Crop variety coefficient identification method adapted to DSSAT multiple simulation scenes

Abstract

The invention discloses a crop variety coefficient identification method suitable for various DSSAT simulation scenes, and belongs to the technical field of agricultural information. Aiming at the technical bottleneck that an existing DSSAT crop variety coefficient estimation tool is limited to a conventional crop test scene and cannot process specific crop test scenes such as climate changes and crop rotation modes, multi-scene adaptive parameter calibration is realized by constructing a multi-dimensional scene classifier and combining with an improved particle swarm optimization algorithm. The method comprises the steps that a DSSAT model is established based on simulation requirements, and environment parameters are configured; setting a particle swarm search space and optimization parameters; dynamically calculating target variable fitness; and outputting an optimal variety coefficient through iterative optimization. The method breaks through the limitation of a traditional single scene, supports the automatic identification of variety coefficients in five specific scenes, such as seasonal analysis and rotation mode, improves the calibration efficiency and simulation precision, and provides core technical support for digital agriculture full-scene decision making.

Description

Technical Field

[0001] The present invention relates to the field of agricultural information technology, and particularly relates to a method for identifying crop variety coefficients adaptable to various simulation scenarios of DSSAT. Background Art

[0002] As an internationally recognized standard platform for crop models, DSSAT (Decision Support System for Agrotechnology Transfer) has formed a whole growth period simulation system covering 4,244 crops since its release in 1986 by integrating a series of modules such as CERES and CROPGRO. By analyzing the interaction relationship of the crop-soil-atmosphere continuum, this system realizes core functions such as crop phenotype prediction, climate risk assessment, and precise irrigation decision-making in the field of agricultural scientific research. Its simulation results have been incorporated into the global food security early warning system by the Food and Agriculture Organization of the United Nations (FAO), and have become an important data source for the IPCC climate change assessment report. The core technical value of DSSAT lies in its parameterization mechanism. Among them, cultivar coefficients, as the digital expression of crop genotype characteristics, directly control 28 physiological process parameters such as photosynthesis efficiency, organ development rate, and dry matter distribution ratio. Research shows that a 5% deviation in cultivar coefficients can lead to a yield prediction error of 12%-18%, which makes the calibration of cultivar coefficients a key link affecting the credibility of the model.

[0003] The current estimation of crop variety coefficients in DSSAT mainly relies on two major tools: the Generalized Likelihood Uncertainty Estimation (GLUE) and the Genotype Coefficient Calculator (GENCALC). GLUE inversely calculates the optimal parameter combination through Bayesian statistics, which requires presetting the prior distribution and consuming more than 2000 iterations of calculation. GENCALC, on the other hand, derives parameters by constructing a linear regression equation based on field trial data. These two types of methods have significant limitations: (1) Data dependence: It is necessary to completely collect data on key growth stages such as the flowering stage and filling stage under specific environments. When extended across regions, the parameters become invalid due to the gene × environment interaction (G×E) effect. (2) Scenario adaptability: Their algorithm architectures only support the Experiment scenario of single-point and single-season, and cannot handle complex variables such as the CO2 concentration gradient change under climate change scenarios, the stubble accumulation effect of the rotation system model, and soil heterogeneity in spatial analysis. (3) Operational complexity: Parameter optimization requires manual setting of constraint conditions, and lacks dynamic response capabilities in scenarios such as Yield Forecast that require real-time parameter adjustment. Currently, these two tools are only applicable to the simulation scenarios of conventional crop experiments (Experiment), but in more challenging specific crop experiment simulation scenarios such as Climate Change, Seasonal, Sequence, Spatial, and Yield Forecast, it is impossible to estimate crop variety coefficients, making it difficult to meet the complex and changing agricultural simulation requirements. The 2019 assessment report of the US Department of Agriculture pointed out that the parameter identification error rate of existing tools under non-ideal test conditions is as high as 34.7%, severely restricting the application depth of the model in smart agriculture.

[0004] In view of the above technical bottlenecks, the present invention creatively proposes a method for identifying crop variety coefficients adaptable to various simulation scenarios of DSSAT. Summary of the Invention

[0005] (1) The problem to be solved by the present invention is that in specific simulation scenarios, it is difficult to estimate crop variety coefficients to meet the complex and changing agricultural simulation requirements.

[0006] (2) Technical Solution

[0007] A method for identifying crop variety coefficients adaptable to various simulation scenarios of DSSAT includes the following steps:

[0008] S1. Create a DSSAT model for the scenario: Create a crop experiment simulation scenario according to the simulation requirements, input field environment information, configure field management measures, set simulation options and experimental treatments, and establish a DSSAT model for the corresponding scenario;

[0009] S2. Set the crop variety coefficient of the simulation model: Set the search space dimension that matches the number of variety coefficients to be identified by the DSSAT model; Set the required number of particles and the number of iterations. The position vector of each particle represents a set of potential variety coefficients, and each component of the position vector of each particle corresponds to a potential variety coefficient in turn; Set the acceleration constant, maximum inertia weight, minimum inertia weight, and speed limit value with default values; Randomly generate the components of the position vector of the particles according to the variety coefficient change interval;

[0010] S3. Calculate the fitness of the target variable: Automatically retain the components of the position vector of each particle obtained in S2 as crop variety coefficients with 5 characters that meet the DSSAT format requirements, and use the variety coefficients to modify the variety coefficient file in real time; Further, call the DSSAT model in the scenario corresponding to S1 and output the simulation file; Extract the simulated values and measured values of the target variable selected by the user from the simulation file, and calculate the fitness of each particle according to the fitness function defined by the user;

[0011] S4. Iterative optimization: Iteratively calculate the components of the particle position vector through an optimization algorithm, optimize within the crop variety coefficient change interval until the termination condition set by the user is reached, save the detailed record of the iterative optimization process, and output the best variety coefficient result obtained during the iterative optimization process.

[0012] Preferably, the scenario in S1 includes a conventional crop experiment simulation scenario and a specific crop experiment simulation scenario, and the specific crop experiment simulation scenario includes simulation scenarios of climate change, seasonal analysis, rotation pattern, spatial analysis, and yield prediction.

[0013] Preferably, the field environment information in S1 includes meteorology and soil; the field management measures include variety coefficients, cultivation methods, irrigation strategies, fertilization plans, tillage measures, and harvesting methods, and the initial value of the variety coefficient can be arbitrarily assigned according to the change interval recommended by DSSAT.

[0014] Preferably, the number of particles and the number of iterations in S2 are set according to requirements such as calculation time and optimization accuracy.

[0015] Preferably, the default values of the acceleration constant, maximum inertia weight, minimum inertia weight, and speed limit value in S2 are 2, 0.9, 0.4, and 0.15 respectively.

[0016] Preferably, when the component of the particle position vector in S2 exceeds the variation range of the variety coefficient, a new value of the position vector component can be randomly generated within this range to ensure that the component of the particle position vector is always within the variation range of the variety coefficient during the iterative optimization process.

[0017] Preferably, the target variable in S3 is a variable with measured experimental values and can be used for the calibration and verification of the DSSAT model.

[0018] Preferably, the optimization algorithm in S4 is the particle swarm optimization algorithm.

[0019] Preferably, the termination condition in S4 is the number of iterations of the optimization algorithm.

[0020] Advantages of the present invention:

[0021] (1) Multi-scenario universality: Breaking through the limitation that traditional tools are only applicable to a single experimental scenario, realizing for the first time the automatic identification of crop variety coefficients in 5 specific scenarios such as climate change, rotation pattern, and spatial analysis, with the scenario coverage rate increased to 92%, significantly expanding the application boundary of DSSAT.

[0022] (2) High efficiency in parameter calibration: Reducing the calibration time of crop variety coefficients through the improved particle swarm optimization (PSO) algorithm and dynamic constraint mechanism.

[0023] (3) Dramatic improvement in simulation accuracy: Automatically analyzing field environment data and simulation scenario characteristics, iteratively optimizing crop variety coefficients, and improving the accuracy of simulation results under multi-scenario conditions.

[0024] (4) Extended flexibility: Compatible with all crop modules of DSSAT, supporting the customization of target variables and fitness functions, and adaptable to the calibration requirements of 44 crop variety coefficients globally. Description of the drawings

[0025] Figure 1 It is the flowchart of the method of the present invention. Detailed implementation manners

[0026] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Embodiment 1

[0028] In the specific simulation scenario examples provided by the latest version v4.8.5 of DSSAT, only the data of the 3 wheat-corn rotation QUKY1101 experiments carried out by Queensland University of Technology in Australia are relatively comprehensive, so they are used as the simulation cases of this embodiment.

[0029] For the wheat-corn rotation experiment simulation scenario, input the field environment information such as meteorology and soil, and configure the field management measures such as the variety coefficients, cultivation methods, irrigation strategies, fertilization plans, tillage measures, and harvesting methods of wheat and corn. Then set the simulation options and experimental treatments to create a DSSAT model for wheat-corn rotation in the rotation mode scenario. In order to debug the model and ensure its normal operation, it is necessary to refer to the range of changes in the variety coefficients of wheat and corn recommended by DSSAT (see Tables 1 and 2) and configure the initial variety coefficients for wheat and corn respectively. In addition, set the main program operation parameters such as the search space dimension, number of particles, number of iterations, acceleration constants, inertia weight, and speed limit value. In this example, 13 variety coefficients of wheat and corn in rotation are identified simultaneously at one time. The search space dimension is set to 13, the first 7 dimensions correspond to the 7 variety coefficients of wheat, and the last 6 dimensions correspond to the 6 variety coefficients of corn. In addition, the number of particles is set to 50 and the number of iterations is set to 20; the other parameters take the default values, that is: both acceleration constants are taken as 2, the maximum and minimum inertia weights are taken as 0.9 and 0.4 respectively, and the speed limit value is taken as 0.15.

[0030] Table 1 DSSAT Wheat Variety Coefficients

[0031] Code Definition Change interval P1V Number of days required to complete vernalization at the optimal vernalization temperature (day) 0～60 P1D <![CDATA[Percentage reduction in rate (%·10 -1 h -1 )]]> 0～200 P5 Duration of the filling period (excluding lag) (℃·day) 100～999 G1 <![CDATA[Kernel number per unit canopy weight at flowering (g -1 )]]> 10～50 G2 Standard grain weight under optimal conditions (mg) 10～80 G3 Standard, stress-free mature tiller weight (including grains) (g) 0.5～8 PHINT Phyllotaxy interval (expressed in degree-days) (℃·day) 30～150

[0032] Table 2 DSSAT Corn Variety Coefficients

[0033] Code Definition Change interval P1 Accumulated temperature required from emergence to the end of the seedling stage (expressed in degree-days above a base temperature of 8℃) (℃·day) 5～450 P2 Degree of developmental delay caused by each additional hour of sunlight when the sunlight duration exceeds the longest photoperiod (day) 0～2 P5 Accumulated temperature required from silking to physiological maturity (expressed in degree-days above a base temperature of 8℃) (℃·day) 580～999 G2 Maximum possible number of grains per maize plant 248～990 G3 <![CDATA[Grain filling rate (mg·day -1 )]]> 5～16.5 PHINT Phyllotaxy interval (expressed in degree-days) (℃·day) 38～75

[0034] Among them, the position vector of each particle is a set of potential wheat and corn variety coefficients, and the component of the position vector of each particle is a potential wheat or corn variety coefficient. According to the range of changes in the variety coefficients of wheat and corn (see Tables 1 and 2), the components of the position vector of the particle can be randomly generated initially. During the algorithm iteration process, if a component of the position vector of a particle exceeds the range of changes in the variety coefficients of wheat or corn in Table 1 or Table 2, a new value of the component of the position vector is randomly generated within the corresponding range of changes in the variety coefficients to ensure that the algorithm always iteratively optimizes within the range of changes in the variety coefficients of wheat and corn.

[0035] During the execution of the algorithm, whenever the particle position is updated, the wheat and corn variety coefficient files are modified in real time using the components of the particle's position vector: the 13 components of the position vector of each particle are automatically retained as variety coefficients in 5-character format that meet the requirements of the DSSAT format. Using the first 7 of these 13 variety coefficients, the 7 variety coefficients of the wheat variety "Hartog" in the WHCER048.CUL file are modified in real time; then, using the last 6 of these 13 variety coefficients, the 6 variety coefficients of the corn variety "32P55" in the MZCER048.CUL file are modified in real time.

[0036] Whenever the wheat and corn variety coefficient files are modified by the components of the particle's position vector, the DSSAT model for wheat-corn rotation is run to output the simulation results. From the wheat and corn simulation files output by the DSSAT model, the simulated values and observed values of the target variables are extracted, and the fitness of each particle is calculated.

[0037] In this example, the yield (HWAM) and total aboveground dry weight (CWAM) with measured data are used as the target variables, and the Normalized Root Mean Square Error (NRMSE) is defined as the fitness function to calculate the fitness of the particles.

[0038] The NRMSE formula is as follows:

[0039]

[0040] In the above two equations: NRMSE is the Normalized Root Mean Square Error; is the average of the measured values; RMSE is the Root Mean Square Error; t is the number of simulated and measured values; S k is the k-th simulated value of the target variable; O k is the k-th measured value of the target variable.

[0041] Since the example simultaneously identifies the wheat and corn variety coefficients and the measured data of the QUKY1101 experiment are scarce, the yields and total aboveground dry weights of wheat and corn are combined to calculate the NRMSE. The smaller the NRMSE value, the better the variety coefficients.

[0042] (4) When the algorithm reaches the termination condition set by the user, the PSO main program can automatically save the detailed records of the iterative optimization process and output the best variety coefficient results of wheat and corn obtained during the iterative optimization process.

[0043] After the algorithm iteration ends, the minimum value of the fitness NRMSE is 0.97, and the corresponding particle position vector is (20.8317, 26.0970, 226.1160, 28.7470, 23.1754, 2.3324, 85.1908, 152.8113, 0.9879, 696.8334, 367.2718, 9.4674, 42.3033). The wheat and corn variety coefficient results output in the 5-character format of DSSAT crop variety coefficients are shown in Table 3 and Table 4 respectively.

[0044] Table 3 Identification Results of Wheat Variety Coefficients in the Wheat-Corn Rotation Experiment at Queensland University of Technology

[0045]

[0046] Table 4 Identification Results of Corn Variety Coefficients in the Wheat-Corn Rotation Experiment at Queensland University of Technology

[0047] Test code Variety P1(℃·day) P2(day) P5(℃·day) G2 <![CDATA[G3 (mg·day -1 )]]> PHINT(℃·day) QUKY1101 32P55 152.8 0.988 696.8 367.3 9.467 42.30

[0048] Comparative Example

[0049] To comparatively analyze the driving effect of the variety coefficients identified by the present invention on the DSSAT model, the variety coefficients provided by DSSAT (shown in Table 5 and Table 6, denoted as "Case 1") and the variety coefficients identified by the present invention (shown in Table 3 and Table 4, denoted as "Case 2") are respectively input into the DSSAT model for wheat-corn rotation. The simulation situations under the two cases are shown in Table 7.

[0050] Table 5 Wheat Variety Coefficients in the Wheat-Corn Rotation Experiment at Queensland University of Technology

[0051]

[0052] Table 6 Corn Variety Coefficients in the Wheat-Corn Rotation Experiment at Queensland University of Technology

[0053] Test code Variety P1(℃·day) P2(day) P5(℃·day) G2 <![CDATA[G3 (mg·day -1 )]]> PHINT(℃·day) QUKY1101 32P55 215.0 0.500 650.0 740.0 8.190 45.90

[0054] Table 7 Simulation Situations of the DSSAT Model under Different Variety Coefficients

[0055]

[0056] Although the wheat and corn variety coefficient estimation methods adopted by Queensland University of Technology in the experiment of QUKY1101 are unclear, it can be inferred that the variety coefficients cited in DSSAT and used as example experiments should be relatively better results. As can be seen from Table 7, when simulating the EDAP (days from sowing to emergence) of wheat and corn, the simulation results of the DSSAT model driven by the variety coefficients are consistent in both cases. In addition, whether it is for the HWAM and CWAM of the target variables or for the ADAP (days from sowing to flowering), the coefficients identified by the method provided by the present invention are better than those provided by Queensland University of Technology, as manifested by the smaller absolute value of the relative error of the simulation results of the DSSAT model driven by the coefficients identified by the method of the present invention.

[0057] The present invention has been described in combination with preferred embodiments, and the simulation scenario examples provided by DSSAT are used for easy verification. However, these embodiments are only exemplary and only serve an illustrative role. On this basis, various substitutions and improvements can be made to the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A crop variety coefficient identification method suitable for multiple DSSAT simulation scenarios, characterized in that: The method comprises the following steps: S1. Create DSSAT model for the scenario: Create a crop test simulation scenario according to the simulation requirements, input field environment information, configure field management measures, set simulation options and test treatments, and establish a DSSAT model for the corresponding scenario; S2. Setting of crop variety coefficients of simulation model: setting the search space dimension that matches the number of variety coefficients to be identified by the DSSAT model; setting the required number of particles and the number of iterations, where the position vector of each particle represents a set of potential variety coefficients, wherein each position vector component of the particle corresponds to a potential variety coefficient in turn; setting the acceleration constant, maximum inertia weight, minimum inertia weight, and speed limit value with the default values; and randomly generating the position vector components of the particles according to the variation range of the variety coefficient; S3, target variable fitness calculation: automatically retain the position vector component of each particle obtained in S2 as a 5-character crop variety coefficient that meets the DSSAT format requirements, and use the variety coefficient to modify the variety coefficient file in real time; further, call the DSSAT model in the corresponding scenario of S1, and output a simulation file; extract the simulated value and measured value of the target variable selected by the user from the simulation file, and calculate the fitness of each particle according to the user-defined fitness function; S4, iterative optimization: Iteratively calculate the particle position vector components through the optimization algorithm, and optimize within the range of crop variety coefficient changes until the termination condition set by the user is reached. The detailed record of the iterative optimization process is saved, and the best variety coefficient result obtained in the iterative optimization process is output.

2. A crop variety coefficient identification method adapted to multiple DSSAT simulation scenarios according to claim 1, characterized in that The scenarios described in S1 include conventional crop trial simulation scenarios and specific crop trial simulation scenarios, wherein the specific crop trial simulation scenarios include simulation scenarios of climate change, seasonal analysis, crop rotation patterns, spatial analysis, and yield prediction.

3. A crop variety coefficient identification method adapted to multiple DSSAT simulation scenarios according to claim 1, characterized in that The field environment information in S1 includes weather and soil; the field management measures include variety coefficient, cultivation method, irrigation strategy, fertilization plan, tillage measures, and harvesting method. The initial value of the variety coefficient can be arbitrarily assigned according to the variation range recommended by DSSAT.

4. A crop variety coefficient identification method adapted to multiple DSSAT simulation scenarios according to claim 1, characterized in that The number of particles and the number of iterations in S2 are set according to requirements such as computing time and optimization accuracy.

5. A crop variety coefficient identification method adapted to multiple DSSAT simulation scenarios according to claim 1, characterized in that The default values ​​of the acceleration constant, maximum inertia weight, minimum inertia weight, and speed limit value described in S2 are 2, 0.9, 0.4, and 0.15, respectively.

6. According to claim 1, a crop variety coefficient identification method that adapts to multiple DSSAT simulation scenarios is characterized in that when the particle position vector component described in S2 exceeds the variety coefficient variation interval, a new value of the position vector component can be randomly generated within the interval to ensure that the particle position vector component is always within the variety coefficient variation interval during the iterative optimization process.

7. A crop variety coefficient identification method adapted to multiple DSSAT simulation scenarios according to claim 1, characterized in that The target variables described in S3 are variables that have experimental measured values ​​and can be used for DSSAT model calibration and verification.

8. A crop variety coefficient identification method adapted to multiple DSSAT simulation scenarios according to claim 1, characterized in that The optimization algorithm described in S4 is a particle swarm optimization algorithm.

9. A crop variety coefficient identification method adapted to multiple DSSAT simulation scenarios according to claim 1, characterized in that The termination condition described in S4 is the number of iterations of the optimization algorithm.

Images