Cultivated land occupation and supplement balance quantitative analysis method based on soil environment perception

Abstract

The invention discloses a cultivated land occupation and supplement balance quantitative analysis method based on soil environment perception, and relates to the technical field of cultivated land data analysis, and the method comprises the steps: collecting ground surface measurement data through a hyperspectral camera carried by an unmanned aerial vehicle, and generating a soil-nutrient-vegetation comprehensive physical feature vector of each land parcel at a moment t; establishing an adaptive yield-ecological prediction model based on the soil-nutrient-vegetation comprehensive physical feature vector in combination with a data driving component, obtaining an initial adaptive weight, and predicting the yield of the land parcel; and when the initial self-adaptive weight does not belong to the set threshold interval, running a self-adaptive and transfer learning strategy, updating the self-adaptive weight, and returning to the self-adaptive yield-ecological prediction model. The method solves the problem that agricultural production modes in different areas are different, and the quality index and the yield model of cultivated land are affected.

Description

Technical Field

[0001] The present invention relates to the technical field of cultivated land data analysis, and in particular to a quantitative analysis method for cultivated land occupation and compensation balance based on soil environment perception. Background Art

[0002] While some progress has been made in both domestic and international research on evaluating farmland occupation and compensation, and predicting yield and ecological benefits, several key technical shortcomings remain. Existing methods often rely on single remote sensing indicators, statistical yearbooks, or sparse field survey data to assess yield and farmland quality. Remote sensing is affected by factors such as cloud formation, observation angle, and vegetation saturation, making it difficult to accurately characterize soil moisture, root zone nutrients, and underground processes. For example, Chinese invention patent CN109214686B discloses a quantitative analysis method for cultivated land occupation and compensation balance, including: determining an analysis area for quantitative analysis; collecting analysis data within the analysis area to obtain analysis data; the analysis data includes land use patch data, cultivated land quality grade patch data, and ground elevation raster data; after obtaining the analysis data, constructing a quantitative analysis model to obtain a quantitative analysis model; and, based on the analysis model, sequentially performing cultivated land occupation and compensation quantity and quality analysis, cultivated land occupation and compensation spatial heterogeneity analysis, and cultivated land occupation and compensation landscape pattern analysis to obtain quantitative analysis results. This method can reduce the interference of human subjective factors during the analysis process, thereby obtaining more accurate information on the spatial morphology and landscape pattern of cultivated land.

[0003] While the aforementioned patents utilize the physical interpretability of physical process models for quantitative analysis, they rely on a large number of parameters that are difficult to calibrate and prone to bias under complex management conditions. Purely data-driven approaches, while flexible, are extremely sensitive to out-of-distribution samples and cross-regional extrapolation, and often lack physical boundary constraints, leading to unrobust predictions when data is scarce or institutional differences are significant. Furthermore, significant regional differences exist in crop seasons, rotation practices, irrigation, and fertilization management. These institutional differences alter the response relationship between the same physical signal and yield. Most existing data models do not explicitly incorporate crop structure and management practices as inputs, or lack domain adaptation strategies based on institutional stratification. This leads to a sharp decline in model performance when migrating from one domain to another, requiring extensive local annotation to restore performance.

[0004] Agricultural production models vary across regions, influencing farmland quality indicators and yield models. Model migration often requires consideration of local crop structure and management practices; otherwise, accurate predictions of food output and ecological benefits after land acquisition and compensation are difficult.

[0005] To this end, the present invention provides a quantitative analysis method for cultivated land occupation and compensation balance based on soil environmental perception. Summary of the Invention

[0006] The purpose of the present invention is to provide a quantitative analysis method for cultivated land occupation and compensation balance based on soil environmental perception, so as to solve the existing problems raised in the above background technology.

[0007] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a quantitative analysis method of cultivated land occupation and compensation balance based on soil environmental perception, comprising the following steps: S1. Deploy soil moisture sensors, soil temperature sensors, soil conductivity sensors, and meteorological sensors on the plots and upload them to the spatiotemporal database through the edge gateway. S2. Use a drone equipped with a hyperspectral camera to collect ground surface measurement data, embed RTK for spatiotemporal registration, and integrate it with multiple sources to generate a soil-nutrient-vegetation comprehensive physical feature vector for each plot at time t. S3. Establishing an adaptive yield-ecological prediction model based on the soil-nutrient-vegetation integrated physical feature vector and a data-driven component to obtain initial adaptive weights and predict the yield of the plot; S4. When the initial adaptive weight does not fall within the set threshold range, execute S5; S5. Establish an adaptive and transfer learning strategy, implement domain alignment, meta-learning small sample fine-tuning, and uncertainty-driven active sampling at the edge, update the adaptive weights, and return to S3.

[0008] A further improvement of the present invention is that the soil-nutrient-vegetation integrated physical characteristic vector in the adaptive yield-ecological prediction model includes a photosynthetically active radiation estimation module, a leaf area index time series module, and a water-nutrient response function module, which is expressed as follows: in, represents photosynthetically active radiation, Expressed as crop light energy utilization efficiency, Expressed as a water-nutrient response function.

[0009] The present invention is further improved in that the data-driven component in the adaptive yield-ecological prediction model combines a graph convolutional network with a long short-term memory network, takes the normalized physical feature vector and the plot-level crop structure vector as input, learns the spatiotemporal dependency of plot adjacency based on the XGBoost model, and outputs data yield estimates. and confidence intervals , where the adjacency matrix of the graph convolutional network is constructed based on the plot centroid distance threshold.

[0010] A further improvement of the present invention is that the adaptive yield-ecological prediction model also includes an adaptive data fusion strategy, which combines the soil-nutrient-vegetation integrated physical feature vector Estimation of data output The weighted fusion is performed to obtain the predicted yield of the plot, where the data yield is estimated The weights of are obtained through the initial adaptive weight function.

[0011] A further improvement of the present invention is that the initial adaptive weight function includes: calculating the relative confidence interval width ,Will Mapping to confidence score through linear truncation ; Secondly, calculate the sample size score , Indicates the number of local calibration samples, Indicates the set half-saturation constant; the sample size score and confidence scores The weighted summation is used to obtain the local reliability score, and the initial adaptive weight is obtained through Sigmoid mapping. .

[0012] A further improvement of the present invention is that the adaptive and transfer learning strategy is based on the plot-level crop structure vector and management practice vector Implementation; the plot-level crop structure vector includes the odd-even season identification, crop rotation sequence code, the proportion of each crop, and the straw treatment method in previous years; the management practice vector includes the irrigation system, fertilization rate, tillage method, sowing density, and input intensity index; the adaptive and transfer learning strategy specifically includes: Unsupervised domain alignment layer, used to reduce the covariance difference between source domain and target domain features, and output the aligned current plot features ; The few-sample fast adaptation layer is used to localize model parameters in the target area with limited calibration samples and output quantitative indicators of adaptability after fine-tuning. ; An online incremental fine-tuning layer for deployment at the edge to perform mini-batch updates using streaming local observation data; The adaptive weight fusion layer is used to receive the updated model parameters output by the online incremental fine-tuning layer and to obtain the updated adaptive weights by weighted summation with the initial adaptive weights. .

[0013] A further improvement of the present invention is that the unsupervised domain alignment layer groups vector samples according to institutional conditions based on the plot crop structure vector and the management practice vector, and calculates the feature covariance matrix of the training data and the current plot for each group k. and , by minimizing the conditional covariance difference, we can achieve conditional unsupervised distribution alignment. The output domain reliability measure of the unsupervised domain alignment layer is expressed as .

[0014] A further improvement of the present invention is that the few-sample fast adaptation layer uses the aligned target domain features , local calibration samples and training data model parameters As input, a 1–K-step fast adaptation is performed locally based on meta-learning to generate preliminary localization parameters , and calculate the quantitative index of adaptability after fine-tuning .

[0015] A further improvement of the present invention is that the online incremental fine-tuning layer is deployed at the edge to receive streaming local observation data and perform incremental parameter updates in small batches, and calculate and output local verification reliability in real time. With fine-tuning the sample size score , and output the updated model parameters 、 、 .

[0016] A further improvement of the present invention is that the predicted value of the adaptive yield-ecological prediction model is expressed as: .

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention first couples physical priors with data-driven models and performs dynamic linear fusion with adaptive weights, solving the problem of single model unreliability when dealing with cross-domain plots or data scarcity. It leverages the advantages of data models when data is sufficient, and relies on physical models to ensure boundary rationality when data is scarce or uncertain, making yield predictions more accurate and less likely to cross boundaries. 2. Secondly, by explicitly encoding block-level crop structure vectors and management practice vectors, and using them for conditional domain alignment, task partitioning, and meta-learning in adaptive units, we address the problem of model transfer failure caused by differences in covariate distributions due to different agricultural systems. System-aware alignment and hierarchical transfer preserve essential differences and adjust only the transferable components, thus avoiding erroneous transfer. 3. The automatic fusion of adaptive and transfer learning strategies may be too radical or lead to the risk of incorrect decisions in abnormal situations; combining human priors with automatic judgment to balance adaptability and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of the quantitative analysis method of cultivated land occupation and compensation balance based on soil environmental perception of the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0020] The term "and / or" is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0021] Example 1 Figure 1 The flowchart of the method for quantitative analysis of cultivated land occupation and compensation balance based on soil environment perception disclosed in this embodiment is shown. The steps are as follows: S1. Deploy soil moisture sensors, soil temperature sensors, soil conductivity sensors, and meteorological sensors on the plots and upload them to the spatiotemporal database through the edge gateway. S2. Use a drone equipped with a hyperspectral camera to collect ground surface measurement data, embed RTK for spatiotemporal registration, and integrate it with multiple sources to generate a soil-nutrient-vegetation comprehensive physical feature vector for each plot at time t. S3. Establishing an adaptive yield-ecological prediction model based on the soil-nutrient-vegetation integrated physical feature vector and a data-driven component to obtain initial adaptive weights and predict the yield of the plot; The soil-nutrient-vegetation integrated physical characteristic vector in the adaptive yield-ecological prediction model includes a photosynthetically active radiation estimation module, a leaf area index time series module, and a water-nutrient response function module, which is expressed as follows: in, Represents photosynthetically active radiation, which is obtained by inverting remote sensing data. Expressed as crop light energy utilization efficiency, Expressed as a water-nutrient response function.

[0022] The calculation formula is expressed as ,in Indicates the moisture content at the wilting point. Indicates the optimal moisture content, Represents the real-time moisture content. Using a saturation function, the soil moisture content can be mapped from the physical level to the crop response. ; A represents the plot area, N represents the soil nitrogen concentration, represents the optimum soil nitrogen concentration; Because farmland is not isolated, graph convolutional networks can model the spatial relationships between neighboring irrigation fields or terrain that affect water flow and microclimate. Furthermore, crop yields are affected by historical dynamics, so LSTM can capture temporal dependencies. The data-driven component of the adaptive yield-ecological prediction model combines a graph convolutional network with a long short-term memory network, takes the normalized physical feature vector and plot-level crop structure vector as input, learns the spatiotemporal dependency of plot adjacencies based on the XGBoost model, and outputs data yield estimates. and confidence intervals , where the adjacency matrix of the graph convolutional network is constructed based on the plot centroid distance threshold; for example, for the adjacency matrix M, if plot i is close enough to plot j, based on the plot centroid distance threshold, then is close to 1, otherwise it is close to 0.

[0023] The adaptive yield-ecological prediction model also includes an adaptive data fusion strategy, which combines the soil-nutrient-vegetation integrated physical feature vector Estimation of data output The weighted fusion is performed to obtain the predicted yield of the plot, where the data yield is estimated The weights of are obtained through the initial adaptive weight function.

[0024] The initial adaptive weight function includes: calculating the relative confidence interval width ,Will Mapping to confidence score through linear truncation ; Secondly, calculate the sample size score , Indicates the number of local calibration samples, represents the set half-saturation constant; since the model may be misleading even if the CI is small if the plot differences are large, the sample size score is set to and confidence scores The weighted summation is used to obtain the local reliability score, and the initial adaptive weight is obtained through Sigmoid mapping. .

[0025] In this case, the narrower the confidence interval width, The smaller, the Increases, when the data model prediction is more "certain", the weight shifts towards the data model, The more local samples, the higher the sample size score. More samples indicate that local calibration is sufficient, which increases the trust in data-driven results. The closer it is to 1, rise.

[0026] When there is a large amount of reliable data in the region, the data model is usually more accurate. should be larger; when data is scarce or unreliable, physical models are more robust. Small strain. Hybridization can combine the advantages of both and reduce the risk of extrapolation.

[0027] For example, if there are only three ground calibration points in a certain area and the confidence interval is very wide, the model relies more on the physical eigenvector If there are 100 calibration points and the error is small, then rely more on the data output estimate .

[0028] S4. When the initial adaptive weight does not fall within the set threshold range, execute S5; S5. Establish adaptive and transfer learning strategies; In decision-making scenarios such as the balance between occupation and compensation, which requires weighing food output and ecological benefits, a single-point prediction value itself cannot reflect the risk and confidence level; most existing systems lack the quantification and utilization of prediction uncertainty, and are unable to conduct active sampling, conservative decision-making or risk control based on uncertainty, thereby increasing the uncertain legal and ecological risks of the implementation of occupation and compensation.

[0029] After implementing domain alignment, meta-learning small sample fine-tuning, and uncertainty-driven active sampling at the edge, the adaptive weights are updated and returned to S3 to output the final plot yield and ecological indicators; It is used to solve the problem of mismatch between the yield prediction after occupation and compensation and the conversion coefficient migration of agricultural production models in different regions.

[0030] The adaptive and transfer learning strategy is based on the plot-level crop structure vector and management practice vector Implementation; the plot-level crop structure vector includes the odd-even season identification, crop rotation sequence code, the proportion of each crop, and the straw treatment method in previous years; the management practice vector includes the irrigation system, fertilization rate, tillage method, sowing density, and input intensity index; the adaptive and transfer learning strategy specifically includes: The unsupervised domain alignment layer is used to reduce the covariance difference between the source and target domain features. The unsupervised domain alignment layer groups vector samples by institutional conditions based on the plot crop structure vector and the management practice vector. This involves first performing regular binning to obtain a seasonal irrigation layer, and then performing secondary clustering of continuous management intensity and crop proportion within each seasonal irrigation layer to obtain the final group k. And for each group k, calculate the feature covariance matrix of the training data and the current plot and , by minimizing the conditional covariance difference, expressed as , to achieve conditional unsupervised distribution alignment, the output domain reliability measure of the unsupervised domain alignment layer is expressed as , output the current plot features after alignment ; Grouping by institutional conditions and aligning the feature covariances of each group in the unsupervised domain alignment layer can retain the essential differences between different institutions and only eliminate the statistical differences of the transferable parts between the source and target, thereby improving the accuracy and security of the alignment; avoiding global alignment from flattening institutional differences that should not be aligned (for example, incorrectly aligning drought-tolerant types with high-input types), thereby reducing the risk of incorrect migration.

[0031] The few-sample fast adaptation layer is used to localize model parameters in the target area with limited calibration samples and output quantitative indicators of adaptability after fine-tuning. ; The few-shot fast adaptation layer uses the aligned target domain features , local calibration samples and training data model parameters As input, a 1–K-step fast adaptation is performed locally based on meta-learning to generate preliminary localization parameters , and calculate the quantitative index of adaptability after fine-tuning , Represents all parameters (weights and biases) of a machine learning model; The few-shot fast adaptation layer provides fast, localized parameter adjustment capabilities through meta-learning or limited-step fine-tuning, enabling the model to achieve significant performance improvements even with very few calibration samples, reducing deployment calibration costs and shortening time to market. It also addresses the difficulty of effectively training or fine-tuning the model when target domain annotations are sparse, avoiding large-scale annotation in each new area.

[0032] An online incremental fine-tuning layer is deployed at the edge to perform small batch updates using streaming local observation data. The online incremental fine-tuning layer is deployed at the edge to receive streaming local observation data and perform incremental parameter updates in small batches. A small batch is to divide the data into W small subsets, and only a small part of them is used to update the parameters each time, such as w / W. The local verification reliability is calculated and output in real time. With fine-tuning the sample size score , and output the updated model parameters 、 、 .

[0033] The online incremental fine-tuning layer absorbs streaming observations in small batches at the edge and updates parameters in real time. It can continuously respond to dynamic changes in the environment / management, reduce the timeliness of the model, and output immediate reliability to support real-time decision-making. It solves the problem of static models experiencing performance degradation due to environmental changes or accumulated implementation errors during long-term operation, and provides online self-calibration capabilities to ensure long-term availability.

[0034] The adaptive weight fusion layer is used to receive the updated model parameters output by the online incremental fine-tuning layer and to obtain the updated adaptive weights by weighted summation with the initial adaptive weights. .

[0035] Then the predicted value of the adaptive yield-ecological prediction model is expressed as: .

[0036] The thresholds, weights and other setting values ​​may be set by default according to the present invention, or may be set by those skilled in the art.

[0037] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0038] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0039] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0040] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0041] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which are all protected by the present invention.

Claims

1. A quantitative analysis method for cultivated land occupation and compensation balance based on soil environmental perception, characterized by: The following steps are involved: S1. Deploy soil moisture sensors, soil temperature sensors, soil conductivity sensors, and meteorological sensors on the plots and upload them to the spatiotemporal database through the edge gateway. S2. Use a drone equipped with a hyperspectral camera to collect ground surface measurement data, embed RTK for spatiotemporal registration, and integrate it with multiple sources to generate a soil-nutrient-vegetation comprehensive physical feature vector for each plot at time t. S3. Establishing an adaptive yield-ecological prediction model based on the soil-nutrient-vegetation integrated physical feature vector and a data-driven component to obtain initial adaptive weights and predict the yield of the plot; S4. When the initial adaptive weight does not fall within the set threshold range, execute S5; S5. Establish an adaptive and transfer learning strategy, implement domain alignment, meta-learning small sample fine-tuning, and uncertainty-driven active sampling at the edge, update the adaptive weights, and return to S3.

2. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 1 is characterized by: The soil-nutrient-vegetation integrated physical characteristic vector in the adaptive yield-ecological prediction model includes a photosynthetically active radiation estimation module, a leaf area index time series module, and a water-nutrient response function module, which is expressed as follows: in, represents photosynthetically active radiation, Expressed as crop light energy utilization efficiency, Expressed as a water-nutrient response function.

3. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 2 is characterized by: The data-driven component of the adaptive yield-ecological prediction model combines a graph convolutional network with a long short-term memory network, takes the normalized physical feature vector and plot-level crop structure vector as input, learns the spatiotemporal dependency of plot adjacencies based on the XGBoost model, and outputs data yield estimates. and confidence intervals , where the adjacency matrix of the graph convolutional network is constructed based on the plot centroid distance threshold.

4. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 3 is characterized by: The adaptive yield-ecological prediction model also includes an adaptive data fusion strategy, which combines the soil-nutrient-vegetation integrated physical feature vector Estimation of data output The weighted fusion is performed to obtain the predicted yield of the plot, where the data yield is estimated The weights of are obtained through the initial adaptive weight function.

5. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 4 is characterized by: The initial adaptive weight function includes: calculating the relative confidence interval width ,Will Mapping to confidence score through linear truncation ; Secondly, calculate the sample size score , Indicates the number of local calibration samples, Indicates the set half-saturation constant; the sample size score and confidence scores The weighted summation is used to obtain the local reliability score, and the initial adaptive weight is obtained through Sigmoid mapping. .

6. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 5 is characterized by: The adaptive and transfer learning strategy is based on the plot-level crop structure vector and management practice vector Implementation; the plot-level crop structure vector includes the odd-even season identification, crop rotation sequence code, the proportion of each crop, and the straw treatment method in previous years; the management practice vector includes the irrigation system, fertilization rate, tillage method, sowing density, and input intensity index; the adaptive and transfer learning strategy specifically includes: Unsupervised domain alignment layer, used to reduce the covariance difference between source domain and target domain features, and output the aligned current plot features ; The few-sample fast adaptation layer is used to localize model parameters in the target area with limited calibration samples and output quantitative indicators of adaptability after fine-tuning. ; An online incremental fine-tuning layer for deployment at the edge to perform mini-batch updates using streaming local observation data; The adaptive weight fusion layer is used to receive the updated model parameters output by the online incremental fine-tuning layer and to obtain the updated adaptive weights by weighted summation with the initial adaptive weights. .

7. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 6 is characterized by: The unsupervised domain alignment layer groups vector samples by institutional conditions based on the plot crop structure vector and the management practice vector, and calculates the feature covariance matrix between the training data and the current plot for each group k and , by minimizing the conditional covariance difference, we can achieve conditional unsupervised distribution alignment. The output domain reliability measure of the unsupervised domain alignment layer is expressed as .

8. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 7 is characterized by: The few-shot fast adaptation layer uses the aligned target domain features , local calibration samples and training data model parameters As input, a 1–K-step fast adaptation is performed locally based on meta-learning to generate preliminary localization parameters , and calculate the quantitative index of adaptability after fine-tuning .

9. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 8 is characterized by: The online incremental fine-tuning layer is deployed at the edge to receive streaming local observation data and perform incremental parameter updates in small batches, computing and outputting local verification reliability in real time. With fine-tuning the sample size score , and output the updated model parameters 、 、 .

10. The method for quantitative analysis of cultivated land occupation and compensation balance based on soil environmental perception according to claim 9, characterized in that: The predicted value of the adaptive yield-ecological prediction model is expressed as: 。

Images