Multi-environmental-factor coupled filipendula angustifolia ecological adaptability prediction system and method thereof
By constructing an ecological adaptability prediction system with multiple environmental factors coupled, and employing a mechanistic-statistical hybrid model and collaborative filtering algorithm, the problem of neglecting the multi-factor coupling effect in traditional methods is solved, achieving high-precision and real-time ecological adaptability assessment and supporting the dynamic management of ecosystems.
Patent Information
- Application Number
- CN202511311801.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-15
- Publication Date
- 2025-12-23
AI Technical Summary
Existing ecological adaptability assessment methods neglect the multi-factor coupling effect, fail to accurately reflect environmental changes, and lack a real-time update mechanism, leading to biased assessment results and inaccurate predictions.
An ecological adaptability prediction system with multiple environmental factors is constructed. A mechanism-statistical hybrid model framework is adopted, combined with real-time monitoring data from the Internet of Things, and a collaborative filtering algorithm is used to handle missing data and dynamically optimize model parameters to achieve multi-factor coupling analysis and high-resolution visualization.
It significantly improves the accuracy and timeliness of ecological adaptability assessment, provides a scientific basis for ecosystem management, and enables timely response to environmental changes and adaptation to climate change and extreme events.
Smart Images

Figure CN121189555A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of ecology and agronomy, and particularly relates to a multi-environment factor coupled ecological adaptability prediction system for Eragrostis poaeoides and a method thereof. BACKGROUND
[0002] In the field of ecology and agricultural engineering, accurate assessment of the ecological adaptability of plants to environmental changes is an important basis for species protection and ecosystem management. Eragrostis poaeoides, as a typical species in the wetland-grassland transition zone, is influenced by multiple environmental factors such as temperature, humidity, light, and soil nutrients, so it is necessary to establish an adaptability prediction model that can accurately reflect the synergistic effect of multiple factors.
[0003] The existing ecological adaptability evaluation methods have obvious limitations: first, traditional methods are mostly based on independent analysis of a single environmental factor (such as temperature or precipitation), ignoring the coupling effects between multiple factors. For example, high humidity can alleviate high temperature stress, but single-factor models cannot capture this interaction, resulting in biased evaluation results. Second, the commonly used static models use fixed parameters, making it difficult to adapt to dynamic environmental conditions under climate change, and unable to effectively respond to the impact of extreme climate events. In addition, statistical models based on historical climate data lack real-time updating mechanisms, and cannot accurately predict the impact of sudden climate events on plant populations, resulting in a large gap between predicted and observed values.
[0004] The present application builds a multi-environment factor coupled ecological adaptability prediction system, adopts a mechanism-statistics hybrid model framework, integrates multi-factor coupling analysis, dynamic parameter updating, and high-resolution spatial visualization technology, and realizes accurate assessment and dynamic prediction of the ecological adaptability of Eragrostis poaeoides. The system integrates real-time monitoring data through the Internet of Things, dynamically optimizes model parameters using data assimilation technology, and uses collaborative filtering algorithms to handle data missing problems, effectively overcoming the limitations of traditional methods, and significantly improving prediction accuracy and practicality. SUMMARY
[0005] In order to overcome the problems raised in the above background art, the present application proposes a multi-environment factor coupled ecological adaptability prediction system for Eragrostis poaeoides and a method thereof.
[0006] The technical solution of the present application is: a multi-environment factor coupled ecological adaptability prediction system for Eragrostis poaeoides, comprising:
[0007] a data acquisition module for acquiring multi-source environmental data of the Eragrostis poaeoides growth area, including real-time monitoring data, historical climate data, and future prediction data;
[0008] The data processing module is configured to clean and fuse the multi-source environmental data acquired by the data acquisition module, and to perform data collaborative inference and generative filling on missing data by using a collaborative filtering compensation algorithm based on the spatiotemporal correlation between environmental factors.
[0009] The coupling analysis modeling module is configured to perform multi-factor coupling analysis on the processed environmental data by using a mechanism-statistics hybrid model framework.
[0010] The dynamic prediction module is configured to receive real-time data streams and dynamically update the model parameters and factor weights of the coupling analysis modeling module.
[0011] The spatial visualization module is configured to receive the ecological adaptability index and generate a high-resolution ecological adaptability spatial distribution map in combination with a geographic information system.
[0012] Preferably, the data processing module, when performing data collaborative inference and generative filling on missing data by using a collaborative filtering compensation algorithm based on the spatiotemporal correlation between environmental factors, comprises:
[0013] A11: a data collaborative inference step, which infers environmental data at a missing position by using environmental data at a known position based on the spatial correlation and the temporal correlation between environmental factors;
[0014] A12: a generative filling step, which generates feature data conforming to a distribution to fill in missing values based on a joint probability distribution of a complete environmental data set learned by a machine learning model.
[0015] Preferably, the data collaborative inference step specifically comprises:
[0016] S11: data input, which receives an environmental factor data set including missing values from multiple geographic positions, wherein the environmental factors include temperature, humidity, illumination, soil pH value, and soil nutrients;
[0017] S12: missing pattern analysis on the acquired environmental factor data set, which identifies spatial missing caused by sparse spatial sampling points and time series missing caused by time series interruption;
[0018] S13: spatial interpolation inference based on spatial correlation, which performs data filling.
[0019] Preferably, when performing spatial interpolation inference based on spatial correlation, the following steps are performed for data identified as spatial missing:
[0020] S21: for a target environmental factor with missing values, calculating an experimental semi-variogram function of known data of the target environmental factor;
[0021] S22: fitting the theoretical model of quantifying the spatial autocorrelation of the target environmental factor to the obtained experimental semi-variogram function;
[0022] S23: based on the fitted semi-variogram function model, the optimal unbiased estimate of the missing value of the environmental factor at the target position is calculated by using the observed values of the environmental factor at the known position points around the target position through the Kriging interpolation method.
[0023] As preferred, when spatial interpolation is performed based on spatial correlation, the following sub-steps are performed for the data identified as missing in the time series:
[0024] S31: for the geographical position points with missing data, the trend, periodicity and randomness components of the known time series data of the target environmental factor are identified by analyzing the known time series data of the target environmental factor;
[0025] S32: according to the obtained trend, periodicity and randomness components, a time series prediction model is established to infer and fill in the environmental factor values in the missing time period;
[0026] S33: output the complete environmental factor data set after time inference and filling.
[0027] The time series prediction model includes an autoregressive integrated moving average model and a long short-term memory network.
[0028] As preferred, the generative filling step specifically includes:
[0029] S41: data input, receiving an environmental factor data set including missing values from multiple geographical positions, wherein the environmental factors include temperature, humidity, light, soil pH and soil nutrients;
[0030] S42: pre-processing the received input data to construct a complete data set sample and a mask matrix containing simulated missing values, wherein the complete data set sample is used for model training;
[0031] S43: constructing a generative machine learning model configured to learn the joint probability distribution of the complete environmental factor data;
[0032] S44: training the constructed generative machine learning model using the complete data set sample, and after training is completed, inputting the input data containing missing values and the corresponding mask matrix into the generative machine learning model;
[0033] S45: the generative machine learning model generates and outputs the filled complete data matrix based on the learned joint probability distribution.
[0034] The execution process of the collaborative filtering compensation algorithm is as follows:
[0035] S51: classify the data missing pattern, and determine as random missing, spatial block missing and time series segment missing;
[0036] S52: select a corresponding compensation method according to different missing patterns: preferentially call a data collaborative inference step for spatial correlation missing, preferentially call a generative filling step for high-dimensional and complex nonlinear missing, or combine the two;
[0037] S53: verify the filling result, evaluate the filling accuracy through an error index including root mean square, and if the accuracy does not reach a preset threshold, reselect a compensation method.
[0038] As preferred, the coupling analysis modeling module specifically comprises:
[0039] A21: a process mechanism model, which is a physiological and ecological process model constructed based on biological characteristics of the moss, and is used for simulating mechanism effects of environmental factors on photosynthesis, transpiration and nutrient absorption processes of the moss;
[0040] A22: a niche statistical model, which is used for mining nonlinear relationships between historical distribution data of the moss and environmental factors;
[0041] A23: a multi-factor coupling analysis unit, which is used for comprehensively outputting the process mechanism model and the niche statistical model, quantifying weights of temperature, humidity, light, soil pH and soil nutrients and interaction effects therebetween, and generating an ecological adaptability index.
[0042] As preferred, the process mechanism model is a physiological and ecological process simulation model constructed based on biological characteristics of the moss, and quantitatively describes mechanism effects of environmental factors on core physiological processes of the moss through mathematical formulas, specifically comprising:
[0043] A31: a photosynthesis sub-model, which is used for simulating photosynthetic rate of the moss, a first input item is environmental factors including light intensity, atmospheric carbon dioxide concentration, leaf temperature and soil water potential, a second input item is physiological parameters defined based on characteristics of the moss including maximum carboxylation rate and maximum electron transfer rate, and an output is net photosynthetic rate;
[0044] A32: a transpiration sub-model, which is used for simulating transpiration rate of the moss, a first input item is environmental factors including air temperature, air humidity, wind speed and solar radiation, a second input item is leaf stomatal conductance from the photosynthesis sub-model, and an output is transpiration rate;
[0045] A33: a nutrient uptake sub-model for simulating the nutrient uptake rate of the moso grass from the soil, the first input being environmental factors including soil temperature, soil pH, nitrogen source concentration and phosphorus source concentration in the soil solution, the second input being root morphological and physiological parameters of the moso grass including root specific surface area and maximum uptake rate, and the output being nitrogen and phosphorus nutrient uptake rate.
[0046] wherein the photosynthesis sub-model adopts the Farquhar photosynthesis biochemical model, and the core calculation formula is:
[0047] An = min{Ac, Aj} - Rd;
[0048] wherein An is the net photosynthetic rate, Ac is the carboxylation rate limited by the activity of the ubisco enzyme, and Aj is the carboxylation rate limited by the RuBP regeneration rate.
[0049] wherein the transpiration sub-model adopts the Penman-Monteith equation, and the core calculation formula is:
[0050] λET = [Δ(Rn - G) + ρacp(es - ea)ga] / [Δ + γ(1 + gs / ga)];
[0051] wherein λET is the heat flux, Rn is the net radiation, G is the soil heat flux, ρa is the air density, cp is the constant pressure specific heat capacity, (es - ea) is the saturated water vapor pressure difference, Δ is the slope of the saturated water vapor pressure-temperature curve, γ is the psychrometric constant, ga is the aerodynamic conductance, and gs is the leaf stomatal conductance.
[0052] wherein the nutrient uptake sub-model adopts the Michaelis-Menten kinetics equation, and the core calculation formula is:
[0053] Uptake = (Imax*[S]) / (Km + [S]);
[0054] wherein Uptake is the nutrient uptake rate, Imax is the maximum uptake rate, [S] is the concentration of a specific nutrient ion in the soil solution, and Km is the half-saturation constant, which is regulated by soil temperature and soil pH.
[0055] As preferred, the dynamic prediction module specifically comprises:
[0056] A41: a real-time data interface unit for receiving a real-time environmental data stream;
[0057] A42: a dynamic update engine connected with the real-time data interface unit and configured to perform the following operations:
[0058] A. Parameter updating, based on real-time environmental data stream, using data assimilation technology, dynamically adjusting the key physiological parameters of the process mechanism model in the coupling analysis modeling module, including the maximum carboxylation rate and leaf stomatal conductance;
[0059] B. Weight optimization, based on real-time environmental data stream, dynamically updating the environmental factor weights and interaction coefficients of the multi-factor coupling analysis unit in the coupling analysis modeling module through an adaptive weight adjustment algorithm;
[0060] A43: a trend prediction unit connected to the dynamic updating engine and configured to:
[0061] A. Short-term prediction, inputting real-time environmental data stream into a time series prediction model to predict environmental factor changes in the near future;
[0062] B. Long-term deduction, accessing future climate scenario data to drive the dynamically updated coupling analysis modeling module to perform simulation and deduction of the long-term trend of the ecological adaptability index of Juncus decipiens;
[0063] A44: a result output unit configured to output real-time evaluation results, short-term prediction results and long-term deduction results to a spatial visualization module.
[0064] The multi-environment factor coupled Juncus decipiens ecological adaptability prediction method comprises the following steps:
[0065] S61: acquiring multi-source environmental data through a data acquisition module;
[0066] S62: cleaning and fusing data through a data processing module, and filling in missing data using a collaborative filtering compensation algorithm;
[0067] S63: processing the fused data through a mechanism-statistics hybrid model in the coupling analysis modeling module, calculating the weights and interaction effects of multiple environmental factors, and generating a baseline ecological adaptability index;
[0068] S64: importing real-time data stream through a dynamic prediction module, dynamically updating model parameters and factor weights, and deducing the adaptability index changes under future environmental trends;
[0069] S65: spatializing the current and future adaptability indexes through a spatial visualization module to generate high-resolution distribution maps.
[0070] The beneficial effects of the present application are:
[0071] 1. Compared with existing technologies that mainly adopt independent evaluation schemes for single environmental factors, which have serious limitations and cannot accurately reflect the complex interactions between multiple environmental elements, often leading to assessment results that deviate from the actual situation, this invention adopts a multi-factor dynamic coupling analysis scheme. By comprehensively quantifying the weights and interaction effects of multi-dimensional environmental factors such as temperature, humidity, light, and soil characteristics, it establishes an evaluation system that comprehensively reflects the comprehensive impact of the environment, significantly improving the integrity and accuracy of ecological adaptability assessment and providing a more scientific and reliable basis for ecosystem management.
[0072] 2. Compared with the existing technologies that generally use static models with fixed parameters, which have obvious lag and inadequacy and cannot effectively cope with constantly changing natural environmental conditions, this invention adopts a dynamic prediction and real-time update scheme. Through data assimilation technology, it realizes dynamic optimization of model parameters and factor weights, enabling the system to respond to various environmental changes in a timely manner, greatly improving the timeliness and reliability of prediction results, and providing strong support for dynamic environmental management.
[0073] 3. Compared with existing technologies that often use a single statistical model or mechanism model, which have obvious defects such as lack of theoretical interpretability or insufficient prediction accuracy, this invention adopts a mechanism-statistical hybrid modeling scheme, which organically integrates the process mechanism model based on biological principles with the data-driven niche statistical model. This maintains the good interpretability of the mechanism model and gives full play to the high precision advantage of the statistical model, achieving a perfect unity of theoretical depth and practical accuracy.
[0074] 4. Compared with the traditional methods of handling missing data by simple interpolation or direct deletion, which have prominent problems such as low data utilization and poor filling accuracy, this invention adopts an intelligent collaborative filtering compensation algorithm, which innovatively combines spatial interpolation technology, time series prediction methods and generative filling and other advanced technologies. It can intelligently identify various missing patterns and automatically select the optimal filling strategy, which greatly improves the quality and integrity of data processing.
[0075] 5. Compared with existing technologies that are limited to current status assessment and lack forward-looking solutions, and have obvious shortcomings in effectively addressing climate change and extreme weather events, this invention adopts a multi-timescale prediction scheme, which has multiple functions such as real-time monitoring and assessment, accurate short-term prediction, and long-term trend extrapolation. It provides a comprehensive scientific basis and decision-making tools for the adaptive management and protection of ecosystems under the background of climate change, and significantly enhances the climate resilience of ecosystems. Attached Figure Description
[0076] Fig. 1 The diagram shown is a schematic representation of the structure of the multi-environmental factor coupled mosquito grass ecological adaptability prediction system of the present invention.
[0077] Fig. 2 The flowchart of the multi-environment factor coupled ecological adaptability prediction method of the invention is shown. DETAILED DESCRIPTION
[0078] The invention will be further described below in conjunction with the drawings and examples.
[0079] Referring to Figs. 1-2 The invention provides an example: a multi-environment factor coupled ecological adaptability prediction system, comprising:
[0080] A data acquisition module for acquiring multi-source environment data of the mosquito grass growing area, including real-time monitoring data, historical climate data and future prediction data;
[0081] A data processing module for cleaning and fusing the multi-source environment data acquired by the data acquisition module, and using a collaborative filtering compensation algorithm based on the spatio-temporal correlation between environment factors to collaboratively infer and generate filling for missing data; comprising:
[0082] A11: a data collaborative inference step, based on the spatial correlation and temporal correlation between environment factors, using known location environment data to infer missing location environment data, specifically including:
[0083] Data input, receiving an environment factor data set including data from multiple geographic locations and containing missing values, wherein the environment factors include temperature, humidity, light, soil pH value and soil nutrients;
[0084] Performing missing pattern analysis on the acquired environment factor data set to identify spatial missing caused by sparse spatial sampling points and time series missing caused by time series interruption;
[0085] Based on the spatial correlation, the spatial interpolation inference is performed to fill in the data;
[0086] A12: a generative filling step, based on a machine learning model to learn the joint probability distribution of the complete environment data set, to generate feature data conforming to the distribution to fill in the missing values, specifically including:
[0087] Data input, receiving an environment factor data set including data from multiple geographic locations and containing missing values, wherein the environment factors include temperature, humidity, light, soil pH value and soil nutrients;
[0088] Preprocessing the received input data to construct a complete data set sample and a mask matrix containing simulated missing, wherein the complete data set sample is used for model training;
[0089] Constructing a generative machine learning model configured to learn the joint probability distribution of the complete environment factor data;
[0090] The generative machine learning model trained by the complete data set sample is used to input the input data containing missing values and the corresponding mask matrix into the generative machine learning model after the training is completed;
[0091] The generative machine learning model generates and outputs the filled complete data matrix based on the learned joint probability distribution.
[0092] In this embodiment, the application receives a multi-dimensional environmental factor data set containing temperature, humidity, light, soil pH value and soil nutrients, etc. First, the data is preprocessed and training samples and mask matrices are constructed; then, the generative machine learning model learns the joint probability distribution of complete environmental data, and the trained model can generate feature data conforming to the statistical distribution of the original data according to the input of real missing data and its mask, and accurately fill in the missing values. Effectively solve the compensation problem of high-dimensional, nonlinear environmental data missing, significantly improve the data quality and integrity, and provide a solid foundation for the accuracy and reliability of subsequent ecological models.
[0093] Among them, when spatial interpolation inference is performed based on spatial correlation, the following steps are performed for the data identified as spatial missing:
[0094] For the target environmental factor with missing data, the experimental semi-variogram function of the known data thereof is calculated;
[0095] The theoretical model quantifying the spatial autocorrelation of the target environmental factor is fitted to the obtained experimental semi-variogram function;
[0096] Based on the fitted semi-variogram function model, the optimal unbiased estimate of the missing value of the environmental factor at the target position is calculated by using the environmental factor observation values of the known position points around the target position through the Kriging interpolation method.
[0097] Among them, when spatial interpolation inference is performed based on spatial correlation, the following steps are performed for the data identified as spatial missing:
[0098] For the geographical position point with missing data, the trend, periodicity and randomness components of the target environmental factor are identified by analyzing the known time series data thereof;
[0099] According to the obtained trend, periodicity and randomness components, a time series prediction model is established to infer and fill in the environmental factor values in the missing time period;
[0100] Output the complete environmental factor data set after time inference filling.
[0101] The time series prediction model used includes an autoregressive integrated moving average model and a long short-term memory network.
[0102] In this embodiment, the application receives multi-source environmental data sets including temperature, humidity, light, soil pH, and nutrients, etc. First, missing mode analysis is performed to intelligently identify spatial missing and time series missing. For spatial missing, the Kriging interpolation technique of geostatistics is used to quantify spatial autocorrelation by calculating the semi-variation function, and the surrounding known points are used to make the optimal unbiased estimation for the missing position. For time series missing, ARIMA and LSTM models are used to analyze the trend, period and random components of time series data to achieve accurate prediction and filling of missing values. The fusion of spatial and temporal dual-dimensional information is realized, which significantly improves the integrity, accuracy and reliability of ecological data reconstruction, and provides a solid data foundation for subsequent ecological adaptability prediction and dynamic modeling.
[0103] In summary, the execution flow of the collaborative filtering compensation algorithm is as follows:
[0104] Classify the data missing mode, and judge it as random missing, spatial block missing and time series fragment missing;
[0105] According to different missing modes, select the corresponding compensation method: for spatial correlation missing, preferentially call the data collaborative inference step, for high-dimensional and complex nonlinear missing, preferentially call the generative filling step, or use both;
[0106] Verify the filling result, evaluate the filling precision through error indicators including root mean square, and if the precision does not reach the preset threshold, reselect the compensation method.
[0107] In this embodiment, the application intelligently identifies data missing modes, including spatial missing, time series missing and complex nonlinear missing, respectively uses Kriging spatial interpolation and ARIMA / LSTM time series prediction model for accurate inference and filling, and combines generative adversarial network to learn environmental factor joint probability distribution to generate data conforming to real statistical characteristics. Finally, through error index verification and self-adaptive optimization, the high-precision reconstruction and deep fusion of multi-source ecological data are realized, which significantly improves the data integrity, accuracy and reliability, and provides a solid data foundation for the ecological system adaptability prediction model.
[0108] The coupling analysis modeling module adopts a mechanism-statistics hybrid model framework to perform multi-factor coupling analysis on the processed environmental data, specifically including:
[0109] A21: Process mechanism model, a physiological and ecological process model based on the biological characteristics of the mosquito grass, used to simulate the mechanism of the influence of environmental factors on the photosynthesis, transpiration and nutrient absorption processes of the mosquito grass.
[0110] A22: a niche statistical model for mining the non-linear relationship between historical distribution data of the mosquito grass and environmental factors;
[0111] A23: a multi-factor coupling analysis unit for quantifying the weights of temperature, humidity, light, soil pH, and soil nutrients and the interaction effects therebetween, and generating an ecological adaptability index.
[0112] In the present embodiment, the present application adopts a mechanism-statistics hybrid model framework, constructs a process mechanism model and a niche statistical model based on the biological characteristics of the mosquito grass, and uses a multi-factor coupling analysis unit to integrate the output results of the two models, quantifies the weights of key environmental factors such as temperature, humidity, light, soil pH, and nutrients and the interaction effects therebetween, and finally generates a precise ecological adaptability index. The present application overcomes the limitations of traditional single models, combines the interpretability of mechanism models and the prediction accuracy of statistical models, significantly improves the comprehensiveness and reliability of the evaluation of the ecological adaptability of the mosquito grass, and provides a scientific basis for the protection of wetland ecosystems and species management.
[0113] The process mechanism model is a physiological and ecological process simulation model constructed based on the biological characteristics of the mosquito grass, which quantitatively describes the mechanism of the influence of environmental factors on the core physiological processes of the mosquito grass through mathematical formulas, and specifically includes:
[0114] A31: a photosynthesis sub-model for simulating the photosynthetic rate of the mosquito grass, the first input item being environmental factors including light intensity, atmospheric carbon dioxide concentration, leaf temperature, and soil water potential, and the second input item being physiological parameters defined based on the characteristics of the mosquito grass including maximum carboxylation rate and maximum electron transport rate, and the output being net photosynthetic rate;
[0115] A32: a transpiration sub-model for simulating the transpiration rate of the mosquito grass, the first input item being environmental factors including air temperature, air humidity, wind speed, and solar radiation, and the second input item being leaf stomatal conductance from the photosynthesis sub-model, and the output being transpiration rate;
[0116] A33: a nutrient absorption sub-model for simulating the absorption rate of soil nutrients by the mosquito grass, the first input item being environmental factors including soil temperature, soil pH, nitrogen source concentration, and phosphorus source concentration in the soil solution, and the second input item being root morphological and physiological parameters of the mosquito grass including root specific surface area and maximum absorption rate, and the output being nitrogen and phosphorus nutrient absorption rate.
[0117] The photosynthesis sub-model adopts the Farquhar photosynthesis biochemical model, and the core calculation formula is:
[0118] An = min{Ac, Aj} - Rd;
[0119] Wherein, An is net photosynthetic rate, Ac is carboxylation rate limited by ubisco enzyme activity, and Aj is carboxylation rate limited by RuBP regeneration rate.
[0120] Wherein, the transpiration sub-model adopts the Penman-Monteith equation, and a core calculation formula thereof is:
[0121] λET = [Δ(Rn-G) + ρacp(es-ea)ga] / [Δ + γ(1 + gs / ga)];
[0122] Wherein, λET is heat flux, Rn is net radiation, G is soil heat flux, ρa is air density, cp is constant pressure specific heat capacity, (es-ea) is saturated water vapor pressure difference, Δ is saturated water vapor pressure-temperature curve slope, γ is psychrometric constant, ga is aerodynamic conductance, and gs is leaf stomatal conductance.
[0123] Wherein, the nutrient absorption sub-model adopts the Michaelis-Menten kinetics equation, and a core calculation formula thereof is:
[0124] Uptake = (Imax*[S]) / (Km + [S]);
[0125] Wherein, Uptake is nutrient absorption rate, Imax is maximum absorption rate, [S] is the concentration of specific nutrient ions in soil solution, and Km is half-saturation constant, and the half-saturation constant is regulated by soil temperature and soil pH value.
[0126] Wherein, the application strictly describes the mechanism relationship between environmental factors and plant physiological processes through constructing a fine process mechanism model based on the biological characteristics of the mosquito grass, and innovatively integrates three core sub-models: the photosynthesis sub-model adopting the Farquhar equation, the transpiration sub-model based on the Penman-Monteith equation and the nutrient absorption sub-model using the Michaelis-Menten kinetics, so that not only the growth state of the mosquito grass is accurately simulated, but also the internal mechanism of multi-factor coupling is revealed, the scientificity and accuracy of ecological adaptability prediction are significantly improved, and a reliable theoretical tool and quantitative basis are provided for wetland plant protection and ecological system management.
[0127] In this embodiment, the present application combines process mechanism models based on the biological characteristics of Kobresia humilis with niche statistical models by constructing a mechanism-statistics hybrid model framework, quantifies the weights of environmental factors such as temperature, humidity, light, soil pH and nutrients and their interaction effects using a multi-factor coupling analysis unit, and finally generates a high-precision ecological adaptability index. The traditional limitations of single models are broken through, and the physiological process mechanism is combined with the high prediction accuracy of data-driven models, significantly improving the scientificity and reliability of the ecological adaptability evaluation of Kobresia humilis, and providing a precise quantitative decision-making tool for wetland ecosystem protection, species diversity management and climate change response.
[0128] A dynamic prediction module for receiving real-time data streams and dynamically updating model parameters and factor weights of the coupling analysis modeling module, specifically including:
[0129] A41: a real-time data interface unit for receiving real-time environmental data streams;
[0130] A42: a dynamic update engine connected with the real-time data interface unit and configured to perform the following operations:
[0131] A. Parameter update, based on the real-time environmental data stream, using data assimilation technology to dynamically adjust the key physiological parameters of the process mechanism model in the coupling analysis modeling module, including the maximum carboxylation rate and leaf stomatal conductance;
[0132] B. Weight optimization, based on the real-time environmental data stream, through an adaptive weight adjustment algorithm, dynamically updating the environmental factor weights and interaction coefficients of the multi-factor coupling analysis unit in the coupling analysis modeling module;
[0133] A43: a trend prediction unit connected with the dynamic update engine and configured to:
[0134] A. Short-term prediction, inputting real-time environmental data streams into a time series prediction model to predict changes in environmental factors in the short term;
[0135] B. Long-term extrapolation, accessing future climate scenario data to drive the dynamically updated coupling analysis modeling module for simulation and operation, and extrapolating the long-term trend of the ecological adaptability index of Kobresia humilis;
[0136] A44: a result output unit configured to output real-time evaluation results, short-term prediction results and long-term extrapolation results to a spatial visualization module;
[0137] In this embodiment, the application realizes real-time evaluation, short-term accurate prediction and long-term trend deduction of the ecological adaptability of Kobresia pygmaea by constructing a dynamic prediction module, integrating a real-time data interface, a dynamic updating engine, a trend prediction unit and a result output unit, dynamically optimizing key physiological parameters of a process mechanism model using data assimilation technology, adjusting the weights and interaction coefficients of multiple environmental factors in real time through an adaptive algorithm, and combining a time series prediction model with future climate scenario data. The application breaks through the limitations of static models, significantly improves the response speed and prediction accuracy of the system to environmental changes, and provides dynamic and reliable decision support for wetland ecosystem adaptability management and climate change response.
[0138] The spatial visualization module is configured to receive the ecological adaptability index and generate a high-resolution ecological adaptability spatial distribution map in combination with a geographic information system. The specific working steps include:
[0139] Data reception: receiving output data from the dynamic prediction module, including the ecological adaptability index of Kobresia pygmaea at the current time point, the short-term and long-term prediction time points in the future, wherein the ecological adaptability index is associated with specific geographic coordinate positions;
[0140] Grid processing: dividing the geographic space of the target research area into geographic grids of a predetermined resolution;
[0141] Spatial interpolation: for each data set at a time point, using the ecological adaptability index of the known coordinate position as input, the spatial interpolation algorithm is used to calculate the ecological adaptability index of each geographic grid center point, and a continuous raster data surface covering the entire target research area is generated;
[0142] Layer superposition and rendering: superimposing the ecological adaptability index raster data obtained after interpolation with the basic geographic information layer, and rendering the raster data according to the preset adaptability level color mapping rule to generate a visual ecological adaptability spatial distribution map;
[0143] Output and deployment: outputting the generated ecological adaptability spatial distribution map as a geographic information system standard format file, and transmitting it to the user terminal for display and interactive analysis.
[0144] In this embodiment, the spatial visualization module in the application associates the multi-time scale ecological adaptability index output by the dynamic prediction module with geographic coordinates, generates a high-resolution raster data surface using gridding processing and spatial interpolation algorithms, and combines with the basic geographic information layer for hierarchical rendering and visual expression, and finally outputs a spatial distribution map in standard GIS format. This scheme realizes the conversion of ecological adaptability evaluation results from abstract data to intuitive spatial atlas, supports 10-meter-level high-precision grid visualization analysis and multi-time dynamic comparison, significantly improves the accuracy and spatial orientation of protected area planning and ecological restoration decision-making, and provides managers with a spatial decision-making tool that is both scientific and practical.
[0145] The multi-environment factor coupled prediction method of Kobresia pygmaea ecological adaptability comprises the following steps:
[0146] Obtain multi-source environmental data through a data acquisition module;
[0147] Clean and fuse the data through a data processing module, and fill in the missing data using a collaborative filtering compensation algorithm;
[0148] Process the fused data through a mechanism-statistics hybrid model in the coupling analysis modeling module, calculate the weights and interaction effects of multiple environmental factors, and generate a baseline ecological adaptability index;
[0149] Import real-time data streams through a dynamic prediction module, dynamically update model parameters and factor weights, and deduce the adaptability index changes under future environmental trends;
[0150] Express the current and future adaptability indexes spatially through a spatial visualization module to generate a high-resolution distribution map.
[0151] In this embodiment, the multi-environment factor coupled prediction method of Kobresia pygmaea ecological adaptability proposed by the application realizes the whole process of accurate prediction from multi-source environmental data acquisition to ecological adaptability spatial distribution map generation through the integration of key technical links such as data acquisition, collaborative filtering compensation data processing, mechanism-statistics hybrid modeling, dynamic parameter updating and high-resolution spatial visualization. This method innovatively uses a dynamic coupling mechanism to solve the problem of multi-factor interaction effect, optimizes model parameters in real time through data assimilation technology, significantly improves the prediction accuracy and timeliness, and finally generates a 10-meter-level high-resolution ecological adaptability spatial distribution map, providing scientific and reliable decision support for Kobresia pygmaea population protection, wetland ecosystem management and climate change response, effectively overcoming the limitations of traditional single-factor static models.
[0152] Embodiment one
[0153] In this embodiment, the system is applied to the Changbai Mountain National Nature Reserve in Northeast China, aiming to accurately delineate the highly adaptable core area of Kobresia pygmaea, to guide the formulation of functional zoning and management strategies in the nature reserve.
[0154] Implementation process:
[0155] 1. Data collection and processing:
[0156] The data collection module mobilized 30 IoT sensor nodes deployed in the nature reserve, continuously collecting real-time monitoring data of temperature, humidity, light, soil pH, and nutrients (nitrogen, phosphorus) for one year. At the same time, historical climate data for the past 20 years and climate prediction data under the RCP 4.5 scenario for the next 50 years were collected.
[0157] The data processing module initiates the collaborative filtering compensation algorithm. It is found that the soil nutrient data has "spatial block missing" due to partial sensor failure, and the humidity data has "time series missing" due to transmission problems at individual time periods. For soil nutrient data, the algorithm is based on spatial correlation and calls for Kriging interpolation: calculate the experimental semi-variogram of known points of soil nitrogen content, fit a spherical model, and successfully estimate the values of missing points. For humidity data, based on temporal correlation, the LSTM time series prediction model is called, which accurately predicts and fills in the missing data at the missing time periods based on the periodicity (daily variation, seasonal variation) and trend of known data. Finally, a complete and high-quality multi-source environmental data set is generated.
[0158] 2. Coupling analysis and benchmark modeling:
[0159] The processed data is input into the coupling analysis and modeling module.
[0160] The process mechanism model in the module starts working: the photosynthesis sub-model simulates the photosynthetic rate of Kobresia pygmaea under different light and temperature conditions; the transpiration sub-model combines meteorological data and stomatal conductance to calculate water dissipation; the nutrient absorption sub-model quantifies the relationship between soil nutrient supply and absorption rate.
[0161] At the same time, the niche statistical model analyzes the relationship between historical distribution data and environmental factors.
[0162] The multi-factor coupling analysis unit integrates the outputs of the above two models, quantifies the weight of each factor (humidity weight is the highest, reaching 36%) through principal component analysis and structural equation modeling, and reveals the synergistic effect of "high temperature and high humidity". Finally, the module generates a benchmark ecological adaptability index under the current climate conditions in the nature reserve.
[0163] 3. Dynamic prediction and long-term extrapolation:
[0164] The dynamic prediction module accesses climate scenario data for the next 50 years.
[0165] Its dynamic updating engine dynamically adjusts the key parameters in the mechanism model (such as the maximum carboxylation rate under high temperature conditions) and the factor weights in the coupling analysis (such as slightly increasing the weight of temperature) based on the prediction of future temperature rise and precipitation pattern changes.
[0166] The trend prediction unit then drives the updated model to perform calculations, long-term extrapolating the spatiotemporal variation trend of the suitability of the protected area under future climate change. The prediction shows that the suitability index in the low-altitude areas of the northwest will decrease significantly.
[0167] 4. Spatial visualization and decision support:
[0168] All current and future suitability indices are transmitted to the spatial visualization module. The module divides the entire protected area into 10m x 10m grids and uses Kriging interpolation to generate continuous, high-resolution raster surfaces from point-based index data.
[0169] The raster data is overlaid and rendered with the protected area's basic geographic layers such as terrain, water systems, and existing functional zoning maps. Using color mapping from red (low suitability) to green (high suitability), a clear spatial distribution map of ecological suitability is generated.
[0170] The results show that the three valley regions in the south and east are the core areas of continuous high suitability (index greater than 0.8). Based on this spatial distribution map, the protected area management bureau accurately delineated a core protected area with a total area of 120 hectares and developed corresponding monitoring and maintenance plans.
[0171] Example Two
[0172] This example demonstrates the application of the system in a wetland park in the Yangtze River Basin that is implementing ecological restoration. The goal is to monitor the ecological suitability of reintroduced populations of K. sibirica in real time and provide emergency management decision support when extreme rainstorm events occur in the summer of 2024.
[0173] Specific implementation process:
[0174] 1. Real-time data collection and anomaly handling:
[0175] The sensor network deployed within the wetland park transmits environmental data streams to the system in real time. During the rainstorm event, the data stream shows that the precipitation has exceeded 65mm for three consecutive days, the soil moisture has increased sharply, and the light intensity has decreased significantly due to continuous overcast weather.
[0176] The data processing module detects a number of sensor data transient losses due to heavy rain. For these "random missing", the module preferentially calls the generative filling step: using a trained generative adversarial network (GAN), it learns the joint probability distribution between various environmental factors (such as humidity and temperature, light and soil temperature), generates data that is completely consistent with the real statistical characteristics, seamlessly fills in the missing values, and ensures the continuity of the data stream.
[0177] 2. Dynamic coupling and real-time warning:
[0178] The real-time data stream during the heavy rain is continuously input into the dynamic prediction module.
[0179] The dynamic update engine responds immediately: based on real-time high humidity data, through an adaptive weight adjustment algorithm, it automatically increases the weight of the humidity factor from the base 35% to 42%, while decreasing the weight of the light factor from 20% to 13%, and adjusts the interaction coefficients between factors accordingly, to more accurately reflect the dominant limiting factor in the current environment.
[0180] The engine simultaneously optimizes parameters such as leaf stomatal conductance in the process mechanism model through data assimilation technology.
[0181] The trend prediction unit makes a short-term prediction, showing that the high humidity environment will last for a week. The updated model calculation shows that although there is insufficient light, the extremely high soil humidity alleviates water stress, and the overall adaptability index of Juncus bufonius is expected to remain above 0.75 (good level).
[0182] 3. Spatial visualization and emergency response:
[0183] The spatial visualization module generates the latest adaptability spatial distribution map every hour. The map clearly shows that the A area (adaptability index 0.65) in the low-lying wetland has dropped to the critical level due to excessive water accumulation, while the B area (adaptability index 0.82) in the high-lying area is still in a highly adaptable state.
[0184] Management personnel view the high-resolution distribution map in real time through the terminal and make quick decisions: start emergency drainage measures in the A area, and strengthen protection in the B area to prevent human disturbance. These measures successfully maintain the survival rate of Juncus bufonius populations under extreme weather conditions above 85%.
[0185] 4. Post-event evaluation and scheme optimization:
[0186] After the heavy rain ends, the system uses the data during the entire event to re-run the coupled analysis model for post-event evaluation. The results show that the survival rate predicted by the model (85%) is highly consistent with the field observation results (83%), verifying the prediction reliability of the system under extreme events.
[0187] The data of this event is incorporated into the historical database for further optimization of the model's parameter and weight settings, thereby improving the system's prediction ability for similar extreme events in the future.
[0188] The embodiments of the present application are described in detail above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the present application.
Claims
1. A multi-environmental factor coupled ecological adaptability prediction system for mosquito grass, characterized in that: include: The data acquisition module is used to acquire multi-source environmental data of the mosquito grass growing area, including real-time monitoring data, historical climate data and future forecast data; The data processing module is used to clean and fuse the multi-source environmental data acquired by the data acquisition module, and to use a collaborative filtering compensation algorithm based on the spatiotemporal correlation between environmental factors to perform collaborative inference and generative filling of missing data. The coupling analysis modeling module adopts a mechanistic-statistical hybrid model framework for multi-factor coupling analysis of processed environmental data. The dynamic prediction module is used to receive real-time data streams and dynamically update the model parameters and factor weights of the coupled analysis and modeling module. The spatial visualization module is used to receive the ecological adaptability index and combine it with the geographic information system to generate a high-resolution spatial distribution map of ecological adaptability.
2. The multi-environmental factor coupled ecological adaptability prediction system for mosquito grass according to claim 1, characterized in that: When the data processing module employs a collaborative filtering compensation algorithm based on the spatiotemporal correlation between environmental factors to perform collaborative inference and generative imputation of missing data, it includes: A11: Data collaborative inference step, based on the spatial and temporal correlations among environmental factors, uses environmental data from known locations to infer environmental data from missing locations; A12: Generative imputation step, which learns the joint probability distribution of the complete environment dataset based on a machine learning model, and generates feature data that conforms to the distribution to fill in the missing values.
3. The multi-environmental factor coupled ecological adaptability prediction system for mosquito grass according to claim 2, characterized in that: The data collaborative inference steps specifically include: S11: Data input, receiving a dataset of environmental factors from multiple geographic locations that includes missing values; S12: Perform missing pattern analysis on the acquired environmental factor dataset to identify spatial missing data caused by sparse spatial sampling points and time series missing data caused by time series interruption. S13: Perform spatial interpolation inference based on spatial correlation to fill in the data.
4. The multi-environmental factor coupled ecological adaptability prediction system for mosquito grass according to claim 3, characterized in that: When performing spatial interpolation inference based on spatial correlation, the following steps are performed on data identified as spatially missing: S21: For target environmental factors with missing data, calculate the experimental semivariogram of their known data; S22: Fit the theoretical model that quantifies the spatial autocorrelation of the target environmental factor to the obtained experimental semivariogram. S23: Based on the fitted semi-variogram model, the optimal unbiased estimate of the missing environmental factor value at the target location is calculated using the Kriging interpolation method and the environmental factor observations of known locations around the target location.
5. The multi-environmental factor coupled ecological adaptability prediction system for mosquito grass according to claim 4, characterized in that: When performing spatial interpolation inference based on spatial correlation, the following sub-steps are performed for data identified as missing in time series: S31: For geographical locations with missing data, identify their trend, periodic and random components by analyzing the known time series data of the target environmental factors; S32: Based on the obtained trend, periodic and random components, establish a time series prediction model to infer and fill in the environmental factor values within the missing time period; S33: Output the complete environmental factor dataset after time inference padding.
6. The multi-environmental factor coupled ecological adaptability prediction system for mosquito grass according to claim 5, characterized in that: The generative filling steps specifically include: S41: Data input, receiving a dataset of environmental factors from multiple geographic locations that includes missing values; S42: Preprocess the received input data to construct a complete dataset sample and a mask matrix containing simulated missing data; S43: Construct a generative machine learning model configured to learn the joint probability distribution of complete environmental factor data; S44: A generative machine learning model is built by training samples from the complete dataset. After training, the input data containing missing values and the corresponding mask matrix are input into the generative machine learning model. S45: Generative machine learning models generate and output a padded complete data matrix based on the learned joint probability distribution.
7. The multi-environmental factor coupled ecological adaptability prediction system for mosquito grass according to claim 6, characterized in that: The coupling analysis modeling module specifically includes: A21: Process mechanism model, used to simulate the mechanistic effects of environmental factors on the photosynthesis, transpiration and nutrient absorption processes of mosquito grass; A22: Niche statistical model, used to explore the nonlinear relationship between historical distribution data of mosquito grass and environmental factors; A23: Multi-factor coupling analysis unit, used to integrate the outputs of process mechanism model and ecological niche statistical model, quantify the weights of multiple environmental factors such as temperature, humidity, light, soil pH and soil nutrients and their interaction effects, and generate ecological adaptability index.
8. The multi-environmental factor coupled ecological adaptability prediction system for mosquito grass according to claim 7, characterized in that: The process mechanism model is a physiological and ecological process simulation model constructed based on the biological characteristics of mosquito grass. It quantitatively describes the mechanistic influence of environmental factors on the core physiological processes of mosquito grass through mathematical formulas, specifically including: A31: Photosynthesis sub-model, used to simulate the photosynthetic rate of mosquito grass. The first input is environmental factors, and the second input is physiological parameters defined based on the characteristics of mosquito grass. Its output is the net photosynthetic rate. A32: Transpiration sub-model, used to simulate the transpiration rate of mosquito grass. The first input is environmental factors, the second input is the leaf stomatal conductance from the photosynthesis sub-model, and the output is the transpiration rate. A33: Nutrient absorption sub-model, used to simulate the absorption rate of soil nutrients by mosquito grass. The first input is environmental factors, the second input is the root morphological and physiological parameters of mosquito grass, and the output is the nitrogen and phosphorus nutrient absorption rate.
9. The multi-environmental factor coupled ecological adaptability prediction system for mosquito grass according to claim 8, characterized in that: The dynamic prediction module specifically includes: A41: Real-time data interface unit, used to receive real-time environmental data streams; A42: Dynamic update engine, connected to the real-time data interface unit, configured to perform the following operations: A. Parameter update: Based on real-time environmental data stream, data assimilation technology is used to dynamically adjust the key physiological parameters of the process mechanism model in the coupled analysis modeling module; B. Weight optimization: Based on real-time environmental data stream, the environmental factor weights and interaction coefficients of the multi-factor coupling analysis unit in the coupling analysis modeling module are dynamically updated through an adaptive weight adjustment algorithm. A43: Trend prediction unit, connected to the dynamic update engine, is configured as follows: A. Short-term forecasting involves inputting real-time environmental data streams into a time series forecasting model to predict changes in environmental factors in the near future. B. Long-term projection: By incorporating future climate scenario data, the dynamically updated coupled analysis and modeling module is driven to perform simulation calculations, and the long-term trend of the ecological adaptability index of mosquito grass is projected. A44: Result Output Unit, configured to output real-time evaluation results, short-term prediction results, and long-term extrapolation results to the spatial visualization module.
10. A method for predicting the ecological adaptability of mosquito grass based on the coupling of multiple environmental factors, characterized in that: Includes the following steps: S61: Acquire multi-source environmental data through the data acquisition module; S62: The data is cleaned and fused through the data processing module, and a collaborative filtering compensation algorithm is used to fill in the missing data. S63: The fused data is processed by the mechanism-statistical hybrid model in the coupled analysis modeling module to calculate the weights and interaction effects of multiple environmental factors and generate a benchmark ecological adaptability index; S64: Import real-time data streams through the dynamic prediction module, dynamically update model parameters and factor weights, and deduce the changes in the adaptability index under future environmental trends; S65: The spatial visualization module spatially represents the current and future adaptability indices, generating a high-resolution distribution map.
