Flood characteristic monitoring method based on detrending flood potential index
By detrending and analyzing GRACE satellite data and calculating the Detrended-FPI index, the problem of seasonal fluctuation interference in existing technologies was solved, and accurate monitoring and risk identification of large-scale floods were achieved.
Patent Information
- Application Number
- CN202510799074.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-16
- Publication Date
- 2025-09-19
AI Technical Summary
Existing flood monitoring technology based on GRACE satellite data fails to fully consider the interference of seasonal fluctuations and residuals, and lacks the ability to effectively monitor large-scale flood disasters.
By preprocessing and decomposing the abnormal land water storage data, the detrended flood potential index (Detrended-FPI) was calculated. The spatiotemporal distribution and trend of floods were analyzed in combination with precipitation, and the improved MK trend detection method was used to identify flood risks.
It effectively eliminates the interference of seasonal fluctuations on flood monitoring, improves the accuracy and identification ability of large-scale flood monitoring, and provides a scientific basis for flood risks.
Smart Images
Figure CN120671098A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of satellite remote sensing flood monitoring, and in particular to a flood characteristic monitoring method based on a detrended flood potential index. Background Art
[0002] Floods are primarily caused by a sudden increase in water volume due to continuous heavy rain, rapid melting of ice and snow, and storm surges. The upward trend in global average temperature will continue until the end of the 21st century, and the frequency of extreme weather events such as heat waves and heavy rainfall will also increase.
[0003] Typically, flood monitoring relies on in-situ monitoring stations that provide direct and timely measurements of key hydrometeorological elements such as precipitation, water level, and discharge. However, these monitoring stations are sparsely and unevenly distributed, posing a challenge to mapping flood-inundated areas. With the development of satellite remote sensing technology, different satellite sensors have been widely used in flood monitoring and forecasting research, and the application of these technologies has helped to reduce, manage, and control increasingly severe disasters around the world. However, these traditional flood monitoring techniques ignore the importance of antecedent soil moisture, which is believed to improve the efficiency of flood monitoring tools and even serve as an alternative data for flood monitoring.
[0004] The Gravity Recovery and Climate Experiment (GRACE) twin satellite mission, launched in 2002, has made it possible to monitor large-scale changes in Terrestrial Water Storage Anomaly (TWSA). Terrestrial water storage (TWS) is defined as the sum of all water above and below Earth's surface, including surface water, groundwater, root zone soil moisture, groundwater, snow cover, glaciers, water stored in vegetation, rivers, and lakes. GRACE-based TWSA has been shown to play an important role in capturing flood formation processes. For example, a Total Storage Deficit Index (TSDI) was constructed using GRACE data to describe flood disasters in the Liaohe River Basin between 2002 and 2016. A Water Storage Deficit Index (WSDI) was also constructed using GRACE data to capture floods in a river basin caused by human activities and climate change.
[0005] The Flood Potential Index (FPI) was proposed and its global applicability was verified by comparing it with flood maps obtained from the Dartmouth Flood Observatory. Subsequently, many studies have applied the FPI to evaluate its potential in flood monitoring in various regions. For example, the FPI calculated based on GRACE data from a certain location from 2003 to 2012 was found to be highly consistent with floods observed at regional and even local scales. The FPI was also used to identify the most severe flood event in a certain river basin in 2010. The FPI's ability to monitor floods of different levels was also verified in a certain river basin. The FPI has become the preferred and effective tool for monitoring large-scale floods using GRACE data.
[0006] However, current research on flood monitoring using GRACE satellite data is typically based on static Terrestrial Water Storage Anomaly (TWSA) data, which fails to fully account for seasonal fluctuations and residual interference. Existing technologies primarily focus on specific river basins or regions, lacking the ability to monitor disasters on a larger scale.
[0007] Currently, no effective solutions have been proposed for the problems in related technologies. Summary of the Invention
[0008] In response to the problems in the related art, the present invention proposes a flood characteristic monitoring method based on a detrended flood potential index to overcome the above-mentioned technical problems existing in the existing related art.
[0009] To this end, the specific technical solutions adopted in the present invention are as follows: A flood characteristic monitoring method based on a detrended flood potential index comprises the following steps: S1. Preprocess the time series of land water storage anomaly data obtained in advance, and decompose and detrend the preprocessed time series; S2. Calculate the detrended flood potential index based on the detrended land water storage anomaly data and the pre-obtained precipitation data; S3, evaluating the detrended flood potential index and the pre-acquired basic index, and calculating the correlation coefficient between the trended flood potential index and the basic index; S4. Calculate the spatiotemporal distribution of floods and flood trends based on the detrended flood potential index and flood potential image data; The step of calculating the spatiotemporal distribution of floods and flood trends based on the detrended flood potential index and the image data of flood potential comprises the following steps: The temporal and spatial distribution of summer floods was analyzed by calculating the pixel averages of the detrended flood potential index and the basic index in summer image data. The long-term temporal and spatial distribution of floods was obtained by combining the temporal and spatial variation characteristics of the flood-affected areas. By calculating the pixel averages of the detrended flood potential index and the basic index image data and combining them with time series for linear regression analysis, the seasonal flood trends were obtained. The modified MK trend test method (Hamed, KH, & Ramachandra Rao, A. (1998)) was used to detect the flood trends in different seasons. Journal of Hydrology , 204 (1–4), 182–196.), to obtain long-term flood trends.
[0010] Optionally, the preprocessing of the pre-acquired time series of abnormal land water storage data and the decomposition and detrending of the preprocessed time series include the following steps: S11. Decompose the processed land water storage anomaly data to obtain long-term trend terms, seasonal trend terms, and residual terms; S12. Remove the seasonal trend term and residual term from the land water storage anomaly data and only retain the long-term trend term.
[0011] Optionally, the calculation formula for the detrended flood potential index is: ; ; Where, Indicates the number of i Year j Monthly flood potential, represents the precipitation at the corresponding time, Indicates that during the monitoring period TWSA′ The maximum value of express Last month's TWSA′ TWSA′ represents the value obtained by detrending the abnormal land water storage data. It represents the maximum value of flood potential during the monitoring period. Indicates the number of i Year j Monthly detrended flood potential index.
[0012] Optionally, the step of evaluating the detrended flood potential index and the pre-acquired basic index and calculating the correlation coefficient between the detrended flood potential index and the basic index comprises the following steps: S31. Classify the flood risk level of the monitoring area based on the detrended flood potential index and the pre-acquired basic index; S32. Based on the basic index, potential flood areas in the monitoring area are identified and marked, and a sample set is constructed. The normal distribution function is used for fitting to generate the numerical range corresponding to each flood risk level; S33. Evaluate the monitoring results of the detrended flood potential index and basic index based on the sample set; S34. Based on the Pearson correlation coefficient formula and combined with the pre-set threshold calculation, the correlation coefficient of the detrended flood potential index and the basic index is generated.
[0013] Optionally, the identifying and marking of potential flood areas in the monitoring area based on the basic index, constructing a sample set, and using normal distribution function fitting to generate numerical intervals corresponding to each flood risk level include the following steps: S321. Based on the flood risk level and combined with the basic index, the pixels in the monitoring area are labeled and a sample set is constructed; S322. Statistically analyze the cumulative frequency distribution of the basic index, use the normal distribution function for fitting, and calculate based on the cumulative distribution function to generate the numerical range corresponding to each flood risk level.
[0014] Optionally, the basic indices include: flood potential index, water storage deficit index, total water storage deficit index and comprehensive climatological deviation index.
[0015] Optionally, the flood risk levels include: non-wet, generally wet, moderately wet, severely wet, extremely wet and abnormally wet.
[0016] Optionally, the evaluating the monitoring results of the detrended flood potential index and the basic index based on the sample set includes the following steps: S331. Based on the sample set, the true positives, false positives, false negatives, and true negatives of the detrended flood potential index and the basic index in flood area identification are counted to establish a confusion matrix for evaluating flood event identification. S332. Based on the confusion matrix, combined with precision, recall and F1 score calculation, the performance of detrended flood potential index and basic index in flood monitoring is evaluated.
[0017] Optionally, the generating of the correlation coefficient between the detrended flood potential index and the basic index based on the Pearson correlation coefficient formula in combination with a preset threshold value calculation comprises the following steps: S341, reading the image data of each time point in the detrended flood potential index and the basic index; S342, classifying the detrended flood potential index and the basic index according to a preset threshold value and converting them into classification values; S343. Based on each time point, a column vector of the detrended flood potential index and the basic index is constructed respectively, and the Pearson correlation coefficient formula is used to calculate and generate a coefficient matrix of the correlation between the detrended flood potential index and the basic index.
[0018] Optionally, the method of calculating the pixel average of the detrended flood potential index and the image data in the basic index, performing linear regression analysis on the time series to obtain flood trends in different seasons, and using the improved MK trend detection method to obtain long-term flood trends includes the following steps: A univariate linear regression model was constructed using the detrended flood potential index or basic index of each month during the monitoring period as the dependent variable and the month number as the independent variable. Based on the univariate linear regression model, the trend slope and significance level of each pixel are calculated; The data of the same season or the same month of each year are extracted and regressed separately to identify the changing trends of different seasons.
[0019] The beneficial effects of the present invention are: The present invention uses TWSA data and precipitation data acquired by the GRACE satellite. By detrending the TWSA data, the interference of seasonal fluctuations on hydrological and climate monitoring is eliminated. A detrended flood potential index (Detrended-FPI) is proposed and constructed, and applied to the extraction of flood characteristics in the region, including the temporal and spatial distribution and trend analysis of floods. In addition, the present invention is of great significance for large-scale flood monitoring, showing significant advantages in the monitoring and identification of flood disasters, and providing a new scientific basis for flood monitoring. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0021] Figure 1 is a flow chart of a flood characteristic monitoring method based on a detrended flood potential index according to an embodiment of the present invention; Figure 2 is a correlation coefficient graph of Detrended-FPI, FPI, CCDI, WSDI, and TSDI according to an embodiment of the present invention; DETAILED DESCRIPTION
[0022] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. By referring to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.
[0023] According to an embodiment of the present invention, a flood characteristic monitoring method based on a detrended flood potential index is provided.
[0024] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 - Figure 2 As shown, according to one embodiment of the present invention, a flood characteristic monitoring method based on a detrended flood potential index is provided, and the flood characteristic monitoring method includes the following steps: S1. Preprocess the time series of land water storage anomaly data obtained in advance, and decompose and detrend the preprocessed time series.
[0025] In one embodiment, the preprocessing of the pre-acquired time series of land water storage anomaly data and the decomposition and detrending of the preprocessed time series include the following steps: S11. Decompose the processed land water storage anomaly data to obtain long-term trend terms, seasonal trend terms, and residual terms; S12. Remove the seasonal trend term and residual term from the land water storage anomaly data and only retain the long-term trend term.
[0026] It should be explained that the present invention relates to a flood characteristic monitoring method based on the Detrended Flood Potential Index (Detrended-FPI for short). The method aims to solve the problem of insufficient accuracy of existing flood monitoring technology based on GRACE (Gravity Recovery and Climate Experiment) satellite observation data under dynamic environmental changes. The present invention optimizes the Flood Potential Index (FPI) through detrending processing to avoid the interference of seasonal fluctuations and residuals on flood monitoring results, thereby more accurately assessing flood risks.
[0027] In addition, the TWSA (Total Water Storage Anomaly) time series of the study area was preprocessed and decomposed and detrended using the STL method; the time series of the study area selected in this example was from May 2002 to June 2017.
[0028] Among them, the STL method can effectively remove seasonal and trend components from TWSA data, thereby obtaining detrended data, which provides a basis for further analysis. The STL method is used to decompose the TWSA time series to obtain long-term trend terms, seasonal trend terms, and residual terms. Only the long-term trend term is retained for subsequent analysis. Its expression is as follows: ; Where, Represents the original TWSA data, represents the long-term trend term, represents the seasonal trend term, represents the residual term.
[0029] S2. Calculate the detrended flood potential index based on the detrended land water storage anomaly data and the pre-obtained precipitation.
[0030] In one embodiment, the calculation formula for the detrended flood potential index is: ; ; Where, Indicates the number of i Year j Monthly flood potential, represents the precipitation at the corresponding time, Indicates that during the monitoring period TWSA′ The maximum value of express Last month's TWSA′ TWSA′ represents the value obtained by detrending the abnormal land water storage data. It represents the maximum value of flood potential during the monitoring period. Indicates the number of i Year j Monthly detrended flood potential index.
[0031] It should be explained that the detrended flood potential index (Detrended-FPI) is calculated for each pixel; this index reflects the flood potential by analyzing the relationship between precipitation and the detrended TWSA value; the Detrended-FPI value ranges from 0 to 1, and the closer the value is to 1, the greater the possibility of flooding.
[0032] S3. Evaluate the detrended flood potential index and the pre-acquired basic index, and calculate the correlation coefficient between the detrended flood potential index and the basic index.
[0033] In one embodiment, evaluating the detrended flood potential index and the pre-acquired basic index and calculating the correlation coefficient between the detrended flood potential index and the basic index comprises the following steps: S31. Based on the detrended flood potential index and combined with the pre-acquired basic index, the flood risk level of the monitoring area is divided.
[0034] S32. Based on the basic index, the potential flood areas in the monitoring area are identified and marked, and a sample set is constructed. The normal distribution function is used for fitting to generate the numerical range corresponding to each flood risk level.
[0035] In one embodiment, the identification and labeling of potential flood areas in the monitoring area based on the basic index, the construction of a sample set, and the generation of numerical intervals corresponding to each flood risk level using a normal distribution function fitting comprise the following steps: S321. Based on the flood risk level and combined with the basic index, the pixels in the monitoring area are labeled and a sample set is constructed; S322. Statistically analyze the cumulative frequency distribution of the basic index, use the normal distribution function for fitting, and calculate based on the cumulative distribution function to generate the numerical range corresponding to each flood risk level.
[0036] S33. Based on the sample set, the monitoring results of the detrended flood potential index and the basic index are evaluated.
[0037] In one embodiment, the evaluation of the monitoring results of the detrended flood potential index and the basic index based on the sample set includes the following steps: S331. Based on the sample set, the true positives, false positives, false negatives, and true negatives of the detrended flood potential index and the basic index in flood area identification are counted to establish a confusion matrix for evaluating flood event identification. S332. Based on the confusion matrix, combined with precision, recall and F1 score calculation, the performance of detrended flood potential index and basic index in flood monitoring is evaluated.
[0038] S34. Based on the Pearson correlation coefficient formula and combined with the pre-set threshold calculation, the correlation coefficient of the detrended flood potential index and the basic index is generated.
[0039] In one embodiment, the generating of the correlation coefficient between the detrended flood potential index and the basic index based on the Pearson correlation coefficient formula and a preset threshold value calculation includes the following steps: S341, reading the image data of each time point in the detrended flood potential index and the basic index; S342, classifying the detrended flood potential index and the basic index according to a preset threshold value and converting them into classification values; S343. Based on each time point, a column vector of the detrended flood potential index and the basic index is constructed respectively, and the Pearson correlation coefficient formula is used to calculate and generate a coefficient matrix of the correlation between the detrended flood potential index and the basic index.
[0040] In one embodiment, the basic indices include: a flood potential index, a water storage deficit index, a total water storage deficit index, and a comprehensive climatological deviation index.
[0041] In one embodiment, the flood risk levels include: non-wet, generally wet, moderately wet, severely wet, extremely wet, and abnormally wet.
[0042] It should be explained that the performance of the Detrended Flood Potential Index (Detrended-FPI) and the Flood Potential Index (FPI) in flood identification was evaluated. Based on the values of Detrended-FPI, FPI, Water Storage Deficit Index (WSDI), Total Water Storage Deficit Index (TSDI), and Comprehensive Climatological Deviation Index (CCDI), the flood risk in the monitoring area was divided into different levels; the levels include non-humid, generally humid, moderately humid, severely humid, extremely humid, and abnormally humid. The calculation formulas for FPI, WSDI, TSDI, and CCDI are as follows: ; ; Where, Indicates the number of i Year j Monthly flood potential, represents the precipitation at the corresponding time, Indicates that during the monitoring period TWSA′ The maximum value of express Last month's TWSA′ TWSA′ represents the value obtained by detrending the abnormal land water storage data. It represents the maximum value of flood potential during the monitoring period. Indicates the number of i Year j Monthly flood potential index.
[0043] ; ; ; ; ; ; Where, Indicates the period from April 2002 to June 2017 j The average value of TWSA in the calendar month, Indicates the period from April 2002 to June 2017 j The minimum value of TWSA in a calendar month, Indicates the i Year j months of TWSA, Indicates the i Year j Months and precipitation, Indicates the period from 2002 to 2017 i Year j Months of abnormal rainfall, Indicates that 2002-2017 j Months of abnormal rainfall, express and The difference in average precipitation, Indicates the water storage deficit at a specific time. represents the total water storage deficit at a specific time, Represents the integrated climate deviation at a specific time.
[0044] The flood season (June to August each year) was selected to identify and label potential flood areas based on the CCDI, WSDI, and TSDI. First, considering the short time series of GRACE satellite data (May 2002 to June 2017), to ensure statistical stability and wide applicability, the cumulative frequency distribution of CCDI, WSDI, and TSDI over the study period was statistically analyzed and fitted with a normal distribution function. Based on this, percentile thresholds were set at 30%, 20%, 10%, 5%, and 2%, and each index was divided into 11 dryness and wetness levels: abnormally dry (D4), extremely dry (D3), severe drought (D2), moderate drought (D1), normal drought (D0), normal (WD), normal wet (W0), moderately wet (W1), severely wet (W2), extremely wet (W3), and abnormally wet (W4). The numerical range corresponding to each level is calculated based on the cumulative distribution function of each index.
[0045] In addition, in order to focus on flood characteristic monitoring, D0-D4 and WD (normal) are further merged into the "Non-Wet Zone" (NW), and the W0 to W4 grade areas are defined as "potential flood areas", see Table 1 for details.
[0046] Table 1 Classification criteria for Detrended-FPI, FPI, WSDI, TSDI, and CCDI
[0047] Specifically, when a pixel meets the wetness level (W0 and above) in any of the three indexes of CCDI, WSDI and TSDI, the pixel is marked as a flood sample, and the sample set is constructed accordingly.
[0048] Data from the flood season (June to August each year) were selected to test the monitoring results of the Detrended-FPI and FPI indices. By comparing them with the sample set, the true positives (TP), false positives (FP), false negatives (FN), and true negatives (TN) of the two indices in flood area identification were counted, and a confusion matrix was established to evaluate the accuracy of flood event identification. The calculation formula is as follows: ; ; ; The confusion matrix and evaluation results in this embodiment are shown in Table 2, as follows: Table 2 Confusion matrix and evaluation results
[0049] Combined with Table 2, based on the confusion matrix, the precision, recall, and F1-score are calculated to quantify the performance of Detrended-FPI and FPI in flood monitoring.
[0050] like Figure 2 As shown in the figure, the correlation coefficients of Detrended-FPI, FPI, CCDI, WSDI, and TSDI were calculated. The calculation process of the correlation coefficients was as follows: First, the image data of each time point of the five indices (Detrended-FPI, FPI, CCDI, WSDI, and TSDI) were read; then, each index was classified according to the set threshold (see Table 1) and converted into a classification value (1 to 6); on this basis, for each time point, the column vectors of Detrended-FPI, FPI, CCDI, WSDI, and TSDI were constructed, and the correlation coefficient matrix between them was calculated using the Pearson correlation coefficient formula; the calculation formula is as follows: ; Where, It represents the correlation coefficient of two indices at a certain time point. Indicates thei The value of a variable at a pixel position, Indicates the i The value of another variable at the pixel location, express The mean of express The mean of n is the total number of valid pixels.
[0051] In addition, this process covers all months, and finally the correlation coefficient matrix of each time point is averaged to obtain the correlation coefficient matrix, which reflects the comprehensive correlation of the five types of indices in space and time.
[0052] S4. Calculate the spatiotemporal distribution of floods and flood trends based on the detrended flood potential index and flood potential image data.
[0053] The step of calculating the spatiotemporal distribution of floods and flood trends based on the detrended flood potential index and the image data of flood potential comprises the following steps: The temporal and spatial distribution of summer floods was analyzed by calculating the pixel averages of the detrended flood potential index and the basic index in summer image data. The long-term temporal and spatial distribution of floods was obtained by combining the temporal and spatial variation characteristics of the flood-affected areas. By calculating the pixel average of the detrended flood potential index and the basic index image data and combining them with time series for linear regression analysis, the flood trends in different seasons are obtained. The long-term flood trends are then obtained using the improved MK trend detection method.
[0054] In one embodiment, the method of calculating the pixel average of the detrended flood potential index and the image data in the basic index, performing linear regression analysis on the time series to obtain the flood trends of different seasons, and using the improved MK trend detection method to obtain the long-term flood trends includes the following steps: A univariate linear regression model was constructed using the detrended flood potential index or basic index of each month during the monitoring period as the dependent variable and the month number as the independent variable. Based on the univariate linear regression model, the trend slope and significance level of each pixel were calculated; The data of the same season or month of each year are extracted and regressed separately to identify the changing trends of different seasons.
[0055] It should be explained that the summer spatiotemporal distribution specifically refers to the spatial distribution of flood areas and intensities during June, July and August each year; the spatiotemporal distribution characteristics of summer floods were analyzed by calculating the pixel average values in June, July and August during the study period; the long-term spatiotemporal distribution refers to the spatiotemporal change characteristics of the flood-affected area over a longer period of time; it reflects the spatial distribution pattern of flood occurrence and the trend of change over time.
[0056] In addition, flood trends include seasonal trends and long-term trends. The seasonal flood trend is calculated by calculating the pixel average values of the four seasons of each year (winter: December, January, and February; spring: March, April, and May; summer: June, July, and August; and autumn: September, October, and November) (taking the average of three months in each season), and then performing linear regression analysis based on the time series. Specifically, the monthly index values (such as Detrended-FPI) from May 2002 to June 2017 are used as the dependent variable, and time (month number) is used as the independent variable. A univariate linear regression model is constructed to calculate the trend slope and significance level of each pixel. In order to obtain seasonal trends, data from the same season (such as spring, summer, autumn, and winter) or each month of each year are extracted and regressed separately to identify the changing trends in different seasons.
[0057] A downward trend (no significant) indicates a downward trend (not significant), an upward trend (no significant) indicates an upward trend (not significant), a significant downward trend (P<0.05) indicates a significant downward trend (P<0.05), a significant upward trend (P<0.05) indicates a significant upward trend (P<0.05), a significant downward trend (P<0.01) indicates a significant downward trend (P<0.01), and a significant upward trend (P<0.01) indicates a significant downward trend. "No trend" indicates no trend. The long-term trend is obtained using the improved Mann-Kendall (MMK) trend test method. The Mann-Kendall mutation test is a nonparametric hypothesis test method used to test trend changes in time series data.
[0058] The modified Mann-Kendall (MK) trend detection method is as follows: For a time series X=x1,x2,x3,......,x n (n>10), in the traditional MK trend detection, the original hypothesis H0 is that the time series X is nThere are independent samples with identical distribution of random variables, and there is no definite upward or downward trend; the alternative hypothesis H1 is a two-sided test, for all k, j ≤n, and k ≠ j , and The distribution of is different. S is the test statistic, Sgn is the sign function; Z is the standard normal distribution statistic, is the variance.
[0059] ; ; ; ; Where, Sgn represents the symbolic function, Z represents the standard normal distribution statistic, represents the variance, S represents the test statistic, Indicates the k samples, Indicates the j samples, n The number of samples representing the identical distribution of independent random variables.
[0060] In the two-sided trend test, at a given α At the confidence level, if , then the null hypothesis is unacceptable, that is, α At the confidence level, the time series data has an obvious upward or downward trend; through the statistical Z The symbol determines the trend type. Z When >0, it is an upward trend. Z <0 is a downward trend, and then through Z The size of the value determines whether the trend is significant; n Represents the number of samples in the time series X; and represents any two distributions in the time series X, where k < j ≤ n .
[0061] However, when the MK test is actually applied, the autocorrelation of the hydrological and meteorological series affects the test results. The modified MK (MMK) trend test method is used to solve the problem of data autocorrelation. The MMK trend test method considers the autocorrelation coefficient (lag Time ) to overcome this problem, these coefficients are significantly different from zero at the 5% level; S Corrected variance The calculation formula is as follows: ; ; Where, represents the correction factor, represents the autocorrelation coefficient.
[0062] By analyzing data from a certain region between May 2002 and June 2017, Detrended-FPI successfully identified multiple flood areas that traditional FPI failed to detect, and performed better than FPI in flood identification indicators such as recall rate and F1-Score. The research results show that Detrended-FPI can effectively capture the temporal and spatial characteristics of flood risk changes and reveal the changing trends of flood risk in different regions and seasons, providing effective technical support for large-scale flood monitoring in the region.
[0063] In summary, with the help of the above-mentioned technical scheme of the present invention, the present invention uses TWSA data and precipitation data acquired by the GRACE satellite, eliminates the interference of seasonal fluctuations on hydrological and climate monitoring by detrending the TWSA data, proposes and constructs a detrended flood potential index (Detrended-FPI), and applies it to the extraction of flood characteristics in the region, including the spatiotemporal distribution and trend analysis of floods; in addition, the present invention is of great significance for large-scale flood monitoring, shows significant advantages in the monitoring and identification of flood disasters, and provides a new scientific basis for flood monitoring.
[0064] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A flood characteristic monitoring method based on detrended flood potential index, characterized in that: The method comprises the following steps: S1. Preprocess the time series of land water storage anomaly data obtained in advance, and decompose and detrend the preprocessed time series; S2. Calculate the detrended flood potential index based on the detrended land water storage anomaly data and the pre-obtained precipitation data; S3. evaluating the detrended flood potential index and the pre-acquired basic index, and calculating the correlation coefficient between the detrended flood potential index and the basic index; S4. Calculate the spatiotemporal distribution and flood trend of floods based on the detrended flood potential index; The step of calculating the spatiotemporal distribution of floods and flood trends based on the detrended flood potential index comprises the following steps: The temporal and spatial distribution of summer floods was analyzed by calculating the pixel average of the detrended flood potential index, and the long-term temporal and spatial distribution of floods was obtained by combining the temporal and spatial variation characteristics of the flood-affected areas. By calculating the pixel average of the detrended flood potential index and combining it with time series for linear regression analysis, the flood trends in different seasons are obtained. The long-term flood trends are then obtained using the improved MK trend detection method.
2. A flood characteristic monitoring method based on detrended flood potential index according to claim 1, characterized in that: The preprocessing of the time series of the pre-acquired land water storage anomaly data and the decomposition and detrending of the preprocessed time series include the following steps: S11. Decompose the processed land water storage anomaly data to obtain long-term trend terms, seasonal trend terms, and residual terms; S12. Remove the seasonal trend term and residual term from the land water storage anomaly data and only retain the long-term trend term.
3. A flood characteristic monitoring method based on detrended flood potential index according to claim 1, characterized in that: The calculation formula for the detrended flood potential index is: ； ； Where, Indicates the number of i Year j Monthly flood potential, represents the precipitation at the corresponding time, Indicates that during the monitoring period TWSA′ The maximum value of express Last month's TWSA′ TWSA′ represents the value obtained by detrending the abnormal land water storage data. It represents the maximum value of flood potential during the monitoring period. Indicates the number of i Year j Monthly detrended flood potential index.
4. A flood characteristic monitoring method based on detrended flood potential index according to claim 1, characterized in that: The step of evaluating the detrended flood potential index and the pre-acquired basic index and calculating the correlation coefficient between the detrended flood potential index and the basic index comprises the following steps: S31. Classify the flood risk level of the monitoring area based on the detrended flood potential index and the pre-acquired basic index; S32. Based on the basic index, potential flood areas in the monitoring area are identified and marked, and a sample set is constructed. The normal distribution function is used for fitting to generate the numerical range corresponding to each flood risk level; S33. Evaluate the monitoring results of the detrended flood potential index and basic index based on the sample set; S34. Based on the Pearson correlation coefficient formula and combined with the pre-set threshold calculation, the correlation coefficient of the detrended flood potential index and the basic index is generated.
5. A flood characteristic monitoring method based on detrended flood potential index according to claim 4, characterized in that: The method of identifying and marking potential flood areas in the monitoring area based on the basic index, constructing a sample set, and using a normal distribution function to fit the data to generate the numerical intervals corresponding to each flood risk level includes the following steps: S321. Based on the flood risk level and combined with the basic index, the pixels in the monitoring area are labeled and a sample set is constructed; S322. Statistically analyze the cumulative frequency distribution of the basic index, use the normal distribution function for fitting, and calculate based on the cumulative distribution function to generate the numerical range corresponding to each flood risk level.
6. A flood characteristic monitoring method based on detrended flood potential index according to claim 4, characterized in that: The basic indices include: flood potential index, water storage deficit index, total water storage deficit index and comprehensive climatological deviation index.
7. A flood characteristic monitoring method based on detrended flood potential index according to claim 4, characterized in that: The flood risk levels include: non-wet, generally wet, moderately wet, severely wet, extremely wet and abnormally wet.
8. The flood characteristic monitoring method based on detrended flood potential index according to claim 4 is characterized in that: The evaluation of the monitoring results of the detrended flood potential index and the basic index based on the sample set includes the following steps: S331. Based on the sample set, the true positives, false positives, false negatives, and true negatives of the detrended flood potential index and the basic index in flood area identification are counted to establish a confusion matrix for evaluating flood event identification. S332. Based on the confusion matrix, combined with precision, recall and F1 score calculation, the performance of detrended flood potential index and basic index in flood monitoring is evaluated.
9. A flood characteristic monitoring method based on detrended flood potential index according to claim 4, characterized in that: The method of generating the correlation coefficient between the detrended flood potential index and the basic index based on the Pearson correlation coefficient formula and a preset threshold value includes the following steps: S341, reading the image data of each time point in the detrended flood potential index and the basic index; S342, classifying the detrended flood potential index and the basic index according to a preset threshold value and converting them into classification values; S343. Based on each time point, a column vector of the detrended flood potential index and the basic index is constructed respectively, and the Pearson correlation coefficient formula is used to calculate and generate a coefficient matrix of the correlation between the detrended flood potential index and the basic index.
10. A flood characteristic monitoring method based on detrended flood potential index according to claim 1, characterized in that: The method of calculating the pixel average of the detrended flood potential index and the basic index image data, performing linear regression analysis on the time series, obtaining the flood trends of different seasons, and using the improved MK trend detection method to obtain the long-term flood trends includes the following steps: A univariate linear regression model was constructed using the detrended flood potential index or basic index of each month during the monitoring period as the dependent variable and the month number as the independent variable. Based on the univariate linear regression model, the trend slope and significance level of each pixel were calculated; The data of the same season or month of each year are extracted and regressed separately to identify the changing trends of different seasons.