A quantitative prediction method for the vegetation recovery duration under the influence of drought events

Abstract

The present invention discloses a method for quantitatively predicting the vegetation recovery duration under the influence of drought events, which comprises the following steps: extracting the time series of the vegetation state index and soil water content at each grid point in the study area on a weekly scale; obtaining the vegetation anomaly conditions at each grid point according to the time series of the vegetation state index; identifying the soil drought events in the study area based on the soil moisture percentile method, and extracting the typical drought characteristics at each grid point; combining the drought events and the corresponding vegetation anomaly conditions, drawing the vegetation cumulative anomaly curve, and quantifying the vegetation recovery duration by identifying the interval between the start and end moments of vegetation recovery; introducing the extreme gradient boosting tree model to train and predict the vegetation recovery duration after the occurrence of drought in the future. The present invention provides a new approach for quantifying the vegetation recovery duration, establishes a prediction method for the vegetation recovery duration under drought influence based on a machine learning algorithm, and provides technical support for evaluating the response of the ecosystem to drought.

Description

Technical Field

[0001] The present invention relates to a method for estimating the duration of vegetation restoration, and particularly to a method for quantitatively predicting the duration of vegetation restoration under the influence of drought events. Background Art

[0002] Drought is a natural disaster event that occurs frequently and lasts for a long time. It affects the hydrological cycle, increases the risk of forest dieback, causes land degradation, and even reduces the biodiversity of terrestrial ecosystems, bringing serious negative effects to the ecological environment, social economy, especially agricultural production. At the same time, global warming has led to an increase in the frequency and intensity of droughts, and the risk of vegetation ecosystems being affected by drought is also rising. In this context, how vegetation responds to and adapts to the increasingly severe drought environment is one of the difficult problems in the current research on drought and ecology at home and abroad.

[0003] Vegetation indices are important parameters for characterizing the growth status and dynamic changes of vegetation, such as the Normalized Difference Vegetation Index ( ), the Enhanced Vegetation Index ( ), the Leaf Area Index ( ), etc. These indices can not only monitor the long-term evolution characteristics of vegetation, but also reflect the recovery of vegetation after experiencing natural disasters such as drought and high temperature. Previous studies have analyzed the response of vegetation to drought based on vegetation indices, especially focusing on the time required for vegetation to recover from an abnormal state to a normal state after drought, that is, the recovery duration. The vegetation recovery duration is an important indicator for evaluating the recovery ability of ecosystems. Especially in areas where drought events occur frequently, accurately predicting the vegetation recovery duration is of great significance for ecological restoration and resource management.

[0004] However, most studies determine the start and end times of vegetation recovery under the influence of drought based on the instantaneous anomalies of vegetation indices. However, they rarely consider the cumulative impact effect of drought on vegetation, and rarely analyze the response of vegetation to drought events from the perspective of vegetation cumulative deficit. In addition, limited studies are based on machine learning models, and there is a lack of models with expected accuracy to simulate and predict the vegetation recovery duration. Summary of the Invention

[0005] Object of the Invention: Aiming at the deficiencies of existing research, the present invention proposes a method for quantitatively predicting the duration of vegetation restoration under the influence of drought events, which can obtain the prediction result of the duration of vegetation restoration under the influence of drought events and meet the requirements of disaster warning in terms of accuracy.

[0006] Technical Solution: The technical solution adopted by the present invention is a method for quantitatively predicting the duration of vegetation restoration under the influence of drought events, including the following steps:

[0007] Step 1: Extract the time series of the grid - point - by - grid - point vegetation condition index and soil water content at the weekly scale in the study area;

[0008] Step 2: After performing "detrending" and "de - seasonalizing" on the time series of the grid - point - by - grid - point vegetation condition index, obtain the anomaly value. Based on the anomaly value, identify the corresponding vegetation anomaly conditions for each grid point;

[0009] For the time series of the grid - point - by - grid - point soil water content, identify the corresponding soil drought events for each grid point based on the soil moisture quantile method, and extract the typical characteristics of drought for each grid point;

[0010] Step 3: Combine the soil drought events and the corresponding vegetation anomaly conditions for each grid point to obtain the cumulative anomaly curve of vegetation. Identify the start and end times of the recovery of the vegetation anomaly state through the cumulative anomaly curve of vegetation, and calculate the vegetation recovery duration ;

[0011] Step 4: According to the typical characteristics of drought events, canopy height , root depth , soil texture and groundwater depth at each grid point in the study area, form the independent variable sequence, and use the vegetation recovery duration as the dependent variable sequence to train the Extreme Gradient Boosting (XGBoost) tree model. Through the trained XGBoost tree model, simulate and predict the vegetation recovery duration in the study area.

[0012] The "de - seasonalizing" process includes: using the difference method, subtracting the multi - year monthly average value from the monthly data of the vegetation condition index;

[0013] The "detrending" process includes: fitting the original data of the vegetation condition index by the least - squares method to obtain the fitting curve, and then subtracting the fitting curve of the original time series from the original time series of the vegetation condition index;

[0014] Identifying the corresponding vegetation anomaly conditions for each grid point based on the anomaly value includes: when the anomaly value is less than 0 for the first time, the corresponding time is defined as the start time of the vegetation negative anomaly event; when the anomaly value returns to not less than 0 again, the corresponding time is defined as the end time of the vegetation negative anomaly event; between this start time and end time, the minimum value of the anomaly value is defined as the peak value of the vegetation negative anomaly event;

[0015] For the time series of soil water content at each grid point, soil drought events corresponding to each grid point are identified based on the soil moisture quantile method, including: splitting the time series of soil water content for each year into several time periods and combining the corresponding time periods for several years; selecting the most suitable distribution function from multiple cumulative probability distribution functions and converting the combined data into a soil moisture quantile sequence; identifying soil drought events according to the soil moisture quantile sequence; the cumulative probability distribution functions include gamma distribution, beta distribution, extreme value distribution, generalized extreme value distribution, log-logistic distribution, logarithmic distribution, and Weibull distribution.

[0016] The typical characteristics of the drought include drought duration, drought intensity, and drought peak; where the drought duration refers to the duration between when the soil moisture quantile is lower than the first threshold and when it recovers or exceeds the first threshold; the drought intensity refers to the average value of the soil moisture quantile during the occurrence of the drought; the drought peak refers to the minimum value of the soil moisture quantile during the drought.

[0017] For the vegetation negative anomaly event that occurs weeks after the occurrence of the drought event, the vegetation anomaly values at each moment of the vegetation negative anomaly event are accumulated and summed moment by moment to obtain the vegetation cumulative anomaly curve.

[0018] The starting moment and ending moment of the recovery of the vegetation anomaly state are identified through the vegetation cumulative anomaly curve, and the vegetation recovery duration is calculated including: the moment when the vegetation cumulative anomaly reaches the local minimum is defined as the moment of the recovery of the vegetation anomaly state , when the vegetation cumulative anomaly recovers or exceeds 0 again and lasts for at least m weeks, it is defined as the ending moment of the recovery of the vegetation anomaly state , calculate the vegetation recovery duration, .

[0019] The vegetation recovery duration of the study area is simulated and predicted through the trained extreme gradient boosting tree model, and the calculation formula is:

[0020] ,

[0021] In the formula, is the vegetation recovery duration output by the extreme gradient boosting tree model, is the prediction result of the th decision tree, is the total number of trees, is the drought duration, is the drought intensity, is the drought peak, is the canopy height, is the root depth, is the soil texture, is the groundwater depth.

[0022] The objective function of the extreme gradient boosting tree model is as follows:

[0023] ,

[0024] In the formula, is the objective function, represents the true value of the th sample, represents the predicted value of the th sample, is the number of samples, is the loss function, which measures the difference between the model's predicted value and the true value ; is the prediction result of the th decision tree, is the total number of trees, is the regularization term.

[0025] 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 above-mentioned quantitative prediction method for the vegetation recovery duration under the influence of drought events is implemented.

[0026] The present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned quantitative prediction method for the vegetation recovery duration under the influence of drought events is implemented.

[0027] The present invention provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the above-mentioned quantitative prediction method for the vegetation recovery duration under the influence of drought events is implemented.

[0028] Beneficial effects: Compared with the prior art, the present invention has the following advantages: Quantifying the vegetation recovery duration under drought disturbance from the perspective of vegetation cumulative deficit, and the cumulative anomaly method can effectively improve the calculation accuracy of the vegetation recovery duration; The present invention can comprehensively consider various factors such as drought characteristics, vegetation, soil, and groundwater. By introducing the extreme gradient boosting (XGBoost) model with efficient feature learning and nonlinear modeling capabilities, it realizes accurate simulation and prediction of the vegetation recovery duration. This method can capture the complex nonlinear relationships during the vegetation recovery process and can significantly improve the prediction accuracy compared with linear models or simple statistical methods, providing a scientific basis for the adaptive evaluation of the vegetation ecosystem to drought and the monitoring and early warning of agricultural disasters. Description of the Drawings

[0029] Figure 1It is a flowchart of the quantitative prediction method for the vegetation recovery duration under the influence of drought events described in the present invention;

[0030] Figure 2 It is a schematic diagram of soil drought identification and typical characteristics described in the present invention;

[0031] Figure 3 It is a schematic diagram of (a) the vegetation cumulative anomaly curve and (b) the vegetation recovery duration described in the present invention;

[0032] Figure 4 It is a framework diagram for predicting the future vegetation recovery duration by the extreme gradient boosting tree model described in the present invention. Detailed implementation manners

[0033] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0034] The quantitative prediction method for the vegetation recovery duration under the influence of drought events described in the present invention has a flowchart as Figure 1 shown. In this embodiment, taking index as an example, the quantitative prediction method for the vegetation recovery duration under drought disturbance is described in detail. This method is also applicable to , and other vegetation indices.

[0035] S1: Extract the time series of the vegetation state index and soil water content at each grid point in the study area on a weekly scale.

[0036] S2: After performing "seasonal detrending" and "trend detrending" on the time series of the vegetation state index at each grid point, the corresponding vegetation anomaly condition is obtained.

[0037] Among them, the "seasonal detrending" process refers to removing the influence of seasonal changes from the original vegetation data to more accurately analyze the long-term trend and changes of the time series. In the present invention, the "seasonal detrending" process adopts the difference method, and its principle is to subtract the monthly average value of multiple years from the monthly vegetation data to remove the seasonal influence of the data.

[0038] The "trend detrending" process refers to eliminating the component that changes with time in the long term in the vegetation data, which is beneficial to accurately analyzing the fluctuation of the data itself. The vegetation data may be affected by long-term climate change or human activities and show an upward or downward trend. The trend detrending process can eliminate these long-term changes and make the data more stable. In the present invention, the "trend detrending" is achieved by subtracting the curve fitted by the least squares method from the original data. The mean value of the detrended vegetation data is zero, which is convenient for statistically analyzing the negative anomaly events of the vegetation.

[0039] The weekly data obtained after detrending and deseasonalizing the vegetation data is the anomaly value relative to the long-term average. When the anomaly value is less than 0 for the first time, it indicates that negative anomalies begin to appear in the vegetation, and this moment is defined as the start time of the vegetation negative anomaly; when the anomaly value returns to 0 or exceeds 0 again, it indicates that the vegetation begins to return to the normal state, and this moment is defined as the end time of the vegetation negative anomaly event. Between these two moments, the minimum value of the vegetation anomaly value is defined as the peak value of this vegetation negative anomaly event. By identifying the start and end times of the vegetation negative anomaly, the vegetation negative anomaly events at each grid point can be obtained.

[0040] S3: Identify the soil drought events in the study area based on the soil moisture quantile method, and extract the typical drought characteristics at each grid point.

[0041] Among them, the "soil moisture quantile method" means splitting the original soil moisture time series into 52 time series (taking 1940 - 2022 as an example, each year can be split into 52 weeks, combining the first week of all years, the second week of all years,..., the 52nd week of all years), so as to eliminate the influence of soil moisture seasonality, and then select a suitable distribution function from a variety of candidate cumulative probability distribution functions (such as gamma distribution, beta distribution, extreme value distribution, generalized extreme value distribution, log-logistic distribution, logarithmic distribution, Weibull distribution, etc.) to convert it into a soil moisture quantile series.

[0042] The "drought event" refers to the agricultural drought event identified based on the soil water content quantile. The discrimination of drought events needs to meet the following conditions: (1) For a given drought event, during the drought event is lower than the 40th percentile, The first moment lower than the 40th percentile indicates the start time of the drought event ( Figure 2 in ), and when returns to or exceeds the 40th percentile, the end time of the event ( Figure 2 in ) is determined; (2) The drought process must include at least one moment of lower than the 20th percentile to ensure that the identified event is indeed a drought event; (3) Given that the impact of short-term drought events on agricultural yields and ecosystems is limited, events with a duration (from the start time to the end time) of less than 12 weeks (about 3 months) are excluded.

[0043] The described "typical drought characteristics" include drought duration ( ), drought intensity ( ), and drought peak ( ). The schematic diagram of soil drought identification and typical characteristics is as Figure 2As shown in the figure. Among them, the drought duration refers to the duration between the moment when the soil moisture percentile is lower than the 40th percentile and the moment when it recovers or exceeds the 40th percentile. The drought intensity refers to the average value of the soil moisture percentile during the drought occurrence; the drought peak refers to the minimum value of the soil moisture percentile during the drought. The specific formulas for drought duration, intensity, and peak can be expressed as:

[0044] ,

[0045] ,

[0046] ,

[0047] In the formula, is the start time of the drought event, is the end time of the drought event; is the soil moisture percentile at time represents the time series of soil moisture percentile during the drought.

[0048] Among them, there is no restriction on the order of steps S2 and S3.

[0049] S4: Combine the drought events at each grid point and the corresponding vegetation anomaly conditions, draw the vegetation cumulative anomaly curve, identify the start and end times of vegetation recovery through this curve, and calculate the vegetation recovery duration .

[0050] Among them, the "vegetation cumulative anomaly curve" refers to the vegetation negative anomaly event that appears after weeks of the occurrence of the drought event. The vegetation anomaly values at each moment of the vegetation negative anomaly event are accumulated and summed moment by moment to obtain the vegetation cumulative anomaly curve, as shown in Figure 3 (a). In the present invention, is set, that is, it is concerned that the vegetation negative anomaly situation within 12 weeks (about 3 months) after the drought event is related to this drought event.

[0051] The "vegetation recovery duration" mentioned above refers to the time interval between the moment when the vegetation reaches the maximum local cumulative loss and the moment when it recovers to the normal state after the drought event. There are two important time nodes involved. Among them, when the drought event occurs, the moment when the vegetation cumulative anomaly reaches the local minimum is defined as the start time of the recovery of the vegetation anomaly state , as shown in Figure 3 (b); and when the vegetation cumulative anomaly gradually increases and recovers or exceeds 0 and lasts for at least 2 weeks, it is defined as the end time of the recovery of the vegetation anomaly state (as shown in Figure 3 ), therefore, the vegetation recovery duration .

[0052] S5: Introduce the extreme gradient boosting tree model to simulate and predict the vegetation recovery duration in the study area.

[0053] Extract the typical characteristics of drought events (drought duration , drought intensity , drought peak ), canopy height , root depth , soil texture and groundwater depth data at grid points in the historical period of the study area to construct the independent variable sequence, and the vegetation recovery duration as the dependent variable sequence. Randomly select 80% of the independent variable data as the training sample and 20% as the test sample.

[0054] Put the training sample into the extreme gradient boosting decision tree (XGBoost) model and set the hyperparameters of the model, including the learning rate, the maximum depth of the tree, the regularization coefficient, etc. Through the gradient boosting algorithm, gradually optimize the objective function:

[0055] ,

[0056] In the formula, represents the true value of the th sample, represents the predicted value of the th sample, is the number of samples, is the loss function, which measures the difference between the predicted value of the model and the true value ; is the prediction result of the th decision tree, is the total number of trees, is the regularization term, which controls the complexity of the model.

[0057] Use the test sample to evaluate the performance of the model. The evaluation metrics are the mean square error and the Nash efficiency coefficient . The former can reflect the overall size of the prediction error, and the latter can reflect the fitting degree between the predicted value and the true value. The formulas are as follows:

[0058] ,

[0059] ,

[0060] In the formula, Represents the average of the true values.

[0061] By monitoring the soil moisture changes at the weekly scale in the study area in the future, identifying soil drought events, extracting typical characteristics such as drought duration, intensity, and peak value, and combining the monitoring data of canopy height, root depth, soil texture, and groundwater depth, and substituting them into the trained extreme gradient boosting decision tree model, the vegetation recovery duration of any grid point can be predicted. , such as Figure 4 shown, the calculation formula is:

[0062] ,

[0063] In the formula, is the prediction result of the th decision tree, and is the total number of trees.

[0064] In one embodiment, a computer device is provided, 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 above-mentioned quantitative prediction method for the vegetation recovery duration under the influence of drought events is implemented.

[0065] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned quantitative prediction method for the vegetation recovery duration under the influence of drought events is implemented.

[0066] In one embodiment, a computer program product is provided, including a computer program / instructions. When the computer program / instructions are executed by a processor, the above-mentioned quantitative prediction method for the vegetation recovery duration under the influence of drought events is implemented.

[0067] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0068] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce a means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0069] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0070] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

Claims

1. A quantitative prediction method for the duration of vegetation recovery under the influence of drought events, characterized in that: The following steps are involved: Step 1: Extract the vegetation status index at the grid point level in the study area and soil moisture content time series; Step 2: After de-seasoning and de-trending the time series of the vegetation status index at each grid point, the anomaly value is obtained, and the abnormal vegetation condition corresponding to each grid point is identified according to the anomaly value; For the time series of soil moisture content at each grid point, the soil drought events corresponding to each grid point are identified based on the soil moisture median method, and the typical drought characteristics of each grid point are extracted; Step 3: Combine the soil drought events at each grid point with the corresponding abnormal vegetation conditions to obtain the vegetation cumulative anomaly curve. Use the vegetation cumulative anomaly curve to identify the start and end time of the abnormal vegetation state recovery and calculate the vegetation recovery time. ; Step 4: Based on the typical characteristics of drought events at each grid point in the study area, canopy height , root depth , soil texture and groundwater depth Construct an independent variable sequence, based on the duration of vegetation recovery As the dependent variable sequence, the extreme gradient boosting tree model is trained, and the vegetation recovery time of the study area is simulated and predicted through the trained extreme gradient boosting tree model; The vegetation cumulative anomaly curve is used to identify the start and end time of vegetation abnormal state recovery and calculate the vegetation recovery time. Including: The moment when the vegetation cumulative anomaly reaches the local minimum is defined as the moment when the abnormal state of vegetation is restored , the vegetation accumulation anomaly recovers or exceeds 0 and lasts for at least m Weekly, defined as the end time of the abnormal vegetation state recovery , calculate the vegetation recovery time, ; The trained extreme gradient boosting tree model is used to simulate and predict the vegetation recovery time in the study area. The calculation formula is: ， In the formula, is the vegetation recovery time output by the extreme gradient boosting tree model, For the The prediction results of a decision tree, is the total number of trees, The duration of drought, The intensity of drought, The peak of drought, is the canopy height, is the root depth, For soil texture, The depth of groundwater.

2. The quantitative prediction method for vegetation recovery time under the influence of drought events according to claim 1 is characterized by: The "deseasonalization" processing includes: using the difference method, the monthly data of the vegetation state index are subtracted from the multi-year monthly average; The "detrending" process includes: fitting the original data of the vegetation state index by the least square method to obtain a fitting curve, and then subtracting the fitting curve of the original time series from the original time series of the vegetation state index; According to the anomaly value, the vegetation anomaly corresponding to each grid point is identified, including: when the anomaly value is less than 0 for the first time, the corresponding time is defined as the start time of the vegetation negative anomaly event; when the anomaly value returns to not less than 0 again, the corresponding time is defined as the end time of the vegetation negative anomaly event; between the start time and the end time, the minimum value of the anomaly value is defined as the peak value of the vegetation negative anomaly event; For the time series of soil moisture content at each grid point, identifying the soil drought events corresponding to each grid point based on the soil moisture median method includes: splitting the time series of soil moisture content in each year into several time periods, and combining the corresponding time periods of several years; selecting a distribution function from a variety of cumulative probability distribution functions, and converting the combined data into a soil moisture median sequence; identifying soil drought events according to the soil moisture median sequence; the cumulative probability distribution function includes gamma distribution, beta distribution, extreme value distribution, generalized extreme value distribution, logarithmic distribution, logarithmic distribution and Weibull distribution.

3. The quantitative prediction method for vegetation recovery time under the influence of drought events according to claim 1 is characterized by: The typical characteristics of drought include drought duration, drought intensity and drought peak; drought duration refers to the duration from when the soil moisture content decreases below a first threshold to when it recovers or exceeds the first threshold; drought intensity refers to the average value of the soil moisture content during the drought; and drought peak refers to the minimum value of the soil moisture content during the drought.

4. The quantitative prediction method for vegetation recovery time under the influence of drought events according to claim 1 is characterized by: In response to drought events For vegetation negative anomaly events that occur after one week, the vegetation anomaly values ​​at each moment of the vegetation negative anomaly event are accumulated moment by moment to obtain the vegetation cumulative anomaly curve.

5. The quantitative prediction method for vegetation recovery time under the influence of drought events according to claim 1, characterized in that: The objective function of the extreme gradient boosting tree model is: ， In the formula, is the objective function, Indicates The true value of the samples, Indicates The predicted value of samples, is the number of samples, is the loss function, which measures the model prediction value and the true value differences; For the The prediction results of a decision tree, is the total number of trees, is the regularization term.

6. 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 quantitatively predicting the duration of vegetation recovery under the influence of a drought event according to any one of claims 1 to 5 is implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for quantitatively predicting the duration of vegetation recovery under the influence of a drought event according to any one of claims 1 to 5 is implemented.

8. 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 quantitatively predicting the duration of vegetation recovery under the influence of a drought event as described in any one of claims 1 to 5 is implemented.

Images