CN110222870A - Assimilate the Regional Fall Wheat yield estimation method of satellite fluorescence data and crop growth model - Google Patents

Abstract

An embodiment of the present invention provides a regional winter wheat yield estimation method assimilating satellite fluorescence data and crop growth models, including: model parameter sensitivity analysis and model parameter calibration; in the outer layer of the SIF data assimilation system, the SIF satellite remote sensing method is used according to the parameter calibration results The data is assimilated with the SCOPE model, and the daily cumulative GPP is obtained from the assimilated model; in the inner layer of the SIF data assimilation system, the daily cumulative GPP obtained from the outer layer of the system is assimilated and the daily cumulative GPP simulated by the WOFOST model; coupled with numerical weather forecast data Obtain the winter wheat yield forecast results of the area to be estimated produced by the WOFOST model. The embodiment of the present invention introduces SIF satellite remote sensing data into the process model, establishes a mechanism relationship with yield from the perspective of photosynthesis, and carries out regional yield forecasting from the optimization of photosynthesis process of crops.

Description

Regional Winter Wheat Yield Estimation Method Based on Assimilation of Satellite Fluorescence Data and Crop Growth Model

technical field

The embodiments of the present invention relate to the field of agricultural technology, and more specifically, relate to a method for estimating regional winter wheat yield by assimilating satellite fluorescence data and crop growth models.

Background technique

In order to estimate the yield of winter wheat, a series of remote sensing and mechanism process model data assimilation systems based on leaf area index, soil moisture, reflectance or vegetation index as assimilation variables have been developed in the prior art. Important results have been achieved in production forecasting. However, these assimilative variables in the prior art cannot establish a mechanistic relationship with yield from the perspective of photosynthesis, so there are limitations in quantitative monitoring of winter wheat growth process and yield forecast.

Contents of the invention

In order to solve the above problems, an embodiment of the present invention provides a regional winter wheat yield estimation method that overcomes the above problems or at least partially solves the above problems by assimilating satellite fluorescence data and crop growth models.

An embodiment of the present invention provides a method for estimating regional winter wheat production by assimilating satellite fluorescence data and crop growth models. Yield estimation area; according to the environmental index data, the area to be estimated yield is divided into multiple planting subregions, and sunlight-induced chlorophyll fluorescence (Sun-Induced chlorophyll Fluorescence, SIF) remote sensing data with a set resolution of the planting subregions are obtained; the WOFOST model is globally Sensitivity analysis to obtain a parameter set sensitive to vegetation gross primary productivity (GrossPrimary Productivity, GPP) and yield; perform a global sensitivity analysis to the SCOPE model to obtain a parameter set sensitive to SIF and GPP; Sensitive parameters in the parameter set are used for parameter calibration and uncertainty assessment; in the outer layer of the SIF data assimilation system, the SIF remote sensing data is assimilated with the SCOPE model after parameter calibration, and the daily cumulative GPP is obtained from the assimilated model; in the SIF data Assimilate the inner layer of the system, assimilate the daily cumulative GPP obtained from the outer layer of the system and the daily cumulative GPP simulated by the WOFOST model to obtain the optimized daily cumulative GPP; based on the optimized GPP, drive the WOFOST model with weather forecast data to obtain WOFOST The forecast results of winter wheat production in the area to be estimated produced by the model.

The method for estimating regional winter wheat yield by assimilating satellite fluorescence data and crop growth model provided by the embodiment of the present invention uses the method of assimilating SIF remote sensing data and mechanism process model, introduces SIF remote sensing data into the process model, and establishes a relationship with yield from the perspective of photosynthesis Mechanistic connection, from optimizing the photosynthesis process of crops to carry out regional yield forecasting.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings that are required in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention, and those skilled in the art can also obtain other drawings based on these drawings without creative efforts.

Fig. 1 is the schematic flow chart of the regional winter wheat production estimation method of assimilating satellite fluorescence data and crop growth model provided by the embodiment of the present invention;

Fig. 2 is the schematic diagram of regional yield ensemble forecast provided by the embodiment of the present invention;

Fig. 3 is a schematic diagram of the probability forecast of yield higher than 8000kg/ha provided by the embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are the Some, but not all, embodiments are invented. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In recent years, the emergence and rapid development of SIF remote sensing products that are closely related to the physiological process of vegetation, as well as the maturation of the process model for dynamically simulating canopy chlorophyll fluorescence, have brought new opportunities for yield forecasting of remote sensing data assimilation—the SIF remote sensing data is introduced into the process model to carry out regional yield forecasting from optimizing the photosynthesis process of crops.

Based on this, an embodiment of the present invention provides a method for estimating regional winter wheat yield by assimilating satellite fluorescence data and crop growth models, as shown in Figure 1. The method includes but is not limited to the following steps:

Step 101: Divide the winter wheat planting area into a plurality of grids, and select the grid whose area ratio of winter wheat is larger than the set ratio as the area to be estimated.

Specifically, before step 101, the multi-temporal Landsat 8 OLI data and Sentinel 2 A/B optical data of the winter wheat planting area during the winter wheat growth period can be obtained. Based on the above optical data, the spatial distribution map of winter wheat with a certain resolution (for example, 30m) in the study area (ie, winter wheat planting area) was obtained by using decision tree classification. Then divide the spatial distribution map of winter wheat into multiple grids with a predetermined resolution (for example, 1 km), calculate the area ratio of winter wheat in each grid (that is, the area ratio of winter wheat), and select the area ratio greater than the set ratio ( For example, 80%) grids are used for production estimation. Among them, the selected grid is the production area to be estimated. Through the above processing, the grid has a higher purity, and the yield estimation result is more accurate.

Step 102: Divide the area to be estimated into a plurality of planting zones according to the environmental index data, and obtain SIF remote sensing data with a set resolution of the planting zones.

Specifically, after the area to be estimated is obtained in step 101, the area to be estimated can be further divided to obtain multiple planting divisions. Wherein, the environmental index data is used to reflect the environmental conditions of the production area to be estimated, and the environmental index may include meteorological conditions, planting patterns, disaster frequencies, yield levels, etc., which are not limited in this embodiment of the present invention. Therefore, the entire area to be estimated can be divided into different planting zones according to the environmental conditions of each area in the area to be estimated. Obtain SIF remote sensing data of the area to be estimated. Among them, SIF data are closely related to the physiological process of vegetation, and can directly reflect the photosynthesis of winter wheat.

Step 103 , perform a global sensitivity analysis on the WOFOST model to obtain a parameter set sensitive to GPP and yield; perform a global sensitivity analysis on the SCOPE model to obtain a parameter set sensitive to SIF and GPP.

Among them, the WOFOST model is a crop growth model, and the SCOPE model can simulate the canopy SIF data. The EFAST sensitivity analysis method can be used to study the global sensitivity of the WOFOST model to GPP and yield, and the global sensitivity of the SCOPE model to SIF and GPP. The union of parameter sets sensitive to GPP and yield of WOFOST model is taken as the sensitive parameter set of WOFOST model, and the union of parameter sets sensitive to SIF and GPP of SCOPE model is taken as the sensitive parameter set of SCOPE model.

Step 104 , performing parameter calibration and uncertainty assessment on the sensitive parameters in the sensitive parameter set according to the planting zones.

Specifically, based on the observation data of agricultural meteorological stations, flux tower data, and measured SIF data in the area to be estimated, for each planting area, the Markov Chain Monte Carlo (MCMC) method is used to estimate the sensitive parameters Perform parameter calibration and uncertainty assessment.

Step 105, in the outer layer of the SIF data assimilation system, assimilate the SIF remote sensing data and the SCOPE model after parameter calibration, and obtain the daily cumulative GPP from the assimilated model; in the inner layer of the SIF data assimilation system, assimilate the outside of the system The daily cumulative GPP obtained by layering with the daily cumulative GPP simulated by the WOFOST model was used to obtain the optimized daily cumulative GPP.

Specifically, according to the SIF data assimilation system of internal and external nesting and parameter transfer, the assimilation is performed in two steps: in the outer layer, the SIF data and SCOPE model obtained in step 102 can be assimilated using the four-dimensional variational (4Dvar) data assimilation method, and the assimilation The subsequent SCOPE model obtains daily cumulative GPP and its uncertainty. In the inner layer, the Ensemble Kalman Filter (EnKF) assimilation method is used to assimilate the daily cumulative GPP obtained from the outer layer and the daily cumulative GPP simulated by WOFOST to obtain the assimilated daily cumulative GPP.

Step 106, based on the optimized GPP, drive the WOFOST model with weather forecast data, and obtain the output forecast result of winter wheat in the region to be estimated produced by the WOFOST model.

Wherein, the weather forecast data is the weather data in the production forecast period. Specifically, after substituting the assimilated GPP into the WOFOST model, the weather forecast data can be used to drive the WOFOST model to perform yield forecasting and obtain winter wheat yield forecasting results.

In this step, the production forecast is mainly driven by meteorological data, but the premise is that the parameters and state variables of the WOFOST model have been calibrated and assimilated. During the forecast period (such as 2 months before the maturity period), the WOFOST model can be brought into the weather forecast data for forecasting every time it is updated with the assimilated GPP. The meteorological data used in the forecast is divided into two parts according to the forecast time: for the meteorological data before the forecast time, the data generated by the surface climate data daily value data set can be used, and for the meteorological data in the forecast period after the current time, it is divided into recent The 15-day forecast (TIGGER weather dataset) and the forecast for the next 15 days (matched with the reanalysis data from 1979-2016 to find the most similar historical weather data to the weather dataset).

Since meteorological data are input into WOFOST model in the form of a set, and its output is a yield forecast set, the median value of the forecast set can be selected as the output forecast result of winter wheat yield. In addition, a certain standard can also be set, for example, (8000 (kg/ha)) is the standard of yield per unit area (per unit yield), and the probability of per unit yield higher than this standard is calculated based on the forecast set.

The regional winter wheat yield estimation method of assimilating satellite fluorescence data and crop growth model provided by the embodiment of the present invention, through the research of assimilating SIF remote sensing data and mechanism process model data, introduces SIF remote sensing data into the process model, and establishes it from the perspective of photosynthesis Mechanistic connection with yield, and regional yield forecast from optimizing photosynthesis process of crops.

Based on the content of the above-mentioned embodiments, as an optional embodiment, a method for dividing the area to be estimated into multiple planting zones according to the environmental index data is provided, including: constructing a Thiessen polygon with the agricultural gas station as the node , and according to the data recorded by the agricultural gas station; according to the recorded data, the Thiessen polygons corresponding to the stations with similar environmental indicators were merged to obtain the partition result; among them, the environmental indicators include meteorological conditions, planting patterns, disaster frequency and production level.

Construct a Thiessen polygon with the agricultural gas station as the node, and combine the Thiessen polygons corresponding to the similar stations for many years of environmental indicators (environmental indicators include meteorological conditions, planting patterns, disaster frequency, production level, etc.) according to the recorded data of the agricultural gas stations, Get the partition result.

Based on the content of the above-mentioned embodiments, as an optional embodiment, a method for obtaining SIF remote sensing data with a set resolution of the planting area is provided, including but not limited to the following steps: setting the resolution of the MODIS data to winter wheat The resolution used for yield forecasting, that is, the set resolution, downscales the TROPOMI SIF remote sensing data with an original resolution of 7km*3.5km during the growth period to the set resolution; wherein, the MODIS data includes leaf Area index LAI data, evapotranspiration ET data and land surface temperature LST data.

In this step, the TROPOMI (Sentinel 5 Satellite monitoring instrument) Sun-Induced chlorophyll Fluorescence (Sun-Inducedchlorophyll Fluorescence, SIF) data is downscaled to SIF data with a set resolution (for example, 1km).

Based on the content of the above-mentioned embodiment, as an optional embodiment, the set resolution is 1km*1km; correspondingly, a method of downscaling the TROPOMI SIF remote sensing data during the growth period to the SIF data with set resolution is provided methods, including but not limited to the following steps:

Step 1. Resampling the sunlight-induced SIF data during the growth period into SIF data with a spatial resolution of 5km*5km.

Specifically, the TROPOMI SIF data are resampled into SIF time series images with a spatial resolution of 5km*5km in the following way: The projection used by the grid is the center point of the Albers equal-area projection with a size of 5km*5km), then the SIF value of the pixel will contribute to the average value of the grid SIF (for example: the center point of a certain grid unit is covered by 2 TROPOMI SIF cell coverage, where the SIF value of one cell is 0.1 and the other is 0.3, then the SIF value of the grid cell is (0.1+0.3)/2=0.2).

Step 2. Reproject the original 1km*1km MODIS data into Albers equal-area projection, and resample into 5km*5km MODIS data by linear aggregation method.

In other words, the adopted MODIS products (LAI, ET, LST) (ie, the original MODIS data) were reprojected into Albers equal-area projection with a spatial resolution of 1km*1km. Then generate 5km*5km MODIS data through linear aggregation. Among them, during aggregation, if there are less than 12 MODIS data, the aggregation will be abandoned.

Step 3. Establish a space-time window for each pixel of SIF data of 5km*5km, bring the MODIS data in the space-time window into the formula, and solve the unknown parameters in the formula.

Wherein, b _i is an unknown parameter, i=1, 2, 3, 4, 5, 6.

Among them, by assuming that SIF data can be converted from MODIS LAI, ET, and LST data, the above formula (1) can be obtained.

Step 4. Based on the solved unknown parameters, bring the original MODIS data into the formula to obtain 5*5 SIF data with a resolution of 1km*1km.

Specifically, before step 4, the unknown parameters in the above formula (1) can be solved first, and the solution can be done in the following way: First, a space-time window is established for each 5km*5km SIF pixel. The time window is set to one month; the setting of the space window is divided into two steps: the first step is to establish a rectangular area composed of 5*5 pixels centered on this pixel (ie 25km*25km). In the second step, 11 pixels adjacent to the target pixel are selected from the rectangular area, plus the target pixel itself, the spatial range corresponding to a total of 12 pixels constitutes the spatial window of the pixel. Then, after the space-time window is established, the MODIS data in the space-time window is brought into the formula (1), and the sum of the squares of its residuals and the SIF data is used as the cost function, and the formula (1) is solved by using the L-BFGS-B optimization algorithm 6 unknown parameters in .

After solving 6 unknown parameters, the original MODIS data before aggregation is brought into the formula (1), and 5*5 SIF data with a resolution of 1km can be obtained directly.

Based on the content of the above-mentioned embodiments, as an optional embodiment, a global sensitivity analysis is performed on the WOFOST model, and before obtaining a parameter set sensitive to GPP and output, it also includes: performing a spatio-temporal analysis on the surface meteorological data of the area to be estimated. Interpolation, obtaining daily meteorological data and hourly meteorological data continuously distributed in time and space; using the daily meteorological data as the weather driving data of the WOFOST model, and using the hourly weather data as the weather driving data of the SCOPE model. Specifically, temporal-spatial interpolation is performed on surface meteorological data to obtain hourly and daily meteorological data continuously distributed in time and space, which are respectively used as meteorological driving data for WOFOST model and SCOPE model.

Specifically, based on the sensitivity analysis results, determine the parameter set to be calibrated; determine the value interval of each sensitive parameter according to the model definition or the actual situation, and define the parameter prior distribution as a uniform distribution on the interval; Calculate the mean and standard deviation of the observed data (yield or leaf area index), and define the likelihood function as the Gaussian distribution of the obtained mean and standard deviation; perform Monte Carlo sampling from the parameter prior distribution, and each sampling result is brought into The model obtains the corresponding model output, puts the output into the likelihood function to obtain the likelihood probability of each sampling, accepts the sampling value according to the Metropolis criterion, and judges whether the Markov chain is convergent according to the variance ratio method, based on the previous sampling Continuously perform new sampling until the Markov chain converges, so as to obtain parameter posterior samples. Finally, the median value of the parameter posterior samples is used as the calibration result of the model parameters, and the root mean square error (RMSE) of the parameter samples is calculated as the quantitative description index of uncertainty.

Based on the content of the above-mentioned embodiments, as an optional embodiment, a method is provided in the outer layer of the SIF data assimilation system to assimilate the SIF remote sensing data and the parameter-calibrated SCOPE model, and obtain the daily accumulation from the assimilated model. GPP; in the inner layer of the SIF data assimilation system, assimilate the daily cumulative GPP obtained by the outer layer of the system and the daily cumulative GPP simulated by the WOFOST model, and obtain the optimized daily cumulative GPP method, including but not limited to the following steps:

Step 1. In the outer layer of the SIF data assimilation system, input hourly meteorological data into the SCOPE model to simulate fluorescence hour by hour to obtain simulated SIF data; establish a four-dimensional variational cost function between simulated SIF data and SIF remote sensing data, and use SCE - The UA algorithm optimizes the cost function to obtain the optimized SCOPE parameter set, which is substituted into the SCOPE model to obtain the assimilated SCOPE model, and the daily cumulative GPP is obtained based on the assimilated SCOPE model.

Specifically, the SIF data assimilation system is divided into an inner layer and an outer layer. In the outer layer, the optimal estimated values of sensitive parameters obtained from MCMC calibration (ie parameter calibration results), other parameters necessary for the model, and hourly meteorological data obtained by interpolation are input into the SCOPE model to simulate fluorescence hour by hour. Establish the 4Dvar cost function of the 1km resolution SIF obtained by TROPOMI downscaling (take the set resolution as 1km as an example) and the SIF simulated by the model during the satellite transit period, and use the SCE-UA algorithm to optimize the cost function. The hourly GPP simulated by the global optimal parameters is used to obtain the daily cumulative GPP and its uncertainty.

Step 2. In the inner layer of the SIF data assimilation system, input the daily meteorological data and the initial parameter set of Gaussian disturbance of the sensitive parameters into the calibrated WOFOST model to generate daily cumulative GPP; The daily cumulative GPP whose certainty is less than the set threshold is assimilated with the daily cumulative GPP simulated by the WOFOST model, and the assimilated daily cumulative GPP is obtained, which is input into the WOFOST model to continue to run forward; If the uncertainty of the daily cumulative GPP obtained by the assimilated SCOPE model is greater than the threshold value, the assimilation will not be performed, and the WOFOST model will continue to forecast until the uncertainty of the daily cumulative GPP obtained by the SCOPE model is smaller than the threshold value. Assimilation, the daily cumulative GPP after assimilation is obtained.

Specifically, in the inner layer, input the daily meteorological data, the optimal value of the general sensitive parameters calibrated by MCMC (that is, the parameter calibration results), and the Gaussian disturbance of the strong sensitive parameters (higher than the threshold index of the set total sensitivity, such as 0.05) The initial parameter set and other parameters required by the model are input into the WOFOST model, and the daily cumulative GPP forecast set is generated day by day. Since SCOPE can output the daily GPP after assimilation, for the WOFOST model, it is equivalent to having "observations" every day, so there is no question of "whether the observations exist" during assimilation. Only select the target daily cumulative GPP with less uncertainty (less than the threshold) after SCOPE assimilation, that is, "effective observation", and perform EnKF assimilation with the daily cumulative GPP simulated by WOFOST to obtain the analysis set of daily cumulative GPP and substitute it into the WOFOST model to continue run forward. If the day-to-day cumulative GPP after the assimilation of the SCOPE model has a large uncertainty (greater than the threshold), the assimilation will not be performed, and the WOFOST model will continue to forecast until the SCOPE model assimilates the GPP with a smaller uncertainty (less than the threshold). Assimilate.

Based on the content of the above-mentioned embodiment, as an optional embodiment, after obtaining the winter wheat yield forecast result of the area to be estimated output output by the WOFOST model, it also includes: running grid by grid, obtaining the forecast set and per unit yield of each grid If the probability is higher than the unit yield threshold, complete the space mapping of winter wheat yield estimation. In other words, run grid by grid, output the forecast set and the probability that the unit yield is higher than 8000 (kg/ha) (that is, the unit yield threshold), and complete the yield spatial mapping; see Figure 2 and Figure 3.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims (8)

1. a kind of Regional Fall Wheat yield estimation method for assimilating satellite fluorescence data and crop growth model characterized by comprising

Winter wheat planting area is divided into multiple grid, select winter wheat area accounting be greater than setting ratio grid as to The yield by estimation region；

According to environmental index data by the region division to be assessed be multiple plantation subregions, obtain plantation subregion setting differentiate The sunlight-induced chlorophyll fluorescence SIF remotely-sensed data of rate；

Global sensitivity analysis is carried out to WOFOST model, obtains the parameter set sensitive to GPP, yield；SCOPE model is carried out Global sensitivity analysis obtains the sensitive parameter set to SIF, GPP；

Parameter calibration and uncertain assessment are carried out to the sensitive parameter that the sensitive parameter is concentrated according to plantation subregion；

In the outer layer of SIF data assimilation system, SIF remotely-sensed data and the SCOPE model after parameter calibration are assimilated, by same SCOPE model after change obtains day accumulation GPP；In the internal layer of SIF data assimilation system, assimilate the system outer layer resulting day The day for accumulating GPP and WOFOST modeling accumulates GPP, and the day after being assimilated accumulates GPP；

The described wait estimate of WOFOST model output is obtained with weather forecast data-driven WOFOST model based on the GPP after assimilation The Prediction For Winter Wheat Production result in producing region domain.

2. Regional Fall Wheat yield estimation method according to claim 1, which is characterized in that described to be incited somebody to action according to environmental index data The region division to be assessed is multiple plantation subregions, comprising:

Multiple Thiessen polygons are constructed by node of agriculture gas website；

Environmental index is established, the corresponding Thiessen polygon of agriculture gas website of close environmental index is merged, is obtained wait assess Regional compartmentalization result.

3. Regional Fall Wheat yield estimation method according to claim 1, which is characterized in that the setting for obtaining plantation subregion The SIF remotely-sensed data of resolution ratio, comprising:

It is resolution ratio used by Prediction For Winter Wheat Production by the resolution setting of MODIS data, that is, sets resolution ratio, will give birth to The TROPOMI SIF remotely-sensed data NO emissions reduction that original resolution in phase is 7km*3.5km is the setting resolution ratio；Wherein, The MODIS data include leaf area index LAI data, evapotranspiration ET data and surface temperature LST data.

4. Regional Fall Wheat yield estimation method according to claim 3, which is characterized in that the resolution ratio that sets is 1km* TROPOMI SIF remotely-sensed data NO emissions reduction in breeding time is the SIF data for setting resolution ratio by 1km, comprising:

It is the SIF data that spatial resolution is 5km*5km by TROPOMI SIF remotely-sensed data resampling in breeding time；

It is osteopetrosis equivalent projection by the MODIS data re-projection of original 1km*1km, and is adopted again by linear polymerization method Sample is the MODIS data of 5km*5km；

Empty window when being established for the pixel of the SIF data of each 5km*5km, by 5km*5km when being located at this in empty window MODIS data bring into formula, solve the unknown parameter in formula；

Wherein, b_iFor unknown parameter to be solved, i=1,2,3,4,5,6；

Based on the unknown parameter solved, the MODIS data of original 1km*1km are brought into formula, obtain 5*5 resolution ratio For the SIF data of 1km*1km.

5. Regional Fall Wheat yield estimation method according to claim 1, which is characterized in that carried out to WOFOST model global quick Perceptual analysis, before obtaining the parameter set sensitive to GPP, yield, further includes:

Temporal-spatial interpolating is carried out to the ground meteorological data in the region to be assessed, obtains the meteorological number day by day of space and time continuous distribution According to the meteorological data by hour；

Using the meteorological data day by day as the meteorological driving data of WOFOST model, and by the meteorological number by hour According to the meteorological driving data as SCOPE model.

6. Regional Fall Wheat yield estimation method according to claim 1, which is characterized in that concentrated to the sensitive parameter quick Feel parameter and carry out parameter calibration and uncertain assessment, comprising:

The prior distribution for determining the value interval of each sensitive parameter, and defining the sensitive parameter is uniformly dividing on section Cloth；

Monte Carlo is carried out in this prior distribution, is brought the sampled result obtained after each sampling into model and is corresponded to Model export result；

It brings the output result into likelihood function and obtains the likelihood probability sampled every time, receive sampling according to Metropolis criterion Value, and judge whether Markov Chain restrains according to Variance ratio method；

New sampling is constantly carried out on the basis of preceding primary sampling until Markov Chain convergence, obtains the sensitive parameter Posteriority sample；

Using the intermediate value of posteriority sample as the parameter calibration result.

7. Regional Fall Wheat yield estimation method according to claim 1, which is characterized in that described in SIF data assimilation system Outer layer, institute's SIF remotely-sensed data and the SCOPE model after parameter calibration are assimilated, obtained by the SCOPE model after assimilating Day accumulation GPP；In the internal layer of SIF data assimilation system, assimilate the resulting day accumulation GPP of the system outer layer and WOFOST model The day of simulation accumulates GPP, and the day after being assimilated accumulates GPP, comprising:

In the outer layer of SIF data assimilation system, meteorological data SCOPE model will be input to by hour simulation fluorescence per hour, Obtain simulation SIF data；The four-dimensional variation cost function of simulation SIF data and SIF remotely-sensed data is established, and is calculated using SCE-UA Method optimizes the cost function, the SCOPE parameter set after being optimized, and is substituted into after SCOPE model obtains assimilation SCOPE model, based on after assimilation SCOPE model obtain day accumulate GPP；

In the internal layer of SIF data assimilation system, by the initial parameter collection of day meteorological data and the Gauss disturbance of the sensitive parameter It is input to WOFOST model after demarcating, generates day accumulate GPP day by day；It will be by the resulting uncertainty of SCOPE model after assimilating Day accumulation GPP less than the day accumulation GPP and WOFOST modeling of given threshold is assimilated, the day accumulation after being assimilated This is input to WOFOST model to continue to move forwards by GPP；In system operation, if the same day is by the SCOPE after assimilating The uncertainty of the resulting day accumulation GPP of model is greater than the threshold value, then without assimilation, WOFOST model continues pre- forward Report is assimilated until being less than the threshold value by the uncertainty of the resulting day accumulation GPP of SCOPE model, again after obtaining assimilation Day accumulate GPP.

8. Regional Fall Wheat yield estimation method according to claim 1, which is characterized in that the acquisition WOFOST model output The region to be assessed Prediction For Winter Wheat Production result after, further includes:

Single-frame net operation, the forecast ensemble and per unit area yield that obtain each grid are higher than the probability of given threshold, complete winter wheat and estimate Produce space mapping.