Yellow River basin drought and flood sudden transition space-time evolution rule research method

Abstract

The invention discloses a Yellow River basin drought and flood sudden change space-time evolution rule research method, and relates to the technical field of water conservancy project research, a long-period drought and flood sudden change index (LDFAI) and a short-period drought and flood sudden change index (SDFAI) are combined, the space-time distribution rule of a basin DTF event and a basin FTD event is analyzed, the drought and flood sudden change type and intensity evolution trend analysis is carried out, and the Yellow River basin drought and flood sudden change time-space evolution rule is obtained. According to the analysis result, key environment factors influencing the LDFAI from June to September, the SDFAI from June to July, the SDFAI from July to August and the SDFAI from August to September of the Yellow River basin are screened out, correlation analysis is carried out, and LDFAI prediction models based on LSTM and SVM are constructed on the basis of the key environment influence factors and correlation analysis. The method aims at providing a scientific basis for disaster prevention and reduction of drought and flood sudden change events in the Yellow River basin and providing a reference for constructing a drought and flood sudden change index prediction model.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of hydraulic engineering research, and more particularly to a method for researching the spatio-temporal evolution law of drought-flood rapid transition in the Yellow River basin. BACKGROUND

[0002] Under the background of global warming, extreme drought and flood events occur frequently. Drought-flood rapid transition events are the continuous alternation of drought and flood events in a short period of time, including two typical cases of drought-to-flood (DTF) and flood-to-drought (FTD), and the cumulative impact degree often exceeds the sum of the impacts of single events. The Yellow River basin is one of the regions with high frequency of drought-flood rapid transition events, and with the continuous emission of greenhouse gases, the frequency and intensity of drought-flood rapid transition events in the basin will increase in the future, so in-depth research is needed.

[0003] The related research on drought-flood rapid transition events focuses on four aspects: spatio-temporal distribution law, evolution trend, influencing factor and prediction model. For the spatio-temporal distribution law of drought-flood rapid transition in the Yellow River basin, Liu Yufeng et al. analyzed the spatio-temporal evolution characteristics of drought-flood rapid transition events in a province by using summer drought-flood rapid transition index; Yu Qun et al. discussed the climate characteristics of drought-flood mutation in a province;

[0004] At present, a large amount of research has been carried out on drought-flood rapid transition events in the Yellow River basin, usually taking a provincial region in the basin as the research scale, and taking some aspects of spatio-temporal distribution law, evolution trend, influencing factor and prediction model of drought-flood rapid transition events as the research object, while the research on the entire Yellow River basin as the research scale, systematically analyzing drought-flood rapid transition events from the four aspects of "distribution law-evolution trend-influencing factor-prediction model" is relatively scarce. Scientifically constructing a drought-flood rapid transition index prediction model is of great significance for drought-flood rapid transition event warning and disaster prevention and reduction, however, the existing research only considers the influence of circulation and sea surface temperature index, ignoring the influence of environmental factors on drought-flood rapid transition index. In addition, there is a complex nonlinear relationship between many influencing factors and drought-flood rapid transition index, and the existing model uses linear regression to construct the prediction equation, which is not completely consistent with the actual situation.

[0005] Therefore, how to improve the accuracy of the evolution of time-varying law and further provide a scientific basis for disaster prevention and reduction of drought-flood rapid transition events in the Yellow River basin is a problem to be solved by those skilled in the art. SUMMARY

[0006] Therefore, the application provides a research method for spatio-temporal evolution law of drought-flood abrupt change in the Yellow River Basin, which combines a long-period drought-flood abrupt change index (LDFAI) and a short-period drought-flood abrupt change index (SDFAI), analyzes the spatio-temporal distribution law of DTF and FTD events in the basin, carries out analysis on the evolution trend of drought-flood abrupt change types and intensity, respectively screens key environmental factors affecting the LDFAI in June-September, the SDFAI in June-July, the SDFAI in July-August and the SDFAI in August-September in the Yellow River Basin, and carries out correlation analysis, and on this basis, respectively constructs an LDFAI prediction model based on LSTM and SVM. The application aims to provide a scientific basis for disaster prevention and reduction of drought-flood abrupt change events in the Yellow River Basin and provide a reference for construction of a drought-flood abrupt change index prediction model.

[0007] In order to achieve the above-mentioned purpose, the application adopts the following technical solutions:

[0008] A research method for spatio-temporal evolution law of drought-flood abrupt change in the Yellow River Basin, comprising:

[0009] Collecting observation data of a plurality of meteorological stations in the basin;

[0010] Combining a long-period drought-flood abrupt change index and a short-period drought-flood abrupt change index, analyzing the spatio-temporal distribution law of drought-flood abrupt change events in the basin based on the observation data, and carrying out analysis on the evolution trend of drought-flood abrupt change types and intensity;

[0011] According to the analysis results, key environmental impact factors affecting the long-period drought-flood abrupt change index in June-September, the short-period drought-flood abrupt change index in June-July, the short-period drought-flood abrupt change index in July-August and the short-period drought-flood abrupt change index in August-September in the Yellow River Basin are respectively screened, and correlation analysis is carried out;

[0012] On the basis of the key environmental impact factors and correlation analysis, a long-period drought-flood abrupt change index prediction model is constructed;

[0013] Based on the constructed long-period drought-flood abrupt change index prediction model, drought-flood abrupt change index prediction results are obtained.

[0014] Optionally, the data source for collecting observation data of a plurality of meteorological stations in the basin is:

[0015] Meteorological data provided by "China Ground Climate Data Daily Value Dataset (V3.0)" of China Meteorological Data Network, which is converted into monthly data;

[0016] Monthly atmospheric circulation indices provided by the National Climate Center of China Meteorological Administration.

[0017] Optionally, the long-period drought-flood abrupt change index is used to study the long-period variation characteristics of drought-flood abrupt change events in June-September, and specifically:

[0018]

[0019] In the formula, LDFAI is a long-period drought-flood abrupt transition index; P 89 and P 67 respectively represent the standardized rainfall in August-September and June-July; (P 89 -P 67 ) is a drought-flood abrupt transition intensity term; (|P 67 |+|P 89 |) is a drought-flood intensity term; is a weight coefficient; when the LDFAI value is higher than 1, a "drought-to-flood" event is considered to occur, and when the LDFAI value is lower than -1, a "flood-to-drought" event is considered to occur, and the greater the absolute value, the higher the drought-flood abrupt transition event intensity.

[0020] Optionally, the calculation formula of the standardized rainfall is as follows:

[0021]

[0022] In the formula, P i is the standardized rainfall; X i is the original rainfall in the ith month; and is the annual average rainfall, and n is the number of months for statistics.

[0023] Optionally, the short-period drought-flood abrupt transition index is used to study the short-period drought-flood abrupt transition change characteristics in June-July, July-August and August-September, respectively, and specifically as follows:

[0024]

[0025] In the formula, P i and P i+1 respectively represent the standardized rainfall in the previous month and the following month; (P i+1 -P i ) is a drought-flood abrupt transition intensity term; (|P i |+|P i+1 |) is a drought-flood intensity term; is a weight coefficient.

[0026] Optionally, the drought-flood abrupt transition type and intensity evolution trend analysis specifically includes using the Mann-Kendall method to analyze the significance of the drought-flood abrupt transition type and intensity evolution trend in the Yellow River Basin, taking the confidence level alpha as 0.2, and the corresponding Z critical value as 1.28, and if |Z|>1.28, it indicates that the research object has a significant increasing or decreasing trend.

[0027] Optionally, the correlation analysis is specifically Pearson correlation coefficient analysis of the correlation between the key environmental impact factor and the long-period and short-period drought-flood rapid transition events, and the p value calculation result is between -1 and 1, a positive value indicates a positive correlation, a negative value indicates a negative correlation, and 0 indicates no correlation.

[0028] Optionally, it further includes using multiple wavelet coherence analysis to analyze the significance of the influence of different environmental factor combinations on the long-period drought-flood rapid transition index from June to September, the short-period drought-flood rapid transition index from June to July, the short-period drought-flood rapid transition index from July to August, and the short-period drought-flood rapid transition index from August to September, and the higher the percentage of significant area is, the more significant the influence is.

[0029] Optionally, the long-period drought-flood rapid transition index prediction model is specifically constructed by using a long short-term memory network or a support vector machine.

[0030] According to the technical solution, compared with the prior art, the present disclosure provides a method for studying the spatio-temporal evolution law of drought-flood rapid transition in the Yellow River Basin, collects observation data of a plurality of meteorological stations in the basin, combines the long-period drought-flood rapid transition index and the short-period drought-flood rapid transition index, analyzes the spatio-temporal distribution law of the drought-flood rapid transition events in the basin based on the observation data, and analyzes the evolution trend of the drought-flood rapid transition type and intensity; according to the analysis result, the key environmental impact factors affecting the long-period drought-flood rapid transition index from June to September, the short-period drought-flood rapid transition index from June to July, the short-period drought-flood rapid transition index from July to August, and the short-period drought-flood rapid transition index from August to September in the Yellow River Basin are selected respectively, and correlation analysis is performed; on the basis of the key environmental impact factors and the correlation analysis, a long-period drought-flood rapid transition index prediction model is constructed; and based on the constructed long-period drought-flood rapid transition index prediction model, a drought-flood rapid transition index prediction result is obtained. The present disclosure aims to provide a scientific basis for disaster prevention and mitigation of drought-flood rapid transition events in the Yellow River Basin, and to provide a reference for the construction of a drought-flood rapid transition index prediction model. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only embodiments of the present disclosure, and those skilled in the art can obtain other drawings according to the provided drawings without creating any inventive labor.

[0032] Figure 1 The method flowchart provided by the present disclosure;

[0033] Figure 2 The 1971-2022 Yellow River Basin DTF and FTD event frequency distribution graph provided by the present disclosure;

[0034] Figure 3 a DTF and FTD event occurrence frequency spatial distribution map provided by the present application in the Yellow River Basin;

[0035] Figure 4 a drought-flood rapid change type evolution trend test Z value distribution map provided by the present application;

[0036] Figure 5 a drought-flood rapid change event intensity trend test Z value distribution map provided by the present application;

[0037] Figure 6 an environmental factor and LDFAI Pearson correlation coefficient distribution map provided by the present application;

[0038] Figure 7 an environmental factor and 6-7 month SDFAI Pearson correlation coefficient distribution map provided by the present application;

[0039] Figure 8 an environmental factor and 7-8 month SDFAI Pearson correlation coefficient distribution map provided by the present application;

[0040] Figure 9 an environmental factor and 8-9 month SDFAI Pearson correlation coefficient distribution map provided by the present application;

[0041] Figure 10 an LDFAI and SDFAI and four environmental factor multiple coherence map provided by the present application;

[0042] Figure 11 a machine learning model prediction result of LDFAI provided by the present application;

[0043] Figure 12 a Yellow River Basin schematic diagram provided by the present application. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0045] The embodiments of the present application disclose a Yellow River Basin drought-flood rapid change spatio-temporal evolution rule research method, as shown in the following formula (1), which comprises the following steps: Figure 1

[0046] Collecting observation data of a plurality of meteorological stations in the basin;

[0047] ​The long-period drought-flood abrupt change index and the short-period drought-flood abrupt change index are combined, the spatiotemporal distribution law of the drought-flood abrupt change event in the basin is analyzed based on the observation data, and the evolution trend of the drought-flood abrupt change type and intensity is analyzed;

[0048] According to the analysis result, the key environmental influence factors affecting the long-period drought-flood abrupt change index of the Yellow River Basin from June to September, the short-period drought-flood abrupt change index from June to July, the short-period drought-flood abrupt change index from July to August and the short-period drought-flood abrupt change index from August to September are screened respectively, and correlation analysis is performed;

[0049] On the basis of the key environmental influence factors and correlation analysis, a long-period drought-flood abrupt change index prediction model is constructed;

[0050] Based on the constructed long-period drought-flood abrupt change index prediction model, the drought-flood abrupt change index prediction result is obtained.

[0051] In one specific embodiment, the observation data of 110 meteorological stations (positions are shown in FIG. 1) in the basin are used, and the interpolation method is used to process the data gap in the downstream. The data sources include: Figure 12

[0052] (1) The meteorological data of rainfall, daily evaporation, relative humidity and temperature provided by the China Meteorological Data Network "China Ground Climate Data Daily Value Dataset (V3.0)" from 1971 to 2022 are converted into monthly data.

[0053] (2) The monthly atmospheric circulation indices such as Arctic Oscillation Index (AO), Sunspot Index (TSNI), NINO3.4 Sea Surface Temperature Anomaly Index (NINO3.4) and North Pacific Decadal Oscillation Index (PDO) provided by the National Climate Center of China Meteorological Administration.

[0054] In one specific embodiment, the long-period drought-flood abrupt change index is used to study the long-period variation characteristics of the drought-flood abrupt change event from June to September, specifically:

[0055]

[0056] In the formula, LDFAI is the long-period drought-flood abrupt change index; P 89 and P 67 respectively represent the standardized rainfall in August-September and June-July; (P 89 -P 67 ) is the drought-flood abrupt change intensity term; (|P 67 |+|P 89 |) is the drought-flood intensity term; is the weight coefficient; when the LDFAI value is higher than 1, it is considered that a "drought to flood" event occurs, and when the LDFAI value is less than -1, it is considered that a "flood to drought" event occurs, and the greater the absolute value, the higher the intensity of the drought-flood abrupt change event. ​

[0057] In one specific embodiment, the calculation formula of the standardized rainfall is:

[0058]

[0059] In the formula, P i is the standardized rainfall; X i is the original rainfall in the i-th month; and is the average annual rainfall, and n is the number of months for statistics.

[0060] In one specific embodiment, the short-period drought-flood abrupt change index is used to study the short-period drought-flood abrupt change characteristics in June-July, July-August and August-September, respectively, and the specific embodiments are as follows:

[0061]

[0062] In the formula, P i and P i+1 respectively represent the standardized rainfall in the previous month and the next month; (P i+1 -P i ) is the drought-flood abrupt change intensity term; (|P i |+|P i+1 |) is the drought-flood intensity term; and is the weight coefficient.

[0063] In one specific embodiment, the evolution trend analysis of the drought-flood abrupt change type and intensity is specifically that the Mann-Kendall method is used to analyze the significance of the evolution trend of the drought-flood abrupt change type and intensity in the Yellow River Basin, the confidence level α is taken as 0.2, the corresponding Z critical value is 1.28, and if |Z|>1.28, it indicates that the research object has a significant increasing or decreasing trend.

[0064] The correlation analysis is specifically that the Pearson correlation coefficient is used to analyze the correlation between the key environmental impact factors and the long-period and short-period drought-flood abrupt change events, and the p value calculation result is between -1 and 1, a positive value indicates a positive correlation, a negative value indicates a negative correlation, and 0 indicates no correlation.

[0065] The multiple wavelet coherence analysis is used to analyze the significance of the influence of different environmental factor combinations on the long-period drought-flood abrupt change index in June-September, the short-period drought-flood abrupt change index in June-July, the short-period drought-flood abrupt change index in July-August and the short-period drought-flood abrupt change index in August-September, and the higher the percentage of significant area value is, the more significant the influence is.

[0066] The construction of the long-period drought-flood abrupt change index prediction model is specifically:

[0067] Long short-term memory network (LSTM) can effectively capture long-term dependencies in complex time series data, and is an ideal choice to solve the time series dependence problem and nonlinear characteristics in drought-flood rapid transition event prediction. Support vector machine (SVM) also has superiority and applicability in dealing with small sample, nonlinearity, high dimensionality and other complex problems. Considering the complex relationship between environmental impact factors and drought-flood rapid transition events, LSTM and SVM are used to build LDFAI prediction models in the study area.

[0068] A specific embodiment is introduced below to further illustrate the method of the present application.

[0069] 1 Spatiotemporal distribution and evolution trend of drought-flood rapid transition in the Yellow River Basin

[0070] 1.1 Temporal distribution of drought-flood rapid transition events

[0071] According to the rainfall data of 110 meteorological stations in the Yellow River Basin from June to September in 1971-2022, combined with formulas (1)-(3), the annual LDFAI and SDFAI were calculated, and the number of meteorological stations with long and short period (June-July, July-August and August-September) DTF and FTD events was counted, as shown in Figure 2 .

[0072] (1) Interannual variation of long-period drought-flood rapid transition events

[0073] Figure 2 It is shown that the long-period DTF events in 1971-1986 showed strong activity, with the number of meteorological stations with DTF events generally exceeding 10, and showing periodic fluctuation characteristics, becoming the core form of drought-flood rapid transition events in the Yellow River Basin during this period. After 2010, DTF events continued to occur intermittently, with the number of meteorological stations with DTF events increasing to more than 30 in some years, far exceeding the frequency intensity of FTD events at the same period, indicating that the extremity of DTF events has significantly increased in recent years. FTD events concentrated in 1986-2003, with the number of meteorological stations with FTD events exceeding 10 every year, which was the main form of long-period drought-flood rapid transition events in the Yellow River Basin during this period. After 2010, FTD events still existed, but the frequency of more than 10 meteorological stations occurring simultaneously showed discontinuous fluctuations, and did not form a sustained high incidence trend. In summary, the long-period drought-flood rapid transition events in the Yellow River Basin were dominated by DTF in the early stage (1971-1986), by FTD in the middle stage (1986-2003), and by DTF in the recent stage (after 2010).

[0074] (2) Interannual variation of short-period drought-flood rapid transition events

[0075] The DTF events in June-July were active from 1971 to 1980, and the FTD events in June-July were frequent from 1980 to 1995, with an average of more than 10 meteorological stations per year, which was the dominant form in this period. After 2000, the frequency of both types of events decreased, with the frequency of DTF events in June-July declining significantly after 2014. The DTF events in July-August showed a significant increase from 1998 to 2011 and from 2016 to 2020, with a peak of 27 meteorological stations per year, highlighting the extreme nature of the events. The FTD events in July-August were more frequent from 1970 to 1980. The DTF events in August-September showed an extreme increase from 1971 to 1975 and from 2008 to 2014, with some years experiencing a significant increase in frequency to more than 30 meteorological stations. The FTD events in August-September were continuously high from 1988 to 2000, with more than 10 meteorological stations per year, and then showed a phase of fluctuation from 2000 onwards, such as from 2003 to 2010 and from 2016 to 2022. In summary, among the short-period drought-flood rapid transition events in the Yellow River Basin, the DTF events showed a "low frequency-high intensity" feature and had a stronger potential for extreme outburst in a short period, while the FTD events were mainly characterized by sustained medium-high frequency.

[0076] 1.2 Spatial distribution of drought-flood rapid transition events

[0077] Based on the frequency of DTF and FTD events in the long period and short periods (June-July, July-August, and August-September) at each meteorological station in the Yellow River Basin, the spatial distribution was analyzed, as shown in Figure 3 Overall, the long-period drought-flood rapid transition events (including DTF and FTD) had a lower frequency in Region III, but a significantly higher frequency in Regions IV-VII, indicating that the middle and lower reaches of the Yellow River were high-frequency regions for long-period drought-flood rapid transition events. The spatial distribution of short-period drought-flood rapid transition events was similar to that of long-period drought-flood rapid transition events. Specifically, long-period DTF events were mainly concentrated in the southern part of Region V, Region VI, and Region VII, while long-period FTD events were concentrated in the eastern part of Region V, Region VI, and Region VII. The core of the DTF events in June-July was located in the junction of the Jingwei River in Region V, while the FTD events in June-July were concentrated in the border between Region VI and Region VII. The DTF events in July-August had the same distribution as the DTF events in June-July, while the FTD events in July-August were concentrated in the middle and eastern parts of Region VII. The DTF events in August-September were concentrated in the two ends of Region VII, while the FTD events in August-September were located in the southern part of Region IV, the eastern part of Region V, and the western part of Region VI.

[0078] 1.3 Evolution trend of drought-flood rapid transition events

[0079] (1) Evolution trend of drought-flood rapid transition types

[0080] The Z value distribution of the long-term and short-term (June-July, July-August, and August-September) drought-flood abrupt transition type evolution trend in the Yellow River Basin was calculated using the Mann-Kendall method, as shown in Table 2. Figure 4 A positive Z value indicates an increasing DTF trend, while a negative Z value indicates an increasing FTD trend, with a confidence level of 80%. Figure 4 The results show that the Z value of the long-term drought-flood abrupt transition type evolution trend is positive in most regions of the basin, with Z value > 1.28 in Region II, indicating that the long-term drought-flood abrupt transition type evolution trend is dominated by DTF. The Z value of the short-term (June-July) drought-flood abrupt transition type evolution trend is mainly negative in the basin, indicating that FTD dominates in June-July. The Z value of the short-term (July-August) drought-flood abrupt transition type evolution trend is positive in a large area of the basin, with Z value > 1.28 in Regions VI and VII, indicating that the short-term (July-August) DTF event is concentrated in the middle and lower reaches of the basin. The Z value of the short-term (August-September) drought-flood abrupt transition type evolution trend is positive in the upper reaches of the basin (Regions II-IV), indicating a DTF trend, while it is negative in the lower reaches of the basin (Regions VI-VII).

[0081] (2) Evolution trend of drought-flood abrupt transition intensity

[0082] The Z value distribution of the long-term and short-term (June-July, July-August, and August-September) drought-flood abrupt transition intensity evolution trend in the Yellow River Basin was calculated, as shown in Table 2. Figure 5 The results show that the Z value of the long-term drought-flood abrupt transition intensity evolution trend exceeds 1.28 in the northern part of Region II, the border between Region IV and Region V, indicating a significant increasing trend in the long-term drought-flood abrupt transition intensity in these regions. The Z value is < -1.28 in the western part of Region V, Region III, and the northern part of Region IV, indicating a significant decreasing trend in the long-term drought-flood abrupt transition intensity. The Z value of the short-term (June-July) drought-flood abrupt transition intensity evolution trend is negative in the basin as a whole, indicating a decreasing trend in intensity. The Z value of the short-term (July-August) drought-flood abrupt transition intensity evolution trend is positive in Regions II, the eastern part of Region V, and Region VI, with Z value > 1.28 in Region II, highlighting the extreme tendency of drought-flood transition in this region. The Z value is negative in Regions III, IV, and the southern part of Region V, but not significant. The short-term (August-September) drought-flood abrupt transition intensity increasing area is concentrated in Regions III and IV (Z value > 1.28), while the decreasing area is located in the southeastern part of Region V and Region VI (Z value < -1.28).

[0083] 2. Prediction results of drought-flood abrupt transition index in the Yellow River Basin based on machine learning

[0084] 2.1 Environmental factors affecting drought-flood abrupt transition events

[0085] The occurrence of drought-flood abrupt events is closely related to multiple environmental factors. According to existing researches [28-29], the environmental factors can be divided into meteorological factors and teleconnection factors. The meteorological factors mainly include temperature (TEP), relative humidity (RHU) and evaporation (EVP), and the teleconnection factors mainly include NINO3.4, PDO, AO and TSNI. The Pearson correlation coefficients between the environmental factors and the LDFAI, the 6-7 month SDFAI, the 7-8 month SDFAI and the 8-9 month SDFAI in the Yellow River Basin were analyzed, and the spatial distribution is shown in Fig. 2, where the red color represents positive correlation and the blue color represents negative correlation. Figures 6-9

[0086] (1) Key environmental factors affecting the LDFAI in the Yellow River Basin

[0087] Figure 6 It is shown that the correlation coefficients between RHU and AO and the LDFAI are the lowest, i.e., they have relatively weak effects on the long-period drought-flood abrupt events in the Yellow River Basin. TEP shows a high positive correlation with the LDFAI in Region II, and a negative correlation in Regions V and VI. EVP shows a high positive correlation with the LDFAI in Regions II, the southern part of Region IV, Regions V, VI and VII, indicating that the evaporation change is an important factor driving the long-period drought-flood abrupt events in these regions. NINO3.4, PDO and TSNI show a negative correlation with the LDFAI in some regions of Regions III-V, and the negative correlation between TSNI and the LDFAI in Region IV is particularly significant. In summary, TEP, EVP, NINO3.4, PDO and TSNI have high correlations with the LDFAI, and are the key environmental factors affecting the long-period drought-flood abrupt events in the Yellow River Basin.

[0088] (2) Key environmental factors affecting the 6-7 month SDFAI in the Yellow River Basin

[0089] Figure 7 It is shown that the correlation coefficients between TEP and RHU and the 6-7 month SDFAI are the lowest. EVP shows a significant negative correlation with the 6-7 month SDFAI in Region I, and a significant positive correlation in Regions V and VI. NINO3.4 shows a negative correlation with the 6-7 month SDFAI in the whole basin, and the negative correlation is particularly significant in Regions III-VII. PDO shows a negative correlation with the 6-7 month SDFAI in the eastern part of Region V, Regions VI and VII, and AO shows a negative correlation with the 6-7 month SDFAI in the southern part of Region I, the western part of Region V and Region IV. TSNI shows a positive correlation with the 6-7 month SDFAI only in the eastern part of Region I and the western part of Region II, and a negative correlation in Region VII. Therefore, the main factors affecting the 6-7 month SDFAI are EVP, NINO3.4, PDO, AO and TSNI.

[0090] (3) Key environmental factors affecting the 7-8 month SDFAI in the Yellow River Basin​

[0091] Figure 8 It is shown that the correlation coefficients of RHU and AO with SDFAI in July-August are relatively low; TEP is positively correlated with SDFAI in the western part of Region II, while it is highly negatively correlated in the southern part of Region III, Region IV and Region V; EVP is positively correlated with SDFAI in the whole basin, especially in Region I, Region II, Region V and Region VI; NINO3.4 is significantly positively correlated with SDFAI in Region V-Region VII, while it is significantly negatively correlated in the northern part of Region IV; PDO is mainly positively correlated with SDFAI in the whole basin, especially in Region V-Region VII; TSNI is negatively correlated with SDFAI in the whole basin, especially in Region IV. Therefore, the main influencing factors of SDFAI in July-August are TEP, EVP, NINO3.4, PDO and TSNI.

[0092] (4) Key environmental factors affecting SDFAI in August-September in the Yellow River Basin

[0093] Figure 9 It is shown that although TEP and RHU are significantly negatively and positively correlated with SDFAI in Region VII in August-September, respectively, the correlation coefficients of TEP and RHU with SDFAI are still relatively low compared with other environmental factors; EVP is highly positively correlated with SDFAI in Region I-Region III, indicating that the relationship between the abrupt transition events and water evaporation in this period is more closely; NINO3.4 is positively correlated with SDFAI in the western and northern parts of the basin; PDO is significantly negatively correlated with SDFAI in the whole basin, except for the northern part of Region I and Region VII, indicating that the Pacific Decadal Oscillation has a wide influence on the abrupt transition events in August-September in the Yellow River Basin; AO is significantly negatively correlated with SDFAI in some stations in Region I, Region III and Region V, while it is positively correlated in the vicinity of Region IV; The correlation between TSNI and SDFAI in August-September is relatively complex, showing a positive correlation in the east and a negative correlation in the west. Therefore, the main influencing factors of the short-period abrupt transition events in August-September are EVP, NINO3.4, PDO, AO and TSNI.

[0094] In summary, the main environmental influencing factors of LDFAI and SDFAI in different periods are summarized in Table 1.

[0095] Table 1 Key environmental factors of different abrupt transition indices in the Yellow River Basin

[0096] PDSI Main environmental factors PDSI Main environmental factors LDFAI TEP, EVP, NINO3.4, PDO, TSNI 7-8 month SDFAI TEP, EVP, NINO3.4, PDO, TSNI 6-7 month SDFAI EVP, NINO3.4, PDO, AO, TSNI 8-9 month SDFAI EVP, NINO3.4, PDO, AO, TSNI

[0097] 2.2 Correlation analysis of environmental factors on abrupt transition events

[0098] Based on Table 1, the effects of four-factor combinations on LDFAI and each SDFAI in the Yellow River Basin were further analyzed using multi-wavelet coherence, as shown in Figure 10 , and the corresponding PASC values were calculated, as shown in Table 2. The results showed that the maximum PASC values of SDFAI from June to July, SDFAI from July to August, SDFAI from August to September, and LDFAI under the combined effects of the four environmental factors were 24.31%, 25.00%, 32.54%, and 49.65%, respectively, corresponding to the environmental factor combinations of EVP, PDO, TSNI, and NINO3.4, TEP, EVP, PDO, and TSNI, EVP, PDO, AO, and NINO3.4, and EVP, PDO, TSNI, and NINO3.4. This result confirmed the important influence of multi-factor synergy on the drought-flood rapid transition event in the Yellow River Basin and provided a basis for constructing a drought-flood rapid transition index prediction model based on machine learning in the Yellow River Basin.

[0099] Table 2 PASC values of LDFAI and SDFAI under the combined effects of environmental factors

[0100]

[0101] 2.3 Long-period drought-flood rapid transition index prediction results based on LSTM and SVM

[0102] To verify the feasibility of constructing a machine learning prediction model for the drought-flood rapid transition index, the key environmental factor combinations (EVP, PDO, TSNI, and NINO3.4) affecting LDFAI in Table 2 were selected as feature values to construct LDFAI prediction models based on LSTM and SVM. Both models randomly selected 42 groups of data from 52 groups of LDFAI data from 1971 to 2022 as the training set, and the remaining 10 groups of data as the test set. The learning efficiency of the LSTM model was 0.001, the gradient descent algorithm was used, and the training times were 500. The correlation coefficient (R2), mean absolute error (MAE), and root mean square error (RMSE) were used as evaluation indicators, and the prediction results of the two models are compared in Figure 11 .

[0103] Figure 11It is shown that the evaluation accuracy of the test set of the LSTM model and the SVM model is higher than that of the prediction set, although the evaluation result of the test set of the SVM model is slightly better than that of the LSTM model, the R2 of the two models is 0.99 and 0.97 respectively, but the evaluation result error of the prediction set of the SVM model is greater than that of the LSTM model, the MAE of the two models is 0.24 and 0.35 respectively, and the average relative error is 36% and 51% respectively. Overall, the LSTM model can better predict the LDFAI of the Yellow River Basin than the SVM model, and the LDFAI machine learning model can be constructed to predict the LDFAI of the basin according to the values of the key environmental factors. Figure 11 It can be seen that the evaluation result of the prediction set of the LSTM model also has certain error, which is mainly caused by two reasons: one is that the number of prediction set samples is too small, only 10 groups of data, and the prediction value and true value of individual data are greatly different; the other is that the direct influencing factor of drought-flood rapid transition index is rainfall, other influencing factors are many and the influence relationship is complex, and the machine learning model only considers four key environmental factors (EVP, PDO, TSNI and NINO3.4). Therefore, in the future, more comprehensive environmental factors can be considered and different machine learning models can be tried to improve the prediction accuracy of drought-flood rapid transition index.

[0104] 3 Conclusion

[0105] The Yellow River Basin is divided into seven hydrological second regions, and the spatio-temporal distribution and evolution trend of drought-flood rapid transition events in the basin are analyzed based on the meteorological data of 52 years and 110 meteorological stations. The key environmental factors of long-period and short-period drought-flood rapid transition index are screened, and the machine learning prediction model of LDFAI is established. The main conclusions are as follows:

[0106] (1) The long-period drought-flood rapid transition event in the Yellow River Basin is dominated by DTF in the early stage (1971-1986), dominated by FTD in the transition stage (1986-2003), and the intensity of DTF rebounds in recent years (after 2010); the short-period drought-flood rapid transition event shows the characteristics of "low frequency-high intensity", and its extreme outbreak potential is stronger, and FTD is mainly sustained and high. The frequency of long-period drought-flood rapid transition event in region III is low, while the frequency in regions IV-VII is significantly increased; the overall spatial distribution of short-period drought-flood rapid transition event is the same as that of long-period drought-flood rapid transition event, that is, the high-frequency area is in the middle and lower reaches of the Yellow River.

[0107] (2) The long-period drought-flood abrupt transition type is mainly DTF in most areas of the basin. The short-period drought-flood abrupt transition type FTD dominates in June-July. The short-period drought-flood abrupt transition type is mainly DTF in July-August and August-September, with the former being highlighted in the southern and eastern parts of the basin and the latter being intensified in the northern and western parts of the basin. The long-period drought-flood abrupt transition intensity shows a significant increasing trend in the transitional zone of the middle and lower reaches of the Yellow River, while it shows a weakening trend in the northern part of the basin. The short-period drought-flood abrupt transition intensity shows a weakening trend in the basin as a whole in June-July. The short-period drought-flood abrupt transition intensity shows a significant increasing trend in the western part of the basin in July-August, and it shows a significant increasing trend in the northern part of the basin and a weakening trend in the southern part of the basin in August-September.

[0108] (3) The four key environmental factors of LDFAI and SDFAI in June-July, July-August, and August-September are EVP, PDO, TSNI, and NINO3.4, EVP, PDO, TSNI, and NINO3.4, TEP, EVP, PDO, and TSNI, EVP, PDO, AO, and NINO3.4, respectively. The PASC values of the corresponding drought-flood abrupt transition indices under the joint action of the four environmental factors are 49.65%, 24.31%, 25.00%, and 32.54%, respectively.

[0109] (4) The evaluation results of the test set of LSTM and SVM models are good, and the R2, MAE, and RMSE of the prediction set are 0.65, 0.24, 0.30 and 0.33, 0.35, 0.41, respectively. The LSTM model can better predict the LDFAI of the Yellow River basin than the SVM model, which can provide a reference for the construction of machine learning-based drought-flood abrupt transition index prediction models in other basins.

[0110] (5) The influencing factors of drought-flood abrupt transition are complex, and the relationship between them is complex. In the future, more measured data can be introduced, more complete environmental factors can be considered, and different machine learning models can be tried to improve the accuracy of the prediction results.

[0111] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0112] The foregoing description of the disclosed embodiments enables a person skilled in the art to make or use the application. Modifications of these embodiments will occur to persons of skill in the art, and that the appended claims are intended to cover all such modifications that do not depart from the true spirit and scope of the application. Therefore, the application is not limited to the embodiments shown but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for researching the spatio-temporal evolution law of sudden changes between drought and flood in the Yellow River Basin, characterized in that, The application comprises the following steps: Collecting observation data of several meteorological stations in a basin; Combining long-period drought-flood abrupt change index and short-period drought-flood abrupt change index, analyzing the spatio-temporal distribution of drought-flood abrupt change events in the basin based on the observation data, and analyzing the evolution trend of drought-flood abrupt change type and intensity; According to the analysis results, the key environmental impact factors affecting the long-period drought-flood abrupt change index of the Yellow River Basin from June to September, the short-period drought-flood abrupt change index from June to July, the short-period drought-flood abrupt change index from July to August, and the short-period drought-flood abrupt change index from August to September are screened, and correlation analysis is performed; Based on the key environmental impact factors and correlation analysis, a long-period drought-flood abrupt change index prediction model is constructed. The data source of the collected observation data of several meteorological stations in the basin is:

2. The method according to claim 1, characterized in that, The meteorological data provided by the "China Ground Climate Data Daily Value Dataset" of the China Meteorological Data Network, which is converted into monthly data; The monthly atmospheric circulation index provided by the National Climate Center of the China Meteorological Administration. The long-period drought-flood abrupt change index is used to study the long-period variation characteristics of drought-flood abrupt change events from June to September, specifically:

3. The method according to claim 1, characterized in that, The calculation formula of the standardized rainfall is: where LDFAI is the long-period drought-flood abrupt transition index; P 89 and P 67 represent the standardized precipitation index for August-September and June-July, respectively; (P 89 -P 67 ) is the drought-flood abrupt transition intensity term; (|P 67 |+|P 89 |) is the drought-flood intensity term; is the weight coefficient; when the LDFAI value is higher than 1, it is considered that a "drought-to-flood" event has occurred, and when it is lower than -1, it is considered that a "flood-to-drought" event has occurred, and the greater the absolute value, the higher the intensity of the drought-flood abrupt transition event.

4. The method according to claim 3, characterized in that, The short-period drought-flood abrupt change index is used to study the short-period drought-flood abrupt change variation characteristics from June to July, from July to August, and from August to September, specifically as follows: In the formula, P i is the standardized rainfall; X i is the original rainfall of the ith month; is the average annual rainfall, and n is the number of months of statistics.

5. The method according to claim 1, characterized in that, The analysis of the evolution trend of drought-flood abrupt change type and intensity is specifically to analyze the significance of the evolution trend of drought-flood abrupt change type and intensity in the Yellow River Basin by using the Mann-Kendall method, and the confidence level α is 0.2, and the corresponding Z critical value is 1.

28. If |Z|>1.28, it indicates that the research object has a significant increasing or decreasing trend. where P i and P i+1 represent the upper and lower month standardized rainfall, respectively; (P i+1 -P i ) is the abrupt change intensity term; (|P i |+|P i+1 |) is the intensity term; and a is the weight coefficient. 6.The method of claim 1, wherein, The correlation analysis is specifically to analyze the correlation between the key environmental impact factors and the long-period and short-period drought-flood abrupt change events by using the Pearson correlation coefficient, and the p value calculation result is between -1 and 1. A positive value indicates a positive correlation, a negative value indicates a negative correlation, and 0 indicates no correlation.

7. The method according to claim 1, characterized in that, It also includes using multiple wavelet coherence analysis to analyze the significance of the influence of different environmental factor combinations on the long-period drought-flood abrupt change index from June to September, the short-period drought-flood abrupt change index from June to July, the short-period drought-flood abrupt change index from July to August, and the short-period drought-flood abrupt change index from August to September. The higher the significant area percentage value, the more significant the influence.

8. The method according to claim 7, characterized in that, The construction of the long-period drought-flood abrupt change index prediction model is specifically to construct the long-period drought-flood abrupt change index prediction model of the research area by using long short-term memory network or support vector machine. 9.The method according to claim 1, characterized in that, ​

Images