Method for predicting agricultural yield under drought disasters

Abstract

The invention discloses an agricultural yield prediction method under drought disasters, and provides an agricultural yield prediction model under drought disaster stress, the formation time of drought events under different meteorological conditions is predicted through a random forest algorithm, and after the drought formation time is converted into a drought formation rate, an agricultural yield prediction model is provided. And the relationship between the drought formation rate and the yield loss rate is constructed, so that the yield loss of the agricultural crops under the drought disaster can be accurately obtained, and more accurate information is provided for early warning of agricultural drought prevention and prediction of the influence of the crops on the drought disaster.

Description

Technical Field

[0001] The present invention relates to a method for predicting agricultural yields, and more particularly to a method for predicting agricultural yields under drought disasters. Background Art

[0002] The growth of crops generally includes three sensitive phenological periods, namely the vegetative period, the reproductive growth period, and the maturity period. The impact of drought in different periods will result in significant differences in the final yield. The existing prediction of the impact of drought on agricultural yields is to predict crop yields through a crop growth model, which solves the problem of predicting the impact of drought in different phenological periods. However, the accuracy of the prediction results of this method depends on the growth model, and it is necessary to build models for different crops. Moreover, driven by multiple influencing factors, the accuracy of the growth model needs to be improved, and it is necessary to track the growth of crops for a long time and use the growth data of crops to correct the prediction results. In addition, this method does not consider the impact of different crop production systems. Different regions have different farming methods, which are mainly divided into two categories: irrigation and rainfed, and their resistance to drought stress is also different. How to improve the accuracy of predicting agricultural yields under drought disasters and obtain a more accurate situation of crops affected by drought disasters is an important topic in agricultural planting, risk assessment, and drought relief. Summary of the Invention

[0003] Object of the Invention: Aiming at the above problems, the present invention proposes a method for predicting agricultural yields under drought disasters, which can better reflect the impact of drought disasters on crops and more accurately predict the yield loss of crops.

[0004] Technical Solution: The technical solution adopted by the present invention is a method for predicting agricultural yields under drought disasters, including the following steps: Step 1, obtain basic data, where the basic data includes agricultural data, meteorological data, and soil water data; continuously monitor the soil water data at the location of the crop to be predicted, and conduct agricultural yield prediction when the soil water data drops to the drought event threshold or it is confirmed that a drought event has occurred; Step 2, for the agricultural yield prediction when the soil water data drops to the drought event threshold, obtain the drought formation time through the drought formation time prediction method according to the meteorological data and the soil water data; for the agricultural yield prediction when it is confirmed that a drought event has occurred, directly calculate the drought formation time according to the soil water data; calculate the drought formation rate according to the drought formation time; Step 3, according to the drought formation rate, the growth stage of the crop at the start time of the drought event, and the planting system applied by the crop, obtain the predicted yield loss rate of the crop through the yield loss prediction model; The yield loss prediction model is: , Wherein, represents the yield loss rate, and n represents the n growth stages of the crop; represents the sensitivity of the crop to drought at each growth stage; a is a constant representing the planting system applied to the crop; represents the drought formation rate; Step 4, calculate the predicted yield value of the crop according to the predicted yield loss rate of the crop.

[0005] The agricultural data is the historical yield data and phenological data of the crop in the research area; the meteorological data includes precipitation, temperature and evapotranspiration data in the research area; the soil water data uses the surface soil water data.

[0006] The growth stages of the crop include: vegetative period, reproductive growth period and maturity period.

[0007] The drought formation time prediction method includes: predicting the drought formation time through a random forest model according to meteorological data and soil water data; the establishment of the random forest model includes: obtaining historical soil water data and performing drought identification to obtain the drought formation time of each drought event; obtaining historical meteorological data to obtain the average precipitation and average temperature and average evapotranspiration value during a period of time before the start time of each drought event, where i represents taking values according to the growth stage of the crop, and y is the year; taking the average precipitation and average temperature and average evapotranspiration value as inputs and the drought formation time as the output, training and validating the random forest model.

[0008] Calculate the drought formation rate according to the drought formation time, and the calculation formula is: , Wherein, represents the drought formation rate, is the drought formation time.

[0009] In the yield loss prediction model, the parameter representing the sensitivity of the crop to drought at each growth stage is calculated by the particle swarm method, and the objective function is to minimize the error between the predicted value and the actual value of the yield loss rate.

[0010] The establishment of the yield loss prediction model includes the following steps: Step 31, obtain basic data, including agricultural data, meteorological data and soil water data; Step 32: Obtain the drought formation time based on the soil water data and calculate the drought formation rate. Step 33: Calculate the actual value of the yield loss rate. The calculation formula is: , where, is the actual value of the yield loss rate, is the yield of the crop in the y-th year when drought occurs, and Y is the theoretical yield of the crop, which is obtained by trend analysis of historical yield data; Step 34: Establish a yield loss prediction model based on the drought formation rate, the growth stage of the crop at the start time of the drought event, and the planting system applied to the crop. Use the actual value of the yield loss rate obtained in Step 33 as the dependent variable of the yield loss prediction model, and obtain the optimal solution of the parameters in the yield loss prediction model through the particle swarm method.

[0011] Calculate the predicted value of the crop yield according to the predicted yield loss rate of the crop. The calculation formula is: , In the formula, is the predicted value of the crop yield, and Y is the theoretical yield of the crop.

[0012] The present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the agricultural yield prediction method under drought disasters as described above is implemented.

[0013] The present invention provides a computer program product, including a computer program and / or instructions. When the computer program and / or instructions are executed by a processor, the agricultural yield prediction method under drought disasters as described above is implemented.

[0014] Beneficial effects: Compared with the prior art, the present invention has the following advantages: The agricultural yield prediction method under drought disasters described in the present invention does not rely on a growth model for the accuracy of its prediction results, nor does it require modeling for different crops. Both the accuracy and timeliness have been greatly improved. Driven by various influencing factors, the prediction model proposed in the present invention is obtained based on a large amount of historical data research. We found that if drought forms in a short period of time and the soil water rapidly decreases, it will have a greater impact on the growth and final yield of crops. It is considered that the formation characteristics of drought should be the primary consideration factor in the yield prediction model. In addition, the method of the present invention also considers the influence of different crop production systems and the multiple effects of multiple small-scale droughts on crop yields during a growth cycle. Description of the Drawings

[0015] Figure 1It is the flowchart of the agricultural yield prediction method under drought disasters described in the present invention. Detailed implementation manners

[0016] The technical solution of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0017] The agricultural yield prediction method under drought disasters described in the present invention has a process as Figure 1 shown.

[0018] Step 1: Obtain basic data, including agricultural data, meteorological data, and soil water data. Continuously monitor the soil water data at the location of the crop to be predicted. When the soil water data drops to the drought event threshold, that is, the soil water content quantile value has dropped to 20%, or it is confirmed that a drought event has entered (the soil water content quantile value has dropped to 40%), agricultural yield prediction is carried out.

[0019] Among the basic data, the agricultural data is the historical yield data and phenological data of the crops in the research area; the meteorological data includes precipitation, temperature, and evapotranspiration data in the research area; the soil water data is the surface soil water.

[0020] The duration of each growth stage (usually three, the vegetative period, the reproductive growth period, and the maturity period) of the crop is obtained from the phenological data of the crop.

[0021] Specifically, the soil water data uses surface soil water. The drought event threshold uses the soil water content quantile value of 40%. The soil water content quantile value is obtained by preprocessing the soil water data and fitting the soil water data through an empirical distribution function.

[0022] Confirming the entry into a drought event includes: preprocessing the soil water data, and converting the soil water data into the soil water content quantile value through fitting by an empirical distribution function. According to the soil water content quantile value, identify the drought event through the following rules: (1) When the soil water content quantile is lower than 40%, it is considered that the drought event starts. (2) There is at least one moment when the soil water content quantile is lower than 20% during a drought event. (3) The drought duration is at least two weeks.

[0023] Step 2: For the agricultural yield prediction when the soil water data drops to the drought event threshold, obtain the drought formation time through the drought formation time prediction method. For the agricultural yield prediction when it is confirmed that a drought event has entered, directly calculate the drought formation time according to the soil water data. Calculate the drought formation rate according to the drought formation time .

[0024] First, define the drought formation time , which means the duration when the soil water content quantile decreases from 40% to 20% in a certain identified drought event. In the drought formation time prediction method, a relatively better solution considering both accuracy and timeliness is as follows: Obtain historical soil water data and conduct drought identification to obtain the drought formation time of each drought event. Obtain historical meteorological data to obtain the average precipitation and average temperature and average evapotranspiration value in the 30 days before the start time of each drought event. i Take values according to the growth stage of the crop. For example, when the crop has three growth stages, i take 1, 2, and 3 in sequence; y is a certain year. Take the average precipitation and average temperature and average evapotranspiration value as independent variables, and the drought formation time as the dependent variable, and build a model through the random forest model. The data is divided into a training set and a validation set at a ratio of 7:3. Finally, obtain the average precipitation and average temperature and average evapotranspiration value in the 30 days before the start time of the drought event corresponding to the current agricultural yield prediction, and output the predicted drought formation event through the trained random forest model.

[0025] It can be foreseen that any other scheme capable of obtaining the drought formation time can be adopted for the drought formation time prediction method. When encountering a severe drought event, a drought formation time prediction method with higher accuracy is preferably selected, such as a drought formation time prediction method using a conditional probability function disclosed in Patent CN2024102894426. Other existing neural network or deep learning models can also be selected to predict the drought formation time.

[0026] When it is confirmed that a drought event has entered, directly calculate the drought formation time based on the soil water data, which is the time from the start of the drought event to when the soil water content quantile is lower than 20% as the drought formation time.

[0027] For the prediction when the soil water data drops to the drought event threshold, the start time of the drought event is the time when the soil water data drops to the drought event threshold.

[0028] Step 3, obtain the predicted yield loss rate of the crop through the yield loss prediction model according to the drought formation rate, the growth stage of the crop at the start time of the drought event, and the planting system applied to the crop.

[0029] The establishment of the yield loss prediction model includes the following steps: Step 31: Obtain basic data, including agricultural data, meteorological data, and soil water data. See the specific content of Step 1 for details.

[0030] Step 32: Obtain the drought formation time and calculate the average drought formation rate. Based on historical soil water data, obtain the drought formation time. See the specific content of Step 2 for its definition.

[0031] After obtaining the drought formation time, calculate the drought formation rate , and use the average drought formation rate formula as follows: , Step 33: Calculate the yield loss rate.

[0032] Based on historical yield data and identified drought events, the actual value of the yield loss rate can be calculated , and the formula is as follows: , where, is the yield of the crop in the yth year when drought occurs, and Y is the theoretical yield of the crop, which is obtained by trend analysis of historical yield data.

[0033] Step 34: Establish a yield loss prediction model based on the drought formation rate, the growth stage of the crop at the start of the drought event, and the planting system applied to the crop. Solve the parameters in the model by the particle swarm method to obtain the yield loss prediction model.

[0034] Establish a yield loss prediction model to calculate the predicted value of the yield loss rate , and the formula is as follows: , where, n represents the nth stage of the crop; represents the sensitivity of the crop to drought at each growth stage, which is obtained by parameter calibration: After calculating the yield loss rate , for , use the random value method and apply the particle swarm method to continuously compare the predicted value of the yield loss rate with the actual value. The comparison index can be variance or other indexes that can evaluate the error between the two. Finally, the purpose of making the predicted value closest to the actual value, that is, the minimum error, is achieved. Obtain the optimal value; a represents different planting systems. When rainfed is adopted, a = 0.2; when a rainfed-irrigation mixture is adopted, a = 0.15; when irrigation is adopted, a = 0.1.

[0035] Step 4: Calculate the predicted crop yield based on the predicted crop yield loss rate.

[0036] Calculate the predicted crop yield, and the calculation formula is: , In the formula, is the predicted value of the crop yield, Y is the theoretical yield of the crop, is the crop yield loss rate output by the yield loss prediction model.

Claims

1. A method for predicting agricultural yield under drought disaster, characterized in that: The following steps are involved: Step 1, obtaining basic data, which includes agricultural data, meteorological data, and soil water data; continuously monitoring the soil water data of the location of the crop to be predicted, and predicting agricultural yield when the soil water data drops to the drought event threshold or is confirmed to have entered a drought event; Step 2: For agricultural yield prediction when soil water data drops to the drought event threshold, the drought formation time is obtained by using the drought formation time prediction method based on meteorological data and soil water data; for agricultural yield prediction when a drought event is confirmed, the drought formation time is directly calculated based on soil water data; and the drought formation rate is calculated based on the drought formation time; Step 3, according to the drought formation rate, the growth stage of the crop at the beginning of the drought event, and the planting system used by the crop, the predicted yield loss rate of the crop is obtained through the yield loss prediction model; The yield loss prediction model is: ， In the formula, represents the yield loss rate, n represents the n growth stage of the crop; represents the sensitivity of crops to drought at various growth stages; a is a constant, representing the cropping system used for the crop; represents the drought formation rate; Step 4, calculating the predicted crop yield value according to the predicted crop yield loss rate.

2. The method for predicting agricultural yield under drought disaster according to claim 1, characterized in that: The agricultural data are the historical yield data and phenological data of crops in the study area; the meteorological data include precipitation, temperature and evapotranspiration data in the study area; and the soil water data use surface soil water data.

3. The method for predicting agricultural yield under drought disaster according to claim 1, characterized in that: The growth stages of crops include: vegetative stage, reproductive growth stage and maturity stage.

4. The method for predicting agricultural yield under drought disaster according to claim 1, characterized in that: The method for predicting the drought formation time includes: predicting the drought formation time through a random forest model based on meteorological data and soil water data; establishing the random forest model includes: obtaining historical soil water data and performing drought identification to obtain the drought formation time of each drought event; obtaining historical meteorological data to obtain the average precipitation for a period of time before the start of each drought event Average temperature The average evapotranspiration ,in i Indicates that the value is taken according to the growth stage of the crop, y is the year; the average precipitation Average temperature The average evapotranspiration As input, the drought formation time As output, a random forest model is trained and validated.

5. The method for predicting agricultural yield under drought disaster according to claim 1, characterized in that: The drought formation rate is calculated based on the drought formation time, and the calculation formula is: ， In the formula, represents the drought formation rate, Time for drought to form.

6. The method for predicting agricultural yield under drought disaster according to claim 1, characterized in that: In the yield loss prediction model, the parameter representing the sensitivity of crops to drought at different growth stages , calculated by particle swarm method, the objective function is to minimize the error between the predicted value and the actual value of the yield loss rate.

7. The method for predicting agricultural yield under drought disaster according to claim 1, characterized in that: The establishment of the yield loss prediction model includes the following steps: Step 31, obtaining basic data, including agricultural data, meteorological data and soil water data; Step 32, obtaining drought formation time according to soil water data, and calculating drought formation rate; Step 33, calculate the actual value of the yield loss rate, the calculation formula is: ， in, is the actual value of the yield loss rate, is the crop yield in the yth year when drought occurs, and Y is the theoretical crop yield; Step 34, based on the drought formation rate, the growth stage of the crop at the beginning of the drought event, and the planting system used by the crop, a yield loss prediction model is established, and the actual value of the yield loss rate obtained in step 33 is used as the dependent variable of the yield loss prediction model. The parameters in the yield loss prediction model are obtained by particle swarm method. The optimal solution of .

8. The method for predicting agricultural yield under drought disaster according to claim 1, characterized in that: The predicted yield value of the crop is calculated based on the predicted yield loss rate of the crop. The calculation formula is: ， In the formula, is the predicted value of crop yield, and Y is the theoretical yield of the crop.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for predicting agricultural yield under drought disaster described in any one of claims 1 to 8 is implemented.

10. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or instruction is executed by a processor, the method for predicting agricultural yield under drought disaster described in any one of claims 1 to 8 is implemented.

Images