Multi-level Bayesian data assimilation method and system for crop yield estimation

Abstract

This invention provides a multi-level Bayesian data assimilation method and system for crop yield estimation, belonging to the field of yield prediction. The method includes: calibrating a crop growth model based on remote sensing LAI time-series data of the region to be estimated and measured crop yield data within a preset range of the region to be estimated, obtaining a posterior sample set of parameters for the crop growth model; determining an initial forecast set for the crop growth model through sampling; using a two-step Bayesian inference algorithm and the LAI remote sensing observation set to determine optimized uncertain parameters, optimize the crop growth model, and optimize the time-series data; using the optimized time-series data to drive further optimization of the crop growth model, obtaining a posterior forecast set; and estimating crop yield in the region to be estimated based on the posterior forecast set. This invention can improve the accuracy of crop yield estimation at the regional scale.

Description

Technical Field

[0001] This invention relates to the field of yield prediction technology, and in particular to a multi-level Bayesian data assimilation method and system for crop yield estimation. Background Technology

[0002] Satellite remote sensing technology provides a reliable basis for regional crop yield estimation. Methods for yield estimation using remote sensing data can be divided into empirical model-based methods and data assimilation-based methods. The accuracy of yield estimation based on empirical model-based methods depends on a large number of high-quality yield samples, which are often difficult to obtain in practical applications. Data assimilation technology, on the other hand, integrates the advantages of remote sensing data and crop growth models, making it the most promising method for regional crop yield estimation.

[0003] However, existing data assimilation yield estimation methods, such as recalibration methods, which construct cost functions for remote sensing observations and crop growth models for each yield estimation unit and use intelligent optimization algorithms to optimize the input parameters of the crop growth model, cannot correct the simulation biases of LAI (leaf area index) and yield caused by structural errors in the crop growth model. Ensemble Kalman filtering algorithms update the state variables of the crop growth model based on remote sensing observation data during the crop growth period, but cannot correct the yield simulation errors caused by parameter biases in the crop growth model. In the regional-scale data assimilation yield estimation problem, due to the high spatial heterogeneity of crop growth environment and management schemes, neither of these two data assimilation methods can fully correct the simulation errors of the crop growth model, resulting in large estimation errors. Summary of the Invention

[0004] This invention provides a multi-level Bayesian data assimilation method and system for crop yield estimation, which addresses the shortcomings of existing crop growth models in terms of large estimation errors.

[0005] This invention provides a multi-level Bayesian data assimilation method for crop yield estimation, comprising:

[0006] Obtain the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated;

[0007] The crop growth model is calibrated based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated, so as to obtain the posterior sample set of the parameters of the crop growth model.

[0008] The posterior sample set of the parameters is sampled, and the initial forecast set of the crop growth model is determined based on the sampling results.

[0009] The initial forecast set is optimized using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters. These optimized uncertain parameters are then substituted into the crop growth model to obtain an optimized crop growth model. Based on this optimized model, simulated LAI time-series data for the crop are determined. The simulated LAI time-series data is further optimized using the Bayesian two-step inference algorithm and the LAI remote sensing observation set to obtain optimized time-series data. The LAI remote sensing observation set is determined based on the crop remote sensing LAI time-series data for the region to be estimated.

[0010] The optimized time series data is used to drive the optimized crop growth model to obtain a posterior prediction set, which includes multiple sets of parameters of the optimized crop growth model.

[0011] Crop yield is estimated for the region to be estimated based on the posterior forecast set.

[0012] Optionally, obtain time-series remote sensing LAI data of the crop in the area to be estimated, including:

[0013] Obtain Sentinel 2 reflectance data during the crop growth period in the area to be estimated;

[0014] Based on the Gaussian process regression model, the reflectance data of Sentinel-2 was inverted to obtain the time series data of crop remote sensing LAI in the area to be estimated.

[0015] Optionally, the crop growth model includes one of the WOFOST model, DSSAT model, and APSIM model.

[0016] Optionally, crop yield estimation for the region to be estimated is performed based on the posterior forecast set, including:

[0017] Multiple sets of optimized crop growth models are determined based on multiple sets of parameters of the optimized crop growth model;

[0018] The crop yield of the region to be estimated is estimated using each set of optimized crop growth models to obtain multiple sets of yield estimates.

[0019] The average value of the multiple yield estimates is determined to obtain the crop yield estimate result for the area to be estimated.

[0020] Optionally, the Sentinel-2 reflectivity data includes B2, B3, B4, B5, B6, B7, B8, B8A, B11, and B12 band data from Sentinel-2's Level-2A data.

[0021] Optionally, the crop growth model is calibrated based on the crop remote sensing LAI time series data of the area to be yielded and the measured crop yield data within a preset range of the area to be yielded, including:

[0022] A log-likelihood function is constructed based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated.

[0023] Based on the log-likelihood function, the parameter space of the crop growth model is sampled using the Markov chain Monte Carlo simulation method to obtain the posterior sample set of the crop growth model parameters.

[0024] Optionally, the LAI remote sensing observation set is determined in the following way:

[0025] Gaussian perturbation is added to the crop remote sensing LAI time series data of the area to be estimated for each time node to obtain the LAI remote sensing observation set; the mean of the Gaussian perturbation is 0, and the standard deviation of the Gaussian perturbation is the standard deviation of the crop remote sensing LAI data at the corresponding time node.

[0026] This invention also provides a multi-level Bayesian data assimilation crop yield estimation system, comprising:

[0027] The data acquisition module is used to acquire the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated.

[0028] The parameter posterior sample set determination module is used to calibrate the crop growth model based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated, so as to obtain the parameter posterior sample set of the crop growth model.

[0029] The initial forecast set determination module is used to sample the parameter posterior sample set and determine the initial forecast set of the crop growth model based on the sampling results.

[0030] The Bayesian optimization module is used to optimize the initial forecast set using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters; substitute the optimized uncertain parameters into the crop growth model to obtain an optimized crop growth model, and determine the crop simulated LAI time series data based on the optimized crop growth model; optimize the crop simulated LAI time series data using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized time series data; the LAI remote sensing observation set is determined based on the crop remote sensing LAI time series data of the area to be estimated.

[0031] The posterior forecast set determination module is used to drive the optimized crop growth model using the optimized time series data to obtain the posterior forecast set, which includes multiple sets of parameters of the optimized crop growth model.

[0032] The crop yield estimation module is used to estimate the crop yield of the area to be estimated based on the posterior forecast set.

[0033] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the multi-level Bayesian data assimilation crop yield estimation method.

[0034] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the multi-level Bayesian data assimilation crop yield estimation method.

[0035] This invention provides a multi-level Bayesian data assimilation method and system for crop yield estimation. It involves acquiring remote sensing LAI time-series data of the crop yield to be estimated area and measured crop yield data within a preset range of the area; calibrating a crop growth model based on the LAI time-series data and the measured crop yield data to obtain a posterior sample set of the crop growth model's parameters; sampling the posterior sample set to determine the initial forecast set of the crop growth model; and optimizing the initial forecast set using a two-step Bayesian inference algorithm and the LAI remote sensing observation set to obtain... The optimized uncertain parameters are used to optimize the crop remote sensing LAI time series data using a Bayesian two-step inference algorithm and a LAI remote sensing observation set, resulting in optimized time series data. The LAI remote sensing observation set is determined based on the crop remote sensing LAI time series data of the region to be yielded. The optimized uncertain parameters are then substituted into the crop growth model to obtain an optimized crop growth model. The optimized time series data is used to drive the optimized crop growth model, resulting in a posterior prediction set, which includes multiple sets of parameters from the optimized crop growth model. Crop yield is estimated for the region to be yielded based on the posterior prediction set. This invention, based on a Bayesian two-step inference algorithm, corrects uncertain parameters and structural errors in the crop growth model, improving the accuracy of crop yield estimation at the regional scale. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0037] Figure 1 This is a flowchart of the multi-level Bayesian data assimilation method for crop yield estimation provided by the present invention;

[0038] Figure 2 This is a flowchart illustrating the Bayesian two-step inference algorithm provided by the present invention;

[0039] Figure 3 This is a graph showing the yield estimation results based on data assimilation provided by the present invention;

[0040] Figure 4 This is a schematic diagram of the multi-level Bayesian data assimilation crop yield estimation system provided by the present invention;

[0041] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0043] This invention is mainly used to solve the problem that existing four-dimensional variational and ensemble Kalman filtering algorithms cannot fully correct the parameter bias and structural error of crop growth models in crop yield estimation applications at the plot or regional scale, resulting in low yield estimation accuracy.

[0044] The following is combined Figures 1-5 This invention describes a multi-level Bayesian data assimilation method and system for crop yield estimation.

[0045] Figure 1 This is a flowchart of the multi-level Bayesian data assimilation method for crop yield estimation provided by the present invention, as follows: Figure 1 As shown, a multi-level Bayesian data assimilation method for crop yield estimation includes:

[0046] Step 101: Obtain the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within the preset range of the area to be estimated.

[0047] The crop can be one of wheat, rice, or corn. The preset range of the area to be estimated can be a portion of the area to be estimated, or a portion of the area near the area to be estimated.

[0048] In one specific embodiment, acquiring the crop remote sensing LAI time series data (hereinafter referred to as remote sensing LAI time series) of the area to be estimated includes:

[0049] Obtain Sentinel 2 reflectance data during the crop growth period in the area to be estimated;

[0050] Based on the Gaussian process regression model, the reflectance data of Sentinel-2 was inverted to obtain the time series data of crop remote sensing LAI in the area to be estimated.

[0051] Specifically, based on the Gaussian process regression model, the Sentinel-2 reflectance data corresponding to the measured LAI data at the site is used as the training sample. The Sentinel-2 reflectance data during the critical growth period of crops in the area to be estimated is inverted into LAI, thus obtaining the time series data of crop remote sensing LAI during the critical growth period of crops in the area to be estimated.

[0052] Furthermore, the specific formula used to invert the Sentinel-2 reflectivity data into LAI time series data is as follows:

[0053]

[0054] Where, μ * and ∑ * These are the predicted LAI and its standard deviation for the sample points; Y is the LAI for the training sample points. Gaussian noise is artificially added to the training samples; I is the identity matrix; x * x represents the Sentinel 2 reflectivity data for the predicted and training sample points, respectively; K() is the kernel function of the Gaussian process regression model.

[0055] The kernel function chosen for this Gaussian process is the Matérn3 / 2 kernel function:

[0056]

[0057] Where, σ 2 Both σ and l are hyperparameters of the kernel function, and σ is a hyperparameter of the kernel function. 2 The maximum value of the covariance is controlled by l, which controls the length scale parameter of the distance between the reflectance data of the sample points; ‖xx′‖ represents the Euclidean distance between sample points x and x′.

[0058] Furthermore, the Sentinel-2 reflectivity data includes B2, B3, B4, B5, B6, B7, B8, B8A, B11, and B12 band data from Sentinel-2's Level-2A data.

[0059] Step 102: Based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within the preset range of the area to be estimated, the crop growth model is calibrated to obtain the posterior sample set of the parameters of the crop growth model.

[0060] In one specific embodiment, the crop growth model is any crop growth model capable of performing daily-scale simulations of the LAI time series and yield of crop growth. The crop growth model includes any one of the following: WOFOST (WOrdFood STudies) model, DSSAT (Decision Support Systems for Agrotechnology Transfer) model, and APSIM (Agricultural Production Systems sIMulator) model.

[0061] In one specific embodiment, step 102 involves calibrating the crop growth model using the Markov chain Monte Carlo simulation method.

[0062] In one specific embodiment, step 102 includes:

[0063] Step 1021: Construct a log-likelihood function based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated.

[0064] The specific formula used is as follows:

[0065]

[0066] Among them, T1 and T n These are the time points of the first and last available remote sensing LAI observations during the critical growth period of crops; and These are the LAI of the crop at time t, simulated by the crop growth model and obtained by inversion from Sentinel-2 remote sensing reflectance data; Y is the standard deviation of the crop's LAI at time t, obtained from the Sentinel 2 remote sensing reflectance data; sim and Y obs These are crop yields simulated by crop growth models and those measured in actual experiments; σ y It is the standard deviation of the measured crop yield.

[0067] Step 1022: Based on the log-likelihood function, use the Markov chain Monte Carlo simulation method to sample the parameter space of the crop growth model to obtain the posterior sample set of the parameters of the crop growth model.

[0068] Specifically, the parameter space of the crop growth model is sampled using the MCMC (Markov Chain Monte Carlo Simulation) method based on the constructed log-likelihood function. After the Markov chain converges, the sampling is continued for 500 times to obtain the posterior sample set of the crop growth model parameters.

[0069] Step 103: Sampling is performed on the posterior sample set of the parameters, and the initial forecast set of the crop growth model is determined based on the sampling results.

[0070] In a specific embodiment, step 103 involves extracting 50 parameter samples at equal intervals from the parameter posterior sample set and substituting these 50 parameter samples into the crop growth model to generate an initial forecast set with 50 members.

[0071] It should be noted that there is no limitation on the number of samplings or the number of members in this invention; only preferred options are described here.

[0072] Step 104: Optimize the initial forecast set using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters; substitute the optimized uncertain parameters into the crop growth model to obtain an optimized crop growth model, and determine the crop simulated LAI time series data based on the optimized crop growth model; optimize the crop simulated LAI time series data using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized time series data; the LAI remote sensing observation set is determined based on the crop remote sensing LAI time series data of the area to be estimated.

[0073] Uncertain parameters refer to crop growth model parameters that exhibit high spatial heterogeneity, including parameters such as initial dry matter weight, accumulated temperature from seedling emergence to flowering, and leaf life at 35°C in the WOFOST model.

[0074] In a specific embodiment, the LAI remote sensing observation set is determined in the following manner:

[0075] Gaussian perturbation is added to the crop remote sensing LAI time series data of the area to be estimated for each time node (each observation time t) to obtain the LAI remote sensing observation set; the mean of the Gaussian perturbation is 0, and the standard deviation of the Gaussian perturbation is the standard deviation of the crop remote sensing LAI data at the corresponding time node.

[0076] Correspondingly, the LAI remote sensing observation set includes 50 members.

[0077] In a specific embodiment, the principle of the Bayesian two-step inference algorithm is as follows: Figure 2 As shown, it specifically includes:

[0078] The uncertain parameters of each member in the initial forecast set are updated sequentially using the LAI remote sensing observation set (hereinafter referred to as the LAI observation set), and the updated uncertain parameters are then substituted back into the crop growth model to obtain the optimized uncertain parameters. The specific formula for parameter update is as follows:

[0079]

[0080] Where, θ a,i and θ f,i These are the optimized and unoptimized uncertain parameters of the i-th member in the forecast set, respectively. When optimizing the uncertain parameters using the LAI observation set at the first time step, θ f,i The uncertain parameter corresponds to the i-th member of the initial forecast set; otherwise, θ f,i It is the uncertain parameter of the i-th member of the forecast set optimized using remote sensing LAI observations from the previous moment; and K represents the LAI value of the i-th member in the observation set at time t and the simulated LAI value of the i-th member in the forecast set at time t, respectively. θ,t This is the Kalman gain used for updating uncertain parameters at time t, and the specific formula is as follows:

[0081]

[0082] in, It is the covariance of the uncertain parameters in the forecast set and the simulated LAI at time t. R is the variance of the LAI at time t simulated in the forecast set. t It is the variance of LAI in the observation set at time t.

[0083] Then, the optimized uncertain parameters are substituted into the crop growth model to obtain the optimized crop growth model, and the crop simulated LAI time series data are determined based on the optimized crop growth model.

[0084] Based on the optimized crop growth model obtained above, and then based on the LAI remote sensing observation set, the crop simulation LAI time series data is optimized using the Bayesian two-step inference algorithm set Kalman filter algorithm to obtain the optimized time series data.

[0085] This invention corrects parameter biases and structural errors in crop growth models when applied to simulate crop yields at the plot or regional scale by employing the aforementioned Bayesian two-step inference algorithm, thereby improving the accuracy of crop yield prediction at the plot or regional scale.

[0086] Furthermore, the uncertain parameters of the crop model are crop growth model parameters with high spatial heterogeneity, including parameters such as the initial dry matter weight of the WOFOST model, the accumulated temperature from seedling emergence to flowering period, and leaf life at 35°C.

[0087] Step 105: Use the optimized time series data to drive the optimized crop growth model to obtain a posterior prediction set, which includes multiple sets of parameters of the optimized crop growth model.

[0088] Specifically, the number of parameter sets in the optimized crop growth model in the posterior forecast set is the same as the number in the initial forecast set. Therefore, the posterior forecast set includes 50 sets of parameters from the optimized crop growth model.

[0089] Step 106: Estimate crop yield for the region to be estimated based on the posterior forecast set.

[0090] In one specific embodiment, step 106 includes:

[0091] Multiple sets of optimized crop growth models are determined based on multiple sets of parameters of the optimized crop growth model;

[0092] The crop yield of the region to be estimated is estimated using each set of optimized crop growth models to obtain multiple sets of yield estimates.

[0093] The average value of the multiple yield estimates is determined to obtain the crop yield estimate result for the area to be estimated.

[0094] According to the above-described scheme of the present invention, the present invention uses the WOFOST model to predict the yield of wheat in a certain region, and the results are as follows. Figure 3 As shown.

[0095] The multi-level Bayesian data assimilation crop yield estimation system provided by this invention is described below. The multi-level Bayesian data assimilation crop yield estimation system described below can be referred to in correspondence with the multi-level Bayesian data assimilation crop yield estimation method described above.

[0096] like Figure 4 As shown, a multi-level Bayesian data assimilation crop yield estimation system includes:

[0097] The data acquisition module 401 is used to acquire the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated.

[0098] In one specific embodiment, acquiring the crop remote sensing LAI time series data of the area to be yielded includes:

[0099] Obtain Sentinel 2 reflectance data during the crop growth period in the area to be estimated;

[0100] Based on the Gaussian process regression model, the reflectance data of Sentinel-2 was inverted to obtain the time series data of crop remote sensing LAI in the area to be estimated.

[0101] In one specific embodiment, the Sentinel-2 reflectivity data includes B2, B3, B4, B5, B6, B7, B8, B8A, B11, and B12 band data from Sentinel-2's Level-2A data.

[0102] The parameter posterior sample set determination module 402 is used to calibrate the crop growth model based on the crop remote sensing LAI time series data of the yield area to be estimated and the measured crop yield data within a preset range of the yield area to be estimated, so as to obtain the parameter posterior sample set of the crop growth model.

[0103] In one specific embodiment, the crop growth model includes one of the WOFOST model, the DSSAT model, and the APSIM model.

[0104] In a specific embodiment, the parameter posterior sample set determination module 402 is specifically used for:

[0105] A log-likelihood function is constructed based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated.

[0106] Based on the log-likelihood function, the parameter space of the crop growth model is sampled using the Markov chain Monte Carlo simulation method to obtain the posterior sample set of the crop growth model parameters.

[0107] The initial forecast set determination module 403 is used to sample the parameter posterior sample set and determine the initial forecast set of the crop growth model based on the sampling results.

[0108] The Bayesian optimization module 404 is used to optimize the initial forecast set using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters; substitute the optimized uncertain parameters into the crop growth model to obtain an optimized crop growth model, and determine the crop simulated LAI time series data based on the optimized crop growth model; optimize the crop simulated LAI time series data using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized time series data; the LAI remote sensing observation set is determined based on the crop remote sensing LAI time series data of the area to be estimated.

[0109] In a specific embodiment, the LAI remote sensing observation set is determined in the following manner:

[0110] Gaussian perturbation is added to the crop remote sensing LAI time series data of the area to be estimated for each time node to obtain the LAI remote sensing observation set; the mean of the Gaussian perturbation is 0, and the standard deviation of the Gaussian perturbation is the standard deviation of the crop remote sensing LAI data at the corresponding time node.

[0111] The posterior prediction set determination module 405 is used to drive the optimized crop growth model using the optimized time series data to obtain a posterior prediction set, wherein the posterior prediction set includes multiple sets of parameters of the optimized crop growth model.

[0112] The crop yield estimation module 406 is used to estimate the crop yield of the area to be estimated based on the posterior forecast set.

[0113] In one specific embodiment, the crop yield estimation module 406 is specifically used for:

[0114] Multiple sets of optimized crop growth models are determined based on multiple sets of parameters of the optimized crop growth model;

[0115] The crop yield of the region to be estimated is estimated using each set of optimized crop growth models to obtain multiple sets of yield estimates.

[0116] The average value of the multiple yield estimates is determined to obtain the crop yield estimate result for the area to be estimated.

[0117] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a multi-level Bayesian data assimilation crop yield estimation method, which includes:

[0118] Obtain the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated.

[0119] The crop growth model is calibrated based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated, so as to obtain the posterior sample set of the parameters of the crop growth model.

[0120] The posterior sample set of the parameters is sampled, and the initial forecast set of the crop growth model is determined based on the sampling results.

[0121] The initial forecast set is optimized using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters. These optimized uncertain parameters are then substituted into the crop growth model to obtain an optimized crop growth model. Based on this optimized model, simulated LAI time-series data for the crop are determined. The simulated LAI time-series data is further optimized using the Bayesian two-step inference algorithm and the LAI remote sensing observation set to obtain optimized time-series data. The LAI remote sensing observation set is determined based on the crop remote sensing LAI time-series data for the region to be estimated.

[0122] The optimized time series data is used to drive the optimized crop growth model to obtain a posterior forecast set, which includes multiple sets of parameters of the optimized crop growth model.

[0123] Crop yield is estimated for the region to be estimated based on the posterior forecast set.

[0124] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is capable of executing a multi-level Bayesian data assimilation crop yield estimation method, the method comprising:

[0126] Obtain the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated.

[0127] The crop growth model is calibrated based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated, so as to obtain the posterior sample set of the parameters of the crop growth model.

[0128] The posterior sample set of the parameters is sampled, and the initial forecast set of the crop growth model is determined based on the sampling results.

[0129] The initial forecast set is optimized using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters. These optimized uncertain parameters are then substituted into the crop growth model to obtain an optimized crop growth model. Based on this optimized model, simulated LAI time-series data for the crop are determined. The simulated LAI time-series data is further optimized using the Bayesian two-step inference algorithm and the LAI remote sensing observation set to obtain optimized time-series data. The LAI remote sensing observation set is determined based on the crop remote sensing LAI time-series data for the region to be estimated.

[0130] The optimized time series data is used to drive the optimized crop growth model to obtain a posterior forecast set, which includes multiple sets of parameters of the optimized crop growth model.

[0131] Crop yield is estimated for the region to be estimated based on the posterior forecast set.

[0132] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a multi-level Bayesian data assimilation crop yield estimation method, the method comprising:

[0133] Obtain the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated.

[0134] The crop growth model is calibrated based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated, so as to obtain the posterior sample set of the parameters of the crop growth model.

[0135] The posterior sample set of the parameters is sampled, and the initial forecast set of the crop growth model is determined based on the sampling results.

[0136] The initial forecast set is optimized using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters. These optimized uncertain parameters are then substituted into the crop growth model to obtain an optimized crop growth model. Based on this optimized model, simulated LAI time-series data for the crop are determined. The simulated LAI time-series data is further optimized using the Bayesian two-step inference algorithm and the LAI remote sensing observation set to obtain optimized time-series data. The LAI remote sensing observation set is determined based on the crop remote sensing LAI time-series data for the region to be estimated.

[0137] The optimized time series data is used to drive the optimized crop growth model to obtain a posterior forecast set, which includes multiple sets of parameters of the optimized crop growth model.

[0138] Crop yield is estimated for the region to be estimated based on the posterior forecast set.

[0139] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0140] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-level Bayesian data assimilation crop yield estimation method, characterized in that, include: Obtain the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated; The crop growth model is calibrated based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated, so as to obtain the posterior sample set of the parameters of the crop growth model. The posterior sample set of the parameters is sampled, and the initial forecast set of the crop growth model is determined based on the sampling results. The initial forecast set is optimized using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters. These optimized uncertain parameters are then substituted into the crop growth model to obtain an optimized crop growth model. Based on this optimized model, simulated LAI time-series data for the crop are determined. The simulated LAI time-series data is further optimized using the Bayesian two-step inference algorithm and the LAI remote sensing observation set to obtain optimized time-series data. The LAI remote sensing observation set is determined based on the crop remote sensing LAI time-series data for the region to be estimated. The optimized time series data is used to drive the optimized crop growth model to obtain a posterior prediction set, which includes multiple sets of parameters of the optimized crop growth model. Crop yield is estimated for the region to be estimated based on the posterior forecast set.

2. The multi-level Bayesian data assimilation crop yield estimation method of claim 1, wherein, The acquisition of crop remote sensing LAI time series data for the area to be estimated includes: Obtain Sentinel-2 reflectance data during the crop growth period in the area to be estimated; Based on the Gaussian process regression model, the reflectance data of Sentinel-2 was inverted to obtain the time series data of crop remote sensing LAI in the area to be estimated.

3. The multi-level Bayesian data assimilation crop yield estimation method according to claim 1 or 2, characterized in that, The crop growth model includes one of the WOFOST model, DSSAT model, and APSIM model.

4. The multi-level Bayesian data assimilation method for crop yield estimation according to claim 1 or 2, characterized in that, Based on the posterior forecast set, crop yield estimation is performed on the region to be estimated, including: Multiple sets of optimized crop growth models are determined based on multiple sets of parameters of the optimized crop growth model; The crop yield of the region to be estimated is estimated using each set of optimized crop growth models to obtain multiple sets of yield estimates. The average value of the multiple yield estimates is determined to obtain the crop yield estimate result for the area to be estimated.

5. The multi-level Bayesian data assimilation method for crop yield estimation according to claim 2, characterized in that, The Sentinel-2 reflectivity data includes Level-2A data from Sentinel-2 in bands B2, B3, B4, B5, B6, B7, B8, B8A, B11, and B12.

6. The multi-level Bayesian data assimilation method for crop yield estimation according to claim 1 or 2, characterized in that, The crop growth model is calibrated based on the crop remote sensing LAI time series data of the area to be yielded and the measured crop yield data within a preset range of the area to be yielded, including: A log-likelihood function is constructed based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated. Based on the log-likelihood function, the parameter space of the crop growth model is sampled using the Markov chain Monte Carlo simulation method to obtain the posterior sample set of the crop growth model parameters.

7. The multi-level Bayesian data assimilation method for crop yield estimation according to claim 1 or 2, characterized in that, The LAI remote sensing observation set is determined in the following way: Gaussian perturbation is added to the crop remote sensing LAI time series data of the area to be estimated for each time node to obtain the LAI remote sensing observation set; the mean of the Gaussian perturbation is 0, and the standard deviation of the Gaussian perturbation is the standard deviation of the crop remote sensing LAI at the corresponding time node.

8. A multi-level Bayesian data assimilation crop yield estimation system, characterized in that, include: The data acquisition module is used to acquire the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated. The parameter posterior sample set determination module is used to calibrate the crop growth model based on the crop remote sensing LAI time series data of the area to be estimated and the measured crop yield data within a preset range of the area to be estimated, so as to obtain the parameter posterior sample set of the crop growth model. The initial forecast set determination module is used to sample the parameter posterior sample set and determine the initial forecast set of the crop growth model based on the sampling results. The initial forecast set is optimized using a Bayesian two-step inference algorithm and a LAI remote sensing observation set to obtain optimized uncertain parameters. These optimized uncertain parameters are then substituted into the crop growth model to obtain an optimized crop growth model. Based on this optimized model, simulated LAI time-series data for the crop are determined. The simulated LAI time-series data is further optimized using the Bayesian two-step inference algorithm and the LAI remote sensing observation set to obtain optimized time-series data. The LAI remote sensing observation set is determined based on the crop remote sensing LAI time-series data for the region to be estimated. The posterior forecast set determination module is used to drive the optimized crop growth model using the optimized time series data to obtain the posterior forecast set, which includes multiple sets of parameters of the optimized crop growth model. The crop yield estimation module is used to estimate the crop yield of the area to be estimated based on the posterior forecast set.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the multi-level Bayesian data assimilation crop yield estimation method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-level Bayesian data assimilation crop yield estimation method as described in any one of claims 1 to 7.

Images