Sugarcane yield remote sensing estimation method and device based on SAFY-Sugar model

Abstract

The invention relates to a remote sensing estimation method and device for sugar cane yield per unit area based on a SAFY-Sugar model, comprising the following steps: constructing a LAI inversion model on the basis of observing the growth potential of sugar cane and measuring biomass; improving the SAFY model to construct a SAFY-Sugar model; assimilating remote sensing inverted LAI data, ground observed stem weight data and the SAFY-Sugar model to calibrate model parameters in combination with an SCE-UA algorithm; and simulating sugar cane LAI, total aboveground biomass and stem weight on a sample plot scale by using the SAFY-Sugar model. The present invention establishes a semi-empirical remote sensing inversion model of sugar cane LAI, analyzes the inversion ability of different vegetation indices on sugar cane LAI, and constructs the SAFY-Sugar model, which is assimilated with remote sensing inversion LAI data and ground-observed stem weight data to simulate the stem weight of sugar cane. The improved SAFY-Sugar model has high simulation accuracy at the sample plot scale, and can accurately capture the phenological changes and spatial growth differences of sugar cane at the regional scale, providing technical support for regional sugar cane growth monitoring and yield estimation applications.

Description

Technical Field

[0001] The invention relates to the field of remote sensing technology, and in particular to a remote sensing estimation method and device for sugar cane yield based on a SAFY-Sugar model. Background Art

[0002] Sugar cane is one of the most important cash crops in the world. More than 100 countries around the world grow sugar cane for the production of sugar and energy. At present, the sugar cane industry has become an important pillar of economic development in major producing areas and a major source of income for farmers. Accurate control of sugar cane growth and yield information is conducive to guiding farmers and cooperatives to choose planting sites and carry out precise field management, and can also serve related companies such as agricultural insurance.

[0003] Optical remote sensing image data commonly used in crop growth research include Sentinel-2, Landsat-8, SPOT-6, RapidEye and WorldView-2. With the continuous development of science and technology, the emergence of multi-source data such as hyperspectral data, synthetic aperture radar data (such as Sentinel-1) and drone image data has provided more data support for crop growth monitoring and yield estimation. Experts and scholars have found in their research that establishing empirical models for spectral reflectance characteristics and biological physiological indicators is an effective means to achieve crop growth monitoring. At the same time, with the continuous in-depth understanding of the crop growth process, the mechanism model for simulating the crop growth process has gradually developed. Crop growth models predict the evolution of crops from sowing to harvesting by simulating photosynthesis, gas exchange between the canopy and the atmosphere, phenology, soil moisture and temperature changes, biomass accumulation and grain formation. These models (such as the AquaCrop model) require input of downwelling shortwave radiation, temperature, precipitation and a series of parameters related to crop varieties, soil conditions and management methods to achieve the prediction of leaf area index, aboveground biomass and grain yield.

[0004] However, crop growth monitoring based on remote sensing technology depends on the quality of ground-based measured data and satellite image data. The use of optical remote sensing data is usually affected by cloudy and rainy weather, and it is often difficult to obtain sufficient high-quality images for growth monitoring and yield prediction during the crop growth period. Early monitoring methods were mainly based on empirical models, which often used vegetation indices with different band information combinations to characterize crop growth and productivity levels. However, methods based on empirical models lack mechanistic significance, have poor universality, are difficult to extend to other applications, and require a large amount of ground observation data to establish empirical relationships, which consumes a lot of manpower and material resources when applied; crop models can generally simulate the response of crop growth to environmental changes (such as temperature, fertilization, and irrigation) well. However, when simulating large-scale crop growth, the input reference of the crop model needs to consider spatial changes, and the uncertainty of soil properties, the spatial distribution of canopy state variables, and meteorological data will cause deviations in the simulation results.

[0005] At present, most crop models of sugar cane are process-based mechanism models, which require a large number of input parameters and data to simulate the growth process of sugar cane, and are difficult to use for large-scale research. Although the sugar cane growth model based on ground observation data is more in line with the actual planting situation, it is difficult to promote and use due to spatial differences. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a remote sensing estimation method and device for sugar cane yield based on the SAFY-Sugar model in view of the deficiencies in the prior art.

[0007] The technical solution of the present invention to solve the above technical problems is as follows:

[0008] A remote sensing estimation method for sugar cane yield based on the SAFY-Sugar model, the method comprising:

[0009] S1. Observe the growth of sugar cane and measure the biomass of sugar cane in the study area; obtain the daily average temperature data and short-wave solar radiation data during the growth period of sugar cane in the study area; obtain the Sentinel-2 image time series and perform preprocessing;

[0010] S2. Using the Sentinel-2 image time series to obtain remote sensing information of the observation plots in the study area, selecting several vegetation indices, constructing a semi-empirical leaf area index LAI inversion model based on the Beer-Lambert law, and solving the model parameters of the semi-empirical LAI inversion model using the least squares method and the global optimal algorithm; based on the ground-observed leaf area index LAI data obtained by observing the growth of sugar cane and measuring the biomass of sugar cane in the study area in step S1 and the cross-validation method, the inversion result of the semi-empirical LAI inversion model is verified for accuracy, the inversion ability of each vegetation index to the semi-empirical LAI inversion model is evaluated, and the optimal LAI inversion model is obtained;

[0011] S3. Improve the SAFY model by combining the growth process of sugar cane and the biomass distribution method of sugar cane stalks, construct the SAFY-Sugar model, and use the accumulated temperature to regulate the biomass accumulation process of sugar cane stalks, so that the model can simulate the daily increase in biomass of sugar cane stalks;

[0012] S4, obtaining the initial parameter value or the value range of the initial parameter of the SAFY-Sugar model, using the daily average temperature data and the shortwave solar radiation data obtained in step S1 as input data of the SAFY-Sugar model, and using the composite hybrid evolutionary algorithm SCE-UA to optimize and calibrate the parameters of the SAFY-Sugar model based on the data assimilation strategy to obtain the optimal model parameters;

[0013] S5. Use the model parameters to run the SAFY-Sugar model to obtain simulation results of the SAFY-Sugar model on the sugar cane LAI, total aboveground biomass and stem weight in the observation plot, use ground measurement data to verify the accuracy of the simulation results, obtain the optimal parameter values ​​of the SAFY-Sugar model, run the SAFY-Sugar model on the study area, and obtain the spatial distribution map of sugar cane yield per unit area in the study area.

[0014] The beneficial effects of the method invention are as follows: a remote sensing estimation method for sugar cane yield is proposed, and on the basis of sugar cane growth observation and biomass measurement, the vegetation index of the corresponding sample plot in the region is calculated using the Sentinel-2 image time series to construct an LAI inversion model based on the Beer-Lambert law; the SAFY model is improved to construct the SAFY-Sugar model, and the sugar cane stalk biomass distribution method is added to the model; the remote sensing inversion LAI data and the ground observation stem weight data are combined with the SAFY-Sugar model to perform data assimilation and calibration of model parameters in combination with the SCE-UA algorithm, and the SAFY-Sugar model is used to simulate the sugar cane LAI, total aboveground biomass and stem weight at the sample plot scale, and the model simulation is further extended to the regional scale. Compared with the prior art, the main advantages of the present invention are as follows: 1. Most of the existing sugar cane LAI inversion models are empirical models constructed using vegetation indices and ground-based observation LAI data. The present invention establishes a semi-empirical remote sensing inversion model for sugar cane LAI based on the Beer-Lambert law, analyzes the inversion capabilities of different vegetation indices for sugar cane LAI, and uses Sentinel-2 time series image data to achieve the inversion of sugar cane LAI. 2. The feasibility of applying the light energy utilization efficiency model to sugar cane growth simulation and yield estimation is proved, and a new stalk biomass allocation method is proposed to improve the SAFY model to generate the SAFY-Sugar model, which significantly improves the simulation capability of the SAFY model for sugar cane stem dry weight. 3. The improved SAFY-Sugar model has a high simulation accuracy at the sample plot scale, and can accurately capture the phenological changes and spatial growth differences of sugar cane at the regional scale, providing technical support for regional sugar cane growth monitoring and yield estimation applications. The sugar cane yield remote sensing estimation method proposed by the present invention has good scalability and is suitable for large-scale research.

[0015] Another technical solution of the present invention to solve the above technical problems is as follows:

[0016] A remote sensing estimation device for sugar cane yield based on the SAFY-Sugar model, the device comprising:

[0017] The acquisition module is used to observe the growth of sugar cane and measure the biomass of sugar cane in the study area; obtain the daily average temperature data and short-wave solar radiation data during the growth period of sugar cane in the study area; obtain the Sentinel-2 image time series and perform preprocessing;

[0018] The first construction module is used to use the Sentinel-2 image time series to obtain the remote sensing information of the observation plots in the study area, select several vegetation indices, construct a semi-empirical leaf area index LAI inversion model based on the Beer-Lambert law, and use the least squares method and the global optimal algorithm to solve the model parameters of the semi-empirical LAI inversion model; based on the ground observation leaf area index LAI data obtained by observing the growth of sugar cane and measuring the biomass of sugar cane in the study area in the acquisition module and the cross-validation method, the inversion result of the semi-empirical LAI inversion model is verified for accuracy, and the inversion ability of each vegetation index to the semi-empirical LAI inversion model is evaluated to obtain the optimal LAI inversion model;

[0019] The second building block is used to improve the SAFY model by combining the sugar cane growth process and the sugar cane stalk biomass allocation method, construct the SAFY-Sugar model, and use the accumulated temperature to regulate the biomass accumulation process of the sugar cane stalk, so that the model can simulate the daily increase in biomass of the sugar cane stalk;

[0020] The first optimization module is used to obtain the initial parameter values ​​or the value range of the initial parameters of the SAFY-Sugar model, use the daily average temperature data and the shortwave solar radiation data obtained in the acquisition module as input data of the SAFY-Sugar model, and use the composite hybrid evolutionary algorithm SCE-UA to optimize and calibrate the parameters of the SAFY-Sugar model based on the data assimilation strategy to obtain the model parameter values;

[0021] The third construction module is used to run the SAFY-Sugar model using the model parameter values ​​to obtain the simulation results of the SAFY-Sugar model on the sugar cane LAI, total aboveground biomass and stem weight in the observation sample plot, use ground measurement data to verify the accuracy of the simulation results, obtain the optimal model parameter values ​​of the SAFY-Sugar model, run the SAFY-Sugar model on the study area, and obtain the spatial distribution map of sugar cane yield per unit area in the study area.

[0022] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 It is a schematic flow chart of a method for remote sensing estimation of sugar cane yield based on the SAFY-Sugar model of the present invention;

[0025] Figure 2 The present invention is a schematic diagram of a module of a remote sensing estimation device for sugar cane yield based on the SAFY-Sugar model.

[0026] Figure 3 A result diagram of the LAI inversion model based on six vegetation indices of the present invention is constructed;

[0027] Figure 4a , 4b It is a simulation comparison diagram of the SAFY-Sugar model of the present invention and the original SAFY model for LAI of sugar cane in some sample plots at the sample plot scale and a LAI simulation accuracy verification result diagram;

[0028] Figure 5a , 5b It is a simulation comparison diagram of the total aboveground biomass of sugar cane in some sample plots by the SAFY-Sguar model of the present invention and the original SAFY model at the sample plot scale, and a result diagram of the total aboveground biomass simulation accuracy verification;

[0029] Figure 6a , 6b It is a simulation comparison diagram of the dry weight of sugarcane stems of some sample plots by the SAFY-Sugar model of the present invention and the original SAFY model at the sample plot scale, and a simulation accuracy verification result diagram of the dry weight of stems;

[0030] Figure 7 This is a diagram showing the simulation results of stem weight at a regional scale of 1 km×1 km using the SAFY-Sugar model of the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0032] like Figure 1The embodiment of the present invention provides a remote sensing estimation method for sugar cane yield based on the SAFY-Sugar model, the method comprising:

[0033] S1. Observe the growth of sugar cane and measure the biomass of sugar cane in the study area; obtain the daily average temperature data and short-wave solar radiation data during the growth period of sugar cane in the study area; obtain the Sentinel-2 image time series and perform preprocessing;

[0034] S2. Using the Sentinel-2 image time series to obtain remote sensing information of the observation plots in the study area, selecting several vegetation indices, constructing a semi-empirical leaf area index LAI inversion model based on the Beer-Lambert law, and solving the model parameters of the semi-empirical LAI inversion model using the least squares method and the global optimal algorithm; based on the ground-observed leaf area index LAI data obtained by observing the growth of sugar cane and measuring the biomass of sugar cane in the study area in step S1 and the cross-validation method, the inversion result of the semi-empirical LAI inversion model is verified for accuracy, the inversion ability of each vegetation index to the semi-empirical LAI inversion model is evaluated, and the optimal LAI inversion model is obtained;

[0035] S3. The SAFY model was improved by combining the growth process of sugar cane and the biomass distribution method of sugar cane stalks, and the SAFY-Sugar (the Simple Algorithm for Yield Estimation model-Sugar) model was constructed. The accumulated temperature was used to regulate the biomass accumulation process of sugar cane stalks, so that the model could simulate the daily increase in biomass of sugar cane stalks.

[0036] S4, obtaining the initial parameter values ​​or the value range of the initial parameters of the SAFY-Sugar model, using the daily average temperature data and the shortwave solar radiation data obtained in the step S1 as the input data of the SAFY-Sugar model, and using the composite hybrid evolution algorithm SCE-UA (shuffled complex evolution method developed at the University of Arizona) to optimize and calibrate the parameters of the SAFY-Sugar model based on the data assimilation strategy to obtain the optimal model parameters;

[0037] S5. Use the model parameters to run the SAFY-Sugar model to obtain simulation results of the SAFY-Sugar model on the sugar cane LAI, total aboveground biomass and stem weight in the observation plot, use ground measurement data to verify the accuracy of the simulation results, obtain the optimal parameter values ​​of the SAFY-Sugar model, run the SAFY-Sugar model on the study area, and obtain the spatial distribution map of sugar cane yield per unit area in the study area.

[0038] Based on the above embodiment, the step S1 specifically includes:

[0039] S11, uniformly selecting a number of sugar cane planting plots of a first preset size in the study area, arranging a plurality of sample plots of a second preset size in each plot, and measuring the multi-stage sugar cane planting density information, longitude and latitude coordinates, natural canopy height, effective stem height, base diameter and LAI according to the growth stage of the sugar cane, wherein the planting density information is the number of sugar canes planted in an area of ​​a third preset size in each sample plot, and after sampling the sugar cane from the plot, measuring the fresh weight and dry weight in the laboratory;

[0040] S12, selecting a meteorological station or a neighboring meteorological station in the study area to obtain daily average temperature data and daily shortwave solar radiation data during the sugar cane growing period;

[0041] S13. Acquire Sentinel-2L2A images covering the sugarcane growing period in the study area and resample them to 20 m resolution using SNAP software.

[0042] Based on the above embodiment, the step S2 specifically includes:

[0043] S21, select the Normalized Difference Vegetation Index NDVI (Normalized Difference Vegetation Index), the Normalized Difference Water Index NDWI1 (Normalized Difference Water Index), the Chlorophyll Index CI green (Chlorophyll Index-Green), Two-band Enhanced Vegetation Index EVI2 (Two-band EnhancedVegetation Index), Red Edge Normalized Difference Vegetation Index NDVI RE (Red-edge Normalized DifferenceVegetation Index) and red-edge chlorophyll vegetation index CI RE(Chlorophyll Index Red-edge), using the pre-processed Sentinel-2 image time series to calculate the value of the vegetation index corresponding to the observation plot, and construct the semi-empirical LAI inversion model based on the Beer-Lambert law;

[0044] Wherein, the normalized difference vegetation index NDVI=(NIR-RED) / (NIR+RED);

[0045] The normalized water index NDWI1=(NIR-SWIR1) / (NIR+SWIR1);

[0046] Chlorophyll Index CI green =NIR / GREEN-1;

[0047] Two-band enhanced vegetation index EVI2 = 2.5 (NIR-RED) / (NIR + 2.4RED + 1);

[0048] Red Edge Normalized Difference Vegetation Index (NDVI) RE =(NIR-RE) / (NIR+RE);

[0049] Red Edge Chlorophyll Index CI RE =NIR / RE-1;

[0050] Among them, RED is the surface reflectance of the B4 red band of the Sentinel-2 image; NIR is the surface reflectance of the B8a near-infrared band of the Sentinel-2 image; RE is the surface reflectance of the B6 red edge band of the Sentinel-2 image; SWIR1 is the surface reflectance of the first short-wave infrared band B11 of the Sentinel-2 image;

[0051] The semi-empirical LAI inversion model expression is:

[0052]

[0053] Where, VI ∞ is the vegetation index when LAI is close to saturation, VI bs K is the vegetation index of bare soil, which is calculated based on the bare soil image reflectance before sugar cane planting according to the growth phenological period. VI is the extinction coefficient of vegetation, ranging from 0.2 to 2.0;

[0054] The least squares method and global optimization algorithm are used to solve the model parameters. The global optimization algorithm starts from different starting points to find multiple local solutions to the problem and obtains the global optimal solution rather than the local optimal solution. The least squares method minimizes the sum of squares of the residuals between the model simulation results and the actual ground observation results by continuously updating the model parameter values ​​after the parameter value range and initial value are given:

[0055]

[0056] In the formula, xdata is the given input data, F(x,xdata) is the output result of the model for the input data, ydata is the real observation result, and xdata is i is the given input data at sample point i, ydata i is the actual observation result at sample point i;

[0057] S22. Use ground-based LAI data to verify the accuracy of model simulation results: Use the 10-fold cross-validation method to randomly divide the ground-based LAI data into 10 parts by non-repeated sampling. Each time, nine parts of the data are randomly selected for model training, and the remaining part is used to verify the model simulation results and evaluate the performance of the current model. Among them, the accuracy verification indicators include the determination coefficient R 2 , root mean square error RMSE, standard root mean square error nRMSE and RPD indicators;

[0058] The calculation formula of each indicator is expressed as:

[0059]

[0060]

[0061]

[0062]

[0063] Among them, yo, i is the ground observation value at sample point i, y m,i is the value simulated by the model at sample point i, is the average value of ground observations, is the mean of the model simulation, n is the number of samples used for validation, and SD is the standard deviation of the ground observations;

[0064] S23. Compare the six semi-empirical LAI inversion models constructed by the vegetation indices, and select the semi-empirical LAI inversion model with the best accuracy verification result as the optimal LAI inversion model.

[0065] Based on the above embodiment, step S3 specifically includes:

[0066] (1) Aboveground biomass simulation

[0067] During the entire growth and development process of crops, the accumulation of aboveground biomass will continue from crop emergence D0 until the crop is completely senescent. The daily accumulated biomass ΔDAM depends on the effective light energy utilization rate ELUE and the photosynthetically active radiation APAR absorbed by the canopy, and is affected by the temperature stress factor F T (T a ) regulates the growth of biomass, and ΔDAM can be expressed by the above determining factors as:

[0068] ΔDAM=ELUE×F T (T a )×APAR

[0069] in,

[0070] APAR=FAPAR×g c ×Rg

[0071] In the formula, photosynthetically active radiation APAR is expressed as light energy interception efficiency FAPAR, meteorological efficiency factor ε c As a function of the daily total solar shortwave radiation Rg, the Beer-Lambert law describes the nonlinear relationship between FAPAR and LAI. FAPAR is expressed by LAI as:

[0072]

[0073] Among them, K ext is the extinction coefficient, LAI t-1 is the leaf area index of the crop on day t-1;

[0074] Considering that different temperatures will affect the accumulation rate of aboveground biomass, the model uses the maximum temperature T max , minimum temperature T min Optimum temperature To pt The average temperature of the day T a The relationship between the temperature stress factor is expressed as:

[0075]

[0076] In the formula, β is the power of the polynomial, usually expressed as a quadratic polynomial; the highest and lowest temperatures are the temperature thresholds for crop growth. When the average temperature exceeds the temperature threshold, the crops will stop growing;

[0077] (2) Simulation of dynamic growth of LAI

[0078] The SAFY-Sugar model divides the crop growth period into the growth stage and the senescence stage. It simulates the dynamic growth process of leaves by increasing and decreasing LAI in steps of days, and divides different growth stages by accumulated temperature SMT. The calculation formula of accumulated temperature SMT is as follows:

[0079]

[0080] The product of biomass ΔDAM, leaf biomass allocation function Pl and specific leaf area SLA is calculated as:

[0081] ΔLAI + =ΔDAM×Pl×SLA

[0082] The leaf biomass allocation function Pl consists of two parameters Pl a and Pl b The composition is expressed in the model as a function of accumulated temperature. The value range of Pl is 0-1. The proportion of biomass allocated to leaves by the model is close to (1-Pl a ), and it decreases with the growth of the crop until the leaves stop growing. The calculation formula of Pl is as follows:

[0083]

[0084] When the accumulated temperature exceeds the accumulated temperature threshold S required for aging TT At , the crop begins to enter the senescence stage, and the daily decrease in leaf area index ΔLAI- is:

[0085]

[0086] Among them, LAI t-1 is the LAI of the crop on the day before the simulation, and Rs is the senescence rate of the leaves;

[0087] Therefore, the leaf area index LAI of the crop is t Leaf Area Index LAI t-1 The change in leaf area index on that day is ΔLAI + , ΔLAI - Adding them together gives:

[0088] LAI t =LAI t-1 +ΔLAI + -ΔLAI -

[0089] (3) Stem weight simulation

[0090] When the sugar cane growth enters the stem extension stage, the daily biomass of the stem is calculated by multiplying the daily increase in aboveground biomass and the daily biomass allocation ratio to the stem:

[0091] ΔDSM=ΔDAM×P stk ,if SMT>S stk

[0092] Among them, ΔDAM is the daily increase in aboveground biomass of sugar cane, P stk is the dynamic allocation ratio of aboveground biomass increment to stems every day, S stk The accumulated temperature threshold required for sugar cane to start stem elongation;

[0093] The start time of sugarcane stem extension can be determined by the number of leaves on the main bud. When the 10th true leaf on the main bud is fully expanded, the internodes at the base begin to joint and elongate rapidly. At this time, the sugarcane begins to enter the stem extension period. The leaf stripping operation at the early stage of sugarcane stem extension causes the plant leaf area to stop growing. Therefore, the accumulated temperature corresponding to the maximum LAI simulated by the model is used as the accumulated temperature required for the sugarcane to start stem extension, P stk Calculation formula P stk =min[0.8,max(0,F stk )], where 0.8 is the maximum proportion of the daily aboveground biomass that can be allocated to the stems, and F stk It is a logarithmic function used to describe the proportion of stem biomass distribution, and its value is regulated by the accumulated temperature;

[0094] F stk The calculation formula is as follows:

[0095] Among them, a is related to the proportion of biomass allocated to the sugar cane stalks. The daily biomass allocated to the stalks will decrease as the value of a increases.

[0096] Based on the above embodiment, step S4 specifically includes:

[0097] S41. The parameters of the SAFY-Sugar model include the meteorological efficiency factor ε c , initial aboveground biomass DAM0, extinction coefficient K ext , specific leaf area SLA, minimum temperature for crop growth T min , maximum temperature T max Optimum temperature To pt ;

[0098] The parameters of the SAFY-Sugar model also include the leaf biomass partition coefficient Pl a and Pl b , the accumulated temperature required for aging S TT, senescence rate Rs, the stalk biomass distribution coefficient a, crop emergence date D0 and effective light energy utilization efficiency ELUE;

[0099] S42, taking the daily average temperature data and the shortwave solar radiation data as model input data, using the composite hybrid evolutionary algorithm SCE-UA to optimize and calibrate the parameters of the SAFY-Sugar model, and determining the optimal parameters of the SAFY-Sugar model based on the data assimilation of the SCE-UA algorithm, so that the LAI simulated by the model and the corresponding remote sensing inversion LAI have the best consistency, that is, the cost function of the calculation of the two is minimized;

[0100] The remote sensing inversion LAI data in step S2 is used as a single variable to perform data assimilation with the SAFY-Sugar model. The model parameter optimization process is divided into two rounds. In the first round of assimilation, five parameters Pl related to crop varieties are determined. a Pl b , S TT , Rs and a values;

[0101] The remote sensing inverted LAI data corresponding to each sample plot were assimilated with the SAFY-Sugar model multiple times to obtain the six parameters Pl corresponding to each sample plot. a Pl b , S TT , Rs, a, D0 and ELUE, for Pl a , Pl b , S TT The optimization results of , Rs and a are summarized and counted, and the median of the statistical results of each parameter is taken as the optimal value of the parameter in the study area;

[0102] In the second round of assimilation, the results of the first round of assimilation are used to a Pl b , S TT The values ​​of Rs and a are fixed, and the remote sensing inversion LAI data corresponding to each sample plot is assimilated with the original model again to determine the D0 and ELUE of each sample plot;

[0103] The cost function F1 used in the assimilation process is calculated as follows:

[0104]

[0105] Among them, LAI i,mod is the LAI at sample point i simulated by the model, LAI i,obs is the LAI at sample point i inverted from remote sensing data, and n is the number of samples of LAI inverted from remote sensing;

[0106] S43, using the remote sensing inversion LAI and the ground measurement stem weight data corresponding to each sample plot as dual variables to optimize the parameters of the SAFY-Sugar model. In the first round of assimilation, some sample plots are randomly selected each time for dual variable data assimilation, and the Pl obtained by assimilation of each sample plot is used. a Pl b , S TT The average value of the five parameters, Rs and a, is taken as the optimal value of the parameter in the study area. a Pl b , S TT The values ​​of Rs and a are fixed, and a second round of assimilation is performed again to determine the D0 and ELUE of each sample plot. The bivariate data assimilation strategy of the SAFY-Sugar model selects three sample plots each time for parameter optimization and calibration, and the cost function F2 calculation formula used is:

[0107]

[0108] Where n1 and n2 are the number of samples of remote sensing inversion of LAI and ground measurement of stem weight, respectively; LAI sim,i is the LAI at sample point i simulated by the SAFY-Sugar model; LAI obs,i is the LAI at sample point i derived from remote sensing;

[0109] DSM sim,i The stem dry weight at sample point i simulated by the SAFY-Sugar model, DSM obs,i is the stem weight measured at sampling point i on the ground; is the average value of LAI inverted by remote sensing in each sample plot, The mean value of stem weight was measured above ground for each plot.

[0110] Based on the above embodiment, step S5 specifically includes:

[0111] S51, respectively using the results of the univariate data assimilation and bivariate data assimilation calibration in step S4 to set the parameters of the SAFY-Sugar model, using the daily average temperature data and the shortwave solar radiation data as model input data to run the SAFY-Sugar model, and simulating the LAI, total aboveground biomass and stem weight of sugar cane in each sample plot at the sample plot scale;

[0112] S52, using ground-measured LAI, total dry weight and stem dry weight data to verify the accuracy of the model quadrat scale simulation results, and obtain the optimal assimilation method and parameters of the SAFY-Sugar model;

[0113] S53, fusing the remote sensing images covering the study area, masking the study area using the sugar cane planting spatial distribution map of the study area, and calculating the LAI time series of the sugar cane growing period in the study area using the optimal LAI inversion model obtained in step S2;

[0114] S54, simulating the study area with the optimal LAI inversion model obtained in step S2, and a Pl b , S TT The parameter values ​​of Rs and a are kept constant in the same study area, and the LAI time series data of the study area are used as a single variable to perform data assimilation with the SAFY-Sugar model, and the values ​​of the seedling emergence date D0 and the effective light energy utilization rate ELUE are calibrated pixel by pixel;

[0115] S55. Use the daily average temperature data and shortwave solar radiation data of the study area and the parameters of the SAFY-Sugar model obtained by optimization and calibration to run the SAFY-Sugar model pixel by pixel to obtain the spatial distribution map of sugar cane yield in the study area.

[0116] It should be understood that crop growth monitoring based on remote sensing technology depends on the quality of ground-based measured data and satellite image data. The use of optical remote sensing data is usually affected by cloudy and rainy weather, and it is often difficult to obtain sufficient high-quality images for growth monitoring and yield prediction during the crop growth period. Early monitoring methods were mainly based on empirical models, which often used vegetation indices with different band information combinations to characterize crop growth and productivity levels. However, methods based on empirical models lack mechanistic significance, have poor universality, are difficult to extend to other applications, and require a large amount of ground observation data to establish empirical relationships, which consumes a lot of manpower and material resources when applied; crop models can generally simulate the response of crop growth to environmental changes (such as temperature, fertilization, and irrigation) well. However, when simulating large-scale crop growth, the input reference of the crop model needs to consider spatial changes, and the spatial distribution of soil properties and canopy state variables and the uncertainty of meteorological data will cause deviations in the simulation results. Therefore, the data assimilation method that combines crop physiological variables estimated by remote sensing data with crop growth models has become a research hotspot in the field of crop yield estimation. Data assimilation simultaneously utilizes the advantages of crop growth models in simulating the physical, chemical and physiological aspects of canopy development and the advantages of large-area observations using remote sensing technology. On the one hand, it can reduce the difficulty of obtaining model parameters in large-area studies while maintaining mechanistic significance, and quantify the growth conditions of crops at the regional scale. On the other hand, it can correct deviations in the model simulation process, thereby improving the accuracy of yield estimates.

[0117] At present, most crop models of sugar cane are process-based mechanism models, which require a large number of input parameters and data to simulate the growth process of sugar cane and are difficult to use for large-scale research. Although the sugar cane growth model constructed based on ground observation data is more in line with the actual planting situation, it is difficult to promote its use due to spatial differences. Compared with the mechanism model, the semi-empirical model (such as SAFY) not only retains the model mechanism and can accurately simulate the main physiological processes of crops, but also simplifies the input data and parameters of the model, thereby reducing the difficulty of obtaining prior information on parameters, making the model easier to optimize and more suitable for combining with remote sensing data for large-scale crop growth monitoring and yield estimation research. It has shown very good application prospects in recent research. However, this type of model cannot simulate the biomass accumulation process during the elongation stage of sugar cane stems, and is still rarely used in sugar cane crop growth simulation. In addition, research on sugar cane yield estimation in China is still in its infancy.

[0118] It should be understood that Figure 3 As shown in Figure 6, a remote sensing estimation method for sugar cane yield based on the SAFY-Sugar model includes the following steps:

[0119] S1 conducts sugarcane growth observation and biomass measurement; obtains daily average temperature data and shortwave solar radiation data during the sugarcane growing period in the corresponding study area; obtains Sentinel-2 image time series and preprocesses them;

[0120] S2 uses Sentinel-2 image time series to obtain observation sample remote sensing information, selects six vegetation indices widely used in crop growth monitoring to build a semi-empirical leaf area index inversion model based on the Beer-Lambert Law, and uses the least squares method and global optimal algorithm to solve the model parameters; based on ground observation LAI data and cross-validation method, the accuracy of the model inversion results is verified, the inversion ability of various vegetation indices to the leaf area index is evaluated, and the optimal leaf area index inversion model is determined;

[0121] S3 improves the SAFY model by combining the growth process of sugar cane and the biomass distribution method of sugar cane stalks, constructs the SAFY-Sugar (the Simple Algorithm for Yield Estimation model-Sugar) model, and uses accumulated temperature to regulate the biomass accumulation process of sugar cane stalks, so that the model can simulate the daily increase in biomass of sugar cane stalks.

[0122] S4 obtains the initial parameter value or value range of the SAFY-Sugar model, takes the daily average temperature data and the shortwave solar radiation data as the model input data, and uses a shuffled complex evolution method developed at the University of Arizona, SCE-UA to optimize and calibrate the model parameters based on a data assimilation strategy;

[0123] S5 used the model parameters obtained by optimized calibration to run the SAFY-Sugar model, obtained the simulation results of the model on the sugar cane LAI, total aboveground biomass and stem weight in the study area at the sample plot scale, and used ground measurement data to verify the accuracy of the simulation results; ran the optimal model at the regional scale to obtain the spatial distribution map of regional sugar cane yield.

[0124] In step S1, the specific steps of acquiring the research area data are:

[0125] S11 14 500m×500m pure pixel sugar cane planting plots were evenly selected in the study area, and 3 30m×30m plots were arranged in each plot to measure the multi-stage sugar cane planting density information, longitude and latitude coordinates, canopy natural height, effective stem height, basal diameter and LAI according to the growth stage of sugar cane; the control test plots were selected to sample sugar cane and measure the fresh weight and dry weight in the laboratory. Among them, the acquisition of planting density information requires the observation of the number of sugar cane plants planted in 3 rows × 10 meters of each sample plot. The plant height and base stem data need to be measured for each sugar cane plant randomly selected in the sample plot; LAI is measured using the LAI-2200 vegetation canopy analyzer. During the observation process, each sample plot needs to be observed multiple times, and the corresponding observation file number, LAI mean and standard deviation are recorded; the measurement of fresh weight and dry weight of sugar cane uses a destructive measurement method. The cut sugar cane is divided into leaves and stems after wiping off the surface attachments. The separated parts of the plant are brought back to the laboratory, and the leaf fresh weight and stem fresh weight of each part are measured separately, and then they are put into paper bags and placed in a drying oven to be baked at 80° for 72 hours until the weight no longer changes. The dry weight of each component of a single sugar cane plant is calculated based on the number of plants recorded in each cluster.

[0126] S12 Select meteorological stations in the study area or nearby meteorological stations to obtain daily average temperature data and daily shortwave solar radiation data during the sugarcane growing period. Among them, the average temperature data during the study period was downloaded from the website of the National Centers for Environmental Information (NCEI) of the National Oceanic and Atmospheric Administration (NOAA) (NOAA-National Centers for Environmental Information, https: / / www.ncei.noaa.gov). This data set includes climate elements such as temperature, air pressure, visibility, wind speed, and precipitation. The shortwave solar radiation data comes from the ERA5 (ECMWF Re-Analysis5) data set of the European Center for Medium-Range Weather Forecasts (ECMWF) (https: / / cds.climate.copernicus.eu / ), which provides the fifth generation of reanalysis data on global climate and weather values ​​over the past 40 to 70 years.

[0127] S13 obtains Sentinel-2L2A images covering the sugar cane growing period in the study area, and resamples to a resolution of 20 meters using SNAP software. The Sentinel-2 images used in the present invention are downloaded from ESA (https: / / scihub.copernicus.eu / ).

[0128] In step S2, the specific steps of remote sensing inversion of sugar cane LAI are:

[0129] S21 selects six vegetation indices that are widely used in crop growth monitoring and yield estimation research: Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI1), Chlorophyll Index-Green (CI green ), Two-band Enhanced Vegetation Index (EVI2), Red-edge Normalized Difference Vegetation Index (NDVI RE ) and the Chlorophyll Index Red-edge (CI RE), the vegetation index values ​​corresponding to the ground observation plots are calculated using the preprocessed remote sensing image data, and a semi-empirical LAI inversion model based on the Beer-Lambert Law is constructed. The calculation formulas for each vegetation index are:

[0130]

[0131] Table 1 Calculation formula of vegetation index

[0132] RED is the surface reflectance of the red band B4 of Sentinel-2MSI; NIR is the surface reflectance of the near-infrared band B8a of Sentinel-2 MSI; RE is the surface reflectance of the red-edge band B6 of Sentinel-2MSI; SWIR1 is the surface reflectance of the first shortwave infrared band B11 of Sentinel-2MSI.

[0133] The semi-empirical LAI inversion model expression is:

[0134]

[0135] Where, VI ∞ is the vegetation index when LAI is close to saturation, VI bs K is the vegetation index of bare soil, which is calculated based on the bare soil image reflectance before sugar cane planting according to the growth phenological period. VI is the extinction coefficient of vegetation, and its value range is set to 0.2-2.0 based on literature research.

[0136] The least squares method and global optimization algorithm are used to solve the model parameters. The global optimization algorithm can find multiple local solutions to the problem from different starting points to obtain the global optimal solution rather than the local optimal solution. The least squares method can minimize the sum of squared residuals between the model simulation results and the actual ground observation results by continuously updating the model parameter values ​​after the parameter value range and initial value are given:

[0137]

[0138] In the formula, xdata is the given input data, F(x,xdata) is the output result of the model for the input data, and ydata is the real observation result.

[0139] S22 uses ground-based LAI data to verify the accuracy of model simulation results: using the ten-fold cross-validation method, the ground-based LAI data is randomly divided into 10 parts by non-repeated sampling, nine parts are randomly selected each time for model training, and the remaining part is used to verify the model simulation results and evaluate the performance of the current model. Among them, the accuracy verification indicators include the determination coefficient (R 2), root mean square error (RMSE), standard root mean square error (nRMSE) and RPD indicators. The calculation formula of each indicator is expressed as:

[0140]

[0141]

[0142]

[0143]

[0144] Among them, y o,i and m,i are the ground observation and model simulation values ​​at sample point i, and are the means of ground observations and model simulations, respectively, n is the number of samples used for validation, and SD is the standard deviation of ground observations.

[0145] S23 compares the LAI inversion models constructed using six vegetation indices and selects the model with the best accuracy verification result as the sugar cane LAI inversion model.

[0146] In step S3, the specific steps of the SAFY-Sugar model are:

[0147] (1) Aboveground biomass simulation

[0148] During the entire growth and development process of crops, the accumulation of aboveground biomass will continue from crop emergence D0 until the crop is completely senescent. The daily accumulated biomass ΔDAM depends on the effective light energy utilization rate ELUE and the photosynthetically active radiation APAR absorbed by the canopy, and is affected by the temperature stress factor F T (T a ) regulates the growth of biomass, and ΔDAM can be expressed by the above determining factors as:

[0149] ΔDAM=ELUE×F T (T a )×APAR

[0150] in,

[0151] APAR=FAPAR×g c ×Rg

[0152] In the formula, photosynthetically active radiation APAR is expressed as light energy interception efficiency FAPAR, meteorological efficiency factor ε c As a function of the daily total solar shortwave radiation Rg, the Beer-Lambert law describes the nonlinear relationship between FAPAR and LAI. FAPAR is expressed by LAI as:

[0153]

[0154] Among them, K ext is the extinction coefficient, LAI t-1 is the leaf area index of the crop on day t-1;

[0155] Considering that different temperatures will affect the accumulation rate of aboveground biomass, the model uses the maximum temperature T max , minimum temperature T min Optimum temperature To pt The average temperature of the day T a The relationship between the temperature stress factor is expressed as:

[0156]

[0157] In the formula, β is the power of the polynomial, usually expressed as a quadratic polynomial; the highest and lowest temperatures are the temperature thresholds for crop growth. When the average temperature exceeds the temperature threshold, the crops will stop growing;

[0158] (2) Simulation of dynamic growth of LAI

[0159] The SAFY-Sugar model divides the crop growth period into the growth stage and the senescence stage. It simulates the dynamic growth process of leaves by increasing and decreasing LAI in steps of days, and divides different growth stages by accumulated temperature SMT. The calculation formula of accumulated temperature SMT is as follows:

[0160]

[0161] The product of biomass ΔDAM, leaf biomass allocation function Pl and specific leaf area SLA is calculated as:

[0162] ΔLAI + =ΔDAM×Pl×SLA

[0163] The leaf biomass allocation function Pl consists of two parameters Pl a and Pl b The composition is expressed in the model as a function of accumulated temperature. The value range of Pl is 0-1. The proportion of biomass allocated to leaves by the model is close to (1-Pl a ), and it decreases with the growth of the crop until the leaves stop growing. The calculation formula of Pl is as follows:

[0164]

[0165] When the accumulated temperature exceeds the accumulated temperature threshold S required for aging TT When the crop begins to enter the senescence stage, the daily decrease in leaf area index ΔLAI- for:

[0166]

[0167] Among them, LAI t-1 is the LAI of the crop on the day before the simulation, and Rs is the senescence rate of the leaves;

[0168] Therefore, the leaf area index LAI of the crop is t Leaf Area Index LAI t-1 The change in leaf area index on that day is ΔLAI + , ΔLAI - Adding them together gives:

[0169] LAI t =LAI t-1 +ΔLAI + -ΔLAI -

[0170] (3) Stem weight simulation

[0171] When the sugar cane growth enters the stem extension stage, the daily biomass of the stem is calculated by multiplying the daily increase in aboveground biomass and the daily biomass allocation ratio to the stem:

[0172] ΔDSM=ΔDAM×P stk , if SMT>S stk

[0173] Among them, ΔDAM is the daily increase in aboveground biomass of sugar cane, P stk is the dynamic allocation ratio of aboveground biomass increment to stems every day, S stk The accumulated temperature threshold required for sugar cane to start stem elongation;

[0174] The start time of sugarcane stem extension can be determined by the number of leaves on the main bud. When the 10th true leaf on the main bud is fully expanded, the internodes at the base begin to joint and elongate rapidly. At this time, the sugarcane begins to enter the stem extension period. The leaf stripping operation at the early stage of sugarcane stem extension causes the plant leaf area to stop growing. Therefore, the accumulated temperature corresponding to the maximum LAI simulated by the model is used as the accumulated temperature required for the sugarcane to start stem extension, P stk Calculation formula P stk =min[0.8,max(0,F stk )], where 0.8 is the maximum proportion of the daily aboveground biomass that can be allocated to the stems, and F stk It is a logarithmic function used to describe the proportion of stem biomass distribution, and its value is regulated by the accumulated temperature;

[0175] F stkThe calculation formula is as follows:

[0176] Wherein, a is related to the biomass allocation ratio of sugar cane stalks, and the biomass allocated to the stalks daily will decrease as the value of a increases. In step S4, the specific steps of optimizing and calibrating the parameters of the SAFY-Sugar model are as follows:

[0177] S41 obtains the initial parameter values ​​or value ranges of the SAFY-Sugar model. Some parameters of the SAFY-Sugar model can be empirically obtained from literature research, including the meteorological efficiency factor ε c , initial aboveground biomass DAM0, extinction coefficient K ext , specific leaf area SLA, minimum, maximum and optimum temperature for crop growth (T min / T max / T opt In addition, some model parameters are difficult to obtain on the ground and need to be obtained using model optimization algorithms. They can be further divided into two categories according to different determining factors: one is related to crop varieties and generally remains constant for the same crop, including the leaf biomass allocation coefficient Pl a and Pl b , the accumulated temperature required for aging S TT , senescence rate Rs, and the stalk biomass distribution coefficient a in the SAFY-Sugar model; another type of parameters is related to environmental information and management measures. Since the parameter values ​​vary with the planting environment and there are spatial differences, the values ​​need to be adjusted separately for each sample plot, including the crop emergence date D0 and the effective light energy utilization rate ELUE. The SAFY-Sugar model parameter values ​​or value ranges are as follows:

[0178]

[0179] Table 2 Model parameter values ​​and value ranges

[0180] S42 model parameter univariate optimization calibration. Daily average temperature data and shortwave solar radiation data are used as model input data, and the model parameters are optimized and calibrated using the shuffled complex evolution method developed at the University of Arizona (SCE-UA). Data assimilation based on the SCE-UA algorithm aims to find the optimal parameters of the model so that the LAI simulated by the model and the corresponding remote sensing inversion LAI have the best consistency, that is, to minimize the cost function of the two calculations.

[0181] The remote sensing inversion LAI data in S2 is used as a single variable to perform data assimilation with the SAFY-Sugar model. The model parameter optimization process is divided into two rounds. In the first round of assimilation, five parameters related to crop varieties need to be determined: a Pl b , S TT The remote sensing inversion LAI data corresponding to each sample plot is assimilated with the SAFY-Sugar model multiple times to obtain the six parameters Pl corresponding to each sample plot. a Pl b , S TT , Rs, a, D0 and ELUE, for Pl a Pl b , S TT The optimization results of , Rs and a are summarized and statistically analyzed, and the median of the statistical results of each parameter is used as the optimal value of the parameter in the study area. In the second round of assimilation, the results of the first round of assimilation are used to optimize Pl a Pl b , S TT The values ​​of Rs and a are fixed, and the remote sensing inverted LAI data corresponding to each sample plot are assimilated with the original model again to determine the D0 and ELUE of each sample plot.

[0182] The cost function F1 used in the assimilation process is calculated as follows:

[0183]

[0184] Among them, LAI i,mod is the LAI simulated by the model; LAI i,obs is the LAI inverted from remote sensing data; n is the number of samples of LAI inverted from remote sensing data.

[0185] S43 Model parameter bivariate optimization calibration. The model parameters were optimized using the remote sensing inversion LAI and the ground-measured stem weight data corresponding to each sample plot as bivariate data. In the first round of assimilation, some sample plots were randomly selected for bivariate data assimilation, and the Pl a Pl b , S TT The average value of the five parameters, Rs and a, is taken as the optimal value of the parameter in the study area. a , Pl b , S TT The values ​​of , Rs and a are fixed, and a second round of assimilation is performed to determine the D0 and ELUE of each sample plot. The bivariate data assimilation strategy of the SAFY-Sugar model selects three samples each time for parameter optimization and calibration, and the cost function F2 calculation formula used is:

[0186]

[0187] Where n1 and n2 are the number of samples of remote sensing inversion of LAI and ground measurement of stem weight, respectively; LAI sim and LAI obs They are the LAI simulated by the SAFY-Sugar model and the LAI inverted by remote sensing; DSM sim and DSM obs are the stem weight simulated by the SAFY-Sugar model and the stem weight measured on the ground, respectively; and They are the average values ​​of LAI retrieved from remote sensing and the average values ​​of stem weight measured on the ground for each plot, respectively.

[0188] In step S5, the specific steps of the SAFY-Sugar model sample plot and regional scale simulation are:

[0189] S51 uses the results of S4 univariate data assimilation and bivariate data assimilation calibration to set the model parameters, and runs the SAFY-Sugar model with daily average temperature data and shortwave solar radiation data as model input data, simulating the LAI, total aboveground biomass and stem weight of sugar cane in each sample plot at the sample plot scale.

[0190] S52 uses ground-measured LAI, total dry weight, and stem dry weight data to verify the accuracy of the model plot-scale simulation results and obtain the optimal assimilation method and model parameters.

[0191] S53 fuses the remote sensing images covering the study area, masks the area using the spatial distribution map of sugar cane planting in the study area, and uses the LAI inversion model obtained in S2 to calculate the LAI time series during the sugar cane growing period in the study area.

[0192] S54 simulates the regional scale using the optimal model parameters obtained in S52. a Pl b , S TT The parameter values ​​of Rs and a were kept constant in the same study area. The S42 method was used to assimilate the LAI time series data of the study area with the SAFY-Sugar model as a single variable, and the values ​​of the seedling emergence date D0 and the effective light energy utilization rate ELUE were calibrated pixel by pixel.

[0193] S55 uses the daily average temperature data and shortwave solar radiation data of the study area and the model parameters obtained by optimization and calibration to run the model pixel by pixel, complete the simulation study at the regional scale, and obtain the spatial distribution map of sugar cane yield.

[0194] like Figure 2As shown, a remote sensing estimation device for sugar cane yield based on the SAFY-Sugar model, the device comprising:

[0195] The acquisition module is used to observe the growth of sugar cane and measure the biomass of sugar cane in the study area; obtain the daily average temperature data and short-wave solar radiation data during the growth period of sugar cane in the study area; obtain the Sentinel-2 image time series and perform preprocessing;

[0196] The first construction module is used to use the Sentinel-2 image time series to obtain the remote sensing information of the observation plots in the study area, select several vegetation indices, construct a semi-empirical leaf area index LAI inversion model based on the Beer-Lambert law, and use the least squares method and the global optimal algorithm to solve the model parameters of the semi-empirical LAI inversion model; based on the ground observation leaf area index LAI data obtained by observing the growth of sugar cane and measuring the biomass of sugar cane in the study area in the acquisition module and the cross-validation method, the inversion result of the semi-empirical LAI inversion model is verified for accuracy, and the inversion ability of each vegetation index to the semi-empirical LAI inversion model is evaluated to obtain the optimal LAI inversion model;

[0197] The second building block is used to improve the SAFY model by combining the sugar cane growth process and the sugar cane stalk biomass distribution method, and construct the SAFY-Sugar (the Simple Algorithm for Yield Estimation model-Sugar) model, which uses the accumulated temperature to regulate the biomass accumulation process of the sugar cane stalk, so that the model can simulate the daily increase in biomass of the sugar cane stalk part;

[0198] The first optimization module is used to obtain the initial parameter values ​​or the value range of the initial parameters of the SAFY-Sugar model, using the daily average temperature data and the short-wave solar radiation data obtained in the acquisition module as the input data of the SAFY-Sugar model, and using the composite hybrid evolution algorithm SCE-UA (shuffled complex evolution method developed at the University of Arizona) to optimize and calibrate the parameters of the SAFY-Sugar model based on the data assimilation strategy to obtain the model parameter values;

[0199] The third construction module is used to run the SAFY-Sugar model using the model parameter values ​​to obtain the simulation results of the SAFY-Sugar model on the sugar cane LAI, total aboveground biomass and stem weight in the observation sample plot, use ground measurement data to verify the accuracy of the simulation results, obtain the optimal model parameter values ​​of the SAFY-Sugar model, run the SAFY-Sugar model on the study area, and obtain the spatial distribution map of sugar cane yield per unit area in the study area.

[0200] Further, the acquisition module is specifically used to: uniformly select a number of sugar cane planting plots of a first preset size in the study area, arrange a plurality of sample plots of a second preset size in each plot, and measure the multi-stage sugar cane planting density information, longitude and latitude coordinates, canopy natural height, effective stem height, base diameter and LAI according to the growth stage of the sugar cane, wherein the planting density information is the number of sugar canes planted in an area of ​​a third preset size in each sample plot, and after sampling the sugar cane from the sample plot, measure the fresh weight and dry weight in the laboratory;

[0201] Select the meteorological station or the adjacent meteorological station in the study area to obtain the daily average temperature data and the daily short-wave solar radiation data during the sugar cane growing period;

[0202] Sentinel-2L2A images covering the sugarcane growing period of the study area were obtained and resampled to a resolution of 20 meters using SNAP software.

[0203] Furthermore, the first construction module is specifically used to select the normalized vegetation index NDVI (Normalized Difference Vegetation Index), the normalized water index NDWI1 (Normalized Difference Water Index), the chlorophyll index CI green (Chlorophyll Index-Green), Two-band Enhanced Vegetation Index EVI2 (Two-band Enhanced Vegetation Index), Red Edge Normalized Difference Vegetation Index NDVI RE (Red-edge NormalizedDifference Vegetation Index) and red-edge chlorophyll vegetation index CI RE (Chlorophyll Index Red-edge), using the pre-processed Sentinel-2 image time series to calculate the value of the vegetation index corresponding to the observation plot, and construct the semi-empirical LAI inversion model based on the Beer-Lambert law;

[0204] Wherein, the normalized difference vegetation index NDVI=(NIR-RED) / (NIR+RED);

[0205] The normalized water index NDWI1=(NIR-SWIR1) / (NIR+SWIR1);

[0206] Chlorophyll Index CI green =NIR / GREEN-1;

[0207] Two-band enhanced vegetation index EVI2 = 2.5 (NIR-RED) / (NIR + 2.4RED + 1);

[0208] Red Edge Normalized Difference Vegetation Index (NDVI) RE =(NIR-RE) / (NIR+RE);

[0209] Red Edge Chlorophyll Index CI RE =NIR / RE-1;

[0210] Among them, RED is the surface reflectance of the B4 red band of the Sentinel-2 image; NIR is the surface reflectance of the B8a near-infrared band of the Sentinel-2 image; RE is the surface reflectance of the B6 red edge band of the Sentinel-2 image; SWIR1 is the surface reflectance of the first short-wave infrared band B11 of the Sentinel-2 image;

[0211] The semi-empirical LAI inversion model expression is:

[0212]

[0213] Where, VI ∞ is the vegetation index when LAI is close to saturation, VI bs K is the vegetation index of bare soil, which is calculated based on the bare soil image reflectance before sugar cane planting according to the growth phenological period. VI is the extinction coefficient of vegetation, ranging from 0.2 to 2.0;

[0214] The least squares method and global optimization algorithm are used to solve the model parameters. The global optimization algorithm starts from different starting points to find multiple local solutions to the problem and obtains the global optimal solution rather than the local optimal solution. The least squares method minimizes the sum of squares of the residuals between the model simulation results and the actual ground observation results by continuously updating the model parameter values ​​after the parameter value range and initial value are given:

[0215]

[0216] In the formula, xdata is the given input data, F(x,xdata) is the output result of the model for the input data, and ydata is the real observation result;

[0217] The ground-based LAI data were used to verify the accuracy of the model simulation results: the 10-fold cross-validation method was used to randomly divide the ground-based LAI data into 10 parts by non-repeated sampling. Nine parts of the data were randomly selected each time for model training, and the remaining part was used to verify the model simulation results and evaluate the performance of the current model. The accuracy verification indicators include the determination coefficient R 2 , root mean square error RMSE, standard root mean square error nRMSE and RPD indicators;

[0218] The calculation formula of each indicator is expressed as:

[0219]

[0220]

[0221]

[0222]

[0223] Among them, y o,i and m,i are the ground observation and model simulation values ​​at sample point i, and are the means of ground observations and model simulations, respectively, n is the number of samples used for validation, and SD is the standard deviation of ground observations;

[0224] The six semi-empirical LAI inversion models constructed by the vegetation indices are compared, and the semi-empirical LAI inversion model with the best accuracy verification result is selected as the optimal LAI inversion model.

[0225] Furthermore, the second building block is specifically used for

[0226] (1) Aboveground biomass simulation

[0227] During the entire growth and development process of crops, the accumulation of aboveground biomass will continue from crop emergence D0 until the crop is completely senescent. The daily accumulated biomass ΔDAM depends on the effective light energy utilization rate ELUE and the photosynthetically active radiation APAR absorbed by the canopy, and is affected by the temperature stress factor F T (T a ) regulates the growth of biomass, and ΔDAM can be expressed by the above determining factors as:

[0228] ΔDAM=ELUE×F T (T a )×APAR

[0229] in,

[0230] APAR=FAPAR×g c ×Rg

[0231] In the formula, photosynthetically active radiation APAR is expressed as light energy interception efficiency FAPAR, meteorological efficiency factor ε c As a function of the daily total solar shortwave radiation Rg, the Beer-Lambert law describes the nonlinear relationship between FAPAR and LAI. FAPAR is expressed by LAI as:

[0232]

[0233] Among them, K ext is the extinction coefficient, LAI t-1 is the leaf area index of the crop on day t-1;

[0234] Considering that different temperatures will affect the accumulation rate of aboveground biomass, the model uses the maximum temperature T max , minimum temperature T min Optimum temperature To pt The average temperature of the day T a The relationship between the temperature stress factor is expressed as:

[0235]

[0236] In the formula, β is the power of the polynomial, usually expressed as a quadratic polynomial; the highest and lowest temperatures are the temperature thresholds for crop growth. When the average temperature exceeds the temperature threshold, the crops will stop growing;

[0237] (2) Simulation of dynamic growth of LAI

[0238] The SAFY-Sugar model divides the crop growth period into the growth stage and the senescence stage. It simulates the dynamic growth process of leaves by increasing and decreasing LAI in steps of days, and divides different growth stages by accumulated temperature SMT. The calculation formula of accumulated temperature SMT is as follows:

[0239]

[0240] The product of biomass ΔDAM, leaf biomass allocation function Pl and specific leaf area SLA is calculated as:

[0241] ΔLAI + =ΔDAM×Pl×SLA

[0242] The leaf biomass allocation function Pl consists of two parameters Pl a and Plb The composition is expressed in the model as a function of accumulated temperature. The value range of Pl is 0-1. The proportion of biomass allocated to leaves by the model is close to (1-Pl a ), and it decreases with the growth of the crop until the leaves stop growing. The calculation formula of Pl is as follows:

[0243]

[0244] When the accumulated temperature exceeds the accumulated temperature threshold S required for aging TT When the crop begins to enter the senescence stage, the daily decrease in leaf area index ΔLAI - for:

[0245]

[0246] Among them, LAI t-1 is the LAI of the crop on the day before the simulation, and Rs is the senescence rate of the leaves;

[0247] Therefore, the leaf area index LAI of the crop is t Leaf Area Index LAI t-1 The change in leaf area index on that day is ΔLAI + , ΔLAI - Adding them together gives:

[0248] LAI t =LAI t-1 +ΔLAI + -ΔLAI -

[0249] (3) Stem weight simulation

[0250] When the sugar cane growth enters the stem extension stage, the daily biomass of the stem is calculated by multiplying the daily increase in aboveground biomass and the daily biomass allocation ratio to the stem:

[0251] ΔDSM=ΔDAM×P stk , if SMT>S stk

[0252] Among them, ΔDAM is the daily increase in aboveground biomass of sugar cane, P stk is the dynamic allocation ratio of aboveground biomass increment to stems every day, S stk The accumulated temperature threshold required for sugar cane to start stem elongation;

[0253] The start time of sugarcane stem extension can be determined by the number of leaves on the main bud. When the 10th true leaf on the main bud is fully expanded, the internodes at the base begin to joint and elongate rapidly. At this time, the sugarcane begins to enter the stem extension period. The leaf stripping operation at the early stage of sugarcane stem extension causes the plant leaf area to stop growing. Therefore, the accumulated temperature corresponding to the maximum LAI simulated by the model is used as the accumulated temperature required for the sugarcane to start stem extension, P stk Calculation formula P stk =min[0.8,max(0,F stk )], where 0.8 is the maximum proportion of the daily aboveground biomass that can be allocated to the stems, and F stk It is a logarithmic function used to describe the proportion of stem biomass distribution, and its value is regulated by the accumulated temperature;

[0254] F stk The calculation formula is as follows:

[0255] Among them, a is related to the proportion of biomass allocated to the sugar cane stalks. The daily biomass allocated to the stalks will decrease as the value of a increases.

[0256] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A remote sensing estimation method for sugar cane yield based on the SAFY-Sugar model, characterized in that: The method comprises: S1. Observe the growth of sugar cane and measure the biomass of sugar cane in the study area; obtain the daily average temperature data and short-wave solar radiation data during the growth period of sugar cane in the study area; obtain the Sentinel-2 image time series and perform preprocessing; S2, using the Sentinel-2 image time series to obtain remote sensing information of the observation plots in the study area, selecting several vegetation indices, constructing a semi-empirical leaf area index LAI inversion model based on the Beer-Lambert law, and solving the model parameters of the semi-empirical leaf area index LAI inversion model using the least squares method and the global optimal algorithm; based on the ground observation leaf area index LAI data obtained by observing the growth of sugar cane and measuring the biomass of sugar cane in the study area in step S1 and the cross-validation method, the inversion result of the semi-empirical leaf area index LAI inversion model is verified for accuracy, and the inversion ability of each vegetation index to the semi-empirical leaf area index LAI inversion model is evaluated to obtain the optimal LAI inversion model; S3. Improve the SAFY model by combining the growth process of sugar cane and the biomass distribution method of sugar cane stalks, construct the SAFY-Sugar model, and use the accumulated temperature to regulate the biomass accumulation process of sugar cane stalks, so that the model can simulate the daily increase in biomass of sugar cane stalks; S4, obtaining the initial parameter value or the value range of the initial parameter of the SAFY-Sugar model, using the daily average temperature data and the shortwave solar radiation data obtained in step S1 as input data of the SAFY-Sugar model, and using the composite hybrid evolutionary algorithm SCE-UA to optimize and calibrate the parameters of the SAFY-Sugar model based on the data assimilation strategy to obtain the optimal model parameters; S5. Use the model parameters to run the SAFY-Sugar model to obtain simulation results of the SAFY-Sugar model on the sugar cane LAI, total aboveground biomass and stem weight in the observation plot, use ground measurement data to verify the accuracy of the simulation results, obtain the optimal parameter values ​​of the SAFY-Sugar model, run the SAFY-Sugar model on the study area based on the optimal parameter values, and obtain a spatial distribution map of sugar cane yield per unit area in the study area; The step S1 specifically includes: S11, uniformly selecting a number of sugar cane planting plots of a first preset size in the study area, arranging a plurality of sample plots of a second preset size in each plot, and measuring the multi-stage sugar cane planting density information, longitude and latitude coordinates, natural canopy height, effective stem height, base diameter and LAI according to the growth stage of the sugar cane, wherein the planting density information is the number of sugar canes planted in an area of ​​a third preset size in each sample plot, and after sampling the sugar cane from the plot, measuring the fresh weight and dry weight in the laboratory; S12, selecting a meteorological station or a neighboring meteorological station in the study area to obtain daily average temperature data and daily shortwave solar radiation data during the sugar cane growing period; S13. Acquire Sentinel-2 L2A images covering the sugarcane growing period in the study area and resample to 20-meter resolution using SNAP software; The step S2 specifically includes: S21, select the normalized vegetation index NDVI, the normalized water index NDWI1, and the chlorophyll index CI green , two-band enhanced vegetation index EVI2, red edge normalized vegetation index NDVI RE and red-edge chlorophyll vegetation index CI RE , using the preprocessed Sentinel-2 image time series to calculate the value of the vegetation index corresponding to the observation plot, and constructing the semi-empirical leaf area index LAI inversion model based on the Beer-Lambert law; Wherein, the normalized difference vegetation index NDVI=(NIR-RED) / (NIR+RED); The normalized water index NDWI1=(NIR-SWIR1) / (NIR+SWIR1); Chlorophyll Index CI green =NIR / GREEN-1; Two-band enhanced vegetation index EVI2 = 2.5 (NIR-RED) / (NIR + 2.4RED + 1); Red Edge Normalized Difference Vegetation Index (NDVI) RE =(NIR-RE) / (NIR+RE); Red Edge Chlorophyll Index CI RE =NIR / RE-1; Among them, RED is the surface reflectance of the B4 red band of the Sentinel-2 image; NIR is the surface reflectance of the B8a near-infrared band of the Sentinel-2 image; RE is the surface reflectance of the B6 red edge band of the Sentinel-2 image; SWIR1 is the surface reflectance of the first short-wave infrared band B11 of the Sentinel-2 image; The semi-empirical leaf area index LAI inversion model expression is: Where, VI ∞ is the vegetation index when the leaf area index LAI is close to saturation, VI bs K is the vegetation index of bare soil, which is calculated based on the bare soil image reflectance before sugar cane planting according to the growth phenological period. VI is the extinction coefficient of vegetation, ranging from 0.2 to 2.0; The least squares method and global optimization algorithm are used to solve the model parameters. The global optimization algorithm starts from different starting points to find multiple local solutions to the problem and obtains the global optimal solution rather than the local optimal solution. The least squares method minimizes the sum of squares of the residuals between the model simulation results and the actual ground observation results by continuously updating the model parameter values ​​after the parameter value range and initial value are given: In the formula, xdata is the given input data, F(x,xdata) is the output result of the model for the input data, ydata is the real observation result, and xdata i is the given input data at sample point i, ydata i is the actual observation result at sample point i; S22. Use ground-based LAI data to verify the accuracy of model simulation results: Use the 10-fold cross-validation method to randomly divide the ground-based LAI data into 10 parts by non-repeated sampling. Each time, nine parts of the data are randomly selected for model training, and the remaining part is used to verify the model simulation results and evaluate the performance of the current model. Among them, the accuracy verification indicators include the determination coefficient R 2 , root mean square error RMSE, standard root mean square error nRMSE and RPD indicators; The calculation formula of each indicator is expressed as: Among them, y o,i is the ground observation value at sample point i, y m,i is the value simulated by the model at sample point i, is the average value of ground observations, is the mean of the model simulation, n is the number of samples used for validation, and SD is the standard deviation of the ground observations; S23, comparing the six semi-empirical leaf area index LAI inversion models constructed by the vegetation indices, and selecting the semi-empirical leaf area index LAI inversion model with the best accuracy verification result as the optimal LAI inversion model; The step S3 specifically includes: (1) Aboveground biomass simulation During the entire growth and development process of crops, the accumulation of aboveground biomass will continue from the crop emergence date D0 until the crop is completely senescent. The daily accumulated biomass ΔDAM depends on the effective light energy utilization rate ELUE and the photosynthetically active radiation APAR absorbed by the canopy, and is affected by the temperature stress factor F. T (T a ) regulates the growth of biomass, and ΔDAM can be expressed by the temperature stress factor as: ΔDAM=ELUE×F T (T a )×APAR in, APAR=FAPAR×ε c ×Rg In the formula, photosynthetically active radiation APAR is expressed as light energy interception efficiency FAPAR, meteorological efficiency factor ε c As a function of the daily total solar shortwave radiation Rg, the Beer-Lambert law describes the nonlinear relationship between FAPAR and LAI. FAPAR is expressed by LAI as: Among them, K ext is the extinction coefficient, LAI t-1 is the leaf area index of the crop on day t-1; Considering that different temperatures will affect the accumulation rate of aboveground biomass, the model uses the maximum temperature T max , minimum temperature T min , optimum temperature T opt The average temperature of the day T a The relationship between the temperature stress factor is expressed as: In the formula, β is the power of the polynomial, usually expressed as a quadratic polynomial; the highest and lowest temperatures are the temperature thresholds for crop growth. When the average temperature exceeds the temperature threshold, the crop will stop growing; (2) Simulation of dynamic growth of LAI The SAFY-Sugar model divides the crop growth period into the growth stage and the senescence stage. It simulates the dynamic growth process of leaves by increasing and decreasing LAI in steps of days, and divides different growth stages by accumulated temperature SMT. The calculation formula of accumulated temperature SMT is as follows: The product of biomass ΔDAM, leaf biomass allocation function Pl and specific leaf area SLA is calculated as: ΔLAI + =ΔDAM×Pl×SLA The leaf biomass allocation function Pl consists of two parameters Pl a and Pl b The composition is expressed in the model as a function of accumulated temperature. The value range of Pl is 0-1. The proportion of biomass allocated to leaves by the model is close to (1-Pl a ), and it decreases with the growth of the crop until the leaves stop growing. The calculation formula of Pl is as follows: When the accumulated temperature exceeds the accumulated temperature threshold S required for aging TT When the crop begins to enter the senescence stage, the daily decrease in leaf area index ΔLAI - for: Among them, LAI t-1 is the LAI of the crop on the day before the simulation, and Rs is the senescence rate of the leaves; Therefore, the leaf area index LAI of the crop is t Leaf Area Index LAI t-1 The change in leaf area index on that day is ΔLAI + , ΔLAI - Adding them together gives: FOR t =FOR t-1 +ΔLAI + -ΔLAI - (3) Stem weight simulation When the sugar cane growth enters the stem extension stage, the daily biomass of the stem is calculated by multiplying the daily increase in aboveground biomass and the daily biomass allocation ratio to the stem: ΔDSM=ΔDAM×P stk ,if SMT>S stk Among them, ΔDAM is the daily increase in aboveground biomass of sugar cane, P stk is the dynamic allocation ratio of aboveground biomass increment to stems every day, S stk The accumulated temperature threshold required for sugar cane to start stem elongation; The start time of sugarcane stem extension can be determined by the number of leaves on the main bud. When the 10th true leaf on the main bud is fully expanded, the internodes at the base begin to joint and elongate rapidly. At this time, the sugarcane begins to enter the stem extension period. The leaf stripping operation at the early stage of sugarcane stem extension causes the plant leaf area to stop growing. Therefore, the accumulated temperature corresponding to the maximum LAI simulated by the model is used as the accumulated temperature required for the sugarcane to start stem extension, P stk Calculation formula P stk =min[0.8,max(0,F stk )], where 0.8 is the maximum proportion of the daily aboveground biomass that can be allocated to the stems, and F stk It is a logarithmic function used to describe the proportion of stem biomass distribution, and its value is regulated by the accumulated temperature; F stk The calculation formula is as follows: Among them, a is related to the proportion of biomass allocated to the sugar cane stalks. The daily biomass allocated to the stalks will decrease as the value of a increases.

2. The method according to claim 1, characterized in that The step S4 specifically includes: S41. The parameters of the SAFY-Sugar model include the meteorological efficiency factor ε c , initial aboveground biomass DAM0, extinction coefficient K ext , specific leaf area SLA, minimum temperature for crop growth T min , maximum temperature T max , optimum temperature T opt ; The parameters of the SAFY-Sugar model also include the leaf biomass partition coefficient Pl a and Pl b , Accumulated temperature required for aging S TT , senescence rate Rs, the stalk biomass distribution coefficient a, crop emergence date D0 and effective light energy utilization efficiency ELUE; S42, taking the daily average temperature data and the shortwave solar radiation data as model input data, using the composite hybrid evolutionary algorithm SCE-UA to optimize and calibrate the parameters of the SAFY-Sugar model, and determining the optimal parameters of the SAFY-Sugar model based on the data assimilation of the SCE-UA algorithm, so that the LAI simulated by the model and the corresponding remote sensing inversion LAI have the best consistency, that is, the cost function of the calculation of the two is minimized; The remote sensing inversion LAI data in step S2 is used as a single variable to perform data assimilation with the SAFY-Sugar model. The model parameter optimization process is divided into two rounds. In the first round of assimilation, five parameters Pl related to crop varieties are determined. a Pl b , S TT , Rs and a values; The remote sensing inverted LAI data corresponding to each sample plot were assimilated with the SAFY-Sugar model multiple times to obtain the six parameters Pl corresponding to each sample plot. a Pl b , S TT , Rs, a, D0 and ELUE, for Pl a Pl b , S TT The optimization results of , Rs and a are summarized and counted, and the median of the statistical results of each parameter is taken as the optimal value of the parameter in the study area; In the second round of assimilation, the results of the first round of assimilation are used to a Pl b , S TT The values ​​of Rs and a are fixed, and the remote sensing inversion LAI data corresponding to each sample plot is assimilated with the SAFY-Sugar model again to determine the D0 and ELUE of each sample plot; The cost function F1 used in the assimilation process is calculated as follows: Among them, LAI i,mod is the LAI at sample point i simulated by the model, LAI i,obs is the LAI at sample point i inverted from remote sensing data, and n is the number of samples of LAI inverted from remote sensing; S43, using the remote sensing inversion LAI and the ground measurement stem weight data corresponding to each sample plot as dual variables to optimize the parameters of the SAFY-Sugar model. In the first round of assimilation, some sample plots are randomly selected each time for dual variable data assimilation, and the Pl obtained by assimilation of each sample plot is used. a Pl b , S TT The average value of the five parameters, Rs and a, is taken as the optimal value of the parameter in the study area. a Pl b , S TT The values ​​of Rs and a are fixed, and a second round of assimilation is performed again to determine the D0 and ELUE of each sample plot. The bivariate data assimilation strategy of the SAFY-Sugar model selects three sample plots each time for parameter optimization and calibration, and the cost function F2 calculation formula used is: Where n1 is the number of samples of LAI inverted by remote sensing, and n2 is the number of samples of stem weight measured on the ground; LAI sim,i is the LAI at sample point i simulated by the SAFY-Sugar model; LAI obs,i is the LAI at sample point i derived from remote sensing; DSM sim,i The stem dry weight at sample point i simulated by the SAFY-Sugar model, DSM obs,i is the stem weight measured at sampling point i on the ground; is the average value of LAI inverted by remote sensing in each sample plot, The mean value of stem weight was measured above ground for each plot.

3. The method according to claim 2, characterized in that The step S5 specifically includes: S51, respectively using the results of the univariate data assimilation and bivariate data assimilation calibration in step S4 to set the parameters of the SAFY-Sugar model, using the daily average temperature data and the shortwave solar radiation data as model input data to run the SAFY-Sugar model, and simulating the LAI, total aboveground biomass and stem weight of sugar cane in each sample plot at the sample plot scale; S52, using ground-measured LAI, total dry weight and stem dry weight data to verify the accuracy of the model quadrat scale simulation results, and obtain the optimal assimilation method and the optimal model parameters of the SAFY-Sugar model; S53, fusing the remote sensing images covering the study area, masking the study area using the sugar cane planting spatial distribution map of the study area, and calculating the LAI time series of the sugar cane growing period in the study area using the optimal LAI inversion model obtained in step S2; S54, simulating the study area with the optimal SAFY-Sugar model obtained in step S52, and a Pl b , S TT The parameter values ​​of Rs and a are kept constant in the same study area, and the LAI time series data of the study area are used as a single variable to perform data assimilation with the SAFY-Sugar model, and the values ​​of the seedling emergence date D0 and the effective light energy utilization rate ELUE are calibrated pixel by pixel; S55. Use the daily average temperature data and shortwave solar radiation data of the study area and the parameters of the SAFY-Sugar model obtained by optimization and calibration to run the SAFY-Sugar model pixel by pixel to obtain the spatial distribution map of sugar cane yield in the study area.

4. A remote sensing estimation device for sugar cane yield based on the SAFY-Sugar model, characterized in that: The device comprises: The acquisition module is used to observe the growth of sugar cane and measure the biomass of sugar cane in the study area; obtain the daily average temperature data and short-wave solar radiation data during the growth period of sugar cane in the study area; obtain the Sentinel-2 image time series and perform preprocessing; The first construction module is used to use the Sentinel-2 image time series to obtain the remote sensing information of the observation plots in the study area, select several vegetation indices, construct a semi-empirical leaf area index LAI inversion model based on the Beer-Lambert law, and use the least squares method and the global optimal algorithm to solve the model parameters of the semi-empirical leaf area index LAI inversion model; based on the ground observation leaf area index LAI data obtained by observing the growth of sugar cane and measuring the biomass of sugar cane in the study area in the acquisition module and the cross-validation method, the inversion result of the semi-empirical leaf area index LAI inversion model is verified for accuracy, and the inversion ability of each vegetation index to the semi-empirical leaf area index LAI inversion model is evaluated to obtain the optimal LAI inversion model; The second building block is used to improve the SAFY model by combining the sugar cane growth process and the sugar cane stalk biomass allocation method, construct the SAFY-Sugar model, and use the accumulated temperature to regulate the biomass accumulation process of the sugar cane stalk, so that the model can simulate the daily increase in biomass of the sugar cane stalk; The first optimization module is used to obtain the initial parameter values ​​or the value range of the initial parameters of the SAFY-Sugar model, use the daily average temperature data and the shortwave solar radiation data obtained in the acquisition module as input data of the SAFY-Sugar model, and use the composite hybrid evolutionary algorithm SCE-UA to optimize and calibrate the parameters of the SAFY-Sugar model based on the data assimilation strategy to obtain the model parameter values; The third construction module is used to run the SAFY-Sugar model using the model parameter values ​​to obtain the simulation results of the SAFY-Sugar model on the sugar cane LAI, total aboveground biomass and stem weight in the observation plot, use ground measurement data to verify the accuracy of the simulation results, obtain the optimal model parameter values ​​of the SAFY-Sugar model, run the SAFY-Sugar model in the study area, and obtain the spatial distribution map of sugar cane yield per unit area in the study area; The acquisition module is specifically used to: uniformly select a number of sugar cane planting plots of a first preset size in the research area, arrange a plurality of sample plots of a second preset size in each sample plot, and measure the multi-stage sugar cane planting density information, longitude and latitude coordinates, canopy natural height, effective stem height, base diameter and LAI according to the growth stage of the sugar cane, wherein the planting density information is the number of sugar canes planted in an area of ​​a third preset size in each sample plot, and after sampling the sugar cane from the sample plot, measure the fresh weight and dry weight in the laboratory; Select the meteorological station or the adjacent meteorological station in the study area to obtain the daily average temperature data and the daily short-wave solar radiation data during the sugar cane growing period; Acquire Sentinel-2 L2A images covering the sugarcane growing period in the study area and resample to 20-meter resolution using SNAP software; The first construction module is specifically used to select the normalized vegetation index NDVI, the normalized water index NDWI1, the chlorophyll index CI green , two-band enhanced vegetation index EVI2, red edge normalized vegetation index NDVI RE and red-edge chlorophyll vegetation index CI RE , using the preprocessed Sentinel-2 image time series to calculate the value of the vegetation index corresponding to the observation plot, and constructing the semi-empirical LAI inversion model based on the Beer-Lambert law; Wherein, the normalized difference vegetation index NDVI=(NIR-RED) / (NIR+RED); The normalized water index NDWI1=(NIR-SWIR1) / (NIR+SWIR1); Chlorophyll Index CI green =NIR / GREEN-1; Two-band enhanced vegetation index EVI2 = 2.5 (NIR-RED) / (NIR + 2.4RED + 1); Red Edge Normalized Difference Vegetation Index (NDVI) RE =(NIR-RE) / (NIR+RE); Red Edge Chlorophyll Index CI RE =NIR / RE-1; Among them, RED is the surface reflectance of the B4 red band of the Sentinel-2 image; NIR is the surface reflectance of the B8a near-infrared band of the Sentinel-2 image; RE is the surface reflectance of the B6 red edge band of the Sentinel-2 image; SWIR1 is the surface reflectance of the first short-wave infrared band B11 of the Sentinel-2 image; The semi-empirical leaf area index LAI inversion model expression is: Where, VI ∞ is the vegetation index when LAI is close to saturation, VI bs K is the vegetation index of bare soil, which is calculated based on the bare soil image reflectance before sugar cane planting according to the growth phenological period. VI is the extinction coefficient of vegetation, ranging from 0.2 to 2.0; The least squares method and global optimization algorithm are used to solve the model parameters. The global optimization algorithm starts from different starting points to find multiple local solutions to the problem and obtains the global optimal solution rather than the local optimal solution. The least squares method minimizes the sum of squares of the residuals between the model simulation results and the actual ground observation results by continuously updating the model parameter values ​​after the parameter value range and initial value are given: In the formula, xdata is the given input data, F(x,xdata) is the output result of the model for the input data, ydata is the real observation result, and xdata i is the given input data at sample point i, ydata i is the actual observation result at sample point i; The ground-based LAI data were used to verify the accuracy of the model simulation results: the 10-fold cross-validation method was used to randomly divide the ground-based LAI data into 10 parts by non-repeated sampling. Nine parts of the data were randomly selected each time for model training, and the remaining part was used to verify the model simulation results and evaluate the performance of the current model. The accuracy verification indicators include the determination coefficient R 2 , root mean square error RMSE, standard root mean square error nRMSE and RPD indicators; The calculation formula of each indicator is expressed as: Among them, y o,i is the ground observation value at sample point i, y m,i is the value simulated by the model at sample point i, is the average value of ground observations, is the mean of the model simulation, n is the number of samples used for validation, and SD is the standard deviation of the ground observations; Comparing the semi-empirical leaf area index (LAI) inversion models constructed by the six vegetation indices, the semi-empirical leaf area index (LAI) inversion model with the best accuracy verification result is selected as the optimal LAI inversion model; The second building block is specifically used for: (1) Aboveground biomass simulation During the entire growth and development process of crops, the accumulation of aboveground biomass will continue from the crop emergence date D0 until the crop is completely senescent. The daily accumulated biomass ΔDAM depends on the effective light energy utilization rate ELUE and the photosynthetically active radiation APAR absorbed by the canopy, and is affected by the temperature stress factor F. T (T a ) regulates the growth of biomass, and ΔDAM can be expressed by the temperature stress factor as: ΔDAM=ELUE×F T (T a )×APAR in, APAR=FAPAR×ε c ×Rg In the formula, photosynthetically active radiation APAR is expressed as light energy interception efficiency FAPAR, meteorological efficiency factor ε c As a function of the daily total solar shortwave radiation Rg, the Beer-Lambert law describes the nonlinear relationship between FAPAR and LAI. FAPAR is expressed by LAI as: Among them, K ext is the extinction coefficient, LAI t-1 is the leaf area index of the crop on day t-1; Considering that different temperatures will affect the accumulation rate of aboveground biomass, the model uses the maximum temperature T max , minimum temperature T min , optimum temperature T opt The average temperature of the day T a The relationship between the temperature stress factor is expressed as: In the formula, β is the power of the polynomial, usually expressed as a quadratic polynomial; the highest and lowest temperatures are the temperature thresholds for crop growth. When the average temperature exceeds the temperature threshold, the crop will stop growing; (2) Simulation of dynamic growth of LAI The SAFY-Sugar model divides the crop growth period into the growth stage and the senescence stage. It simulates the dynamic growth process of leaves by increasing and decreasing LAI in steps of days, and divides different growth stages by accumulated temperature SMT. The calculation formula of accumulated temperature SMT is as follows: The product of biomass ΔDAM, leaf biomass allocation function Pl and specific leaf area SLA is calculated as: ΔLAI + =ΔDAM×Pl×SLA The leaf biomass allocation function Pl consists of two parameters Pl a and Pl b The composition is expressed in the model as a function of accumulated temperature. The value range of Pl is 0-1. The proportion of biomass allocated to leaves by the model is close to (1-Pl a ), and it decreases with the growth of the crop until the leaves stop growing. The calculation formula of Pl is as follows: When the accumulated temperature exceeds the accumulated temperature threshold S required for aging TT When the crop begins to enter the senescence stage, the daily decrease in leaf area index ΔLAI - for: Among them, LAI t-1 is the LAI of the crop on the day before the simulation, and Rs is the senescence rate of the leaves; Therefore, the leaf area index LAI of the crop is t Leaf Area Index LAI t-1 The change in leaf area index on that day is ΔLAI + , ΔLAI - Adding them together gives: FOR t =FOR t-1 +ΔLAI + -ΔLAI - (3) Stem weight simulation When the sugar cane growth enters the stem extension stage, the daily biomass of the stem is calculated by multiplying the daily increase in aboveground biomass and the daily biomass allocation ratio to the stem: ΔDSM=ΔDAM×P stk ,if SMT>S stk Among them, ΔDAM is the daily increase in aboveground biomass of sugar cane, P stk is the dynamic allocation ratio of aboveground biomass increment to stems every day, S stk The accumulated temperature threshold required for sugar cane to start stem elongation; The start time of sugarcane stem extension can be determined by the number of leaves on the main bud. When the 10th true leaf on the main bud is fully expanded, the internodes at the base begin to joint and elongate rapidly. At this time, the sugarcane begins to enter the stem extension period. The leaf stripping operation at the early stage of sugarcane stem extension causes the plant leaf area to stop growing. Therefore, the accumulated temperature corresponding to the maximum LAI simulated by the model is used as the accumulated temperature required for the sugarcane to start stem extension, P stk Calculation formula P stk =min[0.8,max(0,F stk )], where 0.8 is the maximum proportion of the daily aboveground biomass that can be allocated to the stems, and F stk It is a logarithmic function used to describe the proportion of stem biomass distribution, and its value is regulated by the accumulated temperature; F stk The calculation formula is as follows: Among them, a is related to the proportion of biomass allocated to the sugar cane stalks. The daily biomass allocated to the stalks will decrease as the value of a increases.

Images