Flood cascade remote correlation feature mining method, system, equipment and medium

Abstract

The invention discloses a flood cascade remote correlation feature mining method, system and device and a medium, and relates to the technical field of flood disaster assessment, and the method comprises the steps: constructing a cascade flood database through extreme flood events; constructing a plurality of directed complex networks, inputting time information of the cascade flood into the directed complex networks, analyzing out-degree and in-degree by using a single variable pair and a cross variable of the flood, performing flood synchronization between rainfall and runoff, rainfall and flow, and runoff and flow, and representing a flood cascade phenomenon; extracting network indexes of flood propagation characteristics, network clustering and synchronization intensity in the flood cascade phenomenon, and taking the network indexes as flood cascade remote correlation characteristics. According to the method, a complex network analysis method is adopted to clarify flood propagation and synchronization mechanisms including a plurality of hydrological variables such as rainfall, runoff and flow, some remarkable characteristics are researched and revealed, and compared with other networks, the flood network related to the runoff shows higher connectivity.

Description

Technical Field

[0001] The present invention relates to the technical field of flood disaster assessment, and in particular to a method, system, equipment and medium for mining remote correlation characteristics of flood cascades. Background Art

[0002] Globally, floods are one of the most common and costly natural hazards, posing major challenges to societies, infrastructure and ecosystems. The dynamics of floods is a complex and multifaceted phenomenon, influenced by dynamic interrelationships between the atmospheric, surface and subsurface realms. Understanding flood-related interactions between these hydrological realms is essential to gain a deeper understanding of the overall dynamics and connectivity of flood events.

[0003] Extreme event analysis often uses precipitation, runoff, and discharge as proxies for flood characteristics. Precipitation serves as the main driver by generating excess water, which subsequently leads to runoff and discharge. Previous studies have investigated the spatiotemporal interactions of flood generation mechanisms between different flood elements. For example, Yin et al. analyzed the dynamic relationship between extreme precipitation and runoff and found that extreme runoff increased at a rate exceeding the Clausius-Clapeyron scaling, often exceeding the growth rate of extreme precipitation. Sharma et al. studied the association between precipitation and discharge and found that this association changed under climate change.

[0004] However, it is generally assumed that extremes of precipitation, runoff and flow will show correlations over long distances, but the interactions and teleconnection relationships between these mechanisms are still poorly understood, making it difficult to obtain teleconnection characteristics, which in turn leads to poor flood hazard assessment results. Summary of the invention

[0005] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and to provide a method, system, device and medium for mining teleconnection features of flood cascades, so as to solve the problem that the prior art generally assumes that extreme manifestations of precipitation, runoff and flow will show correlation over long distances, but the interactions and teleconnection relationships between these mechanisms are still poorly understood, resulting in difficulty in obtaining teleconnection features, which in turn leads to poor flood disaster assessment results.

[0006] The present invention specifically provides the following technical solutions:

[0007] A method for mining remote correlation features of flood cascades comprises the following steps:

[0008] Obtaining extreme flood events that propagate continuously in sub-basins through precipitation, runoff and flow hydrological variables, and constructing cascading floods through said extreme flood events;

[0009] Multiple directed complex networks are constructed for river basins connected in the incoming or outgoing direction, and the temporal information of cascading floods is input into the directed complex networks. The out-degree and in-degree of single variable pairs and cross-variable pairs in cascading floods are analyzed. When analyzing cross-variable pairs, flood synchronization is performed between precipitation and runoff, precipitation and flow, and runoff and flow. The flood cascade phenomenon in the river basin is characterized by flood synchronization.

[0010] Based on the directed complex network, network indicators about flood propagation characteristics, network clustering and synchronization strength in flood cascade phenomena are extracted, and the network indicators are used as flood cascade teleconnection characteristics.

[0011] Preferably, a cascading flood is constructed through the extreme flood event, wherein the pairing of the cascading flood includes precipitation-precipitation PP, runoff-runoff RR, flow-flow SS, precipitation-runoff PR, precipitation-flow PS and runoff-flow RS.

[0012] Preferably, characterizing the flood cascade phenomenon in a river basin by flood synchronization includes:

[0013] Get precipitation - the time when the precipitation PP floods in sub-basin i And the time when the flood occurs in sub-basin j is obtained as

[0014] If the time lag between two floods occurring in sub-basins i and j is If the floods are within , then these two floods are considered synchronous floods, specifically:

[0015]

[0016] PP_c(i|j) is defined as the number of times a flood occurs at location i after it occurs at location j, specifically:

[0017]

[0018] in:

[0019]

[0020] Quantify synchronization and delay, the specific expression is:

[0021]

[0022] Among them, PP_Q ij It is used to measure the intensity of synchronous floods between sub-basins i and j, while PP_q ij Used to measure the latency behavior between them.

[0023] Preferably, the river basins connected in the incoming or outgoing direction construct multiple directed complex networks, including:

[0024] Based on PP_Q ij and PP_q ij Construct a directed complex network and map precipitation-precipitation PP, runoff-runoff RR, flow-flow SS, precipitation-runoff PR, precipitation-flow PS and runoff-flow RS according to PP_Q ij and PP_q ij Six directed complex networks were constructed.

[0025] Preferably, the PP_Q based ij and PP_q ij When building a directed complex network, it includes:

[0026] PP_Q ij Convert a symmetric binary matrix into an adjacency matrix Specifically:

[0027]

[0028] pass Determine the direction of the link. It is defined as the adjacency matrix, specifically:

[0029]

[0030] Preferably, the network indicators for extracting flood propagation characteristics, network clustering and synchronization strength include:

[0031] When extracting flood propagation characteristics, the level difference data of the river are obtained according to the Strahler method, the river level is given, the statistical data of the propagation direction is obtained according to the direction of all lines from the centroid of the starting basin to the centroid of the ending basin, and the synchronous distance of propagation is determined according to the median of the geographical distance between the connected sub-basins;

[0032] When clustering the network, sub-basins that are directly or indirectly connected are grouped into the same cluster, and five network metrics are obtained from the clustering: the number of clusters, the number of isolated sub-basins, the largest cluster size, the average cluster size, and the sub-basins with the most connections;

[0033] When obtaining the synchronization strength, the total degree is obtained by summing the in-degree and out-degree of the sub-basin, and the total degree is used to obtain the overall number of connections between the sub-basin and other sub-basins in the network.

[0034] The present invention provides a flood cascade teleconnection feature mining system, comprising:

[0035] A collection module for acquiring extreme flood events that are continuously propagated in the sub-basin through precipitation, runoff and flow hydrological variables, and constructing cascading floods through the extreme flood events;

[0036] A model building module is used to construct multiple directed complex networks for river basins connected in the incoming or outgoing direction, and input the time information of cascading floods into the directed complex networks, analyze the out-degree and in-degree of single variable pairs and cross-variable pairs in the cascading floods, and when analyzing cross-variable pairs, perform flood synchronization between precipitation and runoff, precipitation and flow, and runoff and flow, and characterize the flood cascade phenomenon in the river basin through flood synchronization;

[0037] The indicator acquisition module is used to extract network indicators about flood propagation characteristics, network clustering and synchronization strength in flood cascade phenomena based on the directed complex network, and use the network indicators as flood cascade teleconnection characteristics.

[0038] The present invention provides a computer device, comprising a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of the above-mentioned method for mining the telecorrelation characteristics of flood cascades.

[0039] The present invention provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for mining the telecorrelation characteristics of a flood cascade are realized.

[0040] Compared with the prior art, the present invention has the following significant advantages:

[0041] The present invention adopts a directed complex network analysis method to illustrate the flood propagation and synchronization mechanism, including multiple hydrological variables such as precipitation, runoff and flow. By inputting the time information of cascading floods into the directed complex network, the out-degree and in-degree of the river basin can be analyzed, and flood synchronization can be performed to reveal the complex interactions between precipitation, runoff and flow extremes. Higher connectivity can be obtained, and network indicators of flood propagation characteristics, network clustering and synchronization intensity in flood cascade phenomena can be extracted. The network indicators are used as flood cascade teleconnection characteristics. The understanding of flood dynamics is enhanced through flood cascade teleconnection characteristics, and relevant information is provided for enhancing flood risk management strategies, which ultimately helps to improve the effect of assessing flood disasters and make people more resilient to flood impacts. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a propagation diagram of synchronous flood among sub-basins among six pairs of flood variables (i.e., PP, RR, SS, PR, PS and RS) of the present invention;

[0043] Figure 2A detailed diagram of the propagation distance of the link of the present invention;

[0044] Figure 3 The network cluster diagram of synchronous floods between sub-basins among six pairs of flood variables in the Yangtze River Basin is provided by the present invention;

[0045] Figure 4 The present invention is an out-degree graph of synchronous floods in the Yangtze River Basin;

[0046] Figure 5 The present invention is an in-degree diagram of synchronous floods in the Yangtze River Basin;

[0047] Figure 6 The invention is a comprehensive out-degree, in-degree and total degree graph of synchronous flood;

[0048] Figure 7 The present invention is a flow chart of a method for mining teleconnection characteristics of flood cascades. DETAILED DESCRIPTION

[0049] The following is a clear and complete description of the technical solutions of the embodiments of the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0050] The main objectives of this paper are: 1) to reveal the complex interactions between precipitation, runoff, and flow extremes in the Yangtze River Basin using complex network analysis; and 2) to propose a robust framework that can be extended to assess teleconnection relationships between multivariate hydrological events. By applying complex network analysis to multiple proxies of floods, it aims to reveal hidden patterns, measure spatial connectivity, and identify key sub-basins or regions that play a key role in propagating extreme events. The insights from this study can help improve the understanding of flood dynamics and provide valuable information for enhancing flood risk management strategies, which will ultimately help promote communities to become more resilient in the face of increasingly severe flood impacts.

[0051] like Figure 7 As shown, the present invention provides a method for mining remote correlation features of flood cascades, which specifically includes the following steps:

[0052] Step S1: Obtain extreme flood events that propagate continuously in sub-basins through precipitation, runoff and flow hydrological variables, and construct cascading floods through extreme flood events.

[0053] In previous studies, cascade events refer to two or more extreme events that occur continuously or cumulatively without interruption. In this paper, cascade floods are defined as extreme flood events that propagate continuously in sub-basins through hydrological variables such as precipitation, runoff and discharge. Cascade floods are constructed through extreme flood events, where the pairing methods of cascade floods include precipitation-precipitation PP, runoff-runoff RR, discharge-discharge SS, precipitation-runoff PR, precipitation-discharge PS and runoff-discharge RS.

[0054] The peak over threshold (POT) method was used to identify flood events. Based on the POT method, the time series of extreme values ​​of precipitation, runoff, and flow exceeding the selected threshold were obtained. The principle of selecting the threshold was to calculate an average of three floods per year. In order to avoid repeated calculation of the same flood event, only one flood event was calculated within a time window of 15 days. The time information of all flood events was recorded, which would become the input for building the complex network.

[0055] Step S2: Construct multiple directed complex networks for river basins connected in the incoming or outgoing direction, and input the temporal information of cascading floods into the directed complex networks. Use the out-degree and in-degree of the single variable pairs and cross variable pairs of floods to analyze the synchronization of floods between precipitation and runoff, precipitation and flow, and runoff and flow, to characterize the flood cascade phenomenon in the river basin. The river here is the Yangtze River.

[0056] Typically, event-simultaneous analysis is performed based on a single variable; however, here the analysis is performed using both single variable pairs (i.e., PP, RR, SS) and crossed variable pairs (i.e., PR, PS, RS). In the simultaneous flood analysis of crossed variable pairs, the following measures were used: PR_Q ij and PR_q ij For synchronized flooding between precipitation and runoff; PS_Q ij and PS_q ij For synchronous flooding between precipitation and flow; RS_Q ij and RS_q ij Used for synchronized flooding between runoff and flow.

[0057] The event synchronization algorithm is used to characterize the flood cascade phenomenon between sub-basins. The flood cascade phenomenon of the river basin is characterized by:

[0058] Taking PP as an example, obtain the time when the flood occurs in sub-basin i based on precipitation-precipitation PP And the time when the flood occurs in sub-basin j is obtained as PP_s i and PP_s j are the total number of floods occurring in sub-basins i and j, respectively.

[0059] If the time lag between two floods occurring in sub-basins i and j is If the floods are within , then these two floods are considered synchronous floods, specifically:

[0060]

[0061] PP_c(i|j) or PP_c(j|i) is defined as the number of times the flood occurs at location i (or j) after it occurs at location j (or i), specifically:

[0062]

[0063] in:

[0064]

[0065] By quantifying synchronization and delay, the specific expression is:

[0066]

[0067] Among them, PP_Q ij It is used to measure the intensity of synchronous floods between sub-basins i and j, while PP_q ij Used to measure the delay behavior between them. ij It is standardized as 0≤PP_Q≤1, where PP_Q=1 means perfect synchronization. ij is normalized to -1≤PP_q≤1. PP_q ij >0 means that floods in sub-basin i mostly arrive before floods in sub-basin j, PP_q ij =1 indicates that the flood from i always arrives before the flood from j.

[0068] Among them, multiple directed complex networks are constructed by river basins connected in the incoming or outgoing direction, including:

[0069] Based on PP_Q ij and PP_q ij Construct a directed complex network and map precipitation-precipitation PP, runoff-runoff RR, flow-flow SS, precipitation-runoff PR, precipitation-flow PS and runoff-flow RS according to PP_Q ij and PP_q ij Six directed complex networks were constructed.

[0070] Among them, based on PP_Q ij and PP_q ij When building a directed complex network, it includes:

[0071] In order to retain the most persistent connections, a threshold (θQ ), which is the 5% percentile. Taking the construction of PP network as an example, PP_Q ij Convert to a symmetric binary matrix of an adjacency matrix, specifically:

[0072]

[0073] pass Determine the direction of the link, Indicates that the direction of the link is from i to j, not the other way around. It is defined as the adjacency matrix, specifically:

[0074]

[0075] Step S3: Based on the directed complex network, network indicators of flood propagation characteristics, network clustering and synchronization intensity in the flood cascade phenomenon are extracted, and the network indicators are used as flood cascade teleconnection characteristics to explore the cascade phenomenon of mutual influence and mutual reinforcement between floods represented by precipitation, runoff and flow.

[0076] In order to further characterize and analyze the constructed network, network indicators are calculated based on flood propagation characteristics, network clustering and synchronization strength. Among them, the indicators of complex networks are obtained based on flood propagation characteristics, network clustering and synchronization strength, including:

[0077] When extracting flood propagation characteristics, for all propagation paths (i.e., connections), the level difference data of the river are calculated according to the Strahler method to give the river level. For all propagation paths (i.e., connections), the level difference data of the river are obtained according to the Strahler method to give the river level, the statistics of the propagation direction are obtained based on the direction of all lines from the centroid of the starting basin to the centroid of the ending basin, and the synchronous distance of propagation is determined based on the median of the geographical distance between the connected sub-basins. This indicator provides insights into the average potential propagation distance of the physical signal within or between flood variables. The distance between sub-basins is calculated based on the distance between the centroids of two connected sub-basins.

[0078] When clustering the network, it is defined that sub-basins directly connected to at least one other sub-basin belong to one cluster. Directly or indirectly connected sub-basins are divided into the same cluster, and five network indicators are obtained from the clustering: the number of clusters (the number of clusters), the number of isolated sub-basins (sub-basins that do not belong to any cluster), the largest cluster size (the number of sub-basins included in the largest cluster), the average cluster size (the average number of sub-basins included in a cluster), and the sub-basins with the most connections (the number of sub-basins directly connected to the most sub-basins and the number of connections).

[0079] Degree information provides insights into the role, importance, and influence of nodes in complex networks. For degree information, three network metrics were obtained for each pair of flood variable combinations: 1) in-degree; 2) out-degree; 3) total degree. In a directed network, each connection has a specific direction, indicating the propagation path of floods from one sub-basin to another. The in-degree of a sub-basin represents the number of connections pointing to it. In-degree provides insights into the importance or influence of sub-basins in the network. Sub-basins with higher in-degree tend to receive more information, resources, or interactions from other sub-basins, making them potentially central or influential in the network. Out-degree represents the number of connections emanating from a sub-basin and pointing to other sub-basins. To obtain the out-degree of a sub-basin, the number of connections emanating from the sub-basin and pointing to all other sub-basins is calculated. By analyzing the links connected to each sub-basin, the out-degree of each sub-basin can be determined. Sub-basins with higher out-degree may act as flood sources, initiators of flood interactions, or propagators of flood influence.

[0080] When obtaining the synchronization strength, the total degree is obtained by summing the in-degree and out-degree of the sub-basin, and the total degree is used to obtain the overall number of connections of the sub-basin with other sub-basins in the network. This metric provides a measure of the connectivity and participation of the sub-basin in the network. It quantifies the overall connectivity or centrality of the sub-basin in the network, regardless of the direction of the connection. Sub-basins with higher total degrees generally have more connections and have greater potential for interaction with other sub-basins. The total degree is a fundamental metric in network analysis as it helps identify important sub-basins, detect community structure, and understand the resilience and fragility of the network. Matrix Used to calculate degree statistics for each flood variable pair.

[0081] Results and Discussion:

[0082] 1. Analysis of flood propagation characteristics:

[0083] Figure 1 Insights on flood propagation between precipitation, runoff, and flow in the Yangtze River Basin are provided. It is noteworthy that the flow-related flood network has more connections compared to the precipitation- and runoff-dominated networks. In particular, the SS flood network exhibits the highest connectivity, with a total of 147 connections (e.g. Figure 1 c). This indicates that the water flow pattern has strong synchronization. This synchronization may be derived from the combined effects of upstream precipitation and runoff processes converging in the river channel, resulting in a balanced response between different watersheds (Spence et al., 2010). The RS network ranks second with 110 connections ( Figure 1f), indicating the important role of runoff in influencing flow dynamics (Spence, 2007). However, the slightly lower number of connections compared to the SS flood network suggests that local topography and land cover characteristics have a significant impact on the runoff response, resulting in reduced synchrony (Phillips et al., 2011). In contrast, the runoff-dominated networks have fewer connections than the precipitation and streamflow networks, with the RR network having the least number of connections, at 65 ( Figure 1 This difference highlights the moderating effect of local factors (such as land use and soil properties) on runoff behavior, introducing spatial variability and reducing synchrony (Negese, 2021; Niehoff et al., 2002).

[0084] Figure 1 The propagation of synchronized floods between sub-basins for six pairs of flood variables (i.e., PP, RR, SS, PR, PS, and RS) is depicted. The directed connections between each sub-basin in the network represent the propagation of floods from one sub-basin to another. To capture the most persistent connections, only the top 5% synchronized sub-basins were assigned connections. This selective approach ensures that the strongest and most significant connections are retained. The number of connections in the network is as follows: Figure 1 The a in is 73, Figure 1 The b is 65, Figure 1 The c is 147, Figure 1 The d is 82, Figure 1 The e is 87 and Figure 1 f is 110. A gradient color scale from red to blue is applied to the connections, with gradients closer to blue indicating a greater number of connections. The propagation distance statistics of the connections can be found in Figure 2 .

[0085] Figure 1The connections in the figure describe various directions. It is noteworthy that the SS connections show a high consistency of being generally eastward, with the mean, median, and standard deviation being 104.2°, 90.8°, and 61.5°, respectively. Of the 147 connections in the SS, 105 are concentrated in the east direction (between 45° and 135°). The average direction of the SS is consistent with the water flow direction of the main channel of the Yangtze River Basin. In contrast, the average directions of the PP, RR, and PR point to the south, with angles of 176.8°, 181.6°, and 191.3°, respectively. This is mainly because most of the connections in the PP, RR, and PR are mainly oriented to the east (between 45° and 135°) or the west (between 225° and 315°), accounting for 67.1%, 73.9%, and 74.4%, respectively. Therefore, the average of these two relative directions is approximately 180°. On the other hand, the directions of the PS are 221.6° and the RS are 209.0°. The eastward (westward) direction accounted for 17.2% (49.4%) and 17.3% (59.1%) of PS and RS, respectively. This suggests that runoff from the eastern part of the Yangtze River Basin may significantly affect precipitation in the western part of the basin, and vice versa. This phenomenon may be caused by the complex interaction between the atmospheric circulation pattern, topographic characteristics, and surface features within the Yangtze River Basin, which requires more in-depth investigation and rigorous exploration.

[0086] Figure 2 Shown Figure 1 Details of the propagation distances for the links shown in . Boxplot analysis shows the variance of propagation distances between different flood pairs ( Figure 2 a). The median propagation distances are ranked in descending order as follows: PP (610 km) > PS (583 km) > RS (495 km) > PR (465 km) > RR (381 km) > SS (248 km). Individual paths associated with precipitation tend to show longer propagation distances compared to floods dominated by runoff and water flow. This result suggests that the synchronization of precipitation is more significantly affected by large-scale atmospheric phenomena, which have a wider range of effects. In contrast, individual links in runoff and water flow-related networks tend to be shorter, reflecting a greater sensitivity to regional and local scale influences.

[0087] It is worth noting that the cumulative sum ranking based on the total propagation distance ( Figure 2 b) shows the same ranking based on median propagation distance ( Figure 2a) Different patterns. The total propagation distance is ranked in descending order as follows: RS (73,254 km) > PS (58,365 km) > SS (47,178 km) > PP (45,827 km) > PR (41,488 km) > RR (29,464 km). In particular, relatively abundant (110) and long (median 495 km) links contribute to the cumulative distance of the RS network, bringing its total distance to 73,254 km, ranking first. Although the SS network has the highest density of 147 links, they are relatively short (median 248 km), resulting in its total being close to the third place at 47,178 km. It is worth noting that the RR network has the smallest cumulative sum (29,464 km), with sparse links (65) and the second shortest propagation distance of 381 km. The results show that during flood propagation, the extent of flood impact is collectively affected by the distance of the propagation path and its connectivity. Long propagation paths and high connectivity are contributing factors to the widespread spread of floods.

[0088] In addition, the results show that networks associated with water flow tend to have a larger total cumulative distance compared to networks without water flow variables. This indicates that not only floods (SS) propagated by water flow can affect a wide spatial range, but also runoff (RS) and precipitation (PS) can be propagated to considerable distances through water flow. However, the interconnectivity of runoff (RR) between different regions is relatively low, resulting in poor interconnectivity and making it less susceptible to cascading effects on a larger scale. This highlights the importance of water system management for flood management, as its impact is not limited to the local area, but also has far-reaching consequences on a larger scale.

[0089] Figure 3 Box plot statistics of propagation distance and cumulative sum of propagation distance corresponding to six different flood variables are shown. Figure 3 The box plot of a shows Figure 1 Median, quartiles, and outliers of distances for network links are shown in . The distance of each link is determined by the linear distance between the centroids of the two connected sub-basins, which represents their geographic center. Figure 3 The cumulative sum of the b propagation distances is obtained by adding the distances of all network links one by one.

[0090] 2. Flood network cluster analysis:

[0091] Understanding the characteristics of cohesive clusters and critical transmission nodes is essential for effective watershed management and flood mitigation. Figure 3 Network clusters with different degrees of connectivity within and between sub-basins for different flood variables of precipitation, runoff, and streamflow are shown. Figure 3The statistics of the network clusters in are detailed in Table 2 , including five aspects: the number of clusters, the number of isolated sub-basins, the maximum cluster size, the average cluster size, and the most connected sub-basins.

[0092] The number of clusters varied among the networks, ranging from 5 to 13, with the same-variable networks (RR: 13; PP: 11; SS: 9) having a higher number of clusters compared to the different-variable networks (PR: 7; PS: 7; RS: 5). The higher number of clusters in the same-variable networks may be due to the higher similarity and connectivity of the same flood variables. It should be noted that not every sub-basin belongs to a cluster; the number of isolated sub-basins varies among different flood pairings. Specifically, the networks involving water flow have the lowest number of isolated sub-basins, including the RS (38 isolated sub-basins, 30.4%), SS (42 isolated sub-basins, 33.6%), and PS (45 isolated sub-basins, 36.0%) networks. The lower number of isolated sub-basins indicates that more sub-basins are interconnected, suggesting that at the macroscale in the Yangtze River Basin, the connectivity level of water flow floods is higher than that of precipitation and runoff floods.

[0093] according to Figure 3 It can be seen that there are large-scale and small-scale clusters for different flood variable pairs. The size of the largest cluster ranges from 20 to 75, and among the involved networks, the networks related to river flow have larger maximum cluster sizes compared with other networks. Specifically, the sizes of the largest clusters are 75, 68, and 55 in the RS, PS, and SS networks, respectively. This suggests that there are interconnected cohesive sub-basins in river flow that may form super clusters, which may play a key role in regulating hydrological processes and promoting rapid water flow. In addition, the average cluster size shows an opposite pattern to the cluster count, that is, the average cluster size of the within-variable network (RR: 5.7; PP: 6.6; SS: 9.2) is smaller than that of the between-variable network (PR: 10.1; PS: 11.4; RS: 17.4). This is largely because the total number of sub-basins is distributed in more groups in the within-variable network, resulting in a smaller average size of each group.

[0094] In addition, identifying sub-basins with the most connections highlights important sub-basins that serve as key pathways for water movement in the hydrological network. For example, in the SS network, sub-basin 61 stands out with 16 connections (Table 2), indicating that sub-basin 61 is a potential hydrological hub in the Yangtze River Basin hydrological network and requires special attention in flood control management efforts. Overall, the network indicators of cluster statistics provide insights into the hydrological connectivity patterns of the Yangtze River Basin and their impact on water-related risks.

[0095] Table 2 shows Figure 3Statistics for the six network clusters shown in

[0096]

[0097] 3. Analyzing the synchronization intensity of synchronous floods provides additional insights into the spatial distribution and extent of flood regulation.

[0098] Figure 4 The out-degree, in-degree, combined out-degree, combined in-degree, and total degree of synchronous flood are shown.

[0099] Figure 4 Describes the outdegree of synchronous floods in the Yangtze River Basin. Subbasins with high outdegree are more likely to act as flood spreading areas. The spatial distribution of outdegree reveals some interesting patterns across the Yangtze River Basin. Figure 4 a) and runoff synchronization ( Figure 4 The areas with high outflow in b) appear in the upper sub-basins of the Yangtze River Basin, which indicates that these areas serve as water vapor convergence areas and hydrological response initiators. RS synchronous floods occur intensively in the upper and middle reaches ( Figure 4 c), which indicates that these areas are key areas of high risk flood synchronization. The sub-basins with high out-degree in the RS diagram ( Figure 4 f) represent areas that are more susceptible to changes in surface runoff, which have the ability to alter river flows and potentially worsen downstream flood events. Figure 4 c), RS( Figure 4 f), PS( Figure 4 It can also be observed in Figure e) that flood synchronization proceeds from upstream to downstream, which is as expected given the importance of the physical river network in organizing flow patterns.

[0100] Figure 5 The in-degree of synchronous floods in the Yangtze River Basin is depicted. A sub-basin with high in-degree may be affected by other sub-basins during a flood event. In-degree shows different characteristics compared to out-degree. In SS( Figure 5 c) and RS( Figure 5 f) In the network, sub-basins with high in-degree tend to be located in the main area of ​​the Yangtze River Basin, in contrast to sub-basins with high out-degree, which tend to be located on tributaries ( Figure 4 This is due to the cumulative effect of flows from multiple tributaries entering the main channel ( Figure 1 ). In addition, relative to the out-degree pattern of the same network, PR( Figure 5 d) and PS( Figure 5 e) The indegree pattern of the network shifts from downstream to upstream positions, emphasizing that upstream sub-basins are more likely to regulate the runoff and hydrological response of the basin to precipitation dynamics.

[0101] Figure 6 The comprehensive out-degree, in-degree, and total degree of synchronous floods are summarized. The comprehensive out-degree and comprehensive in-degree are determined by summing the out-degree and in-degree of all six flood variable pairs, respectively. The comprehensive total degree is the sum of the comprehensive in-degree and the comprehensive out-degree. There is a clear difference between the out-degree and in-degree. The out-degree tends to be more dispersed, while the in-degree tends to be more concentrated along the main river channel. Sub-basin 61 has the highest out-degree of 15, and sub-basins 8 and 88 have the highest in-degree, both of which are 17. The top 11 sub-basins with the highest comprehensive degrees are 61 (29), 88 (23), 8 (22), 21 (18), 108 (17), 13 (17), 117 (16), 68 (16), 52 (16), 37 (16), and 25 (16). The locations of the sub-basins with the highest comprehensive degrees are shown in Figure 6 c. Based on the comprehensiveness, important flood hubs tend to be concentrated in the central and eastern parts of the Yangtze River Basin. A deeper analysis of the geographical location and characteristics of the sub-basins with the highest comprehensiveness can provide insights into potential flood-vulnerable areas and identify areas that require targeted risk mitigation measures. However, the scope of this issue is beyond the scope of this paper.

[0102] Figure 6 Composite outdegree, indegree, and total degree of synchronous floods. For each sub-catchment, the composite outdegree and composite indegree are calculated by summing the outdegree and indegree of all six groups of flood variables, respectively. The total degree of a sub-catchment is the sum of its indegree and outdegree. The sub-catchment with the highest total degree is labeled.

[0103] The present invention investigates the teleconnection patterns of synchronous propagation of floods in cross-level hydrological domains (atmosphere, land surface, and river networks) in the Yangtze River Basin of China. Daily precipitation, runoff, and flow data from 125 sub-basins from the SWAT model system were used during 1961-2020. The POT method was used to obtain extreme values ​​from the time series of precipitation, runoff, and flow as flood proxies, respectively. Complex network theory and event synchronization were used to characterize the existence of flood cascades between sub-basins. Six different networks (i.e., PP, RR, SS, PR, PS, and RS) were constructed, and three sets of network measures, including propagation characteristics, network clusters, and degree strength, were applied to the flood networks. Based on the results, the following conclusions are emphasized:

[0104] The SS network stands out among all six networks, with the highest number of links (147), the largest average riverine step difference (0.61), the shortest average propagation distance (248 km), and the only eastward propagation direction (104.2°). These features together indicate that teleconnected river floods usually propagate on dense and interconnected pathways along major waterways. The strong connectivity of river floods has a high potential to cause widespread cascading disasters. Therefore, SS cascade floods should be of greatest concern in flood management.

[0105] Compared to the networks dominated by precipitation and runoff (PP and RR), the flood networks involving flow have more links, larger river level differences, longer total propagation distances, fewer isolated sub-basins, and larger maximum cluster sizes. This suggests that not only floods transmitted by streams (SS) can affect a wide spatial range, but also precipitation (PS) and runoff (RS) can be propagated by water transport over considerable distances. Therefore, precipitation and runoff are important factors that widely contribute to large-scale river floods.

[0106] In addition, compared with the networks dominated by flow and runoff, the precipitation-related networks (PP, PR, and PS) tend to be composed of longer respective links. This result reveals that the long-range connections of precipitation are more influenced by large-scale atmospheric phenomena rather than local characteristics. In contrast, the runoff-dominated network (RR) showed the lowest number of links, at 65, indicating greater susceptibility to regional and local scale influences. The analysis showed that long propagation paths and substantial connectivity are key factors in causing widespread spread of cascading floods.

[0107] Surprisingly, it was observed that runoff floods in the lower Yangtze River basin significantly influenced precipitation floods in the upper basin, as indicated by the river order difference of -0.89 and 17.2% (49.4%) in the eastward (westward) direction of the PR network. This finding suggests the existence of a significant large-scale propagation mechanism that significantly mediates the cascading floods between runoff and precipitation within the Yangtze River basin. More comprehensive investigations and rigorous explorations are needed to fully understand and appreciate this phenomenon.

[0108] The intensity of flood synchronicity shows the spatial dynamics of cascading floods. Sub-basins with high outflow degree, mainly located in the upper Yangtze River Basin, play a role in flood diffusion, while sub-basins with high inflow degree, especially along the main river reach, accumulate the impacts from multiple tributaries. Sub-basins 61, 88, 8, 21, 108, 13, 117, 68, 52, 37, and 25, as important flood hubs, tend to be concentrated in the central and eastern parts of the Yangtze River Basin. The differences in spatial patterns highlight the need to consider local and regional factors in cascading flood management.

[0109] Based on the above method, the present invention provides a flood cascade teleconnection feature mining system, including: a collection module, a model building module and an indicator acquisition module.

[0110] Among them, the acquisition module is used to obtain extreme flood events that are continuously propagated in secondary basins through precipitation, runoff and flow hydrological variables, and to construct cascading floods through extreme flood events; the model construction module is used to construct multiple directed complex networks for river basins connected in the incoming or outgoing direction, and input the time information of cascading floods into the directed complex network, analyze the out-degree and in-degree of single variable pairs and cross-variable pairs in the cascading floods, and when analyzing cross-variable pairs, perform flood synchronization on precipitation and runoff, precipitation and flow, and runoff and flow, and characterize the flood cascade phenomenon in the river basin through flood synchronization; the indicator acquisition module is based on the directed complex network, extracts network indicators about flood propagation characteristics, network clustering and synchronization intensity in the flood cascade phenomenon, and uses the network indicators as the teleconnection characteristics of the flood cascade.

[0111] The present invention also provides a computer device, including a memory and a processor. The memory stores a program. When the program is executed by the processor, the processor executes the steps of a method for mining remote correlation characteristics of flood cascades.

[0112] In accordance with the disclosed embodiments, a computing device may communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth communications, etc.), or with any device (e.g., routers, modems, etc.) that enables a computing device to communicate with one or more other computing devices.

[0113] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for mining remote correlation characteristics of flood cascades are implemented.

[0114] According to the disclosed embodiments, the storage medium may be a non-volatile computer-readable storage medium, such as but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0115] The above content is a further detailed description of the present invention in combination with a specific preferred embodiment. For technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for mining teleconnection features of flood cascades, characterized in that: include: Obtaining extreme flood events that propagate continuously in sub-basins through precipitation, runoff and flow hydrological variables, and constructing cascading floods through said extreme flood events; Multiple directed complex networks are constructed for river basins connected in the incoming or outgoing direction, and the temporal information of cascading floods is input into the directed complex networks. The out-degree and in-degree of single variable pairs and cross-variable pairs in cascading floods are analyzed. When analyzing cross-variable pairs, flood synchronization is performed between precipitation and runoff, precipitation and flow, and runoff and flow. The flood cascade phenomenon in the river basin is characterized by flood synchronization. Based on the directed complex network, network indicators about flood propagation characteristics, network clustering and synchronization strength in flood cascade phenomena are extracted, and the network indicators are used as flood cascade teleconnection characteristics.

2. A method for mining remote correlation features of flood cascades according to claim 1, characterized in that: Cascading floods are constructed through the extreme flood events, wherein the pairing modes of cascading floods include precipitation-precipitation PP, runoff-runoff RR, flow-flow SS, precipitation-runoff PR, precipitation-flow PS and runoff-flow RS.

3. A method for mining remote correlation features of flood cascades as claimed in claim 2, characterized in that: The flood cascade phenomenon in the river basin characterized by flood synchronization includes: Get precipitation - the time when the precipitation PP floods in sub-basin i And the time when the flood occurs in sub-basin j is obtained as If the time lag between two floods occurring in sub-basins i and j is If the floods are within , the two floods are considered synchronous floods, specifically: PP_c(i|j) is defined as the number of times a flood appears at location i after it appears at location j, specifically: in: Quantify synchronization and delay, the specific expression is: Among them, PP_Q ij It is used to measure the intensity of synchronous floods between sub-basins i and j, while PP_q ij Used to measure the latency behavior between them.

4. A method for mining remote correlation features of flood cascades as claimed in claim 3, characterized in that: The river basins connected in the incoming or outgoing direction construct multiple directed complex networks, including: Based on PP_Q ij and PP_q ij Construct a directed complex network and transform precipitation-precipitation PP, runoff-runoff RR, flow-flow SS, precipitation-runoff PR, precipitation-flow PS and runoff-flow RS to obtain PP_Q ij and PP_q ij Six directed complex networks were constructed.

5. A method for mining remote correlation features of flood cascades as claimed in claim 4, characterized in that: The PP_Q ij and PP_q ij When building a directed complex network, include: PP_Q ij Convert a symmetric binary matrix into an adjacency matrix Specifically: pass Determine the direction of the link. It is defined as the adjacency matrix, specifically:

6. A method for mining remote correlation features of flood cascades according to claim 1, characterized in that: The network indicators for extracting flood propagation characteristics, network clustering and synchronization strength include: When extracting flood propagation characteristics, the level difference data of the river are obtained according to the Strahler method, the river level is given, the statistical data of the propagation direction is obtained according to the direction of all lines from the centroid of the starting basin to the centroid of the ending basin, and the synchronous distance of propagation is determined according to the median of the geographical distance between the connected sub-basins; When clustering the network, sub-basins that are directly or indirectly connected are grouped into the same cluster, and five network metrics are obtained from the clustering: the number of clusters, the number of isolated sub-basins, the largest cluster size, the average cluster size, and the sub-basins with the most connections; When obtaining the synchronization strength, the total degree is obtained by summing the in-degree and out-degree of the sub-basin, and the total degree is used to obtain the overall number of connections between the sub-basin and other sub-basins in the network.

7. A remote correlation feature mining system for flood cascades, characterized in that: include: A collection module for acquiring extreme flood events that are continuously propagated in the sub-basin through precipitation, runoff and flow hydrological variables, and constructing cascading floods through the extreme flood events; A model building module is used to construct multiple directed complex networks for river basins connected in the incoming or outgoing direction, and input the time information of cascading floods into the directed complex networks, analyze the out-degree and in-degree of single variable pairs and cross-variable pairs in the cascading floods, and when analyzing cross-variable pairs, perform flood synchronization between precipitation and runoff, precipitation and flow, and runoff and flow, and characterize the flood cascade phenomenon in the river basin through flood synchronization; The indicator acquisition module is used to extract network indicators about flood propagation characteristics, network clustering and synchronization strength in flood cascade phenomena based on the directed complex network, and use the network indicators as flood cascade teleconnection characteristics.

8. A computer device, characterized in that: It comprises a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a method for mining remote correlation characteristics of a flood cascade as claimed in any one of claims 1 to 6.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for mining remote correlation characteristics of flood cascades according to any one of claims 1 to 6 are implemented.

Images