A drought-flood abrupt transition prediction method coupling soil moisture regime and precipitation model

Abstract

The application provides a drought-flood sudden change prediction method fusing a soil moisture content and precipitation coupling model, and relates to the technical field of drought-flood sudden change. The method comprises the following steps: collecting layered soil moisture content, high-frequency precipitation and auxiliary data, and preprocessing; constructing soil moisture content-precipitation coupling coordination initial values and correcting; using a standardized soil moisture content-precipitation coupling coordination index sequence, traversing a drought-flood window, and determining an optimal window; performing nonlinear correction, determining a sudden change event according to the corrected threshold value, and outputting drought-flood sudden change prediction results; calculating a drought-flood sudden change duration prediction value T int and a soil moisture content-precipitation mutation index SPI ef , constructing a risk matrix, combining a disaster-bearing body vulnerability to output a warning level and a response suggestion, and outputting warning information. The application determines an optimal window and performs optimization to predict drought-flood sudden change, thereby guaranteeing the scientificity and accuracy of prediction and strengthening practical application value.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of drought-flood rapid transition, and in particular to a drought-flood rapid transition prediction method fusing a soil moisture content and precipitation coupling model. BACKGROUND

[0002] Drought-flood rapid transition event is an extreme combination of two water cycle extreme processes, i.e. drought and flood. The stability of hydrological system is increasingly reduced due to the combined effects of climate change and human activities, leading to the increasing frequency and range of drought-flood rapid transition events in recent years, which has caused serious harm to production, life and ecological environment in many regions.

[0003] Traditional drought-flood prediction methods have a single data source, mostly relying on a few data sources such as ground rain gauges, without integrating weather radar estimated precipitation and satellite precipitation products, which are difficult to fully capture the spatial and temporal distribution and suddenness of precipitation. Moreover, simple processing methods such as mean value filling are often used for data missing, without considering the spatial and temporal correlation of hydrological data and soil texture differences, resulting in uneven data quality and affecting the subsequent prediction accuracy. In terms of coupling relationship construction, linear models are mostly used to describe the correlation between soil moisture content and precipitation, which cannot adapt to the nonlinear dynamic changes between the two, i.e. slow decline of soil moisture content in drought period and rapid rise of soil moisture content in flood period, resulting in a large deviation in the description of coupling coordination degree. In the determination of rapid transition events, fixed windows are mostly used, without the process of selecting the optimal window according to data characteristics, which is easy to misjudge short-term data fluctuations as drought-flood rapid transition. At the same time, traditional methods do not combine with disaster-bearing body vulnerability to carry out risk classification, nor do they provide targeted disaster prevention response suggestions, but only output simple results of whether the rapid transition occurs, which cannot effectively support actual disaster prevention and mitigation decisions, and the overall practical value is limited. SUMMARY

[0004] The present application provides a drought-flood rapid transition prediction method fusing a soil moisture content and precipitation coupling model, which solves the defect that the prior art lacks the process of selecting the optimal window according to data characteristics, which is easy to misjudge short-term data fluctuations as drought-flood rapid transition.

[0005] In one aspect, the present application provides a drought-flood rapid transition prediction method fusing a soil moisture content and precipitation coupling model, comprising:

[0006] S1: constructing a three-dimensional hydrological map of a target region, labeling soil texture distribution and historical drought-flood rapid transition event positions, dividing the region into P sub-regions to be predicted and recording the center coordinates of each sub-region. Collecting layered soil moisture content, high-frequency precipitation and auxiliary data, and outputting a sub-region original drought-flood data set.

[0007] S2: preprocessing the sub-region original drought-flood data set to obtain a standardized data set.

[0008] S3: Calculate the soil moisture content-rainfall coupling coordination initial value according to the standardized data set, and correct it to obtain a standardized soil moisture content-rainfall coupling coordination index sequence.

[0009] S4: Use the standardized soil moisture content-rainfall coupling coordination index sequence to traverse the drought and flood window to determine the optimal window. Nonlinear correction is made in combination with the array rainfall, vegetation and soil texture, the threshold value after correction is determined to determine the sudden change event, and the drought and flood sudden change prediction result is output.

[0010] S5: According to the drought and flood sudden change prediction result and the soil texture coefficient, a drought and flood sudden change duration prediction value T is calculated. int and the soil moisture content-rainfall mutation index SPI ef , a risk matrix is constructed, the warning level and response suggestion are output in combination with the disaster-bearing body vulnerability, and the warning information is output.

[0011] According to the drought and flood sudden change prediction method of the soil moisture content-rainfall coupling model provided by the application, in step S2, the preprocessing includes:

[0012] S21: An isolated forest algorithm is used to identify abnormal values.

[0013] S22: Short missing data is filled by using the correlation interpolation method according to the soil texture difference, and long missing data is filled by using the LSTM-ATTention model.

[0014] S23: The data is converted to the [-1, 1] interval by the Z-score standardization formula.

[0015] According to the drought and flood sudden change prediction method of the soil moisture content-rainfall coupling model provided by the application, in step S3, the specific steps for obtaining the standardized soil moisture content-rainfall coupling coordination index sequence are:

[0016] S31: The coupling coordination initial value is calculated according to the standardized data set.

[0017] S32: In the initial value, the power function and the piecewise function are input, the least square method is used to quantify the soil moisture content-rainfall dynamic relationship, and the quantized initial value is obtained.

[0018] S33: The standardized soil moisture content-rainfall coupling coordination index sequence is obtained by using the Gamma distribution fitting and the inverse normal transformation quantization initial value.

[0019] According to the drought and flood sudden change prediction method of the soil moisture content-rainfall coupling model provided by the application, in step S31, the specific steps for constructing the soil moisture content-rainfall coupling coordination index are:

[0020] S311: The weight values of the soil moisture content index and the rainfall index are determined by the analytic hierarchy process.

[0021] S312: Construct a function expression of the coupling and coordination relationship between soil moisture content and precipitation.

[0022] S313: Input the standardized data set into the function expression to obtain the initial value of the coupling and coordination.

[0023] According to the drought and flood sudden change prediction method of the coupling model of soil moisture content and precipitation provided by the application, in step S33, the specific steps of using Gamma distribution fitting and inverse normal transformation to quantify the initial value are as follows:

[0024] S331: Classify and arrange the quantified initial value to distinguish the soil moisture content quantified initial value and the precipitation quantified initial value.

[0025] S332: Perform Gamma distribution fitting on the soil moisture content quantified initial value and the precipitation quantified initial value by using the probability density function of Gamma distribution to obtain a Gamma distribution model.

[0026] S333: Perform inverse normal transformation on the soil moisture content quantified initial value and the precipitation quantified initial value according to the Gamma distribution model to obtain normalized soil moisture content data and normalized precipitation data.

[0027] S334: Construct a soil moisture content-precipitation coupling coordination index calculation model, input the normalized soil moisture content data and the normalized precipitation data into the index calculation model, and obtain a coupling coordination index.

[0028] S335: Standardize the coupling coordination index to obtain a standardized soil moisture content-precipitation coupling coordination index sequence.

[0029] According to the drought and flood sudden change prediction method of the coupling model of soil moisture content and precipitation provided by the application, in step S4, the specific steps of outputting the determination result are as follows:

[0030] S41: According to the standardized soil moisture content-precipitation coupling coordination index CCIstd sequence, associate the sub-regions with the time nodes to form a sub-region CCIstd value three-dimensional data table.

[0031] S42: According to the three-dimensional data table, set a candidate set of drought and flood window length ranges, extract the window units corresponding to each N value and calculate the change index, determine the optimal window length, output the optimal window length of each sub-region and the CCIstd change characteristic parameters.

[0032] S43: Collect the array precipitation, vegetation, and soil texture data of the optimal window period of each sub-region, associate the CCIstd value in the window to construct a nonlinear correction model, calculate the corrected index CCIstd corr , and output the corrected index sequence of each sub-region.

[0033] S44: Based on the locations of historical drought-flood transition events, the drought threshold T is determined using the percentile method based on the corrected exponential sequence of each sub-region. dry and flood threshold T wet The output is a corrected threshold applicable to all sub-regions.

[0034] S45: Compare the optimal CCIstd within each sub-region's window. corr The sequence trend and the corrected threshold are combined with the optimal window change feature parameters to determine the type of abrupt change, and the prediction results of abrupt changes from drought to flood are output.

[0035] According to the drought-flood transition prediction method based on the integrated soil moisture and precipitation model provided by the present invention, the specific steps in step S42 for outputting the optimal window length and CCIstd change characteristic parameters of each sub-region are as follows:

[0036] S421: Based on the three-dimensional data table, set the range of drought and flood window lengths, clarify the traversal range of each sub-region, and output the candidate set of window lengths for each sub-region.

[0037] S422: For each N value in the candidate set of window length, extract the CCIstd values ​​of the corresponding sub-region for N consecutive time nodes from the 3D data table as window units, calculate the range and linear fitting slope change index, and output the window unit change index dataset.

[0038] S423: Determine the optimal window length based on the window unit change index dataset, and output the optimal window length for each sub-region and the corresponding CCIstd change characteristic parameters.

[0039] According to the drought-flood transition prediction method based on the integrated soil moisture and precipitation model provided by the present invention, the specific steps for determining the optimal window length in step S423 are as follows:

[0040] Based on the CCIstd sequence within the window, the drought-flood transition coupling strength (CSI) is calculated using the following formula:

[0041]

[0042] In the formula, The range of CCIstd within the window. This represents the absolute value of the slope of CCIstd within the window.

[0043] For different candidate window lengths, the CSI is calculated for each window length, and then compared with the historical list of sudden drought-flood events to calculate the matching rate.

[0044] The optimal window length is the one with the highest matching rate and the ability to distinguish non-abrupt events.

[0045] According to the drought and flood sudden change prediction method of the coupling model of soil moisture content and precipitation provided by the application, in step S43, the specific steps of outputting the corrected index sequence of each sub-region are as follows:

[0046] The cumulative frequency R1 of the array precipitation in the calculation window, the maximum intensity R2 of single time, the average value V of the vegetation coverage index NDVI in the window, and the proportion S of the soil clay content of the sub-region are extracted, the data are associated with the CCIstd value in the optimal window of the sub-region, and a nonlinear correction model is constructed;

[0047] The optimal window length and the CCIstd change characteristic parameter of each sub-region are substituted into the nonlinear correction model to obtain the corrected normalized coupling coordination index CCIstd in the optimal window of each sub-region. corr The CCIstd sequence of each sub-region is outputted. corr The CCIstd sequence of each sub-region is outputted.

[0048] According to the drought and flood sudden change prediction method of the coupling model of soil moisture content and precipitation provided by the application, in step S5, the specific steps of outputting the early warning information are as follows:

[0049] S51: According to the drought and flood sudden change prediction result and the soil texture coefficient of the sub-region, the average duration of the same type of sudden change in history is associated, and the drought and flood sudden change duration prediction value T is formed according to the sub-region. int And the soil moisture content precipitation mutation index SPI ef The basic data set is calculated.

[0050] S52: According to the basic data set, T int And SPI ef The core index data set is outputted.

[0051] S53: A matrix is constructed with T int As the vertical axis and SPI ef As the horizontal axis, the data is mapped to a preliminary risk level, and the preliminary risk level data set is outputted.

[0052] S54: The vulnerability data of the disaster-bearing body is collected, the preliminary risk level is corrected according to the preliminary risk level, and the corrected risk level data set is outputted.

[0053] S55: According to the corrected risk level, the warning level is determined, the suggestions are combined according to the sudden change type, the information is integrated, and the structured drought and flood sudden change early warning information report is outputted.

[0054] The drought and flood sudden change prediction method of the coupling model of soil moisture content and precipitation provided by the application calculates the soil moisture content-precipitation coupling coordination initial value, traverses the drought and flood window, determines the optimal window and optimizes it, and predicts the drought and flood sudden change, which has the beneficial effects that:

[0055] By constructing the three-dimensional hydrological map of the target area, not only the distribution of soil texture and the location of the historical drought-flood sudden change events are clearly marked, but also the region is reasonably divided into multiple sub-regions to be predicted and the coordinates are recorded, so that the subsequent analysis can focus on smaller and relatively consistent units, improve the accuracy of prediction, avoid errors caused by general analysis of large areas, and help to understand the characteristics of drought-flood sudden change in different sub-regions. The collection of layered soil moisture, high-frequency precipitation and other auxiliary data such as air temperature, humidity and wind speed comprehensively considers many factors related to drought-flood sudden change, and multiple sources of data complement and verify each other, which can more comprehensively reflect the actual water and heat conditions and trends in the region, and lay a solid foundation for accurate prediction.

[0056] The weights of soil moisture and precipitation indicators are determined by the analytic hierarchy process, and combined with the constructed coupling coordination function expression, the relative importance and interaction of the two in the process of drought-flood sudden change can be scientifically measured, so that the constructed soil moisture-precipitation coupling coordination index is more consistent with the actual physical process, and provides strong support for accurate judgment of drought-flood sudden change. By fusing power function and piecewise function to quantify the dynamic relationship between soil moisture and precipitation, and through a series of operations such as Gamma distribution fitting and inverse normal transformation, the index is finely processed and corrected, so that the standardized soil moisture-precipitation coupling coordination index obtained finally can more accurately reflect the complex coupling coordination between soil moisture and precipitation, and enhance the characterization ability of drought-flood sudden change phenomenon.

[0057] By traversing the drought and flood window, according to the set drought and flood state change core index and the calculation of drought and flood conversion coupling strength CSI, and compared with the historical drought and flood sudden change event to determine the matching rate, the optimal window is selected, which can accurately capture the period most likely to occur drought and flood sudden change in time scale, avoid misjudgment or missed judgment caused by improper window selection, and improve the timeliness and accuracy of drought and flood sudden change prediction. Combined with factors such as array precipitation, vegetation, soil texture, etc. Nonlinear correction, and using the percentile method to determine the drought threshold and flood threshold, fully considering the effect of various actual factors on the determination of drought and flood sudden change, the determination of threshold is more in line with the actual situation of the region, and then the type of drought and flood sudden change in each sub region is more accurately judged, and the reliability of the prediction result is improved. Based on the drought and flood sudden change prediction result, soil texture coefficient, etc. Calculate the drought and flood sudden change duration prediction value Tint and the soil moisture content precipitation mutation index SPIef, construct the risk matrix, and comprehensively evaluate the drought and flood sudden change risk level from the time duration and CCIstd mutation degree, so that the risk assessment result is more comprehensive and objective, and can more accurately reflect the difference in drought and flood sudden change risk faced by different sub regions. Considering the vulnerability of the disaster bearing body, the risk level is corrected, and then the warning level is determined according to the corrected risk level and specific and targeted response suggestions are put forward according to the sudden change type, so that the warning information can not only reflect the situation of drought and flood sudden change itself, but also can be combined with the actual disaster bearing capacity of the region to provide scientific basis for relevant departments to formulate accurate and effective disaster prevention and mitigation measures.

[0058] The present application improves the reliability of data basis, reduces data noise and deviation through multi-source data fusion, targeted abnormal value identification and missing value filling, standardization processing, provides high-quality and high-consistency basic data for drought and flood sudden change prediction, avoids the influence of data problems on prediction accuracy, improves the accuracy of sudden change prediction, quantifies the coupling relationship between soil moisture content and precipitation by combining dynamic function and probability distribution, and then selects the optimal window and nonlinear correction of multiple factors, accurately captures the nonlinear and sudden characteristics of drought and flood sudden change, effectively reduces the problems of misjudgment of short-term fluctuations and missed judgment of rapid sudden change. Realize spatial accurate prediction, divide the sub region matched with hydrological characteristics, customize the coupling modeling, window screening and other parameters for different sub regions, break through the limitation of traditional global general prediction, output sub regional prediction results, and adapt to the differentiated disaster prevention needs of different regions. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0060] Fig. 1This is a flowchart illustrating the method for predicting rapid shifts between drought and flood provided in an embodiment of the present invention, which integrates a coupled model of soil moisture and precipitation.

[0061] Fig. 2 This is a flowchart illustrating the process of outputting the prediction results of sudden shifts between drought and flood in this invention. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0063] The following is combined Figs. 1-2 This invention describes a method for predicting rapid shifts between drought and flood based on a coupled model of soil moisture and precipitation.

[0064] like Figs. 1-2 As shown in the embodiment of the present invention, the method for predicting abrupt shifts between drought and flood by integrating a coupled model of soil moisture and precipitation includes:

[0065] S1: Construct a 3D hydrological map of the target area, marking soil texture distribution and the locations of historical drought-flood transition events. Divide the area into P sub-regions to be predicted and record the center coordinates of each sub-region. Collect stratified soil moisture, high-frequency precipitation, and auxiliary data, and output the original drought-flood datasets for each sub-region.

[0066] Define the administrative and geographical boundaries of the target area, collect 1:50,000 DEM data, soil type distribution maps, records of drought and flood transition events over the past 30 years, and hydrological station distribution data to generate a basic dataset. Use the DEM data to generate a 3D terrain model of the target area through ArcGIS or Surfer, overlay vector layers of hydrological elements such as watershed boundaries, river networks, and reservoirs, set visualization parameters such as elevation and slope, generate a 3D hydrological base map, and output a base map file containing spatial coordinates.

[0067] The soil type distribution map is vectorized, and the soil type is marked in the corresponding position according to its texture. The event occurrence time and the scope of impact are added, and the marked 3D hydrological map is output.

[0068] Based on the soil texture and watershed unit consistency of the 3D hydrological map, the region is divided into N sub-regions of equal area or equal hydrological units using GIS tools. The latitude and longitude coordinates of the geometric center of each sub-region are calculated using the tools, and the sub-region division vector map and reference table are output.

[0069] Collecting the soil moisture data of each layer, high-frequency precipitation, and auxiliary data. According to the sub-regional number, the data is summarized and arranged into the sub-regional original drought and flood data set.

[0070] Soil moisture data: Using hierarchical Internet of Things sensors to capture the slow decline of soil moisture during drought and the rapid rise of soil moisture during flood.

[0071] Precipitation data: Integrating ground rain gauges, weather radar estimated precipitation, and satellite precipitation products, generating 1km x 1km grid hourly precipitation intensity data through Kriging interpolation, adapting to the needs of capturing precipitation bursts in short periods of transition.

[0072] Coupling auxiliary data: including 2m air temperature, relative humidity, wind speed, soil texture, soil compaction, and crop root depth.

[0073] S2: Preprocessing the sub-regional original drought and flood data set to obtain the standardized data set.

[0074] S21: Using the Isolation Forest algorithm to identify outliers, and using a dedicated preprocessing strategy for the temporal and spatial correlation of hydrological data to avoid losing key transition signals.

[0075] S22: Short missing data is filled using correlation interpolation method according to soil texture differences, and long missing data is filled using LSTM-ATTention model. Short time missing data is filled using precipitation-moisture correlation interpolation, such as calculating the moisture change rate according to the same period precipitation intensity and soil texture. Long time missing data is filled using LSTM time series model, inputting previous moisture, precipitation, and evapotranspiration data to generate filling values.

[0076] S23: Convert the data to the [-1, 1] interval by Z-score standardization formula.

[0077] Min-Max standardization is used for all data to eliminate dimension differences, and the formula is:

[0078]

[0079] Where x is the original data, xmin and xmax are the historical extreme values of the data, and xnorm is the standardized data.

[0080] S3: According to the standardized data set, construct the soil moisture-precipitation coupling coordination index, and modify it to obtain the standardized soil moisture-precipitation coupling coordination index sequence.

[0081] S31: According to the standardized data set, construct the soil moisture-precipitation coupling coordination index.

[0082] S311: Determine the weight values of soil moisture and precipitation indicators through AHP.

[0083] S312: Construct a function expression of the coupling coordination relationship between soil moisture content and precipitation.

[0084] S313: Input the standardized data set into the function expression to obtain the initial value of the coupling coordination degree.

[0085] S32: Quantify the dynamic relationship between soil moisture content and precipitation using power function and piecewise function.

[0086] S33: Obtain the standardized soil moisture content-precipitation coupling coordination index using Gamma distribution fitting and inverse normal transformation.

[0087] S331: Classify and organize the quantified initial values to distinguish soil moisture content quantified initial values (denoted as X 初 ) directly related to soil moisture content and precipitation quantified initial values (denoted as Y 初 ) directly related to precipitation, ensuring that the two types of initial values correspond to observation time or observation site one-to-one, forming a paired data set {(X 初i , Y 初 )} (i=1, 2, ⋯, n, n is the sample size).

[0088] S332: Based on the paired data set {(X 初i , Y 初i )}, the soil moisture content quantified initial value X 初i and the precipitation quantified initial value Y 初i are fitted with Gamma distribution respectively. The probability density function of Gamma distribution is expressed as:

[0089]

[0090] In the formula, t≥0, α is the shape parameter, β is the scale parameter, and Γ(α) is the gamma function. The likelihood function for the soil moisture content quantified initial value is constructed, and the formula is:

[0091]

[0092] After taking the logarithm, the optimal parameters α X and β X are determined by the maximum likelihood estimation method, and the Gamma distribution model f X (t) of X 初i is obtained. Similarly, the optimal parameters α Y and β Y are calculated for the precipitation quantified initial value, and the Gamma distribution model f Y (t) of Y 初i is obtained, completing the Gamma distribution fitting of the two types of quantified initial values.

[0093] S333: Based on the fitted Gamma distribution model f X (t) and f Y (t), the initial values ​​of soil moisture quantification and precipitation quantification are respectively subjected to inverse normal transformation using the Box-Cox transformation formula:

[0094] When λ≠0 .

[0095] When λ=0 .

[0096] For Y 初i Combined with its Gamma distribution model f X Based on the characteristics of (t), the optimal λ for the normality of the transformed data was determined using a trial-and-error method combined with the Shapiro-Wilk normality test. X Normalized soil moisture data were obtained. For Y 初i Similarly, determine λ Y We obtain normally normalized precipitation data and achieve inverse normal transformation of the two types of initial quantification values.

[0097] S334: Using the normalized soil moisture data Si and normalized precipitation data Pi obtained by inverse normal transformation, a soil moisture-precipitation coupling coordination index calculation model is constructed. The common coupling coordination index formula is adopted. Each pair of Si and Pi is substituted into the formula to calculate the initial value of the coupling coordination index for the corresponding sample. This initial value is directly generated based on the quantized initial value after fitting and transformation.

[0098] S335: The initial value of the calculated coupling coordination index is standardized by using a normalization method to eliminate the influence of dimensions, and the standard for each sample is calculated to finally obtain the standardized soil moisture-precipitation coupling coordination index.

[0099] S4: Using the standardized soil moisture-precipitation coupling coordination index sequence, the optimal window is determined by traversing drought and flood windows. Nonlinear corrections are applied based on intermittent precipitation, vegetation, and soil texture. Abrupt change events are then identified according to the corrected threshold, and the results are output. The window consists of N consecutive time nodes, treating the soil moisture-precipitation coupling coordination index (CCIstd) within this time period as a whole, and analyzing the changing trends of drought and flood conditions within this time period.

[0100] S41: Organize the standardized soil moisture-precipitation coupling coordination index sequence (CCIstd sequence), clarify the time nodes corresponding to the sequence and the CCIstd value of each time node, and associate it with the N sub-regions to be predicted in S1 to form a three-dimensional data table of sub-region-time node-CCIstd value.

[0101] S42: According to the three-dimensional data table, set the candidate set of the length range of the drought and flood window, extract the window unit corresponding to each N value and calculate the change index, determine the optimal window length, and output the optimal window length of each sub-region and the CCIstd change characteristic parameter.

[0102] S421: Based on the above-mentioned three-dimensional data table, set the length range of the drought and flood window, and the selectable value of N is 3-15 consecutive time nodes, which needs to be determined according to the data acquisition frequency and the typical time scale of the drought and flood rapid transition, and the window length does not exceed the total time node number of the CCIstd sequence. The window traversal range of each sub-region to be predicted is determined to ensure that all possible continuous time periods in the sequence are covered in the traversal process, and the window length candidate set of each sub-region is output.

[0103] S422: For each N value in the window length candidate set of each sub-region, extract the CCIstd values of all consecutive N time nodes of the sub-region from the three-dimensional data table as a window unit, and calculate the drought and flood state change core index in each window unit. The drought and flood state change core index includes the range of CCIstd in the window, the difference between the maximum value and the minimum value, which reflects the drought and flood state change amplitude, the linear fitting slope reflects the change trend direction of CCIstd, the positive slope is the development of the drought state, the negative slope is the development of the drought state, and the number of mutation points is detected by sliding t test to detect whether there is a significant jump in CCIstd, which reflects whether there is a rapid state transition. Output the window unit change index data set corresponding to different N values of each sub-region.

[0104] S423: Determine the optimal window length according to the window unit change index data set, and output the optimal window length of each sub-region and the corresponding CCIstd change characteristic parameter.

[0105] The specific steps for determining the optimal window are as follows:

[0106] Based on the CCIstd sequence in the window, calculate the drought and flood transition coupling strength CSI, and the calculation formula is represented as:

[0107]

[0108] In the formula, is the maximum value of the normalized drought and flood index in the window, is the minimum value of the normalized drought and flood index in the window, is the range of CCIstd in the window, which reflects the drought and flood transition amplitude, and the larger the difference, the greater the span from drought to flood / flood to drought, is the absolute value of the change slope of CCIstd in the window, which reflects the speed of drought and flood transition, and the larger the slope, the faster the transition. The larger the CSI value, the more significant the drought and flood rapid transition characteristics in the window.

[0109] For different candidate window lengths, calculate the CSI under each window length respectively, and compare it with the historical drought and flood sudden change event list to count the matching rate. The matching rate is the proportion of the coincidence of the time of the peak value of the CSI in the window and the time of the historical sudden change event.

[0110] Select the window length with the highest matching rate and capable of distinguishing non-sudden change events as the optimal window. In specific embodiments, if the window length is 3 days with a matching rate of 60%, it means that a large number of non-sudden change short-term fluctuations are included. If the window length is 7 days with a matching rate of 88%, it means that the CSI peak value is highly coincident with the historical sudden change event and can exclude short-term fluctuations. If the window length is 10 days with a matching rate of 75%, it means that part of the rapid sudden change events are diluted by the window, resulting in missed judgment. At this time, 7 days is the optimal window, which can accurately capture the historical sudden change event and avoid misjudgment and missed judgment.

[0111] S43: Collect the data of the array precipitation, vegetation, and soil texture of each sub-region in the optimal window period, correlate the CCIstd value in the window to construct a non-linear correction model, calculate the corrected index, and output the CCIstd of each sub-region corr sequence.

[0112] Collect the auxiliary data of each sub-region in the time period corresponding to the optimal window, i.e., the array precipitation data, calculate the cumulative frequency and single maximum intensity of the array precipitation in the window, denoted as R1 and R2. Extract the average value of the vegetation coverage index NDVI in the window, denoted as vegetation data V, and obtain the proportion of soil clay content in the sub-region, denoted as soil texture data S. Correlate these data with the CCIstd value in the optimal window of the sub-region to construct a non-linear correction model, which can be expressed as:

[0113]

[0114] In the formula, a, b, c, d, and e are correction coefficients fitted by historical drought and flood sudden change sample data. For example, the larger the array precipitation intensity, the higher the values of a and b, which can enhance the response of CCIstd to flood state. The higher the clay content, the higher the value of d, which can increase the sensitivity of CCIstd to changes in soil moisture.

[0115] Substitute the data to calculate the corrected normalized coupling coordination index CCIstd in the optimal window of each sub-region corr sequence, and output the CCIstd corr sequence of each sub-region.

[0116] S44: According to the location of the historical drought and flood sudden change event, use the above CCIstd corr value to determine the drought threshold T dry and the flood threshold T wet, output the revised threshold value applicable to all sub-regions.

[0117] Determine the revised threshold value using the percentile method: take the 90th percentile of CCIstd corr when historical drought-to-flood events occur as the flood threshold T wet , and take the 10th percentile of CCIstd corr when historical flood-to-drought events occur as the drought threshold T dry , output T dry and T wet applicable to all sub-regions.

[0118] S45: Compare the trend of CCIstd corr sequence in the optimal window of each sub-region with the revised threshold value, and determine the abrupt transition type based on the optimal window change characteristic parameter, output the drought and flood abrupt transition prediction results containing sub-region number, time period and determination basis.

[0119] For each sub-region, compare the trend of CCIstd corr sequence in the optimal window with the revised threshold value. If the sequence rapidly rises from below T dry to above T wet , and the absolute value of the slope is greater than or equal to the slope reference value in the optimal window change characteristic parameter, it is predicted that the sub-region will experience a drought-to-flood abrupt transition in the time period corresponding to the optimal window. If the sequence rapidly falls from above T wet to below T dry , and the absolute value of the slope is greater than or equal to the slope reference value, it is predicted that a flood-to-drought abrupt transition will occur. If the sequence fluctuates between T dry and T wet or the change amplitude does not reach the reference value, it is predicted that there is no risk of drought and flood abrupt transition, and finally the drought and flood abrupt transition prediction results containing the sub-region number, optimal window period, abrupt transition type and determination basis are output.

[0120] S5: Calculate the drought and flood abrupt transition duration prediction value T int and the soil moisture condition precipitation mutation index SPI ef based on the drought and flood abrupt transition prediction results and soil texture coefficient, construct a risk matrix, and output the warning information combined with the warning level and response suggestion based on the vulnerability of the disaster-bearing body.

[0121] S51: The drought and flood abrupt transition prediction results, risk level coefficient in the determination basis, and the soil texture coefficient of each sub-region are extracted, and the duration records of the historical drought and flood abrupt transition events in the target region in the past 10 years are associated. These data are aligned one by one according to the sub-region number to form the Tint and SPIef calculation basis data set, which must ensure that there is no missing information for each sub-region.

[0122] S52: Calculate Tint (duration prediction of drought-flood rapid transition) and SPIef (soil moisture precipitation mutation index) based on the basic data set, and the calculation formula of Tint is:

[0123] Tint=historical average duration of the same type of rapid transition × soil texture coefficient.

[0124] The calculation formula of SPIef is:

[0125]

[0126] In the formula, k is the absolute value of the linear fitting slope, is the difference between the maximum and minimum values of CCIstd in the window, and ω is the weight of the rapid transition type.

[0127] In a specific embodiment, the ΔCCIstd of the optimal window of a certain sub-region is 0.7, |k|=0.2 / day, and the drought-flood transition is predicted, then SPIef=0.7×0.2×1.3=0.182. The Tint calculation results and SPIef calculation results of each sub-region are integrated according to the number, and the core index data set is output.

[0128] S53: Construct a matrix with T int as the vertical axis and SPI ef as the horizontal axis, map the data to the preliminary risk level, and output the preliminary risk level data set.

[0129] The vertical axis Tint is divided into three intervals according to the disaster duration impact threshold, short-term risk (Tint≤3 days, short duration, limited impact), medium-term risk (3 days

[0130] The horizontal axis SPIef is divided into three intervals according to the CCIstd mutation intensity threshold, low intensity (SPIef≤0.05, CCIstd slowly fluctuates, no obvious disaster risk in rapid transition), medium intensity (0.05<SPIef≤0.15, CCIstd fluctuates rapidly, rapid transition may cause mild hydrological anomalies such as local waterlogging or mild soil drought), and high intensity (SPIef>0.15, CCIstd rises and falls sharply, rapid transition is easy to cause serious disasters such as large-scale waterlogging or large-area crop withering), each interval corresponds to an intensity risk level. Map the Tint and SPIef values of each sub-region to the matrix cells one by one, determine the preliminary risk level according to the combination of the time risk level and the intensity risk level, and output the preliminary risk level data set.

[0131] S54: Collect the vulnerability data of disaster-affected bodies in each sub-region, which includes three types of core data: first, population density data, obtained through target regional statistical yearbook or remote sensing inversion, divided into high-density, medium-density and low-density areas according to the number of people per square kilometer. Second, main crop type data, obtained through agricultural department planting planning map, divided into flood-sensitive crops, drought-sensitive crops and moderately sensitive crops. Third, infrastructure data, obtained through agricultural department data, focusing on recording whether there are low-lying areas prone to flooding and irrigation facility coverage. Based on these data, the vulnerability level division standard is set: high vulnerability area needs to meet high population density + flood / drought sensitive crops + low-lying area prone to flooding / irrigation coverage < 30%. Medium vulnerability area meets medium population density + moderately sensitive crops + no low-lying area prone to flooding / irrigation coverage 30%-80%. Low vulnerability area meets low population density + moderately sensitive crops + no low-lying area prone to flooding / irrigation coverage > 80%. At the same time, the risk level correction rule is formulated: high vulnerability area will increase the preliminary risk level of the third step by 1 level. Low vulnerability area will reduce the preliminary risk level by 1 level. Medium vulnerability area keeps the preliminary risk level unchanged. Match the vulnerability level of each sub-region with the preliminary risk level, apply the correction rule, output the corrected risk level data set, and ensure that the risk level is more in line with the actual disaster-affected capacity of the region.

[0132] S55: According to the corrected risk level, determine the warning level, combine the sudden change type to make suggestions, integrate information to output structured drought and flood sudden change warning information report. If it is drought to flood, the suggestions focus on preventing internal flooding, protecting drainage, dredging the drainage pipe network of the main and secondary roads in the jurisdiction, relocating residents in low-lying areas to temporary resettlement sites, suspending field irrigation and cleaning the farmland drainage ditch. If it is flood to drought, focus on preventing drought and protecting water supply, start emergency water source reserve, prioritize guaranteeing irrigation of wheat and other drought-sensitive crops in the critical growth period, and restrict high-water industrial water use. If there is no sudden change risk, it is recommended to maintain daily soil moisture monitoring and precipitation warning. Integrated into structured drought and flood sudden change warning information report.

[0133] In summary, the embodiment provides a drought and flood sudden change prediction method combining soil moisture and precipitation coupling model, which calculates the initial value of soil moisture and precipitation coupling coordination, traverses the drought and flood window, determines the optimal window and optimizes it, and predicts the drought and flood sudden change, which has the beneficial effects of:

[0134] The application improves data base reliability, reduces data noise and deviation through multi-source data fusion, targeted abnormal value identification and missing value filling, standardization processing, provides high-quality and high-consistency basic data for drought and flood sudden change prediction, avoids the influence of data problems on prediction accuracy, improves the accuracy of sudden change prediction, quantifies the coupling relationship between soil moisture and precipitation through the combination of dynamic function and probability distribution, and then screens the optimal window and corrects the non-linear factors, accurately captures the non-linear and sudden characteristics of drought and flood sudden change, effectively reduces the problems of misjudgment of short-term fluctuations and missed rapid sudden change. Realize spatial accurate prediction, divide the sub-regions matched with hydrological characteristics, customize the coupling modeling, window screening and other parameters for different sub-regions, break the limitations of traditional global general prediction, output sub-regional prediction results, and adapt to the differentiated disaster prevention needs of different regions.

[0135] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and necessary general hardware platforms, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions for making a computer device (which can be a personal computer, server, or network device, etc.) execute the method of each embodiment or some parts of the embodiment.

[0136] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit it. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting drought-flood abrupt transition by coupling soil moisture regime and precipitation model, characterized in that, The method comprises the following steps: S1: constructing a three-dimensional hydrological map of a target area, marking the distribution of soil texture and the location of historical drought and flood sudden change events, dividing the area into P sub-areas to be predicted and recording the coordinates of the center of each sub-area; collecting layered soil moisture, high-frequency precipitation and auxiliary data, and outputting the original drought and flood data set of the sub-area; S2: preprocessing the original drought and flood data set of the sub-area to obtain a standardized data set; S3: calculating the initial value of soil moisture-precipitation coupling coordination according to the standardized data set and correcting it to obtain a standardized soil moisture-precipitation coupling coordination index sequence; S31: calculating the initial coupling coordination value according to the standardized data set; S311: determining the weight values of the soil moisture index and the precipitation index by the analytic hierarchy process; S312: constructing a function expression of the coupling coordination relationship between soil moisture and precipitation; S313: inputting the standardized data set into the function expression to obtain the initial coupling coordination value; S32: inputting the initial coupling coordination value into a power function and a segmented function, quantifying the dynamic relationship between soil moisture and precipitation by the least square method to obtain a quantized initial value; S33: using Gamma distribution fitting and inverse normal transformation to quantize the initial value to obtain a standardized soil moisture-precipitation coupling coordination index sequence; S331: classifying and arranging the quantized initial value to distinguish the soil moisture quantized initial value and the precipitation quantized initial value; S332: using the probability density function of Gamma distribution to fit the soil moisture quantized initial value and the precipitation quantized initial value by Gamma distribution to obtain a Gamma distribution model; S333: performing inverse normal transformation on the soil moisture quantized initial value and the precipitation quantized initial value according to the Gamma distribution model to obtain normalized soil moisture data and normalized precipitation data; S334: constructing a soil moisture-precipitation coupling coordination index calculation model, inputting the normalized soil moisture data and the normalized precipitation data into the index calculation model to obtain a coupling coordination index; S335: standardizing the coupling coordination index to obtain a standardized soil moisture-precipitation coupling coordination index sequence; S4: using the standardized soil moisture-precipitation coupling coordination index sequence, traversing the drought and flood window, determining the optimal window; making nonlinear correction combined with array precipitation, vegetation and soil texture, determining the sudden change event according to the corrected threshold value, and outputting the drought and flood sudden change prediction result; S41: according to the standardized soil moisture-precipitation coupling coordination index sequence, associating the sub-area with the time node to form a three-dimensional data table of sub-area CCIstd value; S42: according to the three-dimensional data table, setting a candidate set of the length N range of the drought and flood window, extracting the window unit corresponding to each N value and calculating the change index to determine the optimal window length, and outputting the optimal window length and CCIstd change characteristic parameters of each sub-area; S421: according to the three-dimensional data table, setting the length range of the drought and flood window, and clearly defining the traversal range of each sub-area, and outputting the window length candidate set of each sub-area; S422: For each N value in the window length candidate set, extract the CCIstd values of the corresponding sub-region continuous N time nodes from the three-dimensional data table as the window unit, calculate the range and linear fitting slope change index, and output the window unit change index dataset; S423: Determine the optimal window length according to the window unit change index dataset, and output the optimal window length and the corresponding CCIstd change characteristic parameters of each sub-region; Based on the CCIstd sequence within the window, calculate the drought-flood transition coupling strength CSI; For different candidate window lengths, calculate the CSI under each window length, and then compare it with the historical drought-flood sudden transition event list to calculate the matching rate; Select the window length with the highest matching rate and capable of distinguishing non-sudden transition events as the optimal window length; S43: Collect the array precipitation, vegetation, soil texture data of each sub-region optimal window period, correlate the CCIstd value in the window to construct a nonlinear correction model, calculate the corrected index CCIstd corr , output the corrected index sequence of each sub-region; Calculate the cumulative frequency R1 and the maximum intensity R2 of the array precipitation within the window, extract the mean value V of the vegetation cover index NDVI within the window, obtain the soil clay content proportion S, and associate these data with the CCIstd value within the optimal window of the sub-region to construct a nonlinear correction model; The optimal window length of each sub-region and the CCIstd variation characteristic parameter are substituted into a nonlinear correction model to obtain the corrected normalized coupling coordination index CCIstd in the optimal window of each sub-region corr The sequence of CCIstd of each sub-region is output corr The sequence of CCIstd of each sub-region is output S44: According to the historical drought-flood sudden change event position, the drought threshold T is determined by the percentile method according to the modified index sequence of each sub-region dry and the flood threshold T wet , and the modified threshold suitable for all sub-regions is output. S45: Compare CCIstd in the optimal window of each sub-region corr The sequence trend and the corrected threshold value, combined with the optimal window variation characteristic parameter, determine the abrupt change type, and output the drought and flood abrupt change prediction result. S5: calculating a drought-flood rapid change duration prediction value T according to the drought-flood rapid change prediction result and the soil texture coefficient int and the soil moisture content mutation index SPI ef , constructing a risk matrix, combining the disaster-bearing body vulnerability to output the warning level and response suggestions, and outputting the warning information.

2. The method according to claim 1, wherein the method is characterized by, In step S2, the preprocessing includes: S21: Use the isolation forest algorithm to identify outliers; S22: Short missing data is filled by using the correlation interpolation method according to the difference of soil texture, and long missing data is filled by using the LSTM-ATTention model; S23: Convert the data to the [-1, 1] interval by the Z-score standardization formula.

3. The method according to claim 1, wherein the method is characterized by, In step S5, the specific steps of outputting the warning information are: S51: According to the drought-flood sudden change prediction result, the soil texture coefficient of the sub-region, the average duration of the same type of sudden change in history is associated, and the drought-flood sudden change duration prediction value T is formed according to the sub-region int and the soil moisture sudden change precipitation mutation index SPI ef Calculate the basic data set; S52: Calculate T from the base data set int and SPI ef , output core indicator data set; S53: T int as the longitudinal axis, and SPI ef as the transverse axis to construct a matrix, map the data to the preliminary risk level, and output a preliminary risk level dataset; S54: Collect the vulnerability data of the disaster-bearing body, correct according to the preliminary risk level correction, and output the corrected risk level dataset; S55: Determine the warning level according to the corrected risk level, and output the structured drought-flood sudden transition warning information report in combination with the sudden transition type.

Images