A soybean split growth stage irrigation decision method for drought

Abstract

This invention discloses a drought-prone soybean irrigation decision-making method based on growth stages, including: acquiring historical meteorological data and candidate soybean growth data, and calculating the Specialized Physical Requirement (SPEI); calculating the reference crop water requirement and dynamically adjusting the crop coefficient to obtain the planting water requirement for each growth stage; constructing climate-growth stage coupling characteristics and training a water requirement prediction model; using the SHAP method to analyze the contribution of the coupling characteristics and adjusting the weights of the water requirement prediction model according to the growth stage; constructing drought thresholds for different growth stages based on SPEI value changes; triggering irrigation decisions based on the drought thresholds corresponding to each growth stage and real-time SPEI values; determining the irrigation amount based on the predicted water requirement; and revising the irrigation plan based on real-time meteorological data and soil moisture monitoring results. This invention uses the above method to predict water requirement based on the water requirement prediction model and makes irrigation decisions based on SPEI changes and growth stages, improving the accuracy and adaptability of irrigation regulation.

Description

Technical Field

[0001] This invention relates to the field of agricultural information technology, and in particular to a method for decision-making on irrigation of soybeans according to their growth stages in drought conditions. Background Technology

[0002] As an important grain and oil crop, soybean yield is closely related to water supply, especially given the significant differences in water requirements at different growth stages. Under drought or water stress conditions, crop water requirements vary complexly. If irrigation management is not timely or precise, it can easily lead to reduced yield and decreased water resource utilization efficiency. Therefore, implementing precision irrigation management is of great importance.

[0003] Currently, agricultural irrigation decision-making methods mainly include empirical estimation methods based on crop water requirement models and monitoring methods based on meteorological or soil moisture conditions. Among them, crop water requirement models are usually calculated based on the FAO Penman-Monteith formula combined with crop coefficients. However, the crop coefficients in this method often use fixed or empirical values, which are difficult to reflect the dynamic differences of crops at different growth stages and under changing environmental conditions, resulting in a certain deviation between the water requirement estimation results and the actual situation.

[0004] In addition, existing technologies have incorporated machine learning methods for crop water requirement prediction to improve the model's fitting ability. However, these technologies are mostly data-driven, with complex model structures and a lack of consideration for the roles of key meteorological factors and drought indicators at different growth stages. This results in certain deficiencies in the stability and interpretability of the model when applied across different growth stages.

[0005] Therefore, how to construct a crop water requirement identification and irrigation decision-making method based on considering the dynamic changes of drought and the differences in growth stages, and improve the accuracy and adaptability of irrigation regulation under drought conditions, has become an urgent technical problem to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a drought-resistant soybean growth stage irrigation decision-making method to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, this invention provides a drought-resistant soybean growth stage irrigation decision-making method, comprising the following steps: S1. Obtain historical meteorological data and candidate soybean growth data for the target area, and calculate the Standardized Precipitation Evapotranspiration Index (SPEI) for different growth stages; S2. Based on historical meteorological data, calculate the reference crop water requirement, and dynamically adjust the crop coefficient according to different growth stages of soybean to obtain the planting water requirement of soybean at different growth stages. S3. Construct the coupled characteristics of soybean climate-growth stage, and train the water requirement prediction model based on the coupled characteristics and planting requirements. Obtain the predicted water requirement based on the trained water requirement prediction model. S4. The SHAP method is used to analyze the contribution of coupling features, and the contribution change characteristics of factors are analyzed according to the reproductive stage to adjust the weight of the water demand prediction model. S5. Based on the nonlinear response relationship of SHAP contribution value during the change of SPEI value, the SPEI interval corresponding to the significant change of contribution is extracted as the drought sensitive point of each growth stage, and drought thresholds for different growth stages are constructed accordingly. S6. Irrigation decisions are triggered based on drought thresholds and real-time SPEI values ​​corresponding to each growth stage; irrigation amounts are determined based on predicted water demand, water demand sensitivity at each growth stage, and effective precipitation; irrigation plans are dynamically revised based on real-time meteorological data and soil moisture monitoring results to achieve closed-loop updates of irrigation decisions.

[0008] Preferably, historical meteorological data includes precipitation, temperature, net radiation, wind speed, relative humidity, and sunshine duration data, and temperature includes maximum temperature, minimum temperature, and average temperature.

[0009] Preferably, step S2 specifically includes: S21. Obtain temperature, net radiation, wind speed, and air humidity data for the target area, and calculate the reference crop evapotranspiration based on the FAO Penman-Monteith formula. ; S22. Set basic crop coefficients according to different growth stages of soybeans. K c ini , K c dev , K c mid , K c end The basic crop coefficients were corrected by combining wind speed, relative humidity, and plant height parameters in the target area to obtain the meteorologically corrected crop coefficients for each growth stage: ; in, K c / tab The basic crop coefficient for soybeans at different growth stages. K c ini , K c dev , K c mid , K c end These represent the basic crop coefficients for the early growth stage, rapid growth stage, mid-growth stage, and maturity stage of soybean, respectively. The wind speed at a height of 2m; This represents the lowest relative humidity. h This refers to the average plant height of soybeans; S23. Obtain soybean leaf area index (LAI) data, characterize canopy development status based on LAI variation characteristics, calculate canopy interception capacity based on the Beer-Lambert canopy radiation interception law, and adjust the meteorological crop coefficient accordingly. Dynamic adjustments were made to obtain dynamic crop coefficients at different growth stages. : ; in, k The canopy extinction coefficient is determined based on the soybean variety and planting conditions and methods. S24. Dynamic crop coefficients based on different growth stages Compared with reference crop evapotranspiration Calculate the water requirement of soybeans at different growth stages : ; in, The water requirement for soybeans is expressed in mm.

[0010] Preferably, the growth stages include the early growth stage, the rapid growth stage, the mid-growth stage, and the maturity stage; wherein, the early growth stage is the stage from sowing to emergence, the rapid growth stage is the stage from branching to pod formation, the mid-growth stage is the stage of grain filling, and the maturity stage is the stage of maturity. The basic crop coefficient range for different growth stages of soybean is as follows: early growth stage: 0.4~0.7; rapid growth stage: 0.8~1.2; mid-growth stage: 1.0~1.2; maturity stage: 0.5~1.0.

[0011] Preferably, the coupling feature is formed by a combination of meteorological factors, fertility stage and SPEI of different fertility stages. The meteorological factors include precipitation, average temperature, net radiation, wind speed and relative humidity.

[0012] Preferably, the water demand prediction model adopts the random forest model, the XGBoost model, or the long short-term memory network model.

[0013] Preferably, step S4 specifically includes: The SHAP method was used to calculate the marginal contribution values ​​of meteorological factors and SPEI to the output of the water demand prediction model at different soybean growth stages, and the influence weight of each feature on the water demand of soybean planting was obtained. Based on the contribution values ​​of each characteristic at different growth stages, a contribution ranking mechanism is constructed to quantitatively analyze the influence of each characteristic, identify the dominant characteristics of each characteristic at different growth stages on the change of soybean water demand, and adjust the weights of the water demand prediction model according to the ranking results.

[0014] Preferably, step S6 specifically includes: Based on the drought threshold corresponding to each growth stage and the real-time SPEI monitoring value, the current drought status of soybeans is dynamically determined. When the SPEI value enters or exceeds the drought threshold range of the corresponding growth stage, the irrigation decision mechanism is triggered. After the irrigation decision is triggered, the soybean water demand forecast output by the water demand forecasting model is combined with the differences in water demand sensitivity and irrigation efficiency at different growth stages to make a phased weighted correction to determine the unit irrigation amount and the total amount of irrigation per single irrigation. After irrigation is completed, soil moisture content and real-time precipitation data are obtained, and subsequent irrigation amounts and intervals are adjusted based on changes in soil moisture content and irrigation efficiency.

[0015] Therefore, the present invention employs the above-mentioned drought-oriented soybean growth stage irrigation decision-making method, which has the following beneficial effects: (1) This invention modifies the crop coefficient by introducing wind speed, relative humidity and soybean plant height parameters, and combines the soybean leaf area index to reflect canopy changes, so that the crop coefficient can be dynamically adjusted with the growth stage, thereby improving the adaptability of the soybean water requirement calculation results. (2) This invention constructs the coupling characteristics of meteorological factors, growth stage and SPEI, and introduces interactive variables so that the water demand prediction model can use meteorological information and soybean growth stage information at the same time, thereby improving the stability of soybean water demand prediction under different conditions. (3) The present invention uses the SHAP analysis model output, divides the drought threshold according to the contribution change of SPEI in different value ranges, and adjusts the irrigation process in combination with soil moisture data to realize irrigation control based on climate conditions and field feedback.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an embodiment of the soybean growth stage irrigation decision-making method for drought conditions according to the present invention. Figure 2 This is a graph showing the variation of water requirements of soybeans at different growth stages according to an embodiment of the present invention. Figure 3 This is a diagram showing the response relationships of various features of the SHAP-based water demand prediction model according to an embodiment of the present invention. Figure 4 This is a comparison diagram of the SPEI random forest model with different reproductive periods and the SPEI random forest model without different reproductive periods in an embodiment of the present invention, where (a) is with different reproductive periods and (b) is without different reproductive periods; Figure 5 This is a prediction graph from a random forest model in an embodiment of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0020] Example like Figure 1 As shown, this invention provides a method for soybean growth stage irrigation decision-making in drought conditions, comprising the following steps: S1. Obtain historical meteorological data and candidate soybean growth data for the target area, and calculate the Standardized Precipitation Evapotranspiration Index (SPEI) for different growth stages.

[0021] In this embodiment, historical meteorological data includes precipitation, temperature, net radiation, wind speed, relative humidity, and sunshine duration. Temperature includes maximum temperature, minimum temperature, and average temperature.

[0022] The candidate data for soybean growth were obtained from an agricultural ecological experimental station, including plot management records, phenological periods of soybean cultivation, and major weather conditions from 2004 to 2014. Soybeans were sown in May and harvested in October. The growth stages included the early growth stage, the rapid growth stage, the mid-growth stage, and the maturity stage; the early growth stage was from sowing to emergence, the rapid growth stage was from branching to pod formation, the mid-growth stage was the grain-filling stage, and the maturity stage was the maturity stage.

[0023] Based on the calculation results from the site data, inverse distance weighted interpolation was performed using ArcGIS to obtain the SPEI spatial distribution of soybean growing months from May to September in the region over a long time series from 2004 to 2014.

[0024] The calculation process for the Standardized Precipitation Evapotranspiration Index (SPEI) includes: First, the potential evaporation PET was calculated using the FAO Penman-Monteith formula. Secondly, based on precipitation Constructing a moisture balance sequence with PET ,in By performing sliding accumulation on the water balance series at a set time scale, the cumulative water balance series is obtained. Finally, the cumulative water balance sequence was fitted with a three-parameter log-logistic distribution to calculate its cumulative probability distribution function. The cumulative probability was then standardized to obtain the standardized precipitation evapotranspiration index (SPEI).

[0025] S2. Based on historical meteorological data, calculate the reference crop water requirement, and dynamically adjust the crop coefficient according to different soybean growth stages to obtain the planting water requirement for soybeans at different growth stages. Specifically, this includes: S21. Obtain temperature, net radiation, wind speed, and air humidity data for the target area, and calculate the reference crop evapotranspiration based on the FAO Penman-Monteith formula. The calculation formula is: ; In the formula, Δ is the slope of the saturated water vapor pressure curve, kPa / ℃; R n Net radiation at Earth's surface, MJ / m 2 / d; G is soil heat flux, which is 0; γ is dry-wet surface constant, which is 0.067 kPa / ℃; T is average air temperature, ℃; u 2 represents the wind speed at a height of 2m, in m / s; e s -e a The value is the saturated water vapor pressure difference, in kPa. All parameters are taken according to the FAO-56 standard.

[0026] S22. Set basic crop coefficients according to different growth stages of soybeans. K c ini , K c dev , K c mid , K c end The basic crop coefficients were corrected by combining wind speed, relative humidity, and plant height parameters in the target area to obtain the meteorologically corrected crop coefficients for each growth stage: ; in, K c / tab The basic crop coefficient for soybeans at different growth stages. K c ini , K c mid , K c end These represent the basic crop coefficients for the early growth stage, rapid growth stage, mid-growth stage, and maturity stage of soybean, respectively. The wind speed at a height of 2m; This represents the lowest relative humidity. h This represents the average plant height of soybeans.

[0027] S23. Obtain soybean leaf area index (LAI) data, characterize canopy development status based on LAI variation characteristics, calculate canopy interception capacity based on the Beer-Lambert canopy radiation interception law, and adjust the meteorological crop coefficient accordingly. Dynamic adjustments were made to obtain dynamic crop coefficients at different growth stages. : ; in, k This represents the canopy extinction coefficient, which is determined based on the soybean variety and planting conditions and methods.

[0028] S24. Dynamic crop coefficients based on different growth stages Compared with reference crop evapotranspiration Calculate the water requirement of soybeans at different growth stages : ; in, The water requirement for soybeans is expressed in mm.

[0029] like Figure 2 The exhibition showcases the water requirements of soybeans at different growth stages from 2004 to 2014.

[0030] S3. Construct the coupled characteristics of soybean climate-growth stage, and train the water demand prediction model based on the coupled characteristics and planting requirements. Obtain the predicted water demand based on the trained water demand prediction model.

[0031] The coupling characteristics are formed by a combination of meteorological factors, reproductive stages, and SPEI at different reproductive stages. The meteorological factors include precipitation (P), average temperature (T), and net surface radiation. R n Wind speed u 2. Relative humidity (RH).

[0032] One-hot encoding is used to quantify the fertility stages, and the interaction features between climate and fertility stages are constructed to form a climate-fertility stage coupled feature matrix. X : ; in, Stage This indicates the growth stage of soybeans.

[0033] A Random Forest (RF) model was constructed using R language as a water demand prediction model. The Random Forest model was implemented using the ranger package, and the number of trees (num.trees) was set to 1000 to improve model stability and prediction accuracy. Permutation Importance was used to evaluate the importance of variables.

[0034] The constructed coupling feature matrix is ​​used as the model input, taking the water requirement for soybean cultivation as an example. ETc The output labels are used to train the model. The model performance is then validated by splitting the dataset into training and test sets.

[0035] The trained water demand prediction model is used for actual prediction. Real-time meteorological data and the current soybean growth stage are input into the model, and the model outputs the predicted soybean water demand value at the corresponding time. ET c_pred It is used to characterize the water demand of soybeans under the combined effects of current meteorological conditions and growth stage.

[0036] In addition to using the random forest model to construct a water demand prediction model, other models such as the XGBoost model or the long short-term memory network model can also be used.

[0037] S4. The SHAP method is used to analyze the contribution of coupling characteristics, and the contribution changes of factors are analyzed according to the reproductive stage to adjust the weights of the water demand prediction model. Specifically, this includes: The SHAP method was used to calculate the marginal contribution values ​​of meteorological factors and SPEI at different soybean growth stages to the output of the water demand prediction model, thus obtaining the corresponding SHAP contribution values. A generalized additive model (GAM) was used to smoothly fit the SHAP dependency, and a nonlinear response curve was constructed using spline functions. By calculating the point with the largest absolute value of the second derivative of the fitted curve, the response threshold of each environmental factor was identified, and the influence weight of each feature on soybean water demand was obtained based on the corresponding threshold.

[0038] Based on the average SHAP contribution value of each characteristic at different growth stages, a quantitative analysis of the influence of the average SHAP contribution value on each characteristic is conducted to identify the dominant characteristics of temperature, humidity, precipitation, wind speed, SPEI, etc., on the changes in soybean water demand at different growth stages. The weights of the water demand prediction model are adjusted according to the ranking results. Figure 3 As shown, the response relationships of various features of the SHAP-based water demand prediction model are illustrated, with SPEI=-0.54 being the sensitive threshold for soybean water demand response to drought stress.

[0039] S5. Based on the nonlinear response relationship of SHAP contribution value during the SPEI value change process, the SPEI intervals corresponding to significant changes in contribution are extracted as drought-sensitive points for each growth stage, and drought thresholds for different growth stages are constructed accordingly. Specifically: based on the correspondence between SPEI and SHAP contribution values, the SPEI value sequence and its corresponding SHAP contribution value change curve under different growth stages are obtained. The SHAP contribution values ​​corresponding to SPEI in different value intervals are statistically analyzed in segments, and the response relationship between the contribution change and SPEI change is analyzed. When the SHAP increment change rate changes from a stable state to a rapid change state, the SPEI value interval corresponding to the inflection point of the change is determined, and this SPEI interval is used as the drought-sensitive interval for the corresponding growth stage.

[0040] S6. Irrigation decisions are triggered based on drought thresholds and real-time SPEI values ​​corresponding to each growth stage; irrigation amounts are determined based on predicted water demand, water demand sensitivity at each growth stage, and effective precipitation; and the irrigation plan is dynamically revised based on real-time meteorological data, soil moisture, and irrigation efficiency monitoring results, achieving a closed-loop update of irrigation decisions. Specifically, this includes: Based on the drought threshold corresponding to each growth stage and the real-time SPEI monitoring value, the current drought status of soybeans is dynamically determined. When the SPEI value enters or exceeds the drought threshold range of the corresponding growth stage, the irrigation decision mechanism is triggered. After the irrigation decision is triggered, the soybean water demand forecast output by the water demand forecasting model is combined with the differences in water demand sensitivity at different growth stages and effective precipitation. The basic water demand is then adjusted in stages to determine the unit irrigation amount and the total amount of irrigation per session. After irrigation, soil moisture content and real-time precipitation data are acquired, and subsequent irrigation amounts and intervals are adjusted based on changes in soil moisture content and irrigation efficiency. The SPEI (Self-Protected Element for Irrigation) remains unchanged as an exogenous drought background indicator, used only for drought status determination and irrigation trigger control, and does not participate in the feedback update process after irrigation execution. This achieves stratified irrigation regulation based on climate constraints and soil feedback.

[0041] like Figure 4 The figure shows a comparison between a random forest model with different growth stages (SPEI) and one without. The random forest model effectively simulates the changing water demand, but the performance is further improved after introducing SPEI with different growth stages. Compared to the random forest model without SPEI, the random forest model with SPEI with different growth stages shows a higher R1. 2 It increased by 0.02, while RMSE decreased by 1.79.

[0042] like Figure 5The figure shows the predicted values ​​of the random forest model. As can be seen from the figure, the predicted scatter plots are closer to the 1:1 reference line, especially in the high water demand range where the underestimation is reduced, indicating that drought information at different growth stages can improve the model's ability to represent crop water consumption processes and its prediction accuracy. Meanwhile, the central root mean square error (SRMEE) has decreased, indicating that the discrepancy between the predicted and observed sequences has further decreased. Using the SPEI of different growth stages as an input variable helps reduce prediction errors and improves the random forest model's ability to predict water demand.

[0043] Therefore, this invention adopts the above-mentioned soybean growth stage irrigation decision-making method for drought conditions. In terms of water demand calculation, it achieves dynamic prediction of soybean water demand by integrating meteorological factors, growth stage, and water status indicators, thereby improving the stability of water demand estimation at different growth stages. In terms of interpretation and threshold construction, it improves the accuracy of water demand prediction by combining the SPEI index, multi-source meteorological data, and machine learning methods. Based on SHAP analysis, it identifies key influencing factors and their differences in effect at different growth stages and extracts water-sensitive intervals to reduce dependence on empirical parameters. In terms of irrigation decision-making, it combines threshold identification, water demand prediction, and soil moisture feedback to provide technical support for precision irrigation of soybeans under drought conditions.

[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for soybean irrigation decision-making based on growth stages in drought conditions, characterized in that, Includes the following steps: S1. Obtain historical meteorological data and candidate soybean growth data for the target area, and calculate the Standardized Precipitation Evapotranspiration Index (SPEI) for different growth stages; S2. Based on historical meteorological data, calculate the reference crop water requirement, and dynamically adjust the crop coefficient according to different growth stages of soybean to obtain the planting water requirement of soybean at different growth stages. S3. Construct the coupled characteristics of soybean climate-growth stage, and train the water requirement prediction model based on the coupled characteristics and planting requirements. Obtain the predicted water requirement based on the trained water requirement prediction model. S4. The SHAP method is used to analyze the contribution of coupling features, and the contribution change characteristics of factors are analyzed according to the reproductive stage to adjust the weight of the water demand prediction model. S5. Based on the nonlinear response relationship of SHAP contribution value during the change of SPEI value, the SPEI interval corresponding to the significant change of contribution is extracted as the drought sensitive point of each growth stage, and drought thresholds for different growth stages are constructed accordingly. S6. Irrigation decisions are triggered based on drought thresholds and real-time SPEI values ​​corresponding to each growth stage; irrigation amounts are determined based on predicted water demand, water demand sensitivity at each growth stage, and effective precipitation; and irrigation plans are dynamically revised based on real-time meteorological data, soil moisture monitoring results, and irrigation efficiency to achieve closed-loop updates of irrigation decisions.

2. The method for soybean irrigation decision-making based on growth stages in drought conditions according to claim 1, characterized in that: Historical meteorological data includes precipitation, temperature, net radiation, wind speed, relative humidity, and sunshine duration. Temperature data includes maximum, minimum, and average temperatures.

3. The method for soybean irrigation decision-making based on growth stages in drought conditions according to claim 1, characterized in that: Step S2 specifically includes: S21. Obtain temperature, net radiation, wind speed, and air humidity data for the target area, and calculate the reference crop evapotranspiration based on the FAO Penman-Monteith formula. ; S22. Set basic crop coefficients according to different growth stages of soybeans. K c ini , K c dev , K c mid , K c end The basic crop coefficients were corrected by combining wind speed, relative humidity, and plant height parameters in the target area to obtain the meteorologically corrected crop coefficients for each growth stage: ； in, K c / tab The basic crop coefficient for soybeans at different growth stages. K c ini , K c dev , K c mid , K c end These represent the basic crop coefficients for the early growth stage, rapid growth stage, mid-growth stage, and maturity stage of soybean, respectively. The wind speed at a height of 2m; This represents the lowest relative humidity. h This refers to the average plant height of soybeans; S23. Obtain soybean leaf area index (LAI) data, characterize canopy development status based on LAI variation characteristics, calculate canopy interception capacity based on the Beer-Lambert canopy radiation interception law, and adjust the meteorological crop coefficient accordingly. Dynamic adjustments were made to obtain dynamic crop coefficients at different growth stages. : ； in, k The canopy extinction coefficient is determined based on the soybean variety and planting conditions and methods. S24. Dynamic crop coefficients based on different growth stages Compared with reference crop evapotranspiration Calculate the water requirement of soybeans at different growth stages : ； in, The water requirement for soybeans is expressed in mm.

4. The method for soybean irrigation decision-making based on growth stages in drought conditions according to claim 3, characterized in that: The growth stages include the early growth stage, the rapid growth stage, the middle growth stage, and the maturity stage; among them, the early growth stage is from sowing to emergence, the rapid growth stage is from branching to pod formation, the middle growth stage is the grain filling stage, and the maturity stage is the maturity stage.

5. The method for soybean irrigation decision-making based on growth stages in drought conditions according to claim 1, characterized in that: The coupling characteristics are formed by the combination of meteorological factors, fertility stage and SPEI of different fertility stages. Meteorological factors include precipitation, average temperature, net radiation, wind speed and relative humidity.

6. The method for soybean irrigation decision-making based on growth stages in drought conditions according to claim 1, characterized in that: Water demand prediction models employ random forest, XGBoost, or long short-term memory network models.

7. A method for soybean irrigation decision-making based on growth stages in drought conditions, as described in claim 5, is characterized in that... Step S4 specifically includes: The SHAP method was used to calculate the marginal contribution values ​​of meteorological factors and SPEI to the output of the water demand prediction model at different soybean growth stages, and the influence weight of each feature on the water demand of soybean planting was obtained. Based on the contribution values ​​of each characteristic at different growth stages, a contribution ranking mechanism is constructed to quantitatively analyze the influence of each characteristic, identify the dominant characteristics of each characteristic at different growth stages on the change of soybean water demand, and adjust the weights of the water demand prediction model according to the ranking results.

8. A method for soybean irrigation decision-making based on growth stages in drought conditions, as described in claim 7, is characterized in that... Step S6 specifically includes: Based on the drought threshold corresponding to each growth stage and the real-time SPEI monitoring value, the current drought status of soybeans is dynamically determined. When the SPEI value enters or exceeds the drought threshold range of the corresponding growth stage, the irrigation decision mechanism is triggered. After the irrigation decision is triggered, the soybean water demand forecast output by the water demand forecasting model is combined with the differences in water demand sensitivity and irrigation efficiency at different growth stages to make a phased weighted correction to determine the unit irrigation amount and the total amount of irrigation per single irrigation. After irrigation is completed, soil moisture content and real-time precipitation data are obtained, and subsequent irrigation amounts and intervals are adjusted based on changes in soil moisture content and irrigation efficiency.

Images