Pasture plateau adaptability evaluation system based on big data and evaluation method thereof

Abstract

The invention relates to the field of agriculture and animal husbandry, and discloses a pasture plateau adaptability evaluation system and method based on big data, and the system comprises a data acquisition module which is used for collecting meteorological data, soil data and remote sensing data related to a pasture growth environment, and historical pasture growth data; the data preprocessing module is connected with the data acquisition module and is used for performing cleaning, standardization processing and feature selection on the collected data; the model building module is connected with the data preprocessing module and used for building a pasture dynamic growth model based on a partial differential equation according to the preprocessed data; and the numerical solution and optimization module is connected with the model construction module and is used for carrying out numerical solution on the partial differential equation. According to the method, the dynamic growth model based on the partial differential equation is introduced, so that the growth process of the pasture can be continuously and dynamically simulated and predicted.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of agriculture and animal husbandry, and in particular to a highland pasture adaptability evaluation system based on big data and an evaluation method thereof. BACKGROUND

[0002] The grassland ecosystem in the highland region is the foundation and important ecological security barrier for the development of regional animal husbandry, and its health and sustainability are crucial. Therefore, scientific evaluation of the adaptability of pasture in a specific highland environment is a prerequisite for realizing the sustainable use of grassland resources and developing intelligent animal husbandry. However, in the process of applying existing evaluation techniques to actual work, technical personnel will generally encounter a series of operational difficulties and performance bottlenecks resulting from the inherent defects of the techniques.

[0003] Firstly, when running a model based on empirical statistics, the problem lies in the vulnerability of the model and the limitations of its application. For example, a linear regression model of "precipitation-grass yield" based on historical data will produce a large deviation in its prediction results when faced with an atypical weather pattern of "concentrated summer rainfall but extreme drought in spring". Because the model cannot handle this non-linear, time-series distribution-related environmental stress. This results in the model performing well only in regular years with "favorable wind and rain", and once an extreme climate event occurs, the evaluation results lose their reference value and cannot provide effective support for disaster warning and risk management, reflecting its inherent deficiency in robustness.

[0004] Secondly, when deploying a general mechanism model based on processes (such as CENTURY), the core operational obstacle lies in the difficulty of localizing the calibration of parameters. When a technical personnel in a region obtains such a model, they will find that the model contains dozens or even hundreds of biological parameters that need to be set (such as the maximum photosynthetic rate of a specific pasture, the nitrogen absorption efficiency, etc.). In the absence of a localized parameter manual, the technical personnel can only rely on limited literature values or make rough estimates. This "parameter transplantation" results in a significant reduction in the simulation accuracy of the model in the highland region, and the running results often deviate greatly from the measured values. If accurate calibration is required, costly and time-consuming positioning tests or laborious manual "trial-and-error" parameter adjustment are needed, which makes it difficult for the model to be quickly and cost-effectively applied in different pastures, and its actual operational efficiency and economy are poor. SUMMARY

[0005] In order to make up for the above shortcomings, the present application provides a highland pasture adaptability evaluation system based on big data and an evaluation method thereof, aiming to improve the problems of imperfect model mechanism, difficult parameter calibration, and lack of prediction ability in the prior art.

[0006] The first aspect, the present application provides the following technical scheme, a kind of highland adaptability evaluation system of pasture based on big data, comprising: Data acquisition module is used to collect meteorological data, soil data, remote sensing data and pasture growth history growth data related to pasture growing environment; Data preprocessing module is connected with the data acquisition module, for the collected data is cleaned, standardization processing and feature selection; Model construction module is connected with the data preprocessing module, for establishing a pasture dynamic growth model based on partial differential equation according to the data after preprocessing; Numerical solution and optimization module is connected with the model construction module, for the numerical solution of the partial differential equation is carried out, and the preset parameter in the model is optimized; Data output and visualization module is connected with the numerical solution and optimization module, for generating pasture adaptability evaluation data according to the result after solving and optimizing.

[0007] Preferably, the data preprocessing module performs the following operations: data cleaning is performed to handle missing values and outliers, data standardization is performed to unify the dimensions of different data, and principal component analysis is used for feature selection to reduce the dimension of data.

[0008] Preferably, the model construction module establishes a partial differential equation, which is in the form of: the change rate of pasture biomass is determined by a spatial diffusion term and a nonlinear growth term, wherein the nonlinear growth term is a function of pasture growth rate, environmental carrying capacity and environmental variables.

[0009] Preferably, the numerical solution and optimization module uses explicit finite difference method to iteratively calculate the numerical solution of pasture biomass at different time steps by discretizing the partial differential equation in time and space dimensions.

[0010] Preferably, the numerical solution and optimization module further optimizes the preset parameters in the model using genetic algorithm.

[0011] Preferably, the fitness function of the genetic algorithm is defined as the mean square error between the biomass predicted by the model and the measured historical growth data.

[0012] Preferably, the pasture adaptability evaluation data generated by the data output and visualization module includes: spatial distribution prediction of pasture biomass, regional adaptability index and future growth risk level.

[0013] The second aspect, the present application provides the following technical scheme, a kind of highland adaptability evaluation method of pasture based on big data, comprising the following steps: Acquire meteorological data, soil data, remote sensing data and historical growth data related to the growth of forage grasses; Perform data cleaning, data standardization and feature selection processing on the acquired data; Based on the processed data, a forage grass dynamic growth model is established with a partial differential equation as the core; The numerical calculation method is used to solve the forage grass dynamic growth model, and the optimization algorithm is used to optimize the parameters in the model to obtain the model results; Based on the model results, generate and output evaluation data representing the adaptability of forage grasses in the plateau.

[0014] In a third aspect, the present application provides the following technical solution, a computer device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, when the processor executes the computer program, the above-mentioned method for evaluating the adaptability of forage grasses in the plateau based on big data is realized.

[0015] In a fourth aspect, the present application provides the following technical solution, a readable storage medium, the readable storage medium stores a computer program, when the processor executes the computer program, the above-mentioned method for evaluating the adaptability of forage grasses in the plateau based on big data is realized.

[0016] The present application has the following beneficial effects: 1、In the present application, by introducing a dynamic growth model based on partial differential equation, the growth process of forage grasses can be continuously and dynamically simulated and predicted. Compared with the traditional static evaluation method which relies on statistical regression or empirical formula, the present application embeds the time variation rate and spatial variability in the model core, so as to accurately capture the accumulation process and trend of forage grass biomass under the continuous fluctuation of temperature, precipitation and other environmental factors. This solves the technical problem that the existing technology cannot effectively reflect the dynamic time sequence characteristics of the growth process due to the static or quasi-static model, resulting in a large deviation between the evaluation results and the actual growth situation.

[0017] 2、In the present application, by using genetic algorithm to optimize and calibrate the key parameters in the partial differential equation model, the technical problem of "not suitable for the local environment" and low accuracy of general models when applied in specific regions is solved. Traditional methods usually use fixed empirical parameters, which cannot adapt to the unique ecological environment of different plateau regions. The present application uses genetic algorithm to repeatedly iterate and compare the historical measured data of specific regions with the model prediction results, taking the minimum mean square error as the target, and automatically finds a set of optimal localized parameters. This makes the model able to deeply learn and adapt to the environmental characteristics of specific locations, thereby generating highly customized and accurate evaluation results for any target region.

[0018] 3、In the application, by combining partial differential equations with explicit finite difference method, the evaluation area is discretely processed in time and space dimensions, realizing high-resolution and fine evaluation of forage adaptability. The traditional evaluation method usually gives a general and uniform evaluation grade for a large area, ignoring the growth differences caused by micro-topography, soil patch distribution and other factors in the internal region. The numerical solution method of the application can calculate the biomass of each independent grid unit at different time points, improve the evaluation granularity from macro "surface" to micro "point", and intuitively reveal the "golden plot" with the most suitable growth and the "weak area" with growth risks in the internal region in the form of space-time thermal map, solving the technical problems of coarse granularity and inability to support precise agricultural decision-making of the traditional evaluation method. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A system architecture diagram of a forage highland adaptability evaluation system based on big data is provided for the application. Figure 2 A method flowchart of a forage highland adaptability evaluation method based on big data is provided for the application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0021] Embodiment one

[0022] Reference Figure 1 In the first embodiment of the application, the application provides a forage highland adaptability evaluation system based on big data, comprising: A data acquisition module for collecting meteorological data, soil data, remote sensing data and forage growth history data related to forage growth environment; Specifically, the data acquisition module aims to obtain the key meteorological variables driving the physiological and ecological processes of pasture. It accesses the public data service platform provided by the National Meteorological Information Center through a pre-configured server-side interface. The specific workflow is as follows: when the system is initialized, the geographical boundaries of the target evaluation area and the time resolution of the data request are set, such as every hour or every three hours. Then, an automatic scheduler triggers API call events at a pre-set frequency, and the request message contains the geographical range, time point, and required meteorological parameter fields, including at least surface temperature, cumulative precipitation, air relative humidity, and total solar radiation.

[0023] After receiving the data returned in JSON or XML format, the built-in parser immediately extracts the numerical values of the key parameters and stores them in the system's time series database along with their corresponding geographical coordinates and UTC timestamps. At the same time, to obtain soil data representing the physical and chemical properties of the pasture growth substrate, this module uses a hierarchical fusion strategy that balances data coverage and local accuracy. The first layer is the basic background data layer. This unit preloads authoritative soil survey databases, such as the China Soil Database published by the Nanjing Institute of Soil Science of the Chinese Academy of Sciences, to provide macro background information such as soil type and pH value in a wide range. The second layer is the high-precision calibration data layer, which receives and processes laboratory data of field survey soil samples provided by local sites through a standardized data import interface, especially the concentrations of available nitrogen, phosphorus, and potassium.

[0024] After obtaining the two layers of data, the spatial analysis component inside this unit will use the Kriging interpolation method to spatially interpolate and correct the macro background data layer using high-precision point calibration data, generating a comprehensive soil property distribution map covering the entire evaluation area with high resolution and localized calibration. To further provide large-scale and periodic live observations, this module also integrates automatic access and processing capabilities for international mainstream remote sensing data archives and distribution centers. Its workflow includes data retrieval and acquisition, image preprocessing, and vegetation index calculation. Based on the set evaluation area range and time series, this unit automatically submits data retrieval requests to platforms such as NASA's LAADS DAAC or ESA's Copernicus Open Data Center to obtain satellite image products such as MODIS or Sentinel-2.

[0025] The downloaded raw image data will automatically undergo atmospheric correction and geometric precision correction. Then, using the processed near-infrared band and red band surface reflectance, the normalized vegetation index (NDVI) is calculated strictly following the standard formula where DN is the digital number of the pixel, and ​The reflectance of red light band, the final output a series of evaluation area registration, with uniform time interval NDVI time series raster dataset. Finally, in order to ensure the localization calibration and accuracy verification of model parameters, the module also configures the acquisition function for obtaining ground truth data.

[0026] This function is mainly realized by a structured data form import interface, which is used to receive long-term observation sample data or historical yield archives from local animal husbandry scientific research units. The imported data must include accurate longitude and latitude coordinates, specific sampling time recorded to the day, and the biomass dry weight (kg / ha) of the forage sample in unit area after drying treatment.

[0027] The data preprocessing module is connected with the data acquisition module, which is used for cleaning, standardizing and feature selection of the collected data. Data cleaning is used to handle missing values and outliers, data standardization is used to unify the dimensions of different data, and principal component analysis is used for feature selection to reduce the dimension of data.

[0028] Specifically, after receiving meteorological time series data, soil spatial grid data, remote sensing vegetation index sequence and discrete historical growth sample data from the data acquisition module, the module immediately starts an automated processing pipeline. The pipeline first performs data cleaning operation. For missing values that may exist in the data, instead of using simple mean filling, this module applies more targeted interpolation algorithms. For example, for meteorological data with strong time series correlation, linear or spline interpolation method is used to estimate the value of missing points according to the data of several time points before and after.

[0029] For the null points in spatial data, spatial interpolation filling is used with the values of its neighborhood grid cells. Subsequently, the cleaning operation also includes the identification and processing of outliers. This module uses the quartile range (IQR) based detection method to identify any data points below Q1-1.5*IQR or above Q3+1.5*IQR as outliers. For the identified outliers, instead of being deleted directly, the cap method (Winsorizing) is used to replace their values with the upper or lower boundary value of the interval, so as to retain the existence of data points and eliminate the interference of extreme values on the stability of the model. After data cleaning, the pipeline enters the data standardization processing stage.

[0030] Due to the huge difference in the dimension and numerical range of temperature, precipitation, NDVI and other features, in order to avoid the unreasonable amplification or reduction of the weight of some features due to the difference in numerical scale during model training, this module performs Z-score standardization processing on all feature variables involved in modeling. The processing strictly follows the formula , where is the original value, is the mean of this feature, is its standard deviation. This operation converts all feature data into a standard normal distribution with mean 0 and standard deviation 1, ensuring that each feature is on an equal footing in subsequent calculations.

[0031] Finally, to address the possible high correlation between multi-dimensional input variables (i.e., the problem of multicollinearity) and reduce the complexity of model calculation, the module performs a feature selection operation. In this embodiment, the Principal Component Analysis (PCA) technique is specifically used. PCA reconstructs the original highly correlated feature variables into a set of fewer, mutually orthogonal (uncorrelated) new variables, i.e., principal components, through linear transformation.

[0032] The process is as follows: first, calculate the covariance matrix of the standardized data set, then solve the eigenvalues and eigenvectors of this matrix, and these eigenvectors are the directions of the principal components. The module will sort the principal components according to the size of the eigenvalues, and select the first k principal components whose cumulative variance contribution rate exceeds a pre-set threshold (e.g., 95%) as the final feature set input to the model construction module.

[0033] The model construction module is connected to the data preprocessing module and is used to establish a partial differential equation-based dynamic growth model of pasture based on the pre-processed data. The partial differential equation established by the model construction module has the form: the rate of change of pasture biomass is determined by a spatial diffusion term and a nonlinear growth term, where the nonlinear growth term is a function of the growth rate of the pasture itself, the environmental carrying capacity, and the environmental variables.

[0034] Specifically, the core function of this module is to receive the standardized, anomaly-free, and feature-selected high-quality data set generated by the preprocessing module, and based on these data, to establish a mathematical model that can accurately describe the spatio-temporal dynamic evolution process of pasture biomass.

[0035] To overcome the technical limitations of traditional statistical models that cannot inherently express spatial interactions and continuous changes in time, this embodiment discards the static regression analysis method and instead constructs a mechanism model based on partial differential equations (PDE). The model specifically adopts the paradigm of reaction-diffusion equations (Reaction-Diffusion Equation), whose general mathematical form is defined as: .

[0036] In this equation, represents the pasture biomass density at spatial coordinates (x, y) and time point t, which is the core variable of the model solution. The first term on the right side of the equation For the spatial diffusion term, it is used to describe the natural spread and expansion process of the grass population in space, such as the extension of the root system or the seed dispersal with the wind, wherein is the gradient operator, and is the diffusion coefficient, which is not a global constant value, but can be spatially heterogeneous according to factors such as terrain slope and dominant wind direction, so as to more truly reflect the spatial process. The second term on the right side of the equation is a nonlinear growth term (or reaction term) that determines the in-situ growth dynamics of the grass biomass without the influence of spatial diffusion, and its specific form is constructed as a classic logistic growth function in this embodiment, which is coupled with the comprehensive influence of environmental variables.

[0037] Specifically, the growth term is expressed as: , wherein the intrinsic growth rate is no longer a fixed empirical constant, but is defined as a dynamic function related to the environmental factor E, taking the principal components output by the data preprocessing module (for example, variables representing temperature, moisture, light, and other comprehensive environmental conditions) as input, and calculating the promotion or inhibition effect of the current environment on the maximum growth potential of the grass through a preset threshold function or bell-shaped function.

[0038] Similarly, the environmental carrying capacity is also set as a spatial variable, and its value is mainly determined by the basic carrying capacity of the soil properties (such as organic matter, nitrogen, phosphorus, and potassium content) and the real-time weather conditions. By substituting the preprocessed multi-dimensional data into the growth rate function and the carrying capacity function , this module finally forms a complete partial differential equation that can reflect complex ecological processes with dynamic parameters.

[0039] Specifically, the dynamic function of the growth term and the construction method of the environmental carrying capacity function in the present application can be implemented as follows: The "bell-shaped function" is an effective tool for simulating the optimal response of biological processes to environmental factors. In one specific embodiment of the present application, a Gaussian function can be used to quantify this response. For example, if the first principal component output by the data preprocessing module represents the comprehensive temperature conditions, the specific form of the growth term function may be expressed as:

[0040] , wherein is the maximum potential physiological growth rate of the grass; is the temperature principal component score of the current environment. Principal component scores corresponding to the optimal growth temperature; This parameter characterizes the temperature adaptation range of forage grasses. When multiple principal components (such as temperature and moisture) jointly influence growth, the total growth term function can be obtained by multiplying the results of multiple independent Gaussian functions.

[0041] The initial value can be determined in, but is not limited to, the following ways: 1) Consult relevant agricultural science or ecology literature to obtain the recommended physiological growth rate for a specific forage variety (such as alfalfa or ryegrass) under similar climate zones and soil conditions as the starting point for optimization; 2) Obtain preliminary growth rate data through small-scale field trials or greenhouse cultivation, and use the fitted data as the starting point for optimization. The initial estimated value. In this invention, It is not an empirical constant that is fixed in all environments, but rather a variable related to specific environmental factors (such as temperature and light). Therefore, although It represents the "maximum potential," but it is also set as a parameter to be optimized, so that the model can solve the true potential growth rate that best fits the specific pasture environment through subsequent optimization steps based on the actual collected data.

[0042] Similarly, the aforementioned "environmental carrying capacity modulation" can be implemented using a multiplicative correction model. Specifically, spatial points... Environmental bearing capacity at time t It can be represented as:

[0043] in, It is the basic bearing capacity mainly determined by soil properties (such as organic matter, nitrogen, phosphorus and potassium content), and it is a two-dimensional data that varies with space. It is a normalized real-time meteorological adjustment factor, for example, it can be calculated from the difference between the recent cumulative rainfall and the multi-year average; It is a weather regulation coefficient.

[0044] Crucially, the key parameters in the above formula, such as and These are not fixed empirical constants, but rather set as parameters to be optimized. They will be optimized by the numerical solution and optimization module in subsequent steps of the present invention using optimization methods such as genetic algorithms. Based on the goodness of fit (e.g., minimum mean square error) between the model prediction results and measured historical data, automatic optimization and localized calibration are performed. In this way, the problem of difficult-to-determine traditional model parameters is overcome, enabling the model to be fully constructed and implemented.

[0045] This model construction method solves the core technical problem of existing models treating growth parameters as static constants, which makes them unable to respond to real-time environmental changes and leads to distorted evaluation results. It lays a solid theoretical foundation for subsequent numerical solutions and accurate evaluation.

[0046] The numerical solution and optimization module, connected to the model building module, is used to numerically solve partial differential equations and optimize the preset parameters in the model. This module employs the explicit finite difference method, discretizing the partial differential equations in both time and space dimensions to iteratively calculate the numerical solutions for forage biomass at different time steps. The module further utilizes a genetic algorithm to optimize the preset parameters in the model. The fitness function of the genetic algorithm is defined as the mean square error between the model-predicted biomass and the measured historical growth data. The data output and visualization module generates forage adaptability assessment data including: spatial distribution predictions of forage biomass, regional adaptability indices, and future growth risk levels.

[0047] Specifically, this embodiment uses the explicit finite difference method to discretize the continuous PDE. Spatially, the entire evaluation region is divided into a... An orthogonal grid, for any continuous spatial point Discretized into grid nodes In the time dimension, continuous time t is discretized into steps with a step size of ; The time point K.

[0048] Based on this, pasture biomass Represented as a discrete matrix Time partial derivative Using the first-order forward difference approximation, The Laplace operator in the spatial diffusion term Then, the second-order central difference approximation is used as follows: .

[0049] Substituting these discrete forms into the original PDE, an explicit iterative update formula can be derived, allowing the system to directly calculate the biomass at time k+1 based on the state of all grid points at time k. To ensure the stability of the numerical solution, the time step... and spatial step size The choice must satisfy the Courant-Friedrichs-Lewy (CFL) stability condition.

[0050] After completing the numerical solution framework, this module further performs a parameter optimization process to ensure that several key parameters built into the model (e.g., the basic diffusion coefficient) are optimized. and constitutes the dynamic growth rate and load-bearing capacity The inherent parameters in the function can accurately reflect the ecological characteristics of a specific plateau region. Therefore, this embodiment employs a genetic algorithm (GA) for global optimization. This algorithm encodes all parameters to be optimized into a vector, serving as a "chromosome" or "individual." The algorithm starts with a randomly generated population of individuals.

[0051] Next, for each individual in the population, a set of parameters representing it is substituted into the PDE model, and a complete growing season simulation is run using the FDM method described above. After the simulation, the biomass predicted by the model at a specific spatiotemporal point is compared with historical measured growth data for the same location provided by the data acquisition module. The mean squared error (MSE) of both is defined as the fitness function of that individual; the smaller the MSE, the higher the fitness of the individual.

[0052] Subsequently, the algorithm enters an iterative evolutionary loop: through mechanisms such as roulette wheel selection, individuals with the highest fitness are preferentially selected to enter the next generation; the selected individuals generate a new population of offspring through crossover and mutation operations. This process is repeated until the average fitness of the population reaches stability or meets the preset number of iterations. When the algorithm terminates, the parameter set carried by the individual with the highest fitness is considered the optimal parameter set.

[0053] Finally, this module uses this set of optimal parameters to drive the FDM solver, generating the final spatiotemporal distribution prediction results for pasture cattle biomass. It also includes assessment data such as regional adaptability indices and future growth risk levels, and passes these structured data matrices to the data output and visualization module.

[0054] The data output and visualization module, connected to the numerical solution and optimization module, is used to generate pasture adaptability assessment data based on the solution and optimization results.

[0055] Specifically, in this embodiment of the invention, the data output and visualization module serves as the final presentation end of the evaluation system. It is connected to the output port of the numerical solution and optimization module and is responsible for transforming the original numerical results, such as the high-dimensional spatiotemporal matrix of pasture biomass generated by the optimized model, into intuitive and easy-to-read information products with decision support value.

[0056] This module receives the final biomass prediction matrix. Then, it is first processed deeply through an internal data aggregation and analysis engine to generate higher-level evaluation metrics. Specifically, to generate a regional adaptability index, this module performs analysis on each spatial grid. The time series data inside the end of the growing season is integrated or the biological weight final value is calculated, and is normalized, so that a scalar index which can statically characterize the comprehensive growth potential of the plot is obtained.

[0057] To generate the future growth risk level, the module analyzes the coefficient of variation of each grid biomass time series or the duration below a certain preset survival threshold, and maps the continuous risk value to discrete risk levels such as "low", "medium", "high" through predefined classification rules. After the index calculation is completed, the module calls the integrated geographic information system (GIS) rendering engine to visualize the structured data.

[0058] The spatial distribution prediction of forage biomass and the regional adaptability index are rendered into a spatial heat map, in which the color of each grid cell corresponds to the value size, for example, a color ramp from yellow to dark green is used to intuitively show the spatial distribution pattern of biomass from low to high. The future growth risk level is rendered into a risk warning map, with prominent red, yellow and green colors indicating different levels of risk areas.

[0059] In addition, the module also provides interactive chart function, allowing users to click on any grid cell on the visualization interface to dynamically generate the forage biomass growth curve graph over time. All these visualization results are displayed in a unified graphical user interface (GUI), and interactive functions such as map zooming, panning, layer switching and data query are provided, while supporting exporting various graphs to GeoTIF, PNG and other standard image formats, or exporting basic data to CSV files, thus solving the technical problem that the pure digital results of the original model output are difficult for non-professionals to understand and use, and providing direct and effective decision-making basis for precision livestock management.

[0060] Embodiment two: Referring to Figure 2 In the second embodiment of the present application, the present application provides a forage plateau adaptability evaluation method based on big data, comprising the following steps: Obtain meteorological data, soil data, remote sensing data and historical growth data related to forage growth; Perform data cleaning, data standardization and feature selection processing on the obtained data; Based on the processed data, a forage dynamic growth model with partial differential equation as the core is established; Numerical calculation method is used to solve the forage dynamic growth model, and optimization algorithm is used to optimize the parameters in the model to obtain the model results; Based on the model results, evaluation data representing forage plateau adaptability are generated and output.

[0061] Specifically, the method first performs a data acquisition step, for example, for the evaluation needs of a certain pasture in Qinghai Province (central longitude and latitude E98.5°, N35.5°), by calling the meteorological data API, the gridded meteorological data from April 1 to September 30, 2024, with a daily time resolution is obtained, and an example of data of a grid cell in a certain day is as follows: {“date”:“2024-06-15”,“temp_avg”:12.5,“precipitation”:3.2,“solar_rad”:18.5}; The units are Celsius, millimeters, and megajoules per square meter, respectively. At the same time, the 1:1,000,000 soil type map of the region is loaded from the China Soil Database, and the 3 soil sampling point measured data provided by the pasture is imported, such as: {“lon”:98.51,“lat”:35.52,“pH”:7.8,“org_matter”_percent:3.5,“N_mg_kg”:120}; Then, these point data are fused into the background soil map by Kriging interpolation method. Then, the 8-day composite MODIS land surface reflectance product of the region is downloaded from the NASA LAADS DAAC platform, and the NDVI time series data are calculated after atmospheric correction, such as {“date”:“2024-06-12”,“NDVI”:0.58}.

[0062] Finally, the historical yield data are imported as the true value, such as {“lon”:98.51,“lat”:35.52,“date”:“2023-08-20”,“biomass_kg_ha”:6500}。

[0063] After obtaining the data, the method enters the data preprocessing step, for example, in the meteorological data, the temperature data of“2024-06-16”is found to be missing, then linear interpolation method is used to fill it with T_16=(T_15+T_17) / 2; if an abnormal value of temp_avg=35°C is found, then IQR method is used (assuming Q3=15°C, IQR=8°C, then the abnormal threshold is 15+1.5*8=27°C), and the cap method is used to correct its value to 27°C.

[0064] Subsequently, all variables such as temperature (μ = 10, σ = 5) and precipitation (μ = 4, σ = 2) were standardized by Z-score, for example, the temperature of 12.5°C was standardized as (12.5-10) / 5 = 0.5. Finally, the matrix composed of all standardized variables (temperature, precipitation, light, pH, organic matter, N content, etc.) was subjected to principal component analysis (PCA), assuming that the first two principal components whose cumulative contribution rate of the calculated eigenvalues exceeds 95% are PC1 and PC2, and their expressions are PC1 = 0.7*Temp_z + 0.6*Precip_z +..., PC2 = -0.5*Temp_z + 0.4*Soil_N_z +...

[0065] A reaction-diffusion partial differential equation model was established: , where the environmental driving function was specifically defined as and , and the set of parameters to be optimized was .

[0066] Then, the model was solved using the explicit finite difference method, dividing the pasture into 100mx100m grids (1000x1000m2) , setting the time step to 0.1 day (0.1*24h) , and solving according to the iterative formula . To determine the set of parameters to be optimized , the method used a genetic algorithm, setting the population size to 100 and each individual as a specific set of parameters (e.g. [0.02, 0.1, 8000, 0.5, 0.3, 0.4, 0.2]).

[0067] For each individual, running the FDM model obtained the predicted biomass N-pred at the corresponding location on "2023-08-20", and calculated the mean square error MSE = (N-pred-6508)[INDEX]2 with the measured value 6500kg / ha, taking -NSE as the fitness of the individual. After 200 generations of selection (roulette method), crossover (single-point crossover) and mutation (Gaussian disturbance) operations, the individual with the highest fitness was obtained, and the parameters it carried were the optimal parameter set.

[0068] Finally, the optimal parameter set was used to drive the FDM model for a complete simulation, generating the final evaluation data: rendering the biomass matrix at the end of the growing season into a biomass spatial distribution map; calculating the average value of of all grids in a specific area as the adaptability index of that area; calculating the coefficient of variation of the time series of each grid's biological clock, and if it is greater than 0.3, dividing its risk level into "high", thereby generating a risk warning map, achieving the complete transformation from raw data to actionable decision information.

[0069] Embodiment three The third embodiment of the present application is based on the same inventive concept, and the present application provides a computer readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-mentioned embodiment of the method for evaluating the adaptability of pasture to plateau based on big data.

[0070] Embodiment four The fourth embodiment of the present application is based on the same inventive concept, and the present application provides a computer, which comprises a processor, a memory, the processor and the memory being in communication with each other, the memory being configured to store instructions, and the processor being configured to execute the instructions in the memory to implement the above-mentioned embodiment of the method for evaluating the adaptability of pasture to plateau based on big data.

[0071] It should be understood that various parts of the present application can be realized by hardware, software, firmware, or a combination thereof. In the above-described embodiments, a plurality of steps or methods can be realized by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if realized by hardware, and as in another embodiment, it can be realized by any one or a combination of the following technologies known in the art: discrete logic circuit with logic gate circuit for implementing logic functions on data signals, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array (PGA), field programmable gate array (FPGA), etc.

[0072] Finally, it should be noted that: the above only describes the preferred embodiments of the present application and is not used to limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application, should be included in the protection scope of the present application.

Claims

1. A big data-based assessment system for the adaptability of pastureland to plateau conditions, characterized in that, include: The data acquisition module is used to collect meteorological data, soil data, remote sensing data, and historical growth data of pasture related to the pasture growth environment. A data preprocessing module, connected to the data acquisition module, is used to clean, standardize, and select features from the collected soil property data. The model building module, connected to the data preprocessing module, is used to establish a pasture dynamic growth model based on partial differential equations based on the preprocessed data. The partial differential equation established by the model building module is in the form that the rate of change of pasture biomass is determined by a spatial diffusion term and a nonlinear growth term, wherein the nonlinear growth term is a function of the pasture's own growth rate, environmental carrying capacity, and environmental variables. Based on the preprocessed soil property data, the basic bearing capacity of pasture varies with spatial location by weighted summation of the various soil properties. Based on the aforementioned basic carrying capacity and the parameters representing the intrinsic growth rate of pasture, a partial differential equation model is constructed to describe the spatiotemporal dynamic changes of pasture. The numerical solution and optimization module is connected to the model building module and is used to numerically solve the partial differential equations and optimize the preset parameters in the model. The data output and visualization module is connected to the numerical solution and optimization module and is used to generate pasture adaptability assessment data based on the solution and optimization results.

2. The pasture plateau adaptability assessment system based on big data according to claim 1, characterized in that, The data preprocessing module performs the following operations: cleaning the data to handle missing and outlier values, standardizing the data to unify the dimensions of different data, and using principal component analysis to perform feature selection to reduce the dimensionality of the data.

3. The pasture plateau adaptability assessment system based on big data according to claim 1, characterized in that, The numerical solution and optimization module employs the explicit finite difference method, which iteratively calculates the numerical solution of pasture biomass at different time steps by discretizing the partial differential equation in both time and space dimensions.

4. The pasture plateau adaptability assessment system based on big data according to claim 3, characterized in that, The numerical solution and optimization module further employs a genetic algorithm to optimize the preset parameters in the model.

5. The pasture plateau adaptability assessment system based on big data according to claim 4, characterized in that, The fitness function of the genetic algorithm is defined as the mean square error between the biomass predicted by the model and the measured historical growth data.

6. The pasture plateau adaptability assessment system based on big data according to claim 5, characterized in that, The forage adaptability assessment data generated by the data output and visualization module includes: spatial distribution prediction of forage biomass, regional adaptability index, and future growth risk level.

7. A method for assessing the adaptability of pastureland to plateau conditions based on big data, characterized in that, A pasture plateau adaptability assessment system based on big data as described in any one of claims 1-6, comprising the following steps: Acquire meteorological data, soil data, remote sensing data, and historical growth data related to pasture growth; The acquired data undergoes data cleaning, data standardization, and feature selection processing. A dynamic growth model of forage grass with partial differential equations as its core is established based on the processed data. The dynamic growth model of forage grass was solved using numerical calculation methods, and the parameters in the model were optimized using optimization algorithms to obtain the model results. Based on the model results, assessment data characterizing the adaptability of pastureland to plateau conditions are generated and output.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the big data-based pasture plateau adaptability assessment method as described in claim 7.

9. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when executed by a processor, implements the big data-based pasture plateau adaptability assessment method as described in claim 7.

Images