EnKF algorithm and total photosynthetic area index-based assimilation yield estimation method and system

Abstract

The invention belongs to the technical field of agricultural remote sensing, and provides an assimilation yield estimation method based on an EnKF algorithm and a total photosynthetic area index (TPAI), which comprises the following steps: S1, calculating the TPAI which is LAI in a vegetative growth period; the TPAI is the sum of LAI and NFGOAI in the development stage of the non-leaf green organ, and the TPAI is NFGOAI when the non-leaf green organ is in the early stage of grain filling or maturation; s2, calibrating a crop growth model by using TPAI; s3, utilizing the remote sensing data TPAI; and S4, assimilating the inverted TPAI into a crop growth model based on an EnKF algorithm, simulating crop growth and estimating the yield. The invention also correspondingly provides an assimilation estimation system. According to the crop assimilation yield estimation method, the mechanical property and the expansibility of a crop assimilation yield estimation system are improved, the assimilation TPAI can inhibit EnKF filtering divergence, and the crop assimilation yield estimation precision is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of agricultural remote sensing, and particularly relates to an assimilation yield estimation method and system based on the EnKF algorithm and the total photosynthetic area index. Background Art

[0002] Crop assimilation yield estimation is to assimilate the crop canopy or environmental parameters retrieved by remote sensing into the crop growth model to achieve dynamic simulation of the crop growth state and yield estimation. The crop growth model developed based on agronomic knowledge and computer technology can quantitatively describe the dynamic processes such as crop growth, development, and grain formation, and achieve efficient and rapid single-yield simulation and prediction. However, the growth states of field crops and field management measures have regional heterogeneity, making it difficult to accurately obtain model parameters when the crop model is applied to regional crop yield estimation, and the yield estimation accuracy is relatively low. Satellite remote sensing can monitor the state of the regional crop canopy. Introducing remote sensing information into the crop model for data assimilation can not only reveal the internal mechanism of crop growth, development, and yield formation at the single-point scale, but also solve the problems of model parameter acquisition and regionalization, improve the crop model's yield estimation ability, and achieve the spatio-temporal expansion of remote sensing inversion and crop model simulation.

[0003] In the crop model remote sensing data assimilation yield estimation system, the assimilation variables that bridge the crop model and remote sensing observations mainly include canopy parameters (leaf area index, evapotranspiration) and carbon flux-related parameters (gross primary productivity, net primary productivity), etc. Among them, the most widely used and best-yield-estimation-effect assimilation variable is the crop canopy parameter, which is mainly applied to staple crops such as wheat and corn, as well as a few cash crops such as wolfberry and sugarcane. On the one hand, the leaf area index directly affects the amount of light energy intercepted by the population and photosynthetic productivity, and determines the crop growth status and yield formation. On the other hand, the dry matter accumulation of crops such as wheat, rice, and corn mainly depends on the photosynthesis of plant leaves, and there is a significant correlation between the leaf area index at the flowering stage and the crop yield. For these crops, using the leaf area index as the assimilation variable to bridge the model and observations can accurately estimate the crop yield.

[0004] However, leaves are not the only organs in crops that can carry out photosynthesis. The photosynthesis of non-leaf green organs such as siliques, pods, stems, and petioles in crops all contribute to the formation of their yields. Among non-leaf green organs, Brassica napus plants have the characteristics of photosynthetic organ succession and the most active non-leaf green organs. The siliques, which are non-leaf green organs, play a crucial role in yield formation. Research shows that approximately 30% of the rapeseed yield comes from leaf photosynthesis, and 70% comes from silique wall photosynthesis. Different from crops such as wheat, rice, and corn, rapeseed has a special process of photosynthetic organ succession. From the seedling stage to the flowering stage, leaves are the main canopy components of rapeseed; after rapeseed blooms, leaves and siliques together serve as the canopy components of rapeseed and carry out plant photosynthesis together; after the silique stage, rapeseed mainly relies on silique wall photosynthesis to fill the seeds, and rapeseed siliques become the main canopy components. Therefore, the existing research on the assimilation and yield estimation system for crops with active non-leaf green organs using LAI lacks the contribution of non-leaf tissue photosynthesis to yield, cannot describe the changes in photosynthesis during the crop growth period, and cannot accurately estimate crop yields.

[0005] The assimilation algorithm is the key method of the crop assimilation and yield estimation system, which determines whether the observed data can be "reasonably" unfolded as much as possible in the time and space scales and affects the accuracy of the yield estimation results. According to the principle, the assimilation algorithms used in the assimilation system can be divided into two categories - the ensemble filtering method based on estimation theory and the parameter optimization method based on cost function. The EnKF algorithm is a typical representative of the ensemble filtering algorithm, and has achieved many research results in the research on the assimilation of crop growth models and remote sensing data in recent years. There is an existing assimilation framework based on the EnKF algorithm in the prior art. Based on common crop growth models, it combines optical and radar remote sensing data to simulate the yields of wheat, corn, rice, soybean, sugarcane, and other crops, and is used to assimilate SM and CC into the AquaCrop model, and estimates corn yield by integrating field information such as planting date and planting density. There is also prior art that uses the EnKF algorithm to combine three assimilation variables (LAI, SM, and SIF) to achieve the estimation of corn and soybean pairs reflecting crop yields. There is also prior art that couples the WheatGrow and PROSAIL models and combines these models with HJ-CCD and SPOT data to estimate regional wheat yields. However, the assimilation system framework based on the EnKF algorithm lacks the consideration of the spatial heterogeneity of crop yields, resulting in the filtering divergence problem of the EnKF algorithm. Especially when the object of yield estimation is a crop with active non-leaf green organs, but using LAI as the state variable, LAI cannot accurately represent the contribution of non-leaf green organs to yield, which makes the error accumulation of the simulation results exacerbate the filtering divergence of the EnKF algorithm, which may lead to a sharp decline in the reliability of the yield estimation results.

[0006] In summary, on the one hand, existing assimilation yield estimation systems mostly use LAI as a bridging parameter, lacking consideration of the active crop agronomic mechanism of non-leaf green organs and the contribution of non-leaf tissue photosynthesis to yield. On the other hand, the assimilation yield estimation system with LAI as the state variable will lead to the accumulation of simulation state errors of crops with active non-leaf green organs, exacerbate the filtering divergence of the EnKF algorithm, and reduce the accuracy of yield estimation. Summary of the Invention

[0007] In view of the problems in the background technology, the present invention proposes an assimilation yield estimation method and system that uses the Total Photosynthetic Area Index (TPAI) as a bridging parameter, the EnKF algorithm as an assimilation algorithm, and couples a crop growth model and remotely sensed inversion of TPAI. First, calculate TPAI through the photosynthetic organs of different growth stages of the crop; use TPAI to calibrate the crop growth model and invert TPAI using remote sensing data; finally, assimilate the inverted TPAI into the crop growth model based on the EnKF algorithm to achieve crop growth simulation and yield estimation. Based on the above assimilation yield estimation method, the present invention aims to achieve high-precision yield estimation of crops with non-leaf green organs and provide ideas for yield estimation of crops with special plant structures or photosynthesis in the future.

[0008] Specifically, the present invention proposes an assimilation yield estimation method based on the EnKF algorithm and the photosynthetic area index, including: S1, calculating the total photosynthetic area index TPAI, where during the vegetative growth period, TPAI is the leaf area index LAI; during the development stage of non-leaf green organs, TPAI is the sum of the leaf area index LAI and the non-leaf green organ area index NFGOAI, and when the non-leaf green organs are in the grain filling or early maturity stage, TPAI is the non-leaf green organ area index NFGOAI; S2, using the total photosynthetic area index TPAI to calibrate the crop growth model; S3, using remote sensing data to invert the total photosynthetic area index TPAI; S4, assimilating the inverted total photosynthetic area index TPAI into the crop growth model based on the EnKF algorithm to simulate crop growth and estimate yield.

[0009] The present invention also proposes an assimilation yield estimation system based on the EnKF algorithm and the photosynthetic area index, which includes a computer-executable program that, when executed, implements the method of the present invention as described above.

[0010] The present invention fully considers the photosynthesis succession of non-leaf green tissues and their contribution to yield, proposes the bridging parameter TPAI for the succession of crop photosynthetic organs, and connects the various components of the crop assimilation yield estimation system with TPAI as the bridging parameter. The present invention improves the mechanism and expansibility of the crop assimilation yield estimation system, and the assimilated TPAI adopted can inhibit the EnKF filtering divergence and improve the accuracy of crop assimilation yield estimation. Description of the Drawings

[0011] For a better understanding of the present invention, the present invention will be described in more detail by referring to the specific embodiments shown in the drawings. These drawings only depict typical embodiments of the present invention and should not be considered as limiting the scope of protection of the present invention.

[0012] Figure 1 It is a flowchart of the method of the present invention.

[0013] Figure 2 It is a technical roadmap of an embodiment of the method of the present invention.

[0014] Figure 3 It is a schematic diagram for calculating TPAI at different stages of non-leaf green organ active crops.

[0015] Figure 4 It is the time series curve of the photosynthetic area index before assimilating LAI.

[0016] Figure 5 It is the time series curve of the photosynthetic area index after assimilating LAI.

[0017] Figure 6 It is the time series curve of the photosynthetic area index before assimilating TPAI.

[0018] Figure 7 It is the time series curve of the photosynthetic area index after assimilating TPAI.

[0019] Figure 8 It is the time series curve of LAI-TWSO yield before assimilating LAI.

[0020] Figure 9 It is the time series curve of LAI-TAGP yield before assimilating LAI.

[0021] Figure 10 It is the time series curve of LAI-TWSO yield after assimilating LAI.

[0022] Figure 11 It is the time series curve of LAI-TAGP yield after assimilating LAI.

[0023] Figure 12 It is the time series curve of LAI-TWSO yield before assimilating TPAI.

[0024] Figure 13 It is the time series curve of LAI-TAGP yield before assimilating TPAI.

[0025] Figure 14 It is the time series curve of LAI-TWSO yield after assimilating TPAI.

[0026] Figure 15LAI-TAGP production time series curve after assimilating TPAI.

[0027] Figure 16 Results of assimilating LAI into the yield TWSO of rapeseed plants in Hengyang City.

[0028] Figure 17 Results of assimilating LAI into the yield TAGP of rapeseed plants in Hengyang City.

[0029] Figure 18 Results of assimilating TPAI into the yield TWSO of rapeseed plants in Hengyang City.

[0030] Figure 19 Results of assimilating TPAI into the yield TAGP of rapeseed plants in Hengyang City. Specific implementation manners

[0031] The following describes the implementation manners of the present invention with reference to the accompanying drawings, so that those skilled in the art can better understand the present invention and be able to implement it. However, the listed embodiments are not intended to limit the present invention. Without conflict, the following embodiments and the technical features in the embodiments can be combined with each other, and the same components are denoted by the same reference numerals.

[0032] The following refers to Figure 1 to describe an implementation manner of the method of the present invention.

[0033] S1. Calculate the total photosynthetic area index TPAI.

[0034] S2. Use the total photosynthetic area index TPAI to calibrate the crop growth model.

[0035] S3. Invert the total photosynthetic area index TPAI using remote sensing data.

[0036] S4. Based on the EnKF algorithm, assimilate the inverted total photosynthetic area index TPAI into the crop growth model to simulate crop growth and estimate yield.

[0037] The following refers to Figure 2 to describe another implementation manner of the present invention.

[0038] In step S1, calculate TPAI through the photosynthetic organs of the crop at different growth stages. During the entire growth period of gramineous crops, the leaves are the main photosynthetic organs, and the leaf area index of these crops can be regarded as TPAI. Different from crops such as wheat, rice, and corn, crops with active non-leaf green organs have a special process of photosynthetic organ succession, such as Figure 3 shown. Therefore, for plants with active non-leaf green organs, the calculation method of TPAI at different growth stages is different, and there are the following three cases:

[0039] (1) During the vegetative growth stage (before flowering), the leaves are the main source of photosynthesis, and TPAI is the leaf area index LAI at this time.

[0040] TPAI = LAI (1)

[0041] (2) During the gradual development stage of non-leaf green organs, the leaves and non-leaf green organs jointly participate in photosynthesis. For example, during the pod-setting stage of rapeseed, TPAI includes the leaf area index LAI and the non-leaf green organ area index NFGOAI.

[0042] TPAI = LAI + NFGOAI (2)

[0043] (3) When non-leaf green organs become the main contributor to plant photosynthesis during the grain filling or early maturity stage, as the leaves turn yellow and fall off, TPAI at this time is the non-leaf green organ area index NFGOAI of the crop.

[0044] TPAI = NFGOAI (3)

[0045] In step S2, TPAI is used to calibrate the crop growth model, and TPAI is retrieved using remote sensing data. Specifically, step S2 includes steps S21 - S22.

[0046] S21, Calculate the ground measured value of the photosynthetic area index TPAI as the data input for the subsequent single-point scale EnKF assimilation yield estimation.

[0047] S22, Conduct a sensitivity analysis on the crop growth model, and perform parameter calibration steps through the calculated TPAI to obtain the WOFOST model with localized parameters.

[0048] In one embodiment, the TPAI calculation formula is as follows:

[0049] LAI = 0.005 * S leaf +1.318

[0050] NFGOAI = 0.005 * S pods +1.318

[0051] TPAI = LAI + NFGOAI

[0052] Wherein, S leaf and S pods are the leaf area and the surface area of non-leaf green organs respectively.

[0053] The parameter calibration steps include:

[0054] (1) Conduct a sensitivity analysis to screen out the crop model parameters that have a greater impact on the changes in crop TWSO (total dry weight of storage organs) and TAGP (aboveground biomass).

[0055] (2) Combine ground survey data and previous literature to obtain information such as crop sowing, growth cycle, accumulated temperature required for each development stage, and dry matter distribution.

[0056] (3) Adjust parameters such as TDWI, TSUMEM, TSUM1, SPAN, and FLTB so that the simulated curve of TPAI is as close as possible to the measured values at different growth stages of the crop, the peak value and overall change trend of TPAI are consistent with the measured values, and TWSO and TAGP are close to the ground measured values.

[0057] (4) Repeat the operation in step (3) above multiple times to make the effect of parameter adjustment meet certain accuracy requirements.

[0058] S23. Use the calibrated crop growth model to simulate TPAI, and the obtained TPAI is used as the simulated value for EnKF assimilation yield estimation.

[0059] In step S3, use remote sensing data to retrieve TPAI. Step S3 includes S31 - S32.

[0060] S31. Through preprocessing and polarization decomposition of the remote sensing data Sentinel - 1 data, screen the polarization decomposition features.

[0061] Preferably, in the present invention, H / A / Alpha polarization decomposition and three - component polarization decomposition are used. H / A / Alpha polarization decomposition is a non - coherent target decomposition method based on the coherence matrix or covariance matrix. Use the three parameters H, A, and Alpha to describe the full - polarization scattering. The three - component polarization decomposition describes the volume scattering as multiple scattering caused by a layer of randomly oriented particles, and the volume P v , surface P s and dihedral angle scattering P d together constitute the volume scattering.

[0062] Through the comparison of R 2 (Coefficient of determination, which reflects the accuracy of the model fitting data, and the range is 0 - 1) and RMSE (root mean square error) of the inversion results, select R 2Polarization features greater than 0.75 and RMSE less than 1 are used to train the inversion model. After preliminary screening, 11 polarization or backscattering coefficient features related to crop characteristics are finally obtained, which are: vertical-vertical polarization scattering coefficient VV, vertical-horizontal polarization scattering coefficient VH, scattering amplitude differentiation features l1 (λ1 of H / A / Alpha) and l2 (λ2 of H / A / Alpha), polarization information entropy entropy(H), polarization scattering anisotropy anisotropy(A), polarization scattering angle alpha(α), volume scattering component P v , surface scattering component P s .

[0063] S32. Use random forest to rank and select the importance of polarization decomposition features, establish an inversion model and verify it.

[0064] In one embodiment, the feature selection method steps include:

[0065] (1) Conduct importance analysis and ranking of all features through random forest. The more important the feature, the more forward it is, called feature sequence P;

[0066] (2) The current feature set Q is an empty set;

[0067] (3) Take out the most important feature from P and put it into Q. The number of features in P is reduced by 1, and the number of features in Q is increased by 1. Use RF to model the features in Q and analyze the modeling accuracy and speed;

[0068] (4) Loop (3) until P is empty;

[0069] (5) Compare and analyze all the results obtained in the process to obtain the optimal feature set.

[0070] S33. Use the inversion model to invert the TPAI data of the crop research area as the data input for regional-scale EnKF assimilation yield estimation.

[0071] The inversion model is as follows:

[0072] TPAI = M RF (f1, f2, …, f n ) + ε

[0073] where f1, f2, …, f n are the n selected polarization features, M RF is the optimal relationship simulated by random forest, and ε is the inversion error term, representing the random error or noise of the model.

[0074] In step S4, the inverted TPAI is assimilated into the crop growth model based on the EnKF algorithm to simulate crop growth and estimate yield. Step S4 includes S41 - S44. The EnKF algorithm mainly uses the Monte Carlo method to implement ensemble forecasting and ensemble filtering and calculate the error covariance matrix. EnKF mainly includes two parts: forecasting and updating. The steps for yield estimation based on the assimilation of TPAI are as follows.

[0075] As Figure 2 shown, the EnKF assimilation step of the present invention has two inputs, namely the simulated TPAI n and the observed TPAI n , which are respectively reflected in Figure 2 step 3 (simulated TPAI n , called the TPAI prediction ensemble in step 3) and step 1 (observed TPAI n ). In the specific steps, in order to represent the loop process over time t, it is expressed as the state value at a certain moment, such as and other forms. And for the simulated TPAI n and the observed TPAI n above, in order to represent the overall time series of TPAI, the capitalized TPAI n is used to represent, and TPAI n is a set of moments at each growth stage containing tpai (see the formula below). In step 3, there are times t and t + 1, so the concept of "prediction" is used; while in the input of EnKF, it represents the time series curve at each moment, so the word "simulation" is used. Figure 2 The simulated TPAI n and the observed TPAI n in and are respectively represented by

[0076] Simulated

[0077] Observed

[0078] S41, initialization ( Figure 2 step 0 of

[0079] Based on the calibrated crop growth model parameters, the simulated TPAI ensemble is used as the initial state collection of TPAI at (t = 0).

[0080] S42, observation ( Figure 2 step 1 of

[0081] The initial state set of TPAI is run forward in time and simulated. The measured ground values of TPAI (or the regional inversion values of TPAI) are input as TPAI observation values during the critical growth stages of the crop, and the error matrix of the TPAI observation values is calculated.

[0082] S43, update ( Figure 2 Steps 2 and 3): Use the Ensemble Kalman Filter to combine the error covariance matrices of the TPAI observation vector and the state vector, update the state of each ensemble member, and obtain the analysis field ensemble.

[0083] Specifically, first calculate the Kalman gain K t , and then use the state update equation to estimate the current state and update the error covariance of the current estimate.

[0084]

[0085] where K t is the Kalman gain; is the covariance matrix of the TPAI prediction state ensemble member at time t; H t is the observation operator (a linear or non-linear function that maps the state space to the observation space); is the covariance matrix of the TPAI observation state ensemble member at time t; is the updated value of the i-th member of the state variable set of TPAI at time t.

[0086] S44, prediction: Input the ensemble of the initial state variables such as the observed TPAI, crop, meteorological, and soil parameters into the crop growth model to generate the forecast field ensemble of TPAI, which represents the current model state.

[0087]

[0088] where is the predicted value of the i-th member of the state variable set of TPAI at time t; N is the number of TPAI ensemble members; M is the state transition model, which is the WOFOST dynamic model in the present invention.

[0089] S45, calculate the error covariance matrix according to the forecast field ensemble.

[0090]

[0091] where is the expected value of the TPAI prediction state ensemble member.

[0092] S46, calculate the posterior estimate value of the model state as the ensemble mean, that is, the photosynthetic area index of the crop at the growth stage time t.

[0093]

[0094] The present invention also provides an assimilation and yield estimation system based on the EnKF algorithm and the photosynthetic area index, which has a computer-executable program that, when executed, completes the method steps described above.

[0095] The present invention fully considers the photosynthetic succession of non-leaf green tissues and their contribution to yield, and proposes a bridging parameter TPAI for the succession of crop photosynthetic organs, which is used throughout the assimilation and yield estimation method. First, considering the differences between non-leaf green organ-active crops and gramineous crops (such as wheat, rice, and corn) (i.e., the presence of non-leaf photosynthetic tissues causes special photosynthetic changes during different growth stages of crops, and the important contribution of non-leaf green organs to yield), the calculation of crop TPAI is divided into three different stages according to the photosynthetic participation of non-leaf organs and leaves: the vegetative growth stage, the gradual development stage of non-leaf green organs, and the stage of non-leaf green organs during grain filling or the early stage of maturity.

[0096] The present invention proposes to use TPAI as a bridging parameter to connect the various components of the crop assimilation and yield estimation system. The present invention uses TPAI to connect two important components of assimilation and yield estimation, and takes the TPAI simulated by WOFOST and the TPAI retrieved by remote sensing as the data input of the assimilation and yield estimation method. Finally, the yield of crops is estimated by combining the EnKF algorithm and the assimilated TPAI. Although the EnKF has been used in the prior art, the combination of the assimilation method using TPAI and the EnKF has never been proposed. Moreover, TPAI can more accurately describe the growth and development changes of crops. Compared with the error accumulation caused by using LAI, the assimilated TPAI is very helpful for reducing the filtering divergence of the EnKF. The assimilation and yield estimation method and system proposed by the present invention take TPAI as the core clue, and not only make improvements in a certain component or step, but all key links of the entire method and system involve TPAI.

[0097] In terms of technical effects, the present invention improves the mechanistic and extensible nature of the crop assimilation yield estimation method. The method for parameter calibration based on TPAI is based on the characteristics of crop plants and emphasizes the photosynthesis of crops and the succession of photosynthetic organs at different rapeseed growth stages. TPAI not only includes the photosynthetic participation of rapeseed leaves but also fully considers the contribution of the photosynthesis of siliques, the non-leaf green organs of rapeseed, to the yield. Compared with the LAI used in the prior art, the assimilation yield estimation method and system combining the EnKF algorithm and TPAI proposed by the present invention are more applicable to crops with special morphological successions of photosynthetic organs. In addition, compared with LAI, the assimilation method and system based on TPAI and the EnKF algorithm are more universal. Using TPAI instead of LAI for parameter localization can achieve the expansion and extension of the crop growth model. Whether it is gramineous crops such as wheat and corn (special cases where LAI = 0), or other crops with non-leaf green organs such as alfalfa, cotton, sesame, and castor, the growth simulation and yield estimation can be carried out through the assimilation yield estimation method and system proposed by the present invention.

[0098] In the method of the present invention, assimilating TPAI can suppress the EnKF filter divergence. The EnKF algorithm predicts the background error covariance through an ensemble, including the background error covariance that varies with time and space. EnKF initializes the model with the probability distribution of the observation matrix and synchronously predicts to obtain the probability distribution of the prediction result. Although the true state of the crop remains within a bounded region, there is a serious problem of filter divergence in the comprehensive state estimation of the system. There are many factors affecting filter divergence, and the authenticity of the observation data (bridging parameters) is one of the important factors. When the observation data is inaccurate, it will lead to the estimated value of the target state deviating more and more from the actual value. The LAI time series curve ([ Figures 8 - 11 ) simulated by the EnKF algorithm based on LAI of the present invention shows that there is a large gap between the simulated values and the measured values of TAGP and TWSO in the mature stage of rapeseed growth, showing obvious filter divergence. This is because the crop growth model simulates the yield of the crop only under the participation of leaf area in photosynthesis, while the actual rapeseed yield is the combined contribution of the photosynthesis of siliques and leaves. Therefore, the difference between LAI as the observation data and the actual situation leads to the accumulation of errors throughout the growth period of rapeseed, that is, the filter divergence phenomenon of TWSO and TAGP. However, for the TAGP and TWSO time series curves simulated by the EnKF algorithm using TPAI, the error values at this stage are significantly reduced ([ Figures 12 - 15 ), and the simulated values and the observed values show good consistency. Therefore, the parameter TPAI, which can more accurately describe the actual situation of the crop, improves the reliability of the observation data and suppresses the EnKF filter divergence phenomenon to a certain extent.

[0099] The present invention improves the accuracy of crop assimilation yield estimation. First, for the single-point assimilation results of rapeseed, based on different parameter combinations and the EnKF algorithm, the LAI and TPAI are respectively used for rapeseed growth simulation and yield estimation. It is found that for the single-point assimilation yield estimation of rapeseed based on LAI, although the combination of parameters TSUM1, SPAN, CVO, and TDWI performs the best, the simulation accuracies R2 of TWSO and TAGP are only 0.11 and 0.07 respectively, still at a relatively low accuracy; while for the single-point assimilation yield estimation of rapeseed based on TPAI, the simulation accuracies R2 of any group of parameters TWSO and TAGP reach about 0.50 and 0.40 respectively, achieving high-precision estimation of rapeseed single-point yield. First, for the regional assimilation results of rapeseed, from Figure 15 It can be seen that the R2 between the county average value of rapeseed yield assimilating LAI and the statistical data is 0.10, while the R2 between the county average value of rapeseed yield assimilating TPAI and the statistical data is 0.56. Therefore, compared with the yield estimation by LAI assimilation, the assimilation estimation based on TPAI and the EnKF algorithm significantly improves the accuracy of crop growth simulation and yield estimation.

[0100] The method of the present invention is verified for feasibility. The present invention selects the typical representative WOFOST model, which is the most commonly used crop growth general model at present, and takes rapeseed, a crop with active non-leaf green organs, as the research object, and conducts result verification in Hengyang City, Hunan Province. Using the EnKF algorithm and the assimilation bridging parameter TPAI, the growth trend and yield of rapeseed in 2021 are simulated and estimated, and the effects of the method process are verified by comparing the simulation results before and after assimilation and the simulation results of assimilating LAI.

[0101] (1) Single-point assimilation yield estimation of rapeseed based on EnKF and TPAI

[0102] Figures 4 - 7 The time-series curves of the photosynthetic area index before and after assimilating LAI and TPAI respectively are shown. Two sample points are selected for each group of curves to show. The blue curve is before assimilation, and the orange-red curve is the time-series curve after assimilation. The yellow symbols are remote sensing inversion data and error bars. Since during the growth process of rapeseed, the photosynthetic assimilates of siliques have a great impact on the final yield. Therefore, the photosynthesis of rapeseed siliques also needs to be considered additionally in the rapeseed growth simulation, which can more accurately reflect the actual rapeseed photosynthetic area index. As Figure 6 and Figure 7 show, compared with the time-series curve of the photosynthetic area index of assimilating LAI, the time-series curve of assimilating TPAI has a significant increase in the later growth stage, which also conforms to this time trend of the photosynthetic area index.

[0103] Figures 8 - 15Shows the yield time series curves before and after assimilating LAI and TPAI respectively. Two sample points are selected for each group of curves for display. The blue curve is before assimilation, and the orange-red curve is the time series curve after assimilation. The yellow circles are the measured yield data. As Figures 8 - 11 shown in the yield time series diagram of assimilating LAI. It can be seen that regardless of whether the yield value before assimilation is close to the true value, the value of TWSO will decrease after assimilation, while the value of TAGP will increase. Since it is difficult to perfectly reflect the process of photosynthetic production of photoassimilates by rapeseed during growth using LAI, and there are errors in the LAI introduced multiple times in the simulation of rapeseed, the accuracy of the model can no longer be improved with assimilation. And in Figures 12 - 15 , it is the yield time series diagram of assimilating TPAI. Since TPAI takes more comprehensive consideration of the photosynthetic process of rapeseed, even in the simulation of sample points where TWSO is already very close to the true value, assimilating TPAI can prevent its accuracy from decreasing.

[0104] (2) Rapeseed regional assimilation yield estimation based on EnKF and TPAI

[0105] The present invention combines the rapeseed yield data of each county from the National Bureau of Statistics to verify the regional yield results. Since the statistical data is the dry weight of rapeseed, while the model simulation data is the dry weight of rapeseed pods (rapeseed pod peel and rapeseed), they are not exactly equal. Therefore, in this paper, differential coordinate axis joint plotting is performed on the two types of statistical and simulation data. Figures 16 - 19 As can be seen from, although the regional yield results using LAI are not ideal, the regional yield results using TPAI assimilation are relatively accurate. Although rapeseed and rapeseed pods are not exactly the same, relatively accurate yield predictions can still be made.

[0106] To further verify the accuracy of the rapeseed regional yield results based on TPAI assimilation, taking county-level administrative regions as units, the average values of the regional yield assimilation results are calculated and compared with the actual statistical data. From Figure 17 it can be seen that the R2 between the county average value of rapeseed yield assimilating LAI and the statistical data is 0.10, while the R2 between the county average value of rapeseed yield assimilating TPAI and the statistical data is 0.56. For rapeseed crops, compared with LAI, assimilation based on TPAI greatly improves the accuracy of yield estimation.

[0107] The above-described embodiments are only relatively preferred specific implementation manners of the present invention. The present specification uses phrases such as "in one embodiment", "in another embodiment", "in yet another embodiment" or "in other embodiments", which may all refer to one or more of the same or different embodiments according to the present disclosure. Ordinary variations and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. An assimilation and yield estimation method based on the EnKF algorithm and the total photosynthetic area index, characterized in that Including: S1. Calculate the total photosynthetic area index TPAI. During the vegetative growth stage, TPAI is the leaf area index LAI; during the development stage of non-leaf green organs, TPAI is the sum of the leaf area index LAI and the non-leaf green organ area index NFGOAI; when the non-leaf green organs are in the grain filling or early maturity stage, TPAI is the non-leaf green organ area index NFGOAI; S2. Use the total photosynthetic area index TPAI to calibrate the crop growth model; S3. Invert the total photosynthetic area index TPAI using remote sensing data; S4. Based on the EnKF algorithm, assimilate the inverted total photosynthetic area index TPAI into the crop growth model to simulate crop growth and estimate yield.

2. The assimilation yield estimation method according to claim 1, characterized in that Step S2 includes: S21. Calculate the ground measured value of the photosynthetic area index TPAI as the data input for single-point scale EnKF assimilation yield estimation; S22. Conduct a sensitivity analysis on the crop growth model, and perform parameter calibration steps through the calculated total photosynthetic area index TPAI to obtain the WOFOST model with localized parameters; S23. Use the parameter-calibrated crop growth model to simulate TPAI, and the obtained TPAI is used as the simulated value for EnKF assimilation yield estimation.

3. The assimilation yield estimation method according to claim 1, wherein Step S3 includes: S31. Screen the polarization decomposition features by preprocessing and polarization decomposition of remote sensing data; S32. Use random forest to rank and select the importance of the polarization decomposition features, establish an inversion model and verify it; S33. Invert the TPAI data of the crop study area using the inversion model as the data input for regional scale EnKF assimilation yield estimation.

4. The assimilation yield estimation method according to claim 3, characterized in that The feature selection method in step S32 includes: 1) Conduct importance analysis and ranking on all features through random forest, and the more important features are ranked higher, called the feature sequence P; 2) The current feature set Q is an empty set; 3) Take out the most important feature from P and put it into Q, the number of features in P is reduced by 1, and the number of features in Q is increased by 1. Use RF to model the features in Q and analyze the modeling accuracy and speed; 4) Loop step 3) until P is empty; 5) Compare and analyze all the results obtained during the process to obtain the optimal feature set.

5. The assimilation yield estimation method according to claim 1, wherein Step S4 includes: S41. Initialization: Based on the calibrated crop growth model parameters, simulate the obtained TPAI set as the initial state set of TPAI; S42. Observation: The initial state set of TPAI runs forward in time and is simulated. Input the ground measured value of TPAI or the regional inversion value of TPAI as the TPAI observation value at the critical growth stages of the crop, and calculate the error matrix of the TPAI observation value; S43. Update: Use the ensemble Kalman filter to combine the error covariance matrix of the TPAI observation vector and the state vector to update the state of each ensemble member to obtain the analysis field set; S44. Prediction: Input the observed TPAI into the crop growth model to generate the forecast field set of TPAI, and this set represents the current model state; S45. Calculate the error covariance matrix according to the forecast field set; S46. Calculate the posterior estimate of the model state as the ensemble mean, i.e., the photosynthetic area index of the crop at growth stage time t.

6. The assimilation-based yield estimation method according to claim 5, wherein In step S43, calculate the error covariance of the current estimate through the following formula: Among them, K t is the Kalman gain; is the covariance matrix of the members of the TPAI prediction state set at time t; H t is the observation operator; is the covariance matrix of the members of the TPAI observation state set at time t; is the updated value of the i-th member of the state variable set of TPAI at time t.

7. The assimilation yield estimation method according to claim 6, wherein In step S44, the crop growth model is: wherein, is the predicted value of the i-th member of the set of state variables of TPAI at time t; N is the number of members of the TPAI set; M is the state transition model and is the WOFOST dynamic model.

8. The assimilation yield estimation method according to claim 7, characterized in that, In step S45, the error covariance matrix is: wherein, is the expected value of a member of the TPAI prediction status set.

9. The assimilation yield estimation method according to claim 8, characterized in that, In step S46, calculate the photosynthetic area index of the crop at growth stage time t through the following formula:

10. An assimilation yield estimation method and system based on the EnKF algorithm and the total photosynthetic area index, which includes a computer-executable program. When the program is executed, the method described in any one of claims 1-9 is implemented.

Images