A method and device for quantitatively detecting influence of meteorological drought on abnormal irrigation water

Abstract

The application provides a kind of meteorological drought to the quantitative detection method of influence of abnormal irrigation water, comprising: meteorological drought event is determined monthly based on meteorological data set and set drought intensity threshold;Residual sequence is extracted from irrigation water quantity sequence based on time series decomposition, and irrigation water anomaly event is identified monthly;The proportion that irrigation water anomaly event and meteorological drought event occur in the preset time window under the set drought intensity threshold is counted, defined as trigger coincidence rate;Trigger coincidence rate under different drought intensity thresholds is counted, and the region that irrigation water anomaly event is obviously affected by meteorological drought event is identified based on event consistency analysis, defined as affected region;Using time overlay analysis, the residual sequence of irrigation water quantity in the affected region is divided into meteorological drought influence sub-sequence and normal sub-sequence, and the difference is analyzed.The application can effectively capture the change of irrigation water quantity caused by meteorological drought.

Description

Technical Field

[0001] This invention relates to the field of water resources management technology, and in particular to a quantitative detection method and apparatus for the impact of meteorological drought on abnormal irrigation water use. Background Technology

[0002] Agricultural irrigation refers to the process of introducing water that meets quality standards into farmland through artificial facilities to meet the water requirements for crop growth. Irrigation water is crucial for crop growth, especially in areas facing water shortages. Meteorological drought is one of the major disasters threatening crop growth. During its occurrence, a persistent deficit in atmospheric moisture reduces water supply, thus limiting normal crop growth; simultaneously, abnormal changes in meteorological elements can also cause fluctuations in crop water demand. Therefore, irrigation water is highly sensitive to drought changes and often exhibits significant fluctuations during drought events. Against the backdrop of frequent drought events caused by climate change, accurately assessing changes in irrigation water consumption caused by meteorological drought is of great significance for agricultural water resource management.

[0003] Irrigation water use is influenced by both climatic conditions and human management practices, which exhibit spatial heterogeneity across different regions. Current methods for assessing the impact of meteorological drought on irrigation water use typically simply substitute meteorological data into the evapotranspiration formula to extrapolate the increase in crop water demand, and then assume a corresponding increase in irrigation water use, ignoring the difference between water demand and actual water use, as well as the impact of human management practices in different regions. In reality, the threshold for triggering changes in irrigation water use due to meteorological drought may vary across regions. Therefore, identifying the impact of drought based on actual irrigation water use data for each region is necessary to more accurately reflect the differentiated responses between areas.

[0004] In addition, irrigation water consumption is driven by multiple factors such as meteorological conditions, planting structure, and water-saving policies. Changes in irrigation water consumption caused by drought are often masked by long-term trends and seasonal fluctuations, making it difficult for traditional analysis methods to effectively extract drought impact signals from complex backgrounds.

[0005] Furthermore, existing research methods on the impact of drought events mostly remain at the level of spatiotemporal overlap, that is, judging the relationship between drought indicators and affected variables by comparing their consistency in time or space. However, such methods can only reveal correlations and cannot distinguish which anomalies are indeed triggered by drought and which are just accidental co-occurrences, thus lacking causal identification power.

[0006] Therefore, there is an urgent need for an assessment method that can accurately identify changes in irrigation water consumption caused by meteorological drought. Such a method can not only accurately extract the time periods and regions affected by meteorological drought events from actual water use data, but also quantify the specific magnitude and direction of the impact of meteorological drought events on irrigation water use in different regions. Summary of the Invention

[0007] To address the shortcomings of existing quantitative detection methods for the impact of drought on irrigation water, such as insufficient background noise filtering and difficulty in identifying the causal relationship between drought and water use changes, this invention discloses a method for detecting the impact of meteorological drought events on abnormal irrigation water use, combining time series decomposition, machine learning, and event consistency analysis. This method effectively eliminates long-term trends, seasonal fluctuations, and other background noise, identifies the existence and timing of drought impacts based on actual irrigation water use data, and reveals the irrigation water use response characteristics in different regions. This provides technical support for the scientific quantification of the actual impact of meteorological drought events on irrigation water use and regional differences.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] The main objective of this invention is to provide a quantitative detection method for abnormal irrigation water use caused by meteorological drought, so as to capture changes in irrigation water use caused by meteorological drought.

[0010] The first aspect of this invention provides a method for quantitatively detecting the impact of meteorological drought on abnormal irrigation water use, comprising:

[0011] Step S100: Obtain monthly irrigation water consumption data and daily meteorological variable data with the same spatial resolution from the preset database, and construct irrigation water consumption series and meteorological datasets respectively.

[0012] Step S200: Determine meteorological drought events month by month based on the meteorological dataset and the set drought intensity threshold;

[0013] Step S300: Extract residuals from the irrigation water consumption sequence based on the time series decomposition method to obtain the irrigation water consumption residual sequence; perform anomaly detection on irrigation water consumption based on the irrigation water consumption residual sequence; and identify abnormal irrigation water consumption events month by month.

[0014] Step S400: Calculate the proportion of abnormal irrigation water events and meteorological drought events occurring within a preset time window under the set drought intensity threshold, and define it as the trigger overlap rate;

[0015] Step S500: Change the drought intensity threshold, repeat steps S200 to S400, and identify areas where irrigation water abnormal events are significantly affected by meteorological drought events based on the event consistency analysis method, and define them as affected areas.

[0016] Step S600: Using time overlay analysis, the residual sequence of irrigation water consumption in the affected area is divided into a meteorological drought-affected subsequence and a normal subsequence. The difference between the two subsequences is analyzed to quantitatively identify the magnitude of the impact of meteorological drought events on irrigation water anomalies in different areas.

[0017] In some embodiments, step S200 includes:

[0018] Calculate the daily standardized precipitation evapotranspiration index SPEI-1 based on the meteorological dataset;

[0019] If the calculated SPEI-1 for the last day of a month is lower than the baseline drought intensity threshold, a meteorological drought event is determined to have occurred in that month, and the month is recorded as a meteorological drought month; otherwise, no meteorological drought event is determined to have occurred in that month.

[0020] In some embodiments, when extracting residuals from the irrigation water consumption sequence using the time series decomposition method, an improved STL time series decomposition method is employed. This improved STL time series decomposition method uses the following model:

[0021]

[0022] in, For months Irrigation water consumption; For the first Decomposed in the next iteration Trend items in; For the first Decomposed in the next iteration Seasonal items in; For the first Decomposed in the next iteration The residuals in; and These are the smoothing weights used for the trend term and the smoothing weights used for the seasonal term, respectively, and their calculation formulas are as follows:

[0023]

[0024]

[0025]

[0026]

[0027] Among them, months For irrigation water consumption series excluding months In any other month besides This represents the total number of months contained in the irrigation water consumption sequence; and By month Centered on, for other months Apply Gaussian kernel weights to the trend and seasonal terms; and Each month Trend-adaptive bandwidth and seasonal-adaptive bandwidth; The baseline bandwidth; For the current month The median absolute deviation of the residuals within the central window; This represents the absolute deviation of the median of the irrigation water consumption series. and These are adjustment parameters for the trend term and the seasonal term, respectively, used to control the sensitivity of the adaptive bandwidth to the degree of local variation.

[0028] In some embodiments, the step of detecting anomalies in irrigation water volume based on the irrigation water volume residual sequence and identifying abnormal irrigation water events monthly includes:

[0029] A long short-term memory neural network is used to model the residual sequence of irrigation water consumption, and the reconstruction error is calculated; an adaptive threshold is set. The adaptive threshold The reconstruction error distribution of the irrigation water residual sequence modeling based on the long short-term memory neural network is determined;

[0030] If the reconstruction error of a certain month is higher than the adaptive threshold If an abnormal irrigation water event occurs in that month, it is determined that an abnormal irrigation water event has occurred and that month is recorded as an abnormal irrigation water month; otherwise, it is determined that no abnormal irrigation water event has occurred in that month.

[0031] In some embodiments, the Huber loss function is used when training the long short-term memory neural network:

[0032]

[0033] in, For the monthly residual sequence of irrigation water consumption The reconstruction error, i.e., the extracted month The difference between the residual of irrigation water consumption and the predicted value of the long short-term memory neural network; For transition parameters, , This represents the standard deviation of the reconstruction error for all months in the irrigation water consumption residual sequence;

[0034] The adaptive threshold Set it according to the following formula:

[0035]

[0036] in, is the mean of the reconstruction error for all months in the irrigation water consumption residual sequence; c is the threshold coefficient.

[0037] In some embodiments, in step S400, the trigger overlap rate is set to... :

[0038]

[0039]

[0040]

[0041] In the formula, This represents the total number of meteorological drought events statistically analyzed for the aforementioned meteorological dataset. This represents the total number of abnormal irrigation water events statistically analyzed for the irrigation water consumption sequence. for The Middle The timing of meteorological drought events for The Middle The time of occurrence of the abnormal irrigation water event; The set time window length, which is the length of months that allows for a lag between abnormal irrigation water events and drought events; It is an indicator function; It is the Heaviside step function.

[0042] In some embodiments, step S500 includes:

[0043] Step S510: Change the drought intensity threshold And repeat steps S200 to S400 to obtain different drought intensity thresholds. Trigger overlap rate ;

[0044] Step S520: For the drought intensity threshold Overlap rate with trigger The relationship between the two models was analyzed, and both linear and piecewise linear models were fitted. The merits of the linear and piecewise linear models were compared. For a given region, if the piecewise linear fit was superior to the linear fit, it indicated a higher drought intensity threshold. Overlap rate with trigger When there is an inflection point in the relationship between them, that area is identified as the affected area.

[0045] In some embodiments, for drought intensity threshold Overlap rate with trigger The relationship between them can be fitted by the following linear model:

[0046]

[0047] For drought intensity threshold Overlap rate with trigger The relationship between them, and the fitted piecewise linear model is:

[0048]

[0049] in, For the inflection point threshold of the piecewise linear model, These are the regression coefficients, , These are the fitting residuals;

[0050] The linear model and the piecewise linear model are compared using information criteria or statistical testing methods.

[0051] In some embodiments, the quantitative detection method further includes:

[0052] Step S700: Spatial visualization of the quantitative detection results of the impact of meteorological drought on abnormal irrigation water obtained in step S600.

[0053] A second aspect of the present invention provides a quantitative detection device for the impact of meteorological drought on abnormal irrigation water use, comprising:

[0054] The first module is configured to obtain monthly irrigation water consumption data and daily meteorological variable data with the same spatial resolution within the research time range from a preset database, and construct irrigation water consumption series and meteorological datasets respectively.

[0055] The second module is configured to determine meteorological drought events monthly based on the meteorological dataset and a set drought intensity threshold.

[0056] The third module is configured to extract residuals from the irrigation water consumption sequence based on the time series decomposition method to obtain an irrigation water consumption residual sequence, and to perform anomaly detection on irrigation water consumption based on the irrigation water consumption residual sequence to identify abnormal irrigation water consumption events on a monthly basis.

[0057] The fourth module is configured to statistically analyze the proportion of abnormal irrigation water events and meteorological drought events occurring within a preset time window under a set drought intensity threshold, and define it as the trigger overlap rate.

[0058] The fifth module is configured to change the drought intensity threshold, and call the second, third and fourth modules, and identify areas where irrigation water anomalies are significantly affected by meteorological drought events based on the event consistency analysis method, defining them as affected areas;

[0059] The sixth module is configured to use time overlay analysis to divide the residual sequence of irrigation water consumption in the affected area into a meteorological drought-affected subsequence and a normal subsequence, analyze the differences between the two subsequences, and thus quantitatively identify the magnitude of the impact of meteorological drought events on irrigation water anomalies in different areas.

[0060] Features and beneficial effects of the present invention:

[0061] 1. In view of the problem that existing methods generally equate changes in crop evapotranspiration water demand with changes in actual water use when assessing the impact of meteorological drought on irrigation water use, and do not consider the differences in human management in different regions, this invention establishes an assessment framework based on actual irrigation water use data, which can identify the differentiated response characteristics of irrigation water use to drought in different regions under the background of water scarcity.

[0062] 2. To address the problem that irrigation water consumption is driven by multiple factors and that drought impact signals are easily masked by long-term trends and seasonal fluctuations, this invention combines time series decomposition with an anomaly detection method based on LSTM to effectively eliminate interference from trend and seasonal terms, accurately extracting abnormal irrigation water events, and demonstrating accuracy and robustness.

[0063] 3. To address the problem that existing studies often focus on the simple spatiotemporal overlap of meteorological drought and irrigation water anomalies without causal identification, this invention introduces Event Consistency Analysis (ECA). By calculating the trigger overlap rate of drought events and irrigation water anomaly events, and combining it with piecewise linear model fitting to determine the correlation, this invention can identify which irrigation water anomalies are indeed triggered by meteorological drought, thereby improving the causal explanatory power of meteorological drought impact identification.

[0064] 4. This invention not only enables the qualitative identification of irrigation water anomalies triggered by meteorological drought events, but also achieves quantitative detection of the impact of meteorological drought on irrigation water anomalies by calculating the trigger overlap rate and estimating the affected area. Compared with traditional methods that rely on correlation or descriptive statistics, this quantitative detection can output specific amounts of irrigation water impact, making the relationship between meteorological drought and irrigation water comparable across different regions and under different conditions, thus providing a more reliable basis for drought risk management decisions. Attached Figure Description

[0065] Figure 1 This is an overall flowchart of the method of the present invention;

[0066] Figure 2 This is a schematic diagram of the elevation of the Yellow River Basin research area involved in the embodiments of the present invention;

[0067] Figure 3 This refers to the time series, long-term trend, seasonal cycle, and residual sequence of irrigation water consumption in the Yellow River Basin from January 1980 to December 2022, which are involved in the embodiments of the present invention.

[0068] Figure 4 This is a comparison of the autocorrelation coefficients between the time series of irrigation water consumption in the Yellow River Basin from January 1980 to December 2022 and the separated irrigation water residual series involved in this embodiment of the invention.

[0069] Figure 5 This refers to the absolute average error change of the machine learning model for identifying abnormal irrigation water volume based on LSTM, as described in this embodiment of the invention.

[0070] Figure 6 This refers to the distribution of irrigation water consumption affected by drought in the embodiments of the present invention.

[0071] Figure 7 This refers to the impact of meteorological drought from 1980 to 2020 on irrigation water consumption in the Yellow River Basin, as described in this embodiment of the invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of this application clearer, the application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application.

[0073] Conversely, this application covers any alternatives, modifications, equivalent methods, and schemes made within the spirit and scope of this application as defined by the claims. Furthermore, to provide the public with a better understanding of this application, certain specific details are described in detail below. However, this application can be fully understood by those skilled in the art even without these detailed descriptions.

[0074] See Figure 1 The first aspect of this invention provides a method for quantitatively detecting the impact of meteorological drought on abnormal irrigation water use, comprising the following steps:

[0075] Step S100: Obtain monthly irrigation water consumption data and daily meteorological variable data with the same spatial resolution from the preset database, and construct irrigation water consumption series and meteorological datasets respectively.

[0076] Step S200: Determine meteorological drought events month by month based on the above meteorological dataset and the set drought intensity threshold;

[0077] Step S300: Extract residuals from the above irrigation water volume sequence based on the time series decomposition method to obtain the irrigation water volume residual sequence. Based on the irrigation water volume residual sequence, perform anomaly detection on irrigation water volume and identify abnormal irrigation water events month by month.

[0078] Step S400: Calculate the proportion of abnormal irrigation water events and meteorological drought events occurring within a preset time window under the set drought intensity threshold, and define it as the trigger overlap rate;

[0079] Step S500: Change the drought intensity threshold, repeat steps S200 to S400, and identify areas where irrigation water abnormal events are significantly affected by meteorological drought events based on the event consistency analysis method, and define them as affected areas.

[0080] Step S600: Using time overlay analysis, the residual sequence of irrigation water consumption in the affected area is divided into a meteorological drought-affected subsequence and a normal subsequence. The difference between the two subsequences is analyzed to quantitatively identify the magnitude of the impact of meteorological drought events on irrigation water anomalies in different areas.

[0081] In some embodiments, in step S100, daily meteorological observation data collected by meteorological stations set up within the study area are collected, including daily precipitation and daily potential evapotranspiration data; monthly irrigation water data within the study area are collected, including irrigation water volume, and the equivalent water depth is obtained by dividing it by the area of ​​the study area, in millimeters per month; a database is constructed using the collected daily meteorological observation data and monthly irrigation water data. Subsequently, monthly irrigation water volume data and meteorological variable data with the same spatial resolution within the study time range are obtained from the database and arranged in chronological order to construct irrigation water volume series and meteorological datasets, respectively.

[0082] In some embodiments, step S200 includes:

[0083] Step S210: Calculate the daily standardized precipitation evapotranspiration index SPEI-1 based on the meteorological dataset constructed in step S100.

[0084]

[0085]

[0086]

[0087]

[0088] in, It is the cumulative water deficit in the 30 days preceding the nth day of year y; It is the daily water deficit on day i in year y, and is taken as the precipitation on day i in year y. Potential evapotranspiration on day i in year y The difference; It is the standardized precipitation evapotranspiration index for the nth day of year y; Yes The cumulative probability distribution value after adjusting for the probability of a zero-moisture deficit day; q is the logistic distribution function of the cumulative water deficit; q is the probability of a zero water deficit day. It is the inverse cumulative distribution function of the standard normal distribution.

[0089] Step S220: Perform the following judgment:

[0090] When the calculated SPEI-1 for the last day of a month is lower than the baseline drought intensity threshold If a meteorological drought event occurs in a given month, that month is determined to be a meteorological drought month; otherwise, it is determined that no meteorological drought event occurred in that month. Baseline drought intensity threshold. The settings are based on the drought level classification in the "National Standard of the People's Republic of China GB / T 20481-2017". In a specific embodiment of this application, the baseline drought intensity threshold is... Set it to -0.5.

[0091] In some embodiments, step S300 includes:

[0092] Step 310: Use the time series decomposition method based on Loess's Seasonal-Trend Decomposition using LOESS (STL) to separate the interannual variation trend (hereinafter referred to as "trend term"), seasonal cycle (hereinafter referred to as "seasonal term") and residuals from the irrigation water consumption series, so as to eliminate the influence of long-term trends and periodic fluctuations on the identification of irrigation water anomalies.

[0093] Furthermore, this invention improves upon existing STL-based time series decomposition methods. During the decomposition process, the existing STL trickle weight function is replaced with a Gaussian kernel function, and an adaptive bandwidth adjustment mechanism is used to correct the weights of the long-term trend term and the smoothed seasonal term of the irrigation water consumption sequence, thereby obtaining the irrigation water consumption residual sequence after removing the long-term trend term and the smoothed seasonal term. Specifically:

[0094] The basic model of the STL-based time series decomposition method is as follows:

[0095]

[0096] In the formula, For months Irrigation water consumption; for The long-term trend component; for Seasonal cyclical components; for The residuals in the data include disturbances caused by external pressures (such as extreme heat events, meteorological drought events, etc.).

[0097] Iteratively extract trend and seasonal items step by step:

[0098]

[0099] in, For the first Decomposed in the next iteration The long-term trend component; For the first Decomposed in the next iteration Seasonal cyclical components; For the first Decomposed in the next iteration The residuals in; and These are the smoothing weights used for the trend item and the smoothing weights used for the seasonal item, respectively, for the month. With Month The closer the time interval, the more likely it is to occur in a month. The greater the contribution of the trend or seasonal term to the smoothing, the greater the smoothing weight; conversely, the smaller the contribution of the monthly term, the greater the smoothing weight. With Month The greater the time distance, the more it appears in months. The smaller the contribution of the trend or seasonal term to the smoothing, the lower the smoothing weight. and The calculation formula is as follows:

[0100]

[0101]

[0102]

[0103]

[0104] Among them, months For irrigation water consumption series excluding months any other month, i.e. , This represents the total number of months contained in the irrigation water consumption sequence; and By month Centered on, for other months Apply Gaussian kernel weights to the trend and seasonal terms; For the month Trend term adaptive bandwidth, For the month The seasonal adaptive bandwidth represents the effective smoothing scale of the trend term and the seasonal term, respectively. The larger the bandwidth, the greater the weight changes with time distance. The slower the decay and the smaller the bandwidth, the more the weight changes with time distance. The faster the decay; The reference bandwidth is preferably 12; For the current month The median absolute deviation of the residuals within the central window measures the degree to which the local MAD deviates from the overall level. The central window is used to limit the range of data involved in local noise estimation, and is preferably 6. This represents the absolute deviation of the median of the irrigation water consumption series. and These are adjustment parameters for the trend term and the seasonal term, respectively, used to control the sensitivity of the adaptive bandwidth to the degree of local variation. , The larger the value of , the greater the bandwidth scaling when the local MAD deviates from the overall level, and the stronger the adaptive effect. , The smaller the value of , the smoother the adaptive change, approaching a fixed bandwidth. Considering the noise level and extreme event characteristics of the irrigation water consumption sequence, a value between 0.3 and 0.7 is preferred. In this embodiment, is taken as... = 0.5, = 0.4.

[0105] When the following conditions are met: That is, the difference between the sum of the decomposed terms and the irrigation water consumption is less than the threshold for iterative convergence. hour, Pick If the iteration terminates, then the iteration ends.

[0106] It should be noted that as M increases, the estimation of the trend and seasonal terms in the STL decomposition becomes more stable, but the core smoothing window is still determined by the adaptive bandwidth parameter. and The control ensures that the calculation of the residuals will not be biased as M increases. Total number of iterations (i.e. The upper limit of the value is an externally set hyperparameter and is not directly related to M. Its convergence speed is determined by the adaptive bandwidth and weight function, not by M. It is preferable to set the total number of iterations to between 10 and 20 to ensure stable separation between the trend and seasonal terms.

[0107] Step 320: A Long Short-Term Memory (LSTM) neural network is used to model the irrigation water consumption residual sequence obtained in step S310, and the reconstruction error is calculated. Considering the fluctuating nature of irrigation water consumption, an adaptive threshold is set. Adaptive threshold The reconstruction error distribution of irrigation water residual sequence modeling based on LSTM model is determined. If the reconstruction error of a certain month exceeds this adaptive threshold... If an abnormal irrigation water event occurs in that month, it is determined that an abnormal irrigation water event has occurred and that month is recorded as an abnormal irrigation water month; otherwise, it is determined that no abnormal irrigation water event has occurred in that month.

[0108] Furthermore, the Huber loss function is employed during LSTM training to enhance the robustness of LSTM in handling outliers and noise in the irrigation water residual sequence; the Huber loss function is as follows:

[0109]

[0110] in, To target the monthly residual series of irrigation water consumption The reconstruction error, i.e., the true value of the residual. (i.e., the month is finally extracted in step S310) Irrigation water consumption residuals and LSTM simulation values difference; As a transition parameter, the preferred one is... , This represents the standard deviation of the reconstruction error for all months in the irrigation water consumption residual sequence.

[0111] The embodiments of the present invention use the true value of the residual An LSTM was input, and a residual sequence was simulated. Months with large differences between the two were considered to be abnormal months for irrigation water residuals.

[0112] Furthermore, adaptive threshold Based on reconstruction error The overall average level and fluctuation range are adaptively determined, and the calculation formula is as follows:

[0113]

[0114] in, To reconstruct the mean of the error sequence; is the standard deviation of the reconstructed error sequence; c is the threshold coefficient, preferably in the range of 2-3.

[0115] It is understandable that, through residual extraction and anomaly detection in step S300, this invention can identify irrigation water anomalies caused by external factors (such as meteorological drought) on a monthly basis, under the interference of trend changes, seasonal cycles and natural noise, thereby improving the stability and robustness of anomaly identification and providing reliable input data for subsequent quantitative analysis of the impact of meteorological drought on irrigation water anomalies.

[0116] In some embodiments, step S400 includes:

[0117] Based on the irrigation water anomaly events identified in step S320 and the meteorological drought events identified in step S220, if the irrigation water anomaly events and the meteorological drought events occur within a set time window, it is determined that the irrigation water anomaly events and the meteorological drought events coincide in time. For example, if the time window length is set to 1 month, then if an irrigation water anomaly event (in this application, the meteorological drought event is considered a preceding event) occurs in the same month as the meteorological drought event (in this application, the meteorological drought event is considered a triggered event) or in the following month, it is determined that the meteorological drought event and the irrigation water anomaly event coincide (as long as the irrigation water anomaly event occurs within this time window length, whether it is 1 time or 2 times, it is determined that the irrigation water anomaly event and the meteorological drought event coincide in time); if the time window is set to 0 months, then only if an irrigation water anomaly event occurs in the same month as the meteorological drought event is it determined that the meteorological drought event and the irrigation water anomaly event coincide in time. Subsequently, the proportion of all irrigation water anomaly events determined by step S320 that coincide with all meteorological drought events determined by step S220 is statistically analyzed and defined as the trigger overlap rate. :

[0118]

[0119]

[0120]

[0121] In the formula, This is the total number of meteorological drought events counted from the meteorological dataset. This represents the total number of abnormal irrigation water events statistically analyzed based on irrigation water consumption sequences. for The Middle The timing of a meteorological drought event (more specifically, the month). for The Middle The time (more specifically, the month) when the abnormal irrigation water event occurred; The set time window length is the length of months allowed for a lag between abnormal irrigation water events and drought events. ; For the characteristic function, when the first... The meteorological drought event and the first The absolute value of the time difference between the occurrence of each abnormal irrigation water event belongs to Inside, then the first The first irrigation water anomaly event affected the first A single meteorological drought event is counted as 1, otherwise it is counted as 0. For the Heaviside step function, as long as the first... After a meteorological drought event occurred If at least one abnormal irrigation water event occurs within the specified time period, then the event is counted as the [number missing]. The contribution of the first meteorological drought event to the abnormal irrigation water event is 1. If the first... After a meteorological drought event occurred If no abnormal irrigation water event occurs within the specified time period, then the period is counted as the first. The contribution of each meteorological drought event to irrigation water anomalous events is 0. Therefore, each meteorological drought event has a lag effect on its... An abnormal irrigation water event occurring within a given time period can be counted as a maximum of one contribution.

[0122] Understandably, step S400 quantifies the causal triggering relationship between meteorological drought events and abnormal irrigation water events in the time dimension by constructing the trigger overlap rate between the two events.

[0123] In some embodiments, step S500 includes:

[0124] Using event consistency analysis, regions significantly affected by meteorological drought events in irrigation water use were identified and defined as affected areas. The meaning of an affected area is: within a drought intensity threshold... During the upward trend, the trigger overlap rate shows a region where it significantly increases near a certain drought intensity threshold. In other words, when the trigger overlap rate approaches the drought intensity threshold... If there is an inflection point in the relationship, and the overlap rate increases after exceeding this inflection point, then the abnormal irrigation water events in the region are considered to be caused by meteorological drought events; otherwise, the abnormal irrigation water events in the region are considered not to be caused by meteorological drought events.

[0125] Furthermore, step S500 specifically includes:

[0126] Step S510: Calculate the threshold values ​​for different drought intensities. Trigger overlap rate Specifically, changing the drought intensity threshold Repeat steps S200 to S400 to obtain the corresponding trigger overlap rate. Drought intensity threshold According to a fixed step size Gradually from the highest drought intensity threshold Gradually reduce to the lowest drought intensity threshold In one specific embodiment of this application, , , ,Right now Take in order If anomalies in irrigation water use do not show a significant response to meteorological drought events (i.e., the two do not overlap), then The value is close to 0; if anomaly events in irrigation water use show a response in all meteorological drought events, then The value approaches 1.

[0127] Step S520, regarding the drought intensity threshold Overlap rate with trigger The relationship between the two models was analyzed, fitting both linear and piecewise linear models. The merits of the linear and piecewise linear models were compared. For a specific region within the study area, if the piecewise linear fit was superior to the linear fit, it indicated a higher drought intensity threshold. Overlap rate with trigger When there is an inflection point in the relationship between them, that area is identified as the affected area. Among them,

[0128] For drought intensity threshold Overlap rate with trigger The relationship between them can be fitted by the following linear model:

[0129]

[0130] For drought intensity threshold Overlap rate with trigger The relationship between them, and the fitted piecewise linear model is:

[0131]

[0132] in, For the inflection point threshold of the piecewise linear model, These are the regression coefficients, , These are the fitting residuals, respectively.

[0133] Furthermore, information criteria or statistical tests are used to compare the merits of the linear model and the piecewise linear model. Preferably, the Akaike Information Criterion (AIC) is used to compare the merits of the linear model and the piecewise linear model. The formula for calculating the Akaike Information Criterion (AIC) is as follows:

[0134]

[0135] in, The number of regression coefficients in a linear model or a piecewise linear model, for example, for a linear model. For piecewise linear models ; This represents the maximum likelihood estimate for a linear or piecewise linear model. A difference greater than 2 is used as a criterion for the piecewise linear model to be significantly better than the linear model; that is, when the difference between the piecewise linear model and the linear model is greater than 2, the piecewise linear model is significantly better than the linear model. A piecewise linear model is considered superior to a linear model only when the difference is greater than 2.

[0136] Understandably, step S500 robustly identifies spatial regions significantly affected by drought by repeatedly calculating the triggering relationship between meteorological drought and irrigation water anomalies under different drought intensity threshold conditions by changing the drought intensity threshold and using the event consistency analysis method.

[0137] In some embodiments, step S600 includes:

[0138] Within the affected area identified in step S500, firstly, referring to step S300, the irrigation water residual sequence of the affected area is extracted, and abnormal irrigation water events are identified monthly; subsequently, referring to step S400, the temporal overlap relationship between meteorological drought events and abnormal irrigation water events within the affected area is analyzed (wherein, the drought intensity threshold used to determine whether meteorological drought events and abnormal irrigation water events overlap temporally adopts the national meteorological drought standard value, such as...). (etc.) The residual sequence is divided into drought-affected subsequence and normal subsequence. The overlapping part of the residual sequence is taken as the drought-affected subsequence, and the non-overlapping part of the residual sequence is taken as the normal subsequence. The difference in median between the two subsequences is compared and analyzed to quantitatively identify the response amplitude of irrigation water in different regions.

[0139] Understandably, step S600 quantitatively identifies the magnitude of the impact of meteorological drought on irrigation water anomalies by performing a time overlay analysis on the residual series of irrigation water consumption within the affected area and comparing the residual distribution during the drought-affected period with that during the normal period. This method avoids interference from trend and seasonal components and ensures that the analysis is conducted only in areas truly affected by drought through spatial screening, thereby improving the accuracy and regional interpretability of drought impact estimation.

[0140] In some embodiments, the quantitative detection method provided by the present invention further includes, after step S600:

[0141] Step S700: Spatial visualization of the quantitative detection results of the impact of meteorological drought on abnormal irrigation water obtained in step S600, so as to be directly used for regional comparison, risk assessment and water resource management decision-making.

[0142] The following describes Example 1 of a quantitative detection method for the impact of meteorological drought on abnormal irrigation water use, provided by the first aspect of the present invention. This example selects the Yellow River Basin, where water shortages are prominent, as the study area, including parts of Sichuan Province, Gansu Province, Qinghai Province, Inner Mongolia Autonomous Region, Ningxia Hui Autonomous Region, Shaanxi Province, Shanxi Province, Shandong Province, and Henan Province. See [link to relevant documentation]. Figure 2 This embodiment specifically includes the following steps:

[0143] Step S100, Data Collection

[0144] In this embodiment, daily observation data (including daily temperature, precipitation, wind speed, and relative humidity) from meteorological stations of the China Meteorological Administration were collected, covering the period from 1981 to 2022; the high-resolution monthly irrigation water dataset HSWUD for China, publicly available on the Figshare platform, covering the period from 1965 to 2022, with a spatial resolution of 0.1°, was used to construct a pre-defined database. The pre-defined database was constructed using the ArcGIS platform and the ArcPy library of Python software. Based on the boundaries of the Yellow River Basin, the "Extract by mask" tool was used to batch-mask and extract meteorological station and water usage data for the Yellow River Basin from 1981 to 2022. Specifically:

[0145] Step S111: Using the "Create Fishnet" tool in the ArcGIS platform, divide the Yellow River Basin vector file into multiple 0.1°×0.1° grid points;

[0146] Step S112: Based on the ArcPy library, use the "Kriging" interpolation method to unify the spatial resolution of each data variable;

[0147] Step S113: Combine the data of each variable image with the same spatial resolution into an image set, and extract the value of each pixel in the image set pixel by pixel;

[0148] Step S114: Based on ArcGIS software, using the Yellow River Basin vector file, use the "Extract by mask" tool to perform batch masking on the image set to extract the image set of each variable in the Yellow River Basin from 1981 to 2022.

[0149] Step S200: Determine meteorological drought events monthly.

[0150] Step S210, the calculation of potential evapotranspiration, uses the Python software pieto library based on the FAO-56 Penman–Monteith formula. Required inputs include meteorological elements such as air temperature, wind speed, relative humidity, and solar radiation. Specifically:

[0151] Using a climate index calculation library in Python software (such as climate_indices or an equivalent calculation module), calculate the daily potential evapotranspiration from the meteorological dataset constructed in step S100. :

[0152]

[0153] In the formula, For the year The Middle The potential evapotranspiration of the day; For the year The Middle Net radiation of the sky For soil heat flux, This is the hygrometer constant. For the year The Middle The average air temperature of the day, For the year The Middle Wind speed at 2m above the ground For the year The Middle The saturated water vapor pressure of the sky For the year The Middle The actual water vapor pressure of the sky denoted as the slope of the curve relating saturated water vapor pressure to temperature.

[0154] In the climate_indices module, a three-parameter log-symmetric generalized logistic distribution is used to cumulatively fit the difference between precipitation and potential evapotranspiration over one month, and then convert it into a standard normal distribution value to obtain the SPEI-1 sequence.

[0155] Step S220: Using binary logic, months in which SPEI-1 is less than −0.5 are identified as meteorological drought months.

[0156] Step S300: Identify abnormal irrigation water events monthly.

[0157] The Loess-based seasonal-trend decomposition (STL) method was used to perform batch STL processing on monthly irrigation water consumption data from January 1980 to December 2022, obtaining irrigation water consumption residuals. Specifically:

[0158] Step S310: In the Python software, call the STL class of the statsmodels module, select the additive model as the decomposition model, and set the robust option (robust=True) to reduce the impact of outliers on the trend and seasonal terms. An improved STL algorithm replaces the tricube weight function in the traditional STL algorithm with a Gaussian kernel weight function.

[0159] For each month In that month Centered on, for other months The applied Gaussian kernel weights for the trend and seasonal terms are as follows:

[0160]

[0161]

[0162]

[0163]

[0164] The physical meanings of each symbol are as described above and will not be repeated here.

[0165] The improved STL decomposition is performed on a grid-by-grid basis to obtain the trend, seasonal, and residual terms. The residual terms are then used as the irrigation water consumption residual sequence after removing long-term trends and seasonal fluctuations. The output is the residual with the same spatial resolution as the input grid.

[0166] The decomposition results of this embodiment can be found in [link / reference]. Figure 3 . Figure 4 The autocorrelation coefficients are calculated between the time series of total irrigation water consumption in the Yellow River Basin from January 1980 to December 2022 and the separated residual series of irrigation water consumption. Figure 4 This demonstrates that the improved STL decomposition in this embodiment can effectively remove the periodicity and seasonality of irrigation water consumption sequences.

[0167] Step S320: An anomaly detection method based on Long Short-Term Memory (LSTM) networks is used to learn the time-dependent features of the residual sequence and identify irrigation water anomalies by calculating the reconstruction error. Specifically:

[0168] Step S331: In Python software, call the PyTorch deep learning framework to build a single-layer LSTM model. The input is the time window data of the residuals, and the output is the predicted residual value of the next time step.

[0169] Step S332: The main hyperparameters are set as follows: hidden size 64, sequence length 12, training set ratio 80% (remaining 20% ​​for validation), batch size 32, optimizer Adam, initial learning rate 0.001, and epochs 20. The loss function is set to Huber.

[0170]

[0171] The physical meanings of each symbol are as described above and will not be repeated here.

[0172] Step S333: Reconstruct the residuals using the trained LSTM, calculate the reconstruction error (the absolute value of the difference between the original value and the predicted value) at each time step, and calculate the anomaly detection threshold:

[0173]

[0174] in, To reconstruct the mean of the error sequence, To reconstruct the standard deviation of the error sequence, The threshold coefficient is used in this embodiment. .

[0175] See Figure 5 To train a machine learning model based on LSTM for identifying abnormal irrigation water usage, the absolute average error variation is calculated. Figure 5 This indicates that the error gradually decreases during the training process.

[0176] Step S334: When the reconstruction error of the irrigation water residual for a certain month is greater than the threshold... When this occurs, the corresponding month will be identified as a month with abnormal irrigation water usage.

[0177] Step S400: Statistics are performed with the drought intensity threshold set to... Anomalies in irrigation water use and meteorological drought events occurred within a preset time window. The proportion within is defined as the trigger overlap rate, and the calculation formula is as follows:

[0178]

[0179]

[0180]

[0181] Step S500: Change the drought intensity threshold, repeat steps S200-S400, and identify areas where irrigation water anomalies are significantly affected by meteorological drought events based on event consistency analysis, defining these areas as affected areas; specifically:

[0182] Step S510: Calculate the threshold values ​​for different drought intensities. Trigger overlap rate , For each drought intensity threshold With a time window of 0 months, the overlap rate of drought events and abnormal irrigation water events at each drought intensity threshold was calculated. This refers to the proportion of drought events and irrigation water anomalies occurring simultaneously in the same month to the total number of drought events under that drought intensity threshold.

[0183] Step S520: In the Python software, call the OLS method of the statsmodels module to apply drought intensity thresholds respectively. Overlap rate with meteorological drought events and abnormal irrigation water events Fit a linear model to the relationship between them;

[0184] In a Python environment, the pwlf (Piecewise Linear Fit) library is called to apply different values ​​to drought intensity thresholds. Overlap rate with meteorological drought events and abnormal irrigation water events Perform piecewise linear fitting;

[0185] Calculate the Akaike Information Criterion (AIC) for the linear model and the piecewise linear model respectively:

[0186]

[0187] in, The number of model parameters. This represents the maximum likelihood estimate of the model. A difference in AIC greater than 2 is used as the criterion for the piecewise linear model to significantly outperform the linear model.

[0188] Cells with better performance in the piecewise linear model are identified as "affected regions," and the spatial information of these affected regions is written to a netCDF file using the xarray library in Python. See also Figure 6 This is a regional distribution map showing the impact of meteorological drought events on the abnormal irrigation water events identified in this embodiment.

[0189] Step S600: Using time overlay analysis, the residual sequence of irrigation water consumption in the affected area is divided into a meteorological drought-affected subsequence and a normal subsequence. The difference between the two subsequences is analyzed to quantitatively identify the magnitude of the impact of the meteorological drought event on irrigation water anomalies in different regions. Specifically, the netCDF file of the affected area is read using the xarray library in Python software, and based on the temporal overlap between the drought event and the irrigation water anomaly, the residual is divided into a drought-affected subsequence and a normal subsequence, and the median difference between the two subsequences is calculated.

[0190] Step S700: Spatial visualization of the quantitative detection results of the impact of meteorological drought on abnormal irrigation water use obtained in step S600. Specifically, the `imshow` function of the matplotlib library in Python is used for spatial visualization. The visualization results in this embodiment can be found in [reference needed]. Figure 7 The figure shows the impact of meteorological drought on irrigation water consumption in the Yellow River Basin from 1980 to 2020.

[0191] It should be noted that the quantitative detection method for the impact of meteorological drought on abnormal irrigation water provided in the first aspect of the present invention is also applicable to the Northeast China region, which has developed irrigation agriculture, including parts of Jilin Province, Liaoning Province, Heilongjiang Province, and Inner Mongolia Autonomous Region.

[0192] A second aspect of the present invention provides a quantitative detection device for the impact of meteorological drought on abnormal irrigation water use, comprising:

[0193] The first module is configured to obtain monthly irrigation water consumption data and daily meteorological variable data with the same spatial resolution within the research time range from a preset database, and construct irrigation water consumption series and meteorological datasets respectively.

[0194] The second module is configured to determine meteorological drought events monthly based on the meteorological dataset and a set drought intensity threshold.

[0195] The third module is configured to extract residuals from the irrigation water consumption sequence based on the time series decomposition method to obtain an irrigation water consumption residual sequence, and to perform anomaly detection on irrigation water consumption based on the irrigation water consumption residual sequence to identify abnormal irrigation water consumption events on a monthly basis.

[0196] The fourth module is configured to statistically analyze the proportion of abnormal irrigation water events and meteorological drought events occurring within a preset time window under a set drought intensity threshold, and define it as the trigger overlap rate.

[0197] The fifth module is configured to change the drought intensity threshold, and call the second, third and fourth modules, and identify areas where irrigation water anomalies are significantly affected by meteorological drought events based on the event consistency analysis method, defining them as affected areas;

[0198] The sixth module is configured to use time overlay analysis to divide the residual sequence of irrigation water consumption in the affected area into a meteorological drought-affected subsequence and a normal subsequence, analyze the differences between the two subsequences, and thus quantitatively identify the magnitude of the impact of meteorological drought events on irrigation water anomalies in different areas.

[0199] A quantitative detection device for the impact of meteorological drought on abnormal irrigation water use, provided in a second aspect embodiment of the present invention, may further include:

[0200] The seventh module is configured to spatially visualize the quantitative detection results of the impact of meteorological drought on abnormal irrigation water obtained from the sixth module.

[0201] It should be noted that the foregoing explanation of the embodiments of the quantitative detection method also applies to the quantitative detection device of this embodiment, and will not be repeated here.

[0202] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0203] Although embodiments of this disclosure have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this disclosure, the scope of which is defined by the claims and their equivalents.

Claims

1. A quantitative detection method for the impact of meteorological drought on abnormal irrigation water use, characterized in that, include: Step S100: Obtain monthly irrigation water consumption data and daily meteorological variable data with the same spatial resolution from the preset database, and construct irrigation water consumption series and meteorological datasets respectively. Step S200: Determine meteorological drought events month by month based on the meteorological dataset and the set drought intensity threshold; Step S300: Extract residuals from the irrigation water consumption sequence based on the time series decomposition method to obtain the irrigation water consumption residual sequence; perform anomaly detection on irrigation water consumption based on the irrigation water consumption residual sequence; and identify abnormal irrigation water consumption events month by month. Step S400: Calculate the proportion of abnormal irrigation water events and meteorological drought events occurring within a preset time window under the set drought intensity threshold, and define it as the trigger overlap rate; Step S500: Change the drought intensity threshold, repeat steps S200 to S400, and identify areas where irrigation water abnormal events are significantly affected by meteorological drought events based on the event consistency analysis method, and define them as affected areas. Step S600: Using time overlay analysis, the residual sequence of irrigation water consumption in the affected area is divided into a meteorological drought-affected subsequence and a normal subsequence. The difference between the two subsequences is analyzed to quantitatively identify the magnitude of the impact of meteorological drought events on irrigation water anomalies in different areas.

2. The quantitative detection method according to claim 1, characterized in that, Step S200 includes: Calculate the daily standardized precipitation evapotranspiration index SPEI-1 based on the meteorological dataset; If the calculated SPEI-1 for the last day of a month is lower than the baseline drought intensity threshold, a meteorological drought event is determined to have occurred in that month, and the month is recorded as a meteorological drought month; otherwise, no meteorological drought event is determined to have occurred in that month.

3. The quantitative detection method according to claim 1, characterized in that, When extracting residuals from the irrigation water consumption sequence using the time series decomposition method, an improved STL time series decomposition method is employed. This improved STL time series decomposition method uses the following model: in, For months Irrigation water consumption; For the first Decomposed in the next iteration Trend items in; For the first Decomposed in the next iteration Seasonal items in; For the first Decomposed in the next iteration The residuals in; and These are the smoothing weights used for the trend term and the smoothing weights used for the seasonal term, respectively, and their calculation formulas are as follows: Among them, months For irrigation water consumption series excluding months In any other month besides This represents the total number of months contained in the irrigation water consumption sequence; and By month Centered on, for other months Apply Gaussian kernel weights to the trend and seasonal terms; and Each month Trend-adaptive bandwidth and seasonal-adaptive bandwidth; The baseline bandwidth; For the current month The median absolute deviation of the residuals within the central window; This represents the absolute deviation of the median of the irrigation water consumption series. and These are adjustment parameters for the trend term and the seasonal term, respectively, used to control the sensitivity of the adaptive bandwidth to the degree of local variation.

4. The quantitative detection method according to claim 1, characterized in that, The method of detecting anomalies in irrigation water volume based on the residual sequence of irrigation water volume, and identifying abnormal irrigation water events on a monthly basis, includes: A long short-term memory neural network is used to model the residual sequence of irrigation water consumption, and the reconstruction error is calculated; an adaptive threshold is set. The adaptive threshold The reconstruction error distribution of the irrigation water residual sequence modeling based on the long short-term memory neural network is determined; If the reconstruction error of a certain month is higher than the adaptive threshold If an abnormal irrigation water event occurs in that month, it is determined that an abnormal irrigation water event has occurred and that month is recorded as an abnormal irrigation water month; otherwise, it is determined that no abnormal irrigation water event has occurred in that month.

5. The quantitative detection method according to claim 4, characterized in that, The Huber loss function was used when training the Long Short-Term Memory neural network. in, For the monthly residual sequence of irrigation water consumption The reconstruction error, i.e., the extracted month The difference between the residual of irrigation water consumption and the predicted value of the long short-term memory neural network; For transition parameters, , This represents the standard deviation of the reconstruction error for all months in the irrigation water consumption residual sequence; The adaptive threshold Set it according to the following formula: in, is the mean of the reconstruction error for all months in the irrigation water consumption residual sequence; c is the threshold coefficient.

6. The quantitative detection method according to claim 1, characterized in that, In step S400, the trigger overlap rate is set to... : In the formula, This represents the total number of meteorological drought events statistically analyzed for the aforementioned meteorological dataset. This represents the total number of abnormal irrigation water events statistically analyzed for the irrigation water consumption sequence. for The Middle The timing of meteorological drought events for The Middle The time of occurrence of the abnormal irrigation water event; The set time window length, which is the length of months that allows for a lag between abnormal irrigation water events and drought events; It is an indicator function; It is the Heaviside step function.

7. The quantitative detection method according to claim 1, characterized in that, Step S500 includes: Step S510: Change the drought intensity threshold And repeat steps S200 to S400 to obtain different drought intensity thresholds. Trigger overlap rate ; Step S520: For the drought intensity threshold Overlap rate with trigger The relationship between the two models was analyzed, and both linear and piecewise linear models were fitted. The merits of the linear and piecewise linear models were compared. For a given region, if the piecewise linear fit was superior to the linear fit, it indicated a higher drought intensity threshold. Overlap rate with trigger When there is an inflection point in the relationship between them, that area is identified as the affected area.

8. The quantitative detection method according to claim 7, characterized in that, For drought intensity threshold Overlap rate with trigger The relationship between them can be fitted by the following linear model: For drought intensity threshold Overlap rate with trigger The relationship between them, and the fitted piecewise linear model is: in, For the inflection point threshold of the piecewise linear model, These are the regression coefficients, , These are the fitting residuals; The linear model and the piecewise linear model are compared using information criteria or statistical testing methods.

9. The quantitative detection method according to any one of claims 1 to 8, characterized in that, Also includes: Step S700: Spatial visualization of the quantitative detection results of the impact of meteorological drought on abnormal irrigation water obtained in step S600.

10. A quantitative detection device for the impact of meteorological drought on abnormal irrigation water use, characterized in that, include: The first module is configured to obtain monthly irrigation water consumption data and daily meteorological variable data with the same spatial resolution within the research time range from a preset database, and construct irrigation water consumption series and meteorological datasets respectively. The second module is configured to determine meteorological drought events monthly based on the meteorological dataset and a set drought intensity threshold. The third module is configured to extract residuals from the irrigation water consumption sequence based on the time series decomposition method to obtain an irrigation water consumption residual sequence, and to perform anomaly detection on irrigation water consumption based on the irrigation water consumption residual sequence to identify abnormal irrigation water consumption events on a monthly basis. The fourth module is configured to statistically analyze the proportion of abnormal irrigation water events and meteorological drought events occurring within a preset time window under a set drought intensity threshold, and define it as the trigger overlap rate. The fifth module is configured to change the drought intensity threshold, and call the second, third and fourth modules, and identify areas where irrigation water anomalies are significantly affected by meteorological drought events based on the event consistency analysis method, defining them as affected areas; The sixth module is configured to use time overlay analysis to divide the residual sequence of irrigation water consumption in the affected area into a meteorological drought-affected subsequence and a normal subsequence, analyze the differences between the two subsequences, and thus quantitatively identify the magnitude of the impact of meteorological drought events on irrigation water anomalies in different areas.

Images