Method for predicting acid and aluminum resistance of sugarcane based on XGBoost model

Abstract

The invention discloses a sugarcane acid aluminum resistance prediction method based on an XGBoost model, which comprises the following steps: (1) multi-source data acquisition and standardization: acquiring a soil sample, measuring the pH value and the exchange aluminum content, recording the weather condition of the sampling day, and measuring the leaf area and the chlorophyll content; (2) cleaning the data; (3) establishing an XGBoost model; (4) training an XGBoost model, and quantizing the characteristic contribution degree by using a built-in Gain calculation function of the XGBoost model; (5) inputting the feature vectors, adopting a linear regression task mode, and outputting an aluminum acid resistance score on an output layer; (6) score judgment; and giving related management suggestions according to the final score of the output result. According to the method, a static scoring system in the literature is innovatively upgraded into a dynamic scoring model, real-time prediction of the tolerance grade is realized by updating input data every week, and a differentiation decision rule base is constructed based on a prediction result.

Description

Technical Field

[0001] The present invention relates to the technical field of predicting aluminum tolerance in sugarcane, and particularly to a method for predicting aluminum tolerance in sugarcane based on the XGBoost model. Background Art

[0002] As an important sugar crop, the growth of sugarcane is extremely vulnerable to the toxicity of active aluminum (Al3+) in acidic soils, which leads to hindered root development and decreased nutrient absorption efficiency. Traditional methods for evaluating aluminum tolerance rely on laboratory chemical analysis (such as determining soil exchangeable aluminum content) and artificial phenotype observation (such as counting root damage rate), requiring destructive sampling during a specific growth period. The single evaluation cycle is as long as 2 - 3 weeks, and the cost of reagents and consumables is high. Although farmers collect data such as soil pH value and leaf growth status manually every week, these discrete data are only used for empirical judgment, lacking systematic integration and dynamic analysis capabilities, and it is difficult to reflect the changing trend of sugarcane tolerance in real time. In addition, existing management measures (such as applying lime to improve soil) are mostly based on fixed operation manuals and do not make dynamic adjustments in combination with field microclimate fluctuations (such as sudden increase in rainfall leading to leaching of aluminum ions), resulting in unstable implementation effects.

[0003] In the currently disclosed technical solutions, a tolerance grading method based on the soil aluminum ion concentration threshold is proposed. However, it only correlates a single soil index with the sugarcane growth response through a linear regression model, without considering the interaction of multiple indicators (such as the synergistic effect of pH value and organic matter content on aluminum toxicity), and does not introduce machine learning algorithms to handle non - linear relationships. The literature "Comprehensive evaluation of acid - aluminum tolerance traits of 26 sugarcane varieties at the seedling stage" (Jiangsu Agricultural Sciences, 2024, 52(09): 128 - 135) constructs an acid - aluminum resistance scoring system including physiological indicators such as root activity and chlorophyll content. However, its scoring results rely on static laboratory data and cannot be dynamically updated according to continuous field monitoring data, resulting in significant lag in practical applications. For example, when the soil pH changes rapidly due to continuous rainfall, the traditional static model cannot predict the risk of decreased tolerance score in time, delaying the timing of management intervention.

[0004] Aiming at the above - mentioned defects, the present invention proposes to apply the XGBoost algorithm to the dynamic prediction of aluminum tolerance in sugarcane. Compared with traditional methods, XGBoost can efficiently process multi - source heterogeneous data collected manually (including soil physical and chemical indicators, sugarcane physiological parameters, and regional climate variables) through an ensemble learning mechanism, and automatically capture the non - linear associations between features (such as the interaction effect between root activity and the pH change rate in two consecutive weeks). At the same time, the present invention innovatively upgrades the static scoring system in the literature to a dynamic scoring model, realizes real - time prediction of the tolerance level by updating input data weekly, and constructs a differential decision rule library based on the prediction results (such as adjusting the lime application rate according to future rainfall forecasts). Summary of the Invention

[0005] To overcome the above-mentioned disadvantages of the prior art, the present invention proposes a method for predicting aluminum tolerance of sugarcane based on the XGBoost model, which not only solves the problems of poor timeliness and high cost of traditional methods, but also can significantly improve the survival rate and yield stability of sugarcane in acidic soil environments through data-driven management suggestions.

[0006] The technical solution adopted by the present invention to solve its technical problems is: a method for predicting aluminum tolerance of sugarcane based on the XGBoost model, including the following steps:

[0007] (1) Multi-source data collection and standardization: Collect soil samples, measure the pH value and exchangeable aluminum content, and at the same time record the weather conditions on the day of sampling, and measure the leaf area and chlorophyll content;

[0008] (2) Clean the data;

[0009] (3) Establish an XGBoost model; use pH × daily rainfall to construct a multiplicative feature to capture the non-linear interaction between soil indicators and climate factors. The multiplicative feature is used to characterize the dynamic process of aluminum ion activation; calculate the pH change rate; calculate the moving average of aluminum content in the past three weeks as the cumulative effect of aluminum content, which characterizes the cumulative exposure level of aluminum toxicity; calculate the moving average of aluminum content in the past three weeks as the cumulative effect of aluminum content, which characterizes the cumulative exposure level of aluminum toxicity; at the same time, introduce historical data correlation, input the soil aluminum content lagged by 1-3 weeks, the average temperature in the previous week, and the cumulative rainfall in the previous two weeks;

[0010] (4) Train the XGBoost model, use its built-in Gain calculation function to quantify the feature contribution degree, and the feature set includes several feature vectors;

[0011] At the same time, refer to the literature to divide the acid and aluminum resistance of sugarcane into 0-10 points. 0-4 points means high acid and aluminum tolerance, more than 50% of the leaf area is yellowed and the roots are rotten, 5-7 means medium acid and aluminum tolerance, and the leaves are slightly yellowed, and 8-10 points means high acid and aluminum tolerance, and the leaves develop normally; take pictures of the sugarcane in each field every week, and the agricultural technicians score and take the average. Verify the scoring consistency through the Kappa coefficient as the final score;

[0012] (5) Input the above-mentioned feature vectors, adopt the linear regression task mode, the output layer is the acid and aluminum resistance score, and the loss function is set as the mean square error MAE + L2 regularization term to prevent overfitting; the tree structure parameters use Bayesian optimization to find the optimal combination in the parameter space, the maximum depth of the parameter space is 3-10, the learning rate is 0.01-0.3, subsample is 0.6-1, the number of subtrees is 500, and the minimum number of samples in the leaf nodes is 5; adopt 10-fold cross-validation, with R 2MAE is used as the evaluation metric; during training, column sampling rate and L2 weight are set to suppress overfitting;

[0013] (6) Scoring judgment; relevant management suggestions are given according to the final score of the output result.

[0014] Further, in the step (1), collecting soil samples specifically means dividing the sugarcane field into 3×3 grids and collecting soil samples at a fixed time every week at a depth of 0-20 cm.

[0015] Further, in the step (1), recording the weather conditions on the sampling day specifically means: collecting Noaa satellite meteorological data in real time and recording temperature, rainfall and sunshine duration.

[0016] Further, in the step (1), measuring the leaf area and chlorophyll content specifically means taking the third expanded leaf to measure the leaf area and chlorophyll content.

[0017] Further, the specific method of the step (2) is: samples with pH exceeding the range of 3.0-8.5 are marked as abnormal and re-collected after manual review; data with leaf area growth rate exceeding ±50% are regarded as invalid; continuous variables such as pH and aluminum ion exchange capacity are normalized:

[0018] At the same time, discrete weekly data are integrated into a time series, and missing values are filled with linear interpolation of data from the adjacent two weeks.

[0019] Further, the formula for calculating the pH change rate is: The pH change rate reflects the soil acidification rate. When ΔpH < -5%, it indicates that soil acidification is accelerating and priority intervention is needed.

[0020] Further, using its built-in Gain calculation function to quantify the feature contribution degree specifically means:

[0021] where GL and GR are the sum of the first-order gradients of the left and right subtrees after splitting, HL and HR are the sum of the second-order gradients; λ is the L2 regularization coefficient to control the model complexity, γ is the splitting threshold, and the larger its value, the more conservative the splitting; the mean squared error MSE is used as the loss function; the information gain brought by the feature during the splitting of each subtree is statistically calculated, and the average value of all trees is taken as the final weight; features with importance weight > 0.05 are retained, low-contribution features are removed, and at the same time, highly correlated features are removed, and features with higher weights are preferentially retained.

[0022] Further, the feature vector includes soil indicators, physiological indicators, and climate indicators.

[0023] Soil indicators: pH value, exchangeable aluminum content, pH change rate, organic matter content, and moving average of aluminum content.

[0024] Physiological indicators: leaf area growth rate, dry matter accumulation, chlorophyll content.

[0025] Climatic indicators: temperature, daily rainfall, pH × daily rainfall;

[0026] Further, the specific step (6) is as follows: when the score drops by ≥2 points continuously for two weeks, it is automatically marked as a "high-risk field"; query and introduce future weather conditions to establish a decision-rule matching mechanism. When the score <5, if strong rainfall is predicted in the future or the score drops by ≥2 points continuously for two weeks, it is recommended to immediately apply lime and spray 0.3% humic acid foliar fertilizer 48 hours after rainfall; when the score is between 5-7, if the average daily temperature ≥35°C lasts for three days or the weekly increase in soil aluminum content exceeds 20%, then increase the drip irrigation frequency, reduce the single water volume, reduce the application of nitrogen fertilizer and increase the application of potassium silicate, and spray 0.2% magnesium sulfate solution on the leaves every other week; when the score is between 7-8.5, if the cumulative future 7-day sunlight is less than 30 hours, then delay pruning and increase the application of organic fertilizer covered with filter mud, and if conditions permit, turn on the supplementary light for 4 hours every day; give relevant management suggestions according to the output results.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] (1) The present invention realizes the dynamic prediction of the acid aluminum resistance score by fusing multi-source data (soil, physiology, climate) through the XGBoost model. The prediction error (MAE) ≤0.5 points, and the accuracy rate (within ±1 point) reaches 95%, which is 58% higher than that of the traditional regression model (such as the linear regression error of 1.2 points).

[0029] (2) The present invention can timely detect the risk of soil acidification through weekly score updates.

[0030] (3) The present invention corrects the score weight by combining meteorological forecast data, improving the robustness of the model in extreme weather.

[0031] (4) The present invention guides farmers to prioritize the monitoring of core targets through feature importance analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments.

[0033] Figure 1 It is a flowchart of a method for predicting the acid aluminum tolerance of sugarcane based on the XGBoost model according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0035] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention.

[0036] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations. In addition, the technical features involved in different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0037] As Figure 1 shown, a method for predicting aluminum tolerance of sugarcane based on the XGBoost model in an embodiment of the present invention includes the following steps:

[0038] (1) Multi-source data collection and standardization: Collect soil samples, measure the pH value and exchangeable aluminum content, and record the weather conditions on the sampling day at the same time. The pH value and exchangeable aluminum content are measured in the laboratory. The specific method of collecting soil samples is to divide the sugarcane field into a 3×3 grid and collect soil samples at a depth of 0-20 cm at a fixed time every week; record the weather conditions on the sampling day at the same time, specifically, collect NOAA satellite meteorological data in real time and record the temperature, rainfall, and light duration; at the same time, measure the leaf area and chlorophyll content of the third expanded leaf.

[0039] (2) Clean the data. The specific method is as follows: Samples with a pH value exceeding the range of 3.0-8.5 are marked as abnormal and re-collected after manual review. Data with a leaf area growth rate exceeding ±50% is considered invalid. Then, normalize the continuous variables such as pH and aluminum ion exchange capacity:

[0040] At the same time, the discrete weekly data are integrated into a time series, and the missing values are filled by linear interpolation of the adjacent two weeks' data.

[0041] (3) Establish an XGBoost model; use pH × daily rainfall to construct a product-type feature to capture the nonlinear interaction between soil indicators and climate factors. The toxicity of aluminum ions in acidic soils increases with increasing moisture. This product-type feature is used to characterize the dynamic process of aluminum ion activation. When rainfall is high, the activity of aluminum ions in soils with low pH (i.e., strong acidity) increases significantly. The model can identify high-toxicity risk combinations through this product-type feature. Use the formula The pH change rate was calculated to reflect the rate of soil acidification. A ΔpH < -5% indicated accelerated soil acidification and required priority intervention. The moving average of aluminum concentrations over the previous three weeks was calculated as the cumulative effect of aluminum concentration, representing the cumulative exposure level of aluminum toxicity. Historical data correlation was also incorporated, inputting soil aluminum concentrations with a 1-3 week lag, the average temperature of the previous week, and the cumulative rainfall of the previous two weeks.

[0042] (4) Train the XGBoost model and use its built-in Gain calculation function to quantify the feature contribution. The feature set includes several feature vectors.

[0043] Where GL and GR are the sum of the first-order gradients of the left and right subtrees after the split, and HL and HR are the sum of the second-order gradients. λ is the L2 regularization coefficient that controls the complexity of the model, and γ is the split threshold. The larger its value, the more conservative the split. The mean square error (MSE) is used as the loss function. The information gain brought by the features when each subtree is split is calculated, and the average value of all trees is taken as the final weight. Features with importance weights greater than 0.05 are retained, and low-contribution features are removed. At the same time, features with strong correlation are removed, and features with higher weights are retained first: for example, high temperatures are often accompanied by strong light, and the temperature has a higher weight ratio. The temperature feature is retained here. If the weights are similar, the feature with larger variance is retained to improve the sensitivity of the model. The feature vector includes soil indicators, physiological indicators, and climate indicators.

[0044] Soil indicators: pH value, exchangeable aluminum content, pH change rate, organic matter content and moving average aluminum content.

[0045] Physiological indicators: leaf area growth rate, dry matter accumulation, and chlorophyll content.

[0046] Climate indicators: temperature, daily rainfall, pH × daily rainfall (interaction term).

[0047] At the same time, the references classify the aluminum tolerance of sugarcane on a scale of 0-10 points. For 0-4 points, the highly aluminum-tolerant leaves have more than 50% of their area yellowed and the roots are rotten; for 5-7 points, the moderately aluminum-tolerant leaves are slightly yellowed; for 8-10 points, the highly aluminum-tolerant leaves develop normally. After taking pictures of the sugarcane in each field every week, three agricultural technicians score and take the average. The consistency of the scores is verified through the Kappa coefficient (requirement: Kappa > 0.7) as the final score.

[0048] (5) Input the above feature vectors, adopt the linear regression task mode, with the output layer being the aluminum tolerance score. The loss function is set as the mean squared error (MSE) + L2 regularization term (λ = 1) to prevent overfitting. The tree structure parameters use Bayesian optimization to find the optimal combination within the parameter space. The maximum depth of the parameter space is 3-10, the learning rate is 0.01-0.3, subsample is 0.6-1, the number of subtrees is 500, and the minimum number of samples in the leaf nodes is 5. Adopt 10-fold cross-validation, with 2 R and MAE as evaluation indicators. During training, set the column sampling rate (colsample_bytree = 0.8) and L2 weight (reg_lambda = 2.0) to suppress overfitting.

[0049] (6) Score judgment; give relevant management suggestions based on the final score of the output result. When the score drops by ≥2 points for two consecutive weeks, it is automatically marked as a "high-risk field". Then query and introduce future weather conditions to establish a decision-rule matching mechanism. When the score < 5 and it is predicted that there will be heavy rainfall in the future (rainfall greater than 50 mm) or the score drops by ≥2 points for two consecutive weeks, it is recommended to immediately apply lime and spray 0.3% humic acid foliar fertilizer 48 hours after rainfall; when the score is between 5-7 points and the daily average temperature ≥35°C lasts for three days or the weekly increase in soil aluminum content exceeds 20%, then increase the drip irrigation frequency, reduce the single water volume, reduce the application of nitrogen fertilizer and increase the application of potassium silicate, and spray 0.2% magnesium sulfate solution on the leaves every other week; when the score is between 7-8.5 and the cumulative future 7-day sunlight is less than 30 hours, then delay pruning and increase the application of organic fertilizer and cover with filter mud. If conditions permit, turn on the supplementary light for 4 hours every day. Give relevant management suggestions according to the output result.

[0050] Through the description of the above implementation manners, those skilled in the art can clearly understand that each implementation manner can be realized by means of software plus a general hardware platform, and of course, it can also be realized by hardware. Based on such an understanding, the essence of the above technical solution or the part that contributes to the related technology can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application. Whether physically separated or not, the components displayed as units may or may not be physical units, that is, they may be located in one place, or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0052] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a general-purpose hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or rather the part that contributes to the related technology, can be embodied in the form of a software product, and this computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A prediction method for aluminum tolerance of sugarcane based on the XGBoost model, characterized in that, Including the following steps: (1) Multi-source data collection and standardization: Collect soil samples, measure the pH value and exchangeable aluminum content, record the weather conditions on the sampling day, measure the leaf area and chlorophyll content; (2) Clean the data; (3) Establish an XGBoost model; Use pH × daily rainfall to construct a multiplicative feature to capture the non-linear interaction between soil indicators and climate factors. The multiplicative feature is used to characterize the dynamic process of aluminum ion activation; Calculate the pH change rate; Calculate the moving average of aluminum content in the past three weeks as the cumulative effect of aluminum content, characterizing the cumulative exposure level of aluminum toxicity; Calculate the moving average of aluminum content in the past three weeks as the cumulative effect of aluminum content, characterizing the cumulative exposure level of aluminum toxicity; At the same time, introduce historical data relevance, input the soil aluminum content lagged by 1-3 weeks, the average temperature in the previous week, and the cumulative rainfall in the previous two weeks; (4) Train the XGBoost model, and use its built-in Gain calculation function to quantify the feature contribution degree. The feature set includes several feature vectors; At the same time, refer to the literature to divide the acid-aluminum resistance of sugarcane into 0-10 points. 0-4 points means high acid-aluminum tolerance, more than 50% of the leaf area is yellowed and the roots are rotten; 5-7 points means medium acid-aluminum tolerance, and the leaves are slightly yellowed; 8-10 points means high acid-aluminum tolerance, and the leaves grow normally; After taking pictures of the sugarcane in each field every week, the agricultural technicians score and take the average. The consistency of the scores is verified by the Kappa coefficient as the final score; (5) Input the above feature vectors, adopt the linear regression task mode, with the output layer being the anti-acid aluminum property score, and set the loss function as the mean squared error MAE + L2 regularization term to prevent overfitting; use Bayesian optimization to find the optimal combination within the parameter space for the tree structure parameters. The maximum depth of the parameter space is 3 - 10, the learning rate is 0.01 - 0.3, subsample is 0.6 - 1, the number of subtrees is 500, and the minimum number of samples in the leaf nodes is 5; adopt 10-fold cross-validation, with R 2 and MAE as evaluation indicators; set the column sampling rate and L2 weight during training to suppress overfitting; (6) Score judgment; Give relevant management suggestions according to the final score of the output result.

2. The sugarcane aluminum tolerance prediction method based on the XGBoost model according to claim 1, characterized in that In the step (1), the specific method of collecting soil samples is to divide the sugarcane field into a 3×3 grid and collect soil samples at a depth of 0-20 cm at a fixed time every week.

3. A prediction method for aluminum tolerance of sugarcane based on the XGBoost model according to claim 1, characterized in that In the step (1), the specific method of recording the weather conditions on the sampling day is: Collect Noaa satellite meteorological data in real time, and record the temperature, rainfall and sunshine duration.

4. A method for predicting aluminum tolerance of sugarcane based on the XGBoost model according to claim 1, characterized in that In the step (1), the specific method of measuring the leaf area and chlorophyll content is to take the third expanded leaf to measure the leaf area and chlorophyll content.

5. A method for predicting aluminum tolerance of sugarcane based on the XGBoost model according to claim 1, characterized in that, The specific method of the step (2) is: Samples with a pH exceeding the range of 3.0-8.5 are marked as abnormal and re-collected after manual review; Data with a leaf area growth rate exceeding ±50% is considered invalid; Normalize these continuous variables such as pH and aluminum ion exchange capacity: Meanwhile, the discrete weekly data are integrated into a time series, and the missing values are filled by linear interpolation of the data of the adjacent two weeks.

6. A method for predicting aluminum tolerance of sugarcane based on the XGBoost model according to claim 1, characterized in that, The pH change rate is calculated using the formula: The pH change rate reflects the soil acidification rate. When ΔpH < -5%, it indicates that soil acidification is accelerating and priority intervention is needed.

7. A method for predicting aluminum tolerance of sugarcane based on the XGBoost model according to claim 1, characterized in that, The specific method of using its built-in Gain calculation function to quantify the feature contribution degree is: Among them, GL and GR are the sums of the first-order gradients of the left and right subtrees after splitting, and HL and HR are the sums of the second-order gradients; λ is the L2 regularization coefficient to control the model complexity, γ is the splitting threshold, and the larger its value, the more conservative the splitting; the mean squared error MSE is used as the loss function; the information gain brought by the features during the splitting of each subtree is statistically calculated, and the average value among all the trees is taken as the final weight; the features with importance weights > 0.05 are retained, the low-contribution features are removed, and at the same time the features with strong correlation are removed, and the features with higher weights are preferentially retained.

8. A prediction method for aluminum tolerance of sugarcane based on the XGBoost model according to claim 1, characterized in that The feature vectors include soil indicators, physiological indicators, and climate indicators; Soil indicators: pH value, exchangeable aluminum content, pH change rate, organic matter content, and moving average of aluminum content. Physiological indicators: Leaf area growth rate, dry matter accumulation, chlorophyll content. Climate indicators: Temperature, daily rainfall, pH × daily rainfall.

9. A prediction method for aluminum tolerance of sugarcane based on the XGBoost model according to claim 1, characterized in that, The specific content of step (6) is as follows: when the score drops by ≥2 points continuously for two weeks, it is automatically marked as a "high-risk field"; query and introduce future weather conditions to establish a decision-rule matching mechanism. When the score <5 and it is predicted that there will be heavy rainfall in the future or the score drops by ≥2 points continuously for two weeks, it is recommended to immediately apply lime and spray 0.3% humic acid foliar fertilizer 48 hours after rainfall; when the score is between 5-7 and the daily average temperature ≥35°C lasts for three days or the weekly increase in soil aluminum content exceeds 20%, increase the drip irrigation frequency, reduce the single water volume, reduce the application of nitrogen fertilizer and increase the application of potassium silicate, and spray 0.2% magnesium sulfate solution on the leaves every other week; when the score is between 7-8.5 and the cumulative future 7-day light is less than 30 hours, delay pruning and increase the application of organic fertilizer and cover with filter mud. If conditions permit, turn on the supplementary light for 4 hours every day; give relevant management suggestions according to the output results.

Images