction model, provides a prediction of the spatial distribution of vegetation in the target area based on the annual average ecological environment data; the second-layer model, namely the vegetation feature temporal change prediction model, provides a prediction of the dynamic change of vegetation features over time based on the meteorological data of each period. In the spliced ​​comprehensive model, the output of the second-layer vegetation feature temporal change prediction model is the output result of the final vegetation prediction model. This output includes the predicted values ​​of vegetation coverage in space and time, providing a comprehensive prediction of the spatiotemporal distribution of vegetation. Figure 5 The scatter plot schematically shows the remote sensing observation values ​​of vegetation coverage in a certain area and the simulated values ​​of the vegetation characteristic temporal change prediction model. It can be seen that the concentration is high and the prediction accuracy of the vegetation spatial distribution prediction model is good. Figure 6 The diagram schematically shows the annual average change in vegetation coverage of a certain area, where the vertical axis represents the average of vegetation coverage within four years to one year. It can be seen that the error between the predicted value of the vegetation prediction model and the actual observed value is small.

[0075] Further, in the embodiments of the present disclosure, reference is made to Figure 7 As shown, a vegetation prediction method is also provided, the method comprising the following steps: Step S710, obtaining ecological environment data and time-period meteorological data of the area to be predicted; Step S720, inputting the ecological environment data and the meteorological data for each period into a vegetation prediction model to obtain vegetation characteristic change data for each period of the area to be predicted; Wherein, the vegetation prediction model is obtained by the vegetation prediction model construction method as described in any one of the above embodiments.

[0076] It should be noted that, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0077] Next, in the embodiment of the present disclosure, a device for constructing a vegetation prediction model is also provided. Figure 8 As shown in , the construction device 800 of the vegetation prediction model can be composed of a spatial distribution prediction module 801, a temporal change prediction module 802 and a prediction model construction module 803. Among them: the spatial distribution prediction module can be used to train the first random forest model with the ecological environment data of the target area as the dependent variable, and the vegetation spatial distribution data corresponding to the target area as the dependent variable, to obtain the vegetation spatial distribution prediction model; the temporal change prediction module can be used to train the second random forest model with the output of the vegetation parameter spatial distribution prediction model and the meteorological data of each time period as the independent variables, and the vegetation characteristic change data of each time period as the dependent variable, to obtain the vegetation characteristic temporal change prediction model; the prediction model construction module can be used to splice the vegetation spatial distribution prediction model with the vegetation characteristic temporal change prediction model to obtain the vegetation prediction model.

[0078] In the embodiment of the present disclosure, a vegetation prediction device is also provided. Fig. 9 As shown in , the vegetation prediction device 900 may be composed of a data acquisition module 901 and a vegetation prediction module 902. The data acquisition module 901 may be used to acquire ecological environment data and meteorological data of the area to be predicted by time period; the vegetation prediction module 902 may be used to input the ecological environment data and meteorological data by time period into a vegetation prediction model to obtain vegetation characteristic change data of the area to be predicted by time period; the vegetation prediction model is obtained by the vegetation prediction model construction method as described in any one of the above embodiments.

[0079] It should be noted that the specific details of the construction device of the above-mentioned vegetation prediction model and the various parts of the vegetation prediction device have been described in detail in the implementation method of the vegetation prediction model and the vegetation prediction method part. The undisclosed details can be found in the implementation method content of the method part, and will not be repeated here.

[0080] In addition, in an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above-mentioned vegetation prediction model construction method or vegetation prediction method is also provided.

[0081] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods or program products. Therefore, various aspects of the present disclosure may be specifically implemented in the following forms, namely: complete hardware embodiments, complete software embodiments (including firmware, microcode, etc.), or embodiments combining hardware and software aspects, which may be collectively referred to herein as "circuits", "modules" or "systems".

[0082] Refer to the following Fig.10 hereinafter describes the electronic device 1000 according to such an embodiment of the present disclosure. Fig.10 The electronic device 1000 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0083] like Fig.10 As shown, the electronic device 1000 is in the form of a general computing device. The components of the electronic device 1000 may include but are not limited to: the at least one processing unit 1010, the at least one storage unit 1020, a bus 1030 connecting different system components (including the storage unit 1020 and the processing unit 1010), and a display unit 1040.

[0084] The storage unit stores program codes, which can be executed by the processing unit 1010, so that the processing unit 1010 executes the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification.

[0085] The storage unit 1020 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 1021 and / or a cache memory unit 1022 , and may further include a read-only memory unit (ROM) 1023 .

[0086] The storage unit 1020 may also include a program / utility 1024 having a set (at least one) of program modules 1025, such program modules 1025 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0087] Bus 1030 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0088] The electronic device 1000 may also communicate with one or more external devices 1070 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 1000, and / or communicate with any device that enables the electronic device 1000 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed through an input / output (I / O) interface 1050. In addition, the electronic device 1000 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) through a network adapter 1060. As shown, the network adapter 1060 communicates with other modules of the electronic device 1000 through a bus 1030. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1000, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0089] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

[0090] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible embodiments, various aspects of the present disclosure can also be implemented in the form of a program product, which includes a program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of the present specification.

[0091] refer to Fig.11As shown, a program product 1100 for implementing the above-mentioned vegetation prediction model construction method according to an embodiment of the present disclosure is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto, and in this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus, or a device.

[0092] The program product may adopt any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0093] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0094] The program code contained on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, electromagnetic waves, etc., or any suitable combination of the foregoing.

[0095] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0096] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0097] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

[0098] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

[0099] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for constructing a vegetation prediction model, characterized in that: include: The ecological environment data of the target area is used as the independent variable, and the vegetation spatial distribution data corresponding to the target area is used as the dependent variable, and the first random forest model is trained to obtain a vegetation spatial distribution prediction model; The output of the vegetation parameter spatial distribution prediction model and the meteorological data at each time period are used as independent variables, and the vegetation characteristic change data at each time period are used as dependent variables, and the second random forest model is trained to obtain a vegetation characteristic time change prediction model; The vegetation spatial distribution prediction model is combined with the vegetation characteristic temporal change prediction model to obtain a vegetation prediction model.

2. The method for constructing a vegetation prediction model according to claim 1, characterized in that: Before taking the ecological environment data of the target area as the independent variable and the vegetation spatial distribution data corresponding to the target area as the dependent variable, the method further includes: Acquire the ecological environment data consisting of annual cumulative precipitation data, annual accumulated temperature data, annual average direct radiation data, annual average scattered radiation data, altitude data, annual average soil moisture data, soil texture data, soil organic matter data, soil cation exchange capacity data and land use type data of the target area; Using the vegetation coverage data corresponding to the target area as the vegetation spatial distribution data; The temporal and spatial resolutions of the ecological environment data and the vegetation spatial distribution data are unified.

3. The method for constructing a vegetation prediction model according to claim 1, characterized in that: The method uses the output of the vegetation parameter spatial distribution prediction model and the meteorological data at each time period as independent variables, and the vegetation characteristic change data at each time period as the dependent variable, including: The output of the vegetation parameter spatial distribution prediction model and the time period cumulative precipitation data, time period accumulated soil temperature data, time period average surface temperature data, time period average daily diffuse radiation data, and time period average soil moisture data are used as independent variables, and the time period vegetation coverage change data are used as dependent variables.

4. The method for constructing a vegetation prediction model according to claim 1, characterized in that: The training of the first random forest model comprises: Constructing first sample data based on the ecological environment data and the vegetation spatial distribution data; In the node splitting process of each decision tree in the first random forest model, randomly sampling the first sample data to obtain a feature subset; Based on the feature subset, determining a splitting threshold of each of the nodes by using a Gini coefficient; At each of the nodes, a decision tree is recursively constructed by multiple splits until a preset termination condition is reached; The prediction results of each decision tree on the input sample are weighted averaged to obtain the prediction results of the vegetation spatial distribution prediction model.

5. The method for constructing a vegetation prediction model according to claim 4, characterized in that: The randomly sampling the first sample data to obtain a feature subset includes: according to: The first sample data is randomly sampled to obtain a feature subset, where: represents the feature subset matrix, Indicates i The node sample data matrix of a decision tree, represents the number of features, represents a random sampling process; The step of determining the splitting threshold of each node based on the feature subset by using the Gini coefficient includes: according to: Determine the splitting threshold of each of the nodes, where represents the target splitting threshold, represents the candidate split threshold, and Respectively represent the split threshold The two sample subsets after segmentation, Represents the Gini coefficient calculation function; The Gini coefficient calculation function is based on: To determine, Indicates the sample category ratio of the node.

6. The method for constructing a vegetation prediction model according to claim 4, characterized in that: The step of performing weighted averaging on the prediction results of each decision tree for the input sample to obtain the prediction results of the vegetation spatial distribution prediction model includes: according to: The prediction result of the vegetation spatial distribution prediction model is obtained, wherein: represents the prediction result of the vegetation spatial distribution prediction model, represents the number of decision trees in the first random forest model, Indicates i The prediction output of a decision tree, represents a column vector of all ones, A column vector representing the predictions of all decision trees.

7. The method for constructing a vegetation prediction model according to claim 3, characterized in that: The training of the second random forest model comprises: according to: The second random forest model is trained, wherein: represents the predicted value of the vegetation coverage change data for each period, represents the number of decision trees in the second random forest model, Indicates j The prediction function of a decision tree, represents the output of the vegetation parameter spatial distribution prediction model, represents the cumulative precipitation data for each period, represents the soil accumulated temperature data for each period, represents the average surface temperature data for each period, represents the average daily diffuse radiation data for each period, It represents the average soil moisture data for each time period.

8. A vegetation prediction method, characterized in that: include: Obtain ecological environment data and time-based meteorological data for the area to be predicted; Inputting the ecological environment data and the meteorological data for each period into the vegetation prediction model to obtain the vegetation characteristic change data for each period of the area to be predicted; Wherein, the vegetation prediction model is obtained by the vegetation prediction model construction method as described in any one of claims 1-7.

9. A device for constructing a vegetation prediction model, characterized in that: include: A spatial distribution prediction module is used to train the first random forest model using the ecological environment data of the target area as an independent variable and the vegetation spatial distribution data corresponding to the target area as a dependent variable to obtain a vegetation spatial distribution prediction model; A temporal change prediction module is used to train a second random forest model using the output of the vegetation parameter spatial distribution prediction model and the meteorological data for each period as independent variables and the vegetation characteristic change data for each period as dependent variables to obtain a vegetation characteristic temporal change prediction model; The prediction model building module is used to combine the vegetation spatial distribution prediction model with the vegetation feature temporal change prediction model to obtain the vegetation prediction model.

10. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; Wherein, the processor is configured to execute the method for constructing a vegetation prediction model according to any one of claims 1 to 7 by executing the executable instructions.

Images