A method for analyzing the spatial transmission effect of watershed flood control resilience

Abstract

This invention provides a method for analyzing the spatial transmission effect of flood control resilience in watersheds, involving the interdisciplinary fields of watershed flood control analysis, spatial econometric analysis, and water resources management. The method includes: obtaining a set of indicators to be used based on multi-source basic spatial data; acquiring the flood control resilience influencing factor vector and flood control resilience value for each administrative unit; correlating and mapping the flood control resilience influencing factor vector and flood control resilience value at the administrative unit scale to obtain the flood control resilience influencing factor vector and flood control resilience characterization value for each sub-watershed; constructing a spatial weight matrix and a spatial econometric model to identify target sub-watersheds with spatial transmission effects; and characterizing the impact of spatial transmission effects at the administrative unit scale based on the model parameter estimation results for the target sub-watersheds and all sub-watersheds. This invention can clearly identify the main driving factors of flood control resilience within a watershed and their spatial transmission paths.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of watershed flood control analysis, spatial quantitative analysis and water resources management, and in particular to a method for analyzing the spatial transmission effect of watershed flood control resilience. Background Technology

[0002] Flood resilience is a crucial indicator for measuring a region's ability to resist, absorb, recover from, and adapt to flood disturbances, and it is a key research area in watershed flood control, disaster reduction, and water resource management. In watershed flood resilience research, the research paradigm is gradually shifting from static indicator evaluation to a comprehensive analysis integrating "evaluation-transmission." This shift means not only the need to propose scientifically sound evaluation indicators but also to deeply reveal the spatial transmission characteristics and influencing mechanisms of flood resilience. Currently, multi-scale nested analysis frameworks are frequently used to analyze the spatial transmission effects of watershed flood resilience. This framework allows the construction scale of indicators to be separated from the construction scale of spatial weight matrices and spatial econometric models, thus overcoming the inherent limitations of single analytical scales. For example, using only watershed-scale indicators can mask the significant differences in socio-economic and natural conditions within the region, leading to indicator distortion; conversely, using only regional scales such as administrative units often results in spatial weight matrices and spatial econometric models that fail to accurately characterize resilience transmission paths based on hydrological processes.

[0003] However, the multi-scale nested analysis framework still faces significant challenges in practical applications, primarily due to substantial deficiencies in the handling of the connection between analyses at different scales. Firstly, methods that rely on traditional geographical adjacency or Euclidean distance weighting for data transformation have clear limitations. These methods ignore the directionality of flood propagation along the hydrological network, making it difficult to accurately depict the true transmission path of flood resilience between upstream and downstream areas. Secondly, while using descriptive statistical indicators to bridge scale differences can summarize basin characteristics, as derivative and generalized statistics, the underlying driving mechanisms are obscured. Directly introducing such indicators as independent or dependent variables into spatial econometric models easily leads to results that remain merely at the correlation level, failing to clearly identify the specific driving factors and pathways of flood resilience formation and its spatial transmission, thus reducing the guiding significance of the analysis results for basin flood control and zonal management. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for identifying the spatial transmission effect of watershed flood control resilience. This method solves the problem that existing technologies, when analyzing the spatial transmission characteristics and influencing mechanisms of flood control resilience using a multi-scale nested analysis framework, only output overall conclusions due to the different analysis scales used at different levels, ignoring the spatial heterogeneity within the watershed, and thus failing to clearly identify the fundamental path affecting the formation and transmission of resilience.

[0005] According to an embodiment of the present invention, a method for analyzing the spatial transmission effect of watershed flood control resilience is proposed, comprising:

[0006] Acquire the hydrological network structure and elevation data of the target watershed, divide the target watershed into hydrological zones, and identify the hydrological connectivity between sub-watersheds;

[0007] The target watershed is divided into multiple administrative units, and a set of indicators to be used is constructed and selected based on multi-source basic spatial data to generate a vector of flood control resilience influencing factors for each administrative unit and calculate the flood control resilience value for each administrative unit; wherein, the target watershed includes multiple sub-watersheds, and each administrative unit includes at least one sub-watershed;

[0008] Based on the spatial superposition relationship between administrative units and sub-basins, the flood control resilience influencing factor vector and flood control resilience value at the administrative unit scale are associated and mapped to each sub-basin included in the administrative unit, so as to obtain the flood control resilience influencing factor vector and flood control resilience characterization value of each sub-basin;

[0009] Based on the hydrological connectivity between the sub-basins, a spatial weight matrix based on hydrological connectivity is constructed, and a spatial econometric model is built on this basis. The flood resilience characterization value of each sub-basin is used as the dependent variable, and the flood resilience influencing factor vector after scale mapping is used as the explanatory variable. The model parameters are estimated and significance is tested to identify the target sub-basins with spatial transmission effects.

[0010] Based on the model parameter estimation results of the target sub-basin and all sub-basins, scale transformation, effect decomposition and action path identification are performed to characterize the impact of spatial transmission effects at the administrative unit scale.

[0011] Optionally, the step of constructing and filtering the set of indicators to be used based on multi-source basic spatial data includes:

[0012] Construct an initial set of candidate indicators for the target watershed;

[0013] The initial candidate indicator set includes at least land use and cover data and socio-economic statistics data that are spatially integrated according to administrative units based on registration, and preset terrain data that are overlaid according to administrative units based on registration; wherein, the multi-source basic spatial data includes at least the land use and cover data, the socio-economic statistics data, and the preset terrain data;

[0014] The initial candidate index set is subjected to significance screening to remove indices that have no significant impact on flood control resilience; based on the results of significance screening, importance screening is performed to identify indices that contribute significantly to flood control resilience; and based on the results of importance screening, stability screening is performed to obtain indices that function stably in different spatial units, thereby obtaining the set of indicators to be used for subsequent flood control resilience evaluation and spatial transmission effect analysis.

[0015] Optionally, the saliency screening of the initial candidate indicator set includes:

[0016] For each administrative unit n, calculate:

[0017] ;

[0018] in, This represents the intensity of economic loss per unit area due to flooding in the nth administrative unit. This represents the direct economic loss caused by flooding in the nth administrative unit. This represents the area of ​​the nth administrative unit;

[0019] Calculate the overall variance σ based on the unit area flood economic loss intensity of N administrative units. 2 Each indicator in the initial candidate indicator set is classified into levels, and each indicator X after stratification is further classified. p Based on the classification results, administrative units are divided into several levels, and the overall variance of the economic loss intensity per unit area due to flooding within each level is calculated. and the corresponding number of administrative units ;

[0020] Calculate the p-th index X of the initial candidate index set. p Statistical explanatory power for the spatial differentiation of economic loss intensity per unit area caused by flooding The formula is:

[0021] ;

[0022] in, This represents the overall variance of the intensity of economic losses per unit area caused by flooding. This represents the overall variance of the intensity of economic losses per unit area from flooding within the h-th level. H represents the number of administrative units included in the h-th level for the p-th indicator. p This represents the number of hierarchical levels formed after the p-th indicator is processed.

[0023] Based on the explanatory power statistic, a significance test is performed on all indicators in the initial candidate indicator set, and L indicators with significance test results greater than a preset significance threshold are retained as the results of the significance-based screening, where p is a positive integer less than or equal to P, P is the number of indicators in the initial candidate indicator set, L is less than P, and h is less than or equal to H. p Positive integers.

[0024] Optionally, the importance screening based on the results of saliency screening includes:

[0025] The result X based on the significance screening l Using the intensity of economic loss per unit area from flooding as the input variable and the intensity of economic loss per unit area from flooding as the output variable, a random forest model is constructed.

[0026] Calculate the change in prediction error I of the random forest model. l The formula is:

[0027] ;

[0028] in, This represents the number of decision trees in the random forest model. This represents the original prediction error of the t-th decision tree on the out-of-bag samples; X represents the prediction error corresponding to random perturbations of the input variable. l The l-th indicator representing the significance screening result, the change in prediction error I l X represents the l-th indicator. l The extent to which the predicted intensity of economic losses per unit area from flooding is affected.

[0029] The importance of the L indicators is normalized and sorted in descending order of importance. The top M indicators in the sorting results are taken as the importance-based screening results, where L is a positive integer less than P, P is the number of indicators in the initial candidate indicator set, l is a positive integer less than or equal to L, and t is a positive integer less than or equal to T.

[0030] Optionally, the stability screening based on the importance screening results includes:

[0031] The result X of the importance-based screening m As input, a predictive model is constructed to predict the intensity of economic losses per unit area from flooding, and the additive contribution of each indicator to the prediction result is calculated based on the SHAP interpretation method. The additive contribution satisfies the following relationship:

[0032] ;

[0033] Where f(X) represents the predicted value of the predictive model for the dependent variable, which is the intensity of economic loss per unit area due to flooding. As the baseline output of the prediction model, The m-th indicator X represents the result of the importance screening. m SHAP contribution value;

[0034] Calculate the statistical characteristics of the SHAP contribution values ​​of M indicators to characterize the spatial stability of the influence of each indicator.

[0035] Based on the statistical characteristics, stability screening is performed on the M indicators, and K indicators whose SHAP contribution value dispersion is lower than the preset stability threshold are selected as the stability screening results. The K indicators constitute the set of indicators to be used, where M is a positive integer less than L, M is the number of indicators in the results of the importance-based screening, L is the number of indicators in the results of the significance-based screening, m is a positive integer less than or equal to M, and K is a positive integer less than or equal to M.

[0036] Optionally, the calculation of the flood resilience value of each administrative unit includes:

[0037] Based on the correlation between the values ​​of K indicators and the SHAP contribution value of each indicator, the direction of influence of each indicator on the dependent variable is obtained, and the K indicators are divided into positive indicators and negative indicators. The dependent variable is the intensity of economic loss from flooding per unit area.

[0038] The K indicators are positively oriented to unify the influence direction of each indicator on the flood resilience evaluation results. Based on this, the indicators after unifying the influence direction are dimensionless to obtain a standardized indicator matrix, as shown in the formula:

[0039] Z= ;

[0040] Among them, Z kn This represents the standard value of the k-th indicator in the n-th administrative unit, where K represents the number of indicators in the set of indicators to be used obtained from the initial candidate indicator set in the n-th administrative unit, and N represents the total number of administrative units.

[0041] Based on the preset objective weight determination rules and subjective weight determination rules, the comprehensive weight ω of each indicator is obtained. k The standardized index matrix is ​​then weighted and summed to obtain the flood resilience value R of the nth administrative unit. n The formula is:

[0042] ;

[0043] Where, ω kThis represents the overall weight of the k-th indicator, where k is a positive integer less than or equal to K, and n is a positive integer less than or equal to N.

[0044] Optionally, based on the spatial overlay relationship between administrative units and sub-basins, the flood resilience value at the administrative unit scale is associated with and mapped to each sub-basin included in the administrative unit, serving as the flood resilience characterization value for the corresponding sub-basin, including:

[0045] Determine the spatial overlap relationship between each sub-basin and the administrative unit;

[0046] If the sub-basin is located in at least two administrative units, the flood resilience values ​​of the at least two administrative units are weighted and summed according to the area ratio of the sub-basin in each administrative unit to obtain the flood resilience characterization value of the sub-basin.

[0047] Optionally, the construction of a spatial weight matrix based on hydrological connectivity based on the hydrological connectivity relationships between the sub-basins includes:

[0048] Extract weighted terms to characterize the spatial transmission characteristics of floods, wherein the weighted terms include at least flood generation capacity index, rainfall characteristics, flood propagation distance, energy attenuation characteristics, and river channel structure difference correction coefficient;

[0049] Based on the aforementioned weighting term, the asymmetric spatial weight between any two sub-basins is calculated. This spatial weight represents the intensity of the flood's influence along the hydrological network from the upstream sub-basin to the downstream sub-basin, and its calculation formula is as follows:

[0050] ;

[0051] in, Sub-basin The flood impact weight of sub-basin j Sub-basin Rainfall, Indicators representing flood generation capacity, The distance of flood propagation is represented by γ, which is equal to the hydraulic path distance from sub-basin i to sub-basin j calculated in GIS. γ is the distance attenuation coefficient, used to represent the energy attenuation characteristics of the flood during its propagation. This represents the watershed difference correction coefficient;

[0052] The spatial weight matrix W is obtained based on the asymmetric spatial weights of all sub-basins, W= , i and j represent different sub-basins.

[0053] Optionally, the spatial econometric model is constructed based on the spatial weight matrix, and the spatial econometric model is as follows:

[0054] ;

[0055] in, R represents the flood resilience characterization value of sub-basin i, R=( ) T , represents the column vector consisting of the flood resilience characterization values ​​of all sub-basins within the target watershed, and α represents the spatial transmission coefficient; Let X represent the spatial weight matrix, X represent the flood resilience influencing factor vector of the sub-basin, which is obtained by spatial mapping from the flood resilience influencing factor vector at the administrative unit scale; β represent the direct effect coefficient of the explanatory variable, θ represent the effect coefficient of the spatially lagged explanatory variable, and ɛ represent the random disturbance term.

[0056] Optionally, based on the model parameter estimation results of the target sub-basin and all sub-basins, scale transformation, effect decomposition, and action path identification are performed to characterize the impact of spatial transmission effects at the administrative unit scale, including:

[0057] Based on the area proportion of the sub-basin within its administrative unit, the flood resilience characterization value at the sub-basin scale and its spatial transmission effect are weighted and summarized to achieve scale conversion from the sub-basin scale to the administrative unit scale, thereby obtaining the flood resilience impact effect characterization result of the administrative unit.

[0058] Based on the model parameter estimation results of the target sub-basin, the effects of the flood control resilience impact characterization results are divided into direct effects, indirect effects, and total effects;

[0059] Based on the spatial weight matrix and the model parameter estimation results of all sub-basins, the significant transmission direction and intensity of the flood control resilience influencing factor vector of each sub-basin between upstream and downstream sub-basins are identified, and key sub-basin nodes in the hydrological network structure are selected accordingly.

[0060] Using the key sub-basin nodes as anchor points, the complete spatial path of the flood control resilience influencing factor vector of each sub-basin is output along the hydrological network structure, and all upstream related sub-basins that have an impact on the target sub-basin and the transmission links including the target sub-basin and the related sub-basins are extracted from it;

[0061] Based on the aforementioned transmission link, the directionality, asymmetry, and attenuation law of the flood control resilience effect characterization results between the target sub-basin and related sub-basins are analyzed.

[0062] Compared to existing technologies, this invention offers the following advantages: It constructs an asymmetric spatial weight matrix and a spatial econometric model based on the hydrological connectivity between all sub-basins within the target watershed, using this as the core physical constraint. Then, it integrates multi-source basic spatial data to construct and select a set of indicators to obtain the flood resilience influencing factor vector and flood resilience value for each administrative unit. The flood resilience value of each administrative unit is then associated with each sub-basin included within that unit, serving as the flood resilience characterization value for each sub-basin. This characterization value is then used as the foundational data driving the spatial econometric model, accurately identifying whether spatial transmission effects exist within the sub-basins of the administrative unit and obtaining the impact of these effects at the administrative unit scale. In this process, the scale connection between the spatial weight matrix and the spatial econometric model—two basic spatial carriers at different analytical scales—is achieved, enabling the mapping and transmission analysis of flood resilience evaluation results, including but not limited to flood resilience values ​​and flood influencing factor vectors, across different spatial scales. Compared to traditional methods that rely on geographical adjacency or Euclidean distance weights, this invention introduces a spatial weighting method constrained by hydrological connectivity in the analysis of spatial transmission effects. This method can more fully characterize the spatial dependencies between sub-basins, thereby improving the interpretability of the results of the spatial transmission analysis of flood control resilience. It overcomes the limitations of insufficient characterization of hydrological processes and provides a more scientific, precise, and operable decision-making basis for basin flood control planning, engineering layout optimization, and cross-regional collaborative governance. Attached Figure Description

[0063] Figure 1 This is a schematic diagram illustrating the implementation process of the spatial transmission effect analysis method for watershed flood control resilience according to an embodiment of the present invention. Detailed Implementation

[0064] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0065] like Figure 1 As shown in the figure, this invention proposes a method for analyzing the spatial transmission effect of watershed flood control resilience, including but not limited to the following steps:

[0066] S101. Obtain the hydrological network structure and elevation data of the target watershed, divide the target watershed into hydrological zones, and identify the hydrological connectivity between sub-watersheds.

[0067] S102. Divide the target watershed into multiple administrative units, and construct and select a set of indicators to be used based on multi-source basic spatial data to generate a vector of flood control resilience influencing factors for each administrative unit and calculate the flood control resilience value for each administrative unit; wherein, the target watershed includes multiple sub-watersheds, and each administrative unit includes at least one sub-watershed.

[0068] S103. Based on the spatial superposition relationship between administrative units and sub-basins, the flood control resilience influencing factor vector and flood control resilience value at the administrative unit scale are associated and mapped to each sub-basin included in the administrative unit, so as to obtain the flood control resilience influencing factor vector and flood control resilience characterization value of each sub-basin.

[0069] S104. Based on the hydrological connectivity between the sub-basins, construct a spatial weight matrix based on hydrological connectivity, and on this basis, construct a spatial econometric model. The flood resilience characterization value of each sub-basin is the explained variable, and the flood resilience influencing factor vector after scale mapping is the explained variable. Perform model parameter estimation and significance test to identify the target sub-basins with spatial transmission effects.

[0070] S105. Based on the model parameter estimation results of the target sub-basin and all sub-basins, scale transformation, effect decomposition and action path identification are performed to characterize the impact of spatial transmission effects at the administrative unit scale.

[0071] In step S101 above, the hydrological network structure and digital elevation model can be used to identify sub-basins within the target watershed and to identify hydrological connectivity, such as upstream and downstream relationships and hydrological connection paths between sub-basins, providing a data foundation for the construction of the spatial weight matrix and spatial econometric model. Furthermore, in step S104 above, a spatial weight matrix is ​​constructed based on the upstream and downstream relationships and hydrological connection paths between sub-basins identified in step S101. The spatial weight matrix is ​​used to characterize the spatial influence relationships between sub-basins based on hydrological connectivity, and its weight values ​​vary with hydrological paths, propagation distances, and watershed characteristics.

[0072] For example, the above step S101 can be implemented as follows: obtain the hydrological network structure and elevation data of the target analysis area, use hydrological analysis tools to extract the natural river network and automatically divide it into sub-basins, and these sub-basins are the basic spatial carriers of the spatial weight matrix and spatial econometric model in the embodiments of the present invention.

[0073] Regarding step S102 above, this embodiment of the invention uses administrative units as the basic spatial carrier to construct and screen a set of indicators to be used. The basic data for construction and screening is multivariate basic spatial data, and the screening is divided into three levels to ensure the accuracy and rationality of each indicator in the set of indicators for flood control resilience evaluation. The implementation method is as follows:

[0074] Construct an initial set of candidate indicators for the target watershed;

[0075] The initial candidate index set includes at least land use and cover data and socio-economic statistics that are spatially integrated according to administrative units based on registration, and preset topographic data that are overlaid according to administrative units based on registration.

[0076] The initial candidate indicator set is subjected to significance screening, then importance screening, and finally stability screening to obtain the indicator set to be used.

[0077] It should be noted that multi-source basic spatial data includes at least land use and cover data, socio-economic statistics, and pre-defined topographic data. All of these data are publicly available or obtainable through conventional survey methods. For example, land use and cover data can be information on precipitation, topography, land use, vegetation cover, and river system distribution obtained through remote sensing imagery products, digital elevation models, or conventional meteorological and hydrological observation data. Socio-economic statistics can include information on population size, economic output, industrial structure, and construction land use, obtainable through statistical yearbooks, census data, or publicly available data; and information on flood control projects and facilities, such as flood dikes, reservoirs, pumping stations, flood storage areas, and related dispatching rules, obtainable through engineering data, planning documents, or publicly available data from management departments. Pre-defined topographic data can be obtained by spatially interpolating or resampling digital elevation model data and overlaying it with administrative boundaries to extract information such as average elevation, slope distribution characteristics, and the proportion of high-slope areas within each administrative unit, used to characterize the impact of watershed topography on runoff generation and confluence processes.

[0078] It should be noted that the above registration can be achieved through spatial integration or overlay analysis methods such as coordinate system one, spatial resolution matching, and boundary clipping, so as to ensure that the multivariate basic spatial data can be expressed in a standardized manner at the administrative unit scale in the initial candidate indicator set.

[0079] In this embodiment of the invention, the significance screening aims to calculate the statistical explanatory power of the indicators in the initial candidate indicator set for the spatial differentiation of flood losses, thereby determining whether the above indicators have the statistical ability to explain flood differences, and eliminating indicators that do not significantly affect flood resilience. For example, indicator A is "percentage of impermeable ground," indicator B is "standard deviation of elevation," and indicator C is "average annual temperature." Through significance screening, indicators such as indicator C, which are unrelated to the spatial differentiation of flood losses, are excluded, while indicators such as indicators A and B, which can distinguish between high-loss and low-loss areas, are retained.

[0080] In one embodiment, saliency filtering is achieved using a geographic detector, including:

[0081] For each administrative unit n, calculate:

[0082] ;

[0083] in, This represents the intensity of economic loss per unit area due to flooding in the nth administrative unit. This represents the direct economic loss caused by flooding in the nth administrative unit. This represents the area of ​​the nth administrative unit;

[0084] Calculate the overall variance σ based on the unit area flood economic loss intensity of N administrative units. 2 Each indicator in the initial candidate indicator set is classified into levels, and each indicator X after stratification is further classified. p Based on the classification results, administrative units are divided into several levels, and the overall variance of the economic loss intensity per unit area due to flooding within each level is calculated. and the corresponding number of administrative units ;

[0085] Calculate the p-th index X of the initial candidate index set. p Statistical explanatory power for the spatial differentiation of economic loss intensity per unit area caused by flooding The formula is:

[0086] ;

[0087] in, This represents the overall variance of the intensity of economic losses per unit area caused by flooding. This represents the overall variance of the intensity of economic losses per unit area from flooding within the h-th level. H represents the number of administrative units included in the h-th level for the p-th indicator. p This represents the number of hierarchical levels formed after the p-th indicator is processed.

[0088] Based on the explanatory power statistic, a significance test is performed on all indicators in the initial candidate indicator set, and L indicators with significance test results greater than a preset significance threshold are retained as the results of the significance-based screening, where p is a positive integer less than or equal to P, P is the number of indicators in the initial candidate indicator set, L is less than P, and h is less than or equal to H. p Positive integers.

[0089] Based on indicators A1, B1, and C1 in the initial candidate indicator set shown above, indicator D1 is also proposed. Indicator C1 is "average annual rainfall," and the application of the significance screening using the geographic detector is explained. Indicator A1 is "percentage of impervious surface," indicator B1 is "standard deviation of elevation," indicator C1 is "average annual temperature," and indicator D1 is "average annual rainfall." Significance screening is performed using the geographic detector model, calculating the q-statistic for each indicator. For example, with a preset significance threshold of 0.05, the q-value for indicator A1, "percentage of impervious surface," is 0.55. This indicates that indicator A1 can explain 55% of the spatial differentiation of flood losses within the region, distinguishing between high-loss and low-loss areas. That is, administrative units with a high percentage of impervious surface tend to have higher flood losses per unit area, and vice versa. Similarly, for indicator B1, the q-value for "standard deviation of elevation" is 0.42, indicating a significant correlation between areas with large topographic relief and loss distribution; for indicator C1, the q-value for "average annual temperature" is 0.04, indicating that indicator C1 is not related to the spatial differentiation of flood losses; and for indicator D1, "average annual rainfall" has a q-value of 0.38, indicating that basic climatic conditions are also an important factor affecting loss differentiation. Therefore, through the above steps, indicators with statistical power to explain flood differences can be selected, ensuring that subsequent analysis focuses on core driving factors.

[0090] In this embodiment of the invention, the importance screening is also based on the intensity of economic loss per unit area from flooding, but the aim is to quantify the contribution of the above indicators to the accurate prediction of flood losses, in order to identify indicators that contribute more to flood control resilience. For example, when all indicators from the above significance screening results are used together, does a particular indicator still possess irreplaceable value? Taking indicator A2 as reservoir storage capacity, indicator B2 as river flood discharge capacity, and indicator C2 as emergency material reserve density as an example, through importance screening, it can be determined that indicator A2 is more important than indicators B2 and C2 in accurately predicting flood losses.

[0091] In one embodiment, importance filtering is implemented using a random forest model, including:

[0092] The result X based on the significance screening l Using the intensity of economic loss per unit area from flooding as the input variable and the intensity of economic loss per unit area from flooding as the output variable, a random forest model is constructed.

[0093] Calculate the change in prediction error I of the random forest model. l The formula is:

[0094] ;

[0095] in, This represents the number of decision trees in the random forest model. This represents the original prediction error of the t-th decision tree on the out-of-bag samples; X represents the prediction error corresponding to random perturbations of the input variable. l The l-th indicator representing the significance screening result, the change in prediction error I l X represents the l-th indicator. l The degree of impact on the predicted intensity of economic losses per unit area from flooding;

[0096] The importance of the L indicators is normalized and sorted in descending order of importance. The top M indicators in the sorting results are taken as the importance-based screening results, where L is a positive integer less than P, P is the number of indicators in the initial candidate indicator set, l is a positive integer less than or equal to L, and t is a positive integer less than or equal to T.

[0097] Based on the significance-based screening results shown above for indicators A2, B2, and C2, in the constructed random forest prediction model, when these indicators work together, disturbances in indicator A2, "reservoir regulation capacity," lead to a significant increase in model prediction error, resulting in a significantly higher normalized importance score. This reflects that in watersheds with aging flood control infrastructure and significant differences in engineering effectiveness, engineering regulation capacity has a decisive impact on the final loss outcome, and its information cannot be completely replaced by other hydrological or social indicators. In contrast, while indicator B2, "river channel flood discharge capacity," is a basic condition, its importance may be secondary due to the relatively stable natural river conditions or limited room for improvement. Indicator C2, "emergency material reserve density," mainly affects the post-disaster recovery speed, and its contribution to the prediction of direct economic losses is relatively indirect; therefore, its importance ranking may be lower. This screening result confirms that within a specified watershed area, namely traditional agricultural irrigation areas with aging flood control facilities and significant differences in engineering conditions, core engineering indicators have irreplaceable key predictive value.

[0098] It is conceivable that, depending on a different specific context within another river basin, such as a newly developed urban area located in a river delta with a very high urbanization rate and highly hardened surface, indicators reflecting social disaster resilience and emergency response capabilities, such as "coverage rate of emergency shelters," "density of medical resources," and "coverage rate of intelligent early warning systems," usually have more irreplaceable key predictive value.

[0099] In this embodiment of the invention, the stability screening aims to verify the directional consistency and impact stability of the indicators in the importance screening results under the condition of multi-indicator coupling, so as to ensure that the constructed flood control resilience evaluation index system has clear physical meaning and reliable interpretability.

[0100] In one embodiment, stability screening aims to obtain indicators that function stably across different spatial units. This is evaluated and implemented using the output of the SHAP model. Therefore, indicators with poor stability refer to those whose SHAP contribution values ​​exhibit highly discrete or inconsistent characteristics across different administrative units, such as significantly higher variance or coefficient of variation than other indicators. Taking indicator A3 as "percentage of cultivated land area," indicator B3 as "percentage of impermeable ground," and indicator C3 as "road density" as an example, stability screening can identify and eliminate indicators A3 and C3, which have significantly unstable effects.

[0101] The implementation steps of stability screening based on the SHAP model include:

[0102] The result X of the importance-based screening m As input, a predictive model is constructed to predict the intensity of economic losses per unit area from flooding, and the additive contribution of each indicator to the prediction result is calculated based on the SHAP interpretation method. The additive contribution satisfies the following relationship:

[0103] ;

[0104] Where f(X) represents the predicted value of the predictive model for the dependent variable, which is the intensity of economic loss per unit area due to flooding. As the baseline output of the prediction model, The m-th indicator X represents the result of the importance screening. m SHAP contribution value;

[0105] Calculate the statistical characteristics of the SHAP contribution values ​​of M indicators to characterize the spatial stability of the influence of each indicator.

[0106] Based on the statistical characteristics, stability screening is performed on the M indicators, and K indicators whose SHAP contribution value dispersion is lower than the preset stability threshold are selected as the stability screening results. The K indicators constitute the set of indicators to be used, where M is a positive integer less than L, M is the number of indicators in the results of the importance-based screening, L is the number of indicators in the results of the significance-based screening, m is a positive integer less than or equal to M, and K is a positive integer less than or equal to M.

[0107] Based on the importance-based screening results shown above for indicators A3, B3, and C3, SHAP model analysis reveals that the contribution value of indicator B3, "proportion of impervious surfaces," is highly consistent across different administrative units, with a significantly negative mean and a small standard deviation, indicating a clear and stable physical meaning as a vulnerability factor. However, the SHAP value of indicator A3, "proportion of cultivated land area," shows a positive contribution in some counties dominated by plain irrigated agriculture, as cultivated land may play a role in flood storage, while in counties dominated by sloping cultivated land, it shows a negative contribution, potentially exacerbating soil erosion and runoff risks, leading to a chaotic contribution direction and extremely high coefficient of variation. Similarly, the contribution direction of indicator C3, "road density," also shows positive and negative differentiation depending on the degree of improvement of the regional drainage system. Indicators with such high context dependence and inability to form a unified physical explanation, if retained in the system, will severely weaken the reliability and universality of the evaluation results. Therefore, stability screening eliminates such indicators by quantifying and comparing the dispersion of SHAP values, ultimately ensuring that each indicator constituting the "set of indicators to be used" has a robust and consistent meaning in terms of the direction and intensity of its influence throughout the entire study area.

[0108] Based on the indicators shown above, for example, for a target watershed, the initial candidate indicator set can be {indicator A, indicator B, indicator C, indicator D, indicator A1, indicator B1, indicator C1, indicator A2, indicator B2, indicator C2, indicator A3, indicator B3, indicator C3}. After three rounds of screening, the set of indicators to be used can be {A1, B1, D1, A2, B3}.

[0109] In step S102 above, the selected set of indicators is used to evaluate flood resistance resilience based on administrative units as the basic spatial carrier. It is understood that in this embodiment of the invention, the set of indicators corresponds to flood resistance resilience influencing factors and is represented in vector form. Based on the set of indicators, or the vector set of flood resistance resilience influencing factors, the flood resistance resilience of administrative units can be evaluated to obtain their flood resistance values.

[0110] In one embodiment, calculating the flood resilience value of each administrative unit includes:

[0111] Based on the correlation between the values ​​of K indicators and the SHAP contribution value of each indicator, the direction of influence of each indicator on the dependent variable is obtained, and the K indicators are divided into positive indicators and negative indicators. The dependent variable is the intensity of economic loss from flooding per unit area.

[0112] The K indicators are positively oriented to unify the influence direction of each indicator on the flood resilience evaluation results. Based on this, the indicators after unifying the influence direction are dimensionless to obtain a standardized indicator matrix, as shown in the formula:

[0113] Z= ;

[0114] Among them, Z kn This represents the standard value of the k-th indicator in the n-th administrative unit, where K represents the number of indicators in the set of indicators to be used obtained from the initial candidate indicator set in the n-th administrative unit, and N represents the total number of administrative units.

[0115] Based on the preset objective weight determination rules and subjective weight determination rules, the comprehensive weight ω of each indicator is obtained. k The standardized index matrix is ​​then weighted and summed to obtain the flood resilience value R of the nth administrative unit. n The formula is:

[0116] ;

[0117] Where, ω k This represents the overall weight of the k-th indicator, where k is a positive integer less than or equal to K, and n is a positive integer less than or equal to N.

[0118] In the above steps, the direction of influence of each flood resilience influencing factor vector is first unified, and then a combined weighting method is introduced to determine the weight of each vector, which is then used to determine the flood resilience value of each administrative unit within the target watershed.

[0119] Among them, positive indicators indicate that the larger the indicator value, the higher the flood control resilience, while negative indicators indicate that the larger the indicator value, the lower the flood control resilience. Therefore, unifying the influence direction of the K indicators in the above steps can be achieved by reversing the direction of the negative indicators, so that all indicators satisfy the unified evaluation direction that the larger the indicator value, the stronger the flood control resilience.

[0120] The dimensionless processing can be implemented using standardization methods such as X, Y, and Z, aiming to eliminate differences in dimensions, value ranges, and orders of magnitude among different indicators. For example, the original value of the k-th indicator in the n-th administrative unit... Its standardization results Represented as:

[0121] For positive indicators: ;

[0122] For contrarian indicators: ;

[0123] in, These represent the maximum and minimum values ​​of the k-th indicator across all administrative units, respectively.

[0124] In this embodiment of the invention, the combined weighting method is used to combine objective weights and subjective weights to obtain a comprehensive weight, thereby taking into account both the distribution characteristics of the indicator data and the model interpretation results.

[0125] Among them, the objective weights are constructed based on information entropy and include:

[0126] For indicators Its information entropy Represented as:

[0127] ;

[0128] in, a= , where 'a' is the correction coefficient for information entropy, and 'N' is the total number of evaluation objects, which is equal to the total number of administrative units within the target watershed.

[0129] Determine objective weights: Calculate the coefficient of difference for each indicator based on its information entropy value. : And based on this, determine the objective weight of the indicator. : The larger the coefficient of variation of an indicator, the stronger its ability to distinguish between spatial units, and the greater its objective weight.

[0130] The supervisor's weight is determined based on the contribution of the indicator, including:

[0131] Combining the importance of random forest features and the average absolute contribution value of the SHAP model in step S102 above, the importance of each indicator is quantitatively characterized, and this is used as the basis for determining the subjective weight of the indicator. For the k-th indicator... Its subjective weight The formula is:

[0132] ;

[0133] in, Indicators Feature importance in random forest models.

[0134] A comprehensive weighting is constructed based on objective and subjective weightings, including:

[0135] For the k-th indicator Its combined weight for:

[0136] ;

[0137] in: Subjective weighting; α is the objective weight; α' is the weight adjustment coefficient, with a value range of [0,1].

[0138] It should be noted that the flood resilience values ​​of each administrative unit can also be input into a pre-built flood resilience cloud model to calculate its membership degree to different resilience levels, thereby obtaining the flood resilience level discrimination results. This allows for rapid risk identification, hierarchical classification of management, and optimal allocation of resources. For example, based on the needs of flood resilience evaluation, flood resilience values ​​can be divided into four levels: low resilience, relatively low resilience, medium resilience, relatively high resilience, and high resilience. Thus, for an administrative unit classified as "low resilience," its weak link in the flood control system can be quickly identified, allowing for priority investment in its renovation and enhanced monitoring and early warning. For areas classified as "high resilience," their successful experiences can be summarized and their protection benefits to downstream areas can be assessed.

[0139] Step S103 above is a key step in achieving the connection between analyses at different scales in this embodiment of the invention. It unifies the basic spatial carrier—the administrative unit used for flood resilience values—with the basic spatial carrier—the sub-basin used for the spatial weight matrix and spatial econometric model. In step S103, considering the overlap between sub-basins and administrative units, such as when a sub-basin exists in at least two administrative units, the aforementioned spatial overlap needs to be adjusted. Therefore, in one embodiment, one implementation of step S103 includes:

[0140] Determine the spatial overlap relationship between each sub-basin and the administrative unit;

[0141] If the sub-basin is located in at least two administrative units, the flood resilience values ​​of the at least two administrative units are weighted and summed according to the area ratio of the sub-basin in each administrative unit to obtain the flood resilience characterization value of the sub-basin.

[0142] It is conceivable that when the sub-basin is entirely located within a single administrative unit, the flood resilience value of that administrative unit can be directly used as the flood resilience characterization value of the sub-basin.

[0143] Based on this, this embodiment of the invention utilizes the spatial overlay relationship between sub-basins and administrative units to achieve spatial mapping of flood resilience evaluation results at different scales, enabling flood resilience values ​​and flood resilience influencing factor vectors to participate in subsequent analysis within a unified spatial framework. Building upon this scale mapping, this embodiment introduces a hydrological network structure to characterize the interaction relationships between sub-basins and analyzes the spatial transmission process of flood resilience (including but not limited to flood resilience values ​​and flood influencing factor vectors) between sub-basins. Furthermore, unlike directly establishing the driving relationship between flood resilience values ​​and multiple flood influencing factor vectors, and then using machine learning models such as random forests to achieve scale connectivity across different basic spatial frameworks, this embodiment is not data-driven but mechanism-driven. It can differentiate allocations based on the topological location and functional importance of each sub-basin within a specified portion of the hydrological network structure, where the specified portion refers to the range corresponding to the administrative unit within the hydrological network structure.

[0144] In step S104 above, the spatial weight matrix is ​​constructed based on watershed features with directional differences. These watershed features are identified based on the elevation data and the hydrological network structure, including upstream and downstream relationships and hydrological connectivity paths between sub-watersheds, as well as weight terms for calculating spatial weights. These weight terms include at least a flood generation capacity index, flood propagation distance, flood energy attenuation characteristics, and a watershed anomaly correction coefficient. The flood generation capacity index characterizes the potential runoff generation and confluence capacity of the sub-watershed and can be constructed based on the confluence area, topographic conditions, and underlying surface features of the sub-watershed. The flood propagation distance is the shortest hydraulic path length calculated along the river network from the outlet of sub-watershed i to the inlet of sub-watershed j in a GIS (Geographic Information System). The flood energy attenuation characteristics are calibrated based on historical flood evolution data for the study area. If the flood energy attenuation characteristic is assumed to be 1.5, it indicates that the flood impact decreases non-linearly with increasing distance. The watershed anomaly correction coefficient distinguishes the differences in flood propagation efficiency between the main channel and secondary channels and can be obtained based on the channel classification and the structural characteristics of the main and tributary streams.

[0145] Therefore, in this embodiment of the invention, an asymmetric spatial weight matrix is ​​constructed using watershed features with directional differences. For example, constructing a spatial weight matrix based on hydrological connectivity based on the hydrological connectivity relationships between the sub-watersheds includes:

[0146] Extract weighted terms to characterize the spatial transmission characteristics of floods, wherein the weighted terms include at least flood generation capacity index, rainfall characteristics, flood propagation distance, energy attenuation characteristics, and river channel structure difference correction coefficient;

[0147] Based on the aforementioned weighting term, the asymmetric spatial weight between any two sub-basins is calculated. This spatial weight represents the intensity of the flood's influence along the hydrological network from the upstream sub-basin to the downstream sub-basin, and its calculation formula is as follows:

[0148] ;

[0149] in, Sub-basin The flood impact weight of sub-basin j Sub-basin Rainfall, Indicators representing flood generation capacity, The distance of flood propagation is represented by γ, which is equal to the hydraulic path distance from sub-basin i to sub-basin j calculated in GIS. γ is the distance attenuation coefficient, used to represent the energy attenuation characteristics of the flood during its propagation. This represents the watershed difference correction coefficient;

[0150] The spatial weight matrix W is obtained based on the asymmetric spatial weights of all sub-basins, W= , i and j represent different sub-basins.

[0151] In step S104 above, the spatial econometric model is constructed based on the spatial weight matrix. A better implementation is to introduce the spatial weight matrix into the spatial lag terms of the explained variable and the spatial lag terms of the explanatory variables. In this way, the spatial econometric model can simultaneously focus on the spatial dependence of the explained variable itself, that is, the mutual influence of resilience levels between regions, and the spillover effect of the explanatory variables on adjacent regions, that is, the spatial transmission of influencing factors.

[0152] For example, the spatial measurement model has the following formula:

[0153] ;

[0154] Where R=( ) T , representing a column vector consisting of the flood resilience characterization values ​​of all sub-basins within the target watershed. α represents the flood resistance characterization value of sub-basin i, and α represents the spatial conduction coefficient; Let X represent the spatial weight matrix, X represent the vector of flood control resilience influencing factors for all administrative units, β represent the direct effect coefficient of the explanatory variables, θ represent the coefficient of the spatially lagged explanatory variables, and ɛ represent the random disturbance term.

[0155] Based on the above spatial econometric model, if the flood resilience characterization value of a sub-basin is used as the explained vector, and the spatial transmission coefficient in the model parameters of the spatial econometric model obtained by the maximum likelihood method is significantly non-zero, then the sub-basin is the target sub-basin, and there is a spatial transmission effect in the target basin.

[0156] It should be noted that, according to step S104 above, when analyzing a certain administrative unit, multiple target sub-basins may be obtained. In this case, the spatial transmission effect of each target sub-basin is analyzed sequentially in this embodiment of the invention. For the case where a sub-basin is located in multiple administrative units as shown in the above embodiment, according to step S103 above, the flood resilience characterization value of the sub-basin is a synthesis of the values ​​of multiple administrative units. If the spatial transmission coefficient is significantly non-zero when the flood resilience characterization value of the sub-basin is used as the explained vector, then the conclusion based on the model parameter estimation results, such as the flood resilience impact effect characterization results in the following steps, is applicable to all subordinate administrative units.

[0157] For step S105 above, scaling transformation is used to obtain the flood control resilience impact effect characterization results of administrative units based on the model parameter estimation results, effect decomposition is used to divide the effect of the flood control resilience impact effect characterization results, and action path identification is used to obtain the related sub-basins affecting the target sub-basin, as well as the spatial spillover effect of the flood control resilience impact effect characterization results in the target sub-basin and related sub-basins.

[0158] In one embodiment, the above-mentioned scale conversion includes: weighting and summing the flood resilience characterization values ​​and their spatial transmission effects at the sub-basin scale according to the area proportion of the sub-basin within its administrative unit, thereby realizing the scale conversion from the sub-basin scale to the administrative unit scale and obtaining the flood resilience impact effect characterization results of the administrative unit.

[0159] Taking upstream forest cover as an example, the model parameter estimation results of the spatial econometric model show a significant positive spatial spillover effect. For the target sub-basin, the intensity of this effect is an average increase of 0.06 units in the flood control resilience of its downstream area. Since this target sub-basin accounts for 85% of the area of ​​its administrative unit, after scale transformation, the spatial spillover effect of the forest cover of this administrative unit on downstream flood control resilience has a strength of approximately 0.051 units (i.e., 0.06 × 0.85). This is the characterization result of the flood control resilience impact effect of this administrative unit on this flood control resilience characterization value.

[0160] In one embodiment, the above-mentioned effect decomposition includes: dividing the effect of the flood control resilience influence effect characterization result into direct effects, indirect effects and total effects based on the model parameter estimation results of the target sub-basin.

[0161] The direct effects include local impacts, and the indirect effects include spatial spillover effects.

[0162] In one embodiment, the above-mentioned path identification includes: based on the spatial weight matrix and the model parameter estimation results of all sub-basins, identifying the significant transmission direction and intensity of the flood resilience influencing factor vector of each sub-basin between upstream and downstream sub-basins, and thereby screening out the key sub-basin nodes in the hydrological network structure.

[0163] Using the key sub-basin nodes as anchor points, the complete spatial path of the flood control resilience influencing factor vector of each sub-basin is output along the hydrological network structure, and all upstream related sub-basins that have an impact on the target sub-basin and the transmission links including the target sub-basin and the related sub-basins are extracted from it;

[0164] Based on the aforementioned transmission link, the directionality, asymmetry, and attenuation law of the flood control resilience effect characterization results between the target sub-basin and related sub-basins are analyzed.

[0165] It should be noted that the basis for selecting key sub-basin nodes is to retain sub-basin nodes in the hydrological network that have a significant amplification or control effect.

[0166] For example, taking the "river network density" index as an example, its spatial lag coefficient θ is significantly positive. The complete transmission chain from key sub-basin 2 → key sub-basin 3 → key sub-basin 5 can be traced, and the attenuation pattern of the index's influence along this chain can be calculated. In practical applications, to improve the resilience of key sub-basin 5, in addition to local measures, attention should also be paid to the river network conditions of upstream key sub-basins 2 and 3 (which may belong to other administrative units).

[0167] Through steps S101 to S105 above, the spatial transmission effect analysis method for watershed flood control resilience provided in this embodiment of the invention uses hydrological connectivity as the core physical constraint, integrates multi-source spatial data, and constructs a technical framework that integrates index screening, comprehensive resilience evaluation, and spatial transmission mechanism analysis. Based on the quantitative evaluation of flood control resilience of administrative units, this method innovatively introduces an asymmetric spatial weight matrix based on hydrological network construction and couples it with a spatial econometric model. This realizes a paradigm shift in flood control resilience analysis from static evaluation of administrative divisions to identification of dynamic transmission processes in the watershed hydrological system, and can quantitatively reveal the path, intensity, and spatial spillover effects of resilience transmission along the river network. Compared with traditional methods that rely on geographical adjacency or Euclidean distance weights, this invention effectively overcomes the limitations of insufficient characterization of hydrological processes, ensures the hydrological rationality of spatial dependency modeling, and thus provides a more scientific, refined, and operable decision-making basis for watershed flood control planning, engineering layout optimization, and cross-regional collaborative governance.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for analyzing the spatial transmission effect of watershed flood control resilience, characterized in that, include: Acquire the hydrological network structure and elevation data of the target watershed, divide the target watershed into hydrological zones, and identify the hydrological connectivity between sub-watersheds; The target watershed is divided into multiple administrative units, and a set of indicators to be used is constructed and selected based on multi-source basic spatial data to generate a vector of flood control resilience influencing factors for each administrative unit and calculate the flood control resilience value for each administrative unit; wherein, the target watershed includes multiple sub-watersheds, and each administrative unit includes at least one sub-watershed; Based on the spatial superposition relationship between administrative units and sub-basins, the flood control resilience influencing factor vector and flood control resilience value at the administrative unit scale are associated and mapped to each sub-basin included in the administrative unit, so as to obtain the flood control resilience influencing factor vector and flood control resilience characterization value of each sub-basin; Based on the hydrological connectivity between the sub-basins, a spatial weight matrix based on hydrological connectivity is constructed, and a spatial econometric model is built on this basis. The flood resilience characterization value of each sub-basin is used as the dependent variable, and the flood resilience influencing factor vector after scale mapping is used as the explanatory variable. The model parameters are estimated and significance is tested to identify the target sub-basins with spatial transmission effects. Based on the model parameter estimation results of the target sub-basin and all sub-basins, scale transformation, effect decomposition and action path identification are performed to characterize the impact of spatial transmission effects at the administrative unit scale. The construction of a spatial weight matrix based on hydrological connectivity based on the hydrological connectivity relationships between the sub-basins includes: Extract weighted terms to characterize the spatial transmission characteristics of floods, wherein the weighted terms include at least flood generation capacity index, rainfall characteristics, flood propagation distance, energy attenuation characteristics, and river channel structure difference correction coefficient; Based on the aforementioned weighting term, the asymmetric spatial weight between any two sub-basins is calculated. This asymmetric spatial weight represents the intensity of the flood's influence along the hydrological network from the upstream sub-basin to the downstream sub-basin, and its calculation formula is as follows: ； in, Sub-basin The flood impact weight of sub-basin j Sub-basin Rainfall, Indicators representing flood generation capacity The distance of flood propagation is represented by γ, which is equal to the hydraulic path distance from sub-basin i to sub-basin j calculated in GIS. γ is the distance attenuation coefficient, used to represent the energy attenuation characteristics of the flood during its propagation. This represents the correction factor for watershed differences; The spatial weight matrix W is obtained based on the asymmetric spatial weights of all sub-basins, W= i and j represent different sub-basins.

2. The method as described in claim 1, characterized in that, The set of indicators to be used, constructed and selected based on multi-source basic spatial data, includes: Construct an initial set of candidate indicators for the target watershed; The initial candidate indicator set includes at least land use and cover data and socio-economic statistics data that are spatially integrated according to administrative units based on registration, and preset terrain data that are overlaid according to administrative units based on registration; wherein, the multi-source basic spatial data includes at least the land use and cover data, the socio-economic statistics data, and the preset terrain data; The initial candidate indicator set is subjected to significance screening; based on the results of significance screening, importance screening is performed; and based on the results of importance screening, stability screening is performed to obtain the indicator set to be used.

3. The method as described in claim 2, characterized in that, The step of performing significance screening on the initial candidate indicator set includes: For each administrative unit n, calculate: ； in, This represents the intensity of economic loss per unit area due to flooding in the nth administrative unit. This represents the direct economic loss caused by flooding in the nth administrative unit. This represents the area of ​​the nth administrative unit; Calculate the overall variance σ based on the unit area flood economic loss intensity of N administrative units. 2 Each indicator in the initial candidate indicator set is classified into levels, and each indicator X after stratification is further classified. p Based on the classification results, administrative units are divided into several levels, and the overall variance of the economic loss intensity per unit area due to flooding within each level is calculated. and the corresponding number of administrative units ; Calculate the p-th index X of the initial candidate index set. p Statistical explanatory power for the spatial differentiation of economic loss intensity per unit area caused by flooding The formula is: ； in, This represents the overall variance of the intensity of economic losses per unit area caused by flooding. This represents the overall variance of the intensity of economic losses per unit area from flooding within the h-th level. H represents the number of administrative units included in the h-th level for the p-th indicator. p This represents the number of hierarchical levels formed after the p-th indicator is processed. Based on the explanatory power statistic, a significance test is performed on all indicators in the initial candidate indicator set, and L indicators with significance test results greater than a preset significance threshold are retained as the results of the significance-based screening, where p is a positive integer less than or equal to P, P is the number of indicators in the initial candidate indicator set, L is less than P, and h is less than or equal to H. p Positive integers.

4. The method as described in claim 2, characterized in that, The importance screening based on the results of saliency screening includes: The result X based on the significance screening l Using the intensity of economic loss per unit area from flooding as the input variable and the intensity of economic loss per unit area from flooding as the output variable, a random forest model is constructed. Calculate the change in prediction error I of the random forest model. l The formula is: ； in, This represents the number of decision trees in the random forest model. This represents the original prediction error of the t-th decision tree on the out-of-bag samples; X represents the prediction error corresponding to random perturbations of the input variable. l The l-th indicator representing the significance screening result, the change in prediction error I l X represents the l-th indicator. l The degree of impact on the predicted intensity of economic losses per unit area from flooding; The importance of the L indicators is normalized and sorted in descending order of importance. The top M indicators in the sorting results are taken as the importance-based screening results, where L is a positive integer less than P, P is the number of indicators in the initial candidate indicator set, l is a positive integer less than or equal to L, and t is a positive integer less than or equal to T.

5. The method as described in claim 2, characterized in that, The stability screening based on the importance screening results includes: The result X of the importance-based screening m As input, a predictive model is constructed to predict the intensity of economic losses per unit area from flooding, and the additive contribution of each indicator to the prediction result is calculated based on the SHAP interpretation method. The additive contribution satisfies the following relationship: ； Where f(X) represents the predicted value of the predictive model for the dependent variable, which is the intensity of economic loss per unit area due to flooding. As the baseline output of the prediction model, The m-th indicator X represents the result of the importance screening. m SHAP contribution value; Calculate the statistical characteristics of the SHAP contribution values ​​of M indicators to characterize the spatial stability of the influence of each indicator. Based on the statistical characteristics, stability screening is performed on the M indicators, and K indicators whose SHAP contribution value dispersion is lower than the preset stability threshold are selected as the stability screening results. The K indicators constitute the set of indicators to be used, where M is a positive integer less than L, M is the number of indicators in the results of the importance-based screening, L is the number of indicators in the results of the significance-based screening, m is a positive integer less than or equal to M, and K is a positive integer less than or equal to M.

6. The method as described in claim 1, characterized in that, The calculation of the flood resilience value of each administrative unit includes: Based on the correlation between the values ​​of K indicators and the SHAP contribution value of each indicator, the direction of influence of each indicator on the dependent variable is obtained, and the K indicators are divided into positive indicators and negative indicators. The dependent variable is the intensity of economic loss from flooding per unit area. The K indicators are positively oriented to unify the influence direction of each indicator on the flood resilience evaluation results. Based on this, the indicators after unifying the influence direction are dimensionless to obtain a standardized indicator matrix, as shown in the formula: Z= ； Among them, Z kn This represents the standard value of the k-th indicator in the n-th administrative unit, where K represents the number of indicators in the set of indicators to be used obtained from the initial candidate indicator set in the n-th administrative unit, and N represents the total number of administrative units. Based on the preset objective weight determination rules and subjective weight determination rules, the comprehensive weight ω of each indicator is obtained. k The standardized index matrix is ​​then weighted and summed to obtain the flood resilience value R of the nth administrative unit. n The formula is: ； Where, ω k This represents the overall weight of the k-th indicator, where k is a positive integer less than or equal to K, and n is a positive integer less than or equal to N.

7. The method as described in claim 1, characterized in that, Based on the spatial overlay relationship between administrative units and sub-basins, the flood resilience value at the administrative unit scale is correlated and mapped to each sub-basin included in the administrative unit, serving as the flood resilience characterization value for the corresponding sub-basin, including: Determine the spatial overlap relationship between each sub-basin and the administrative unit; If the sub-basin is located in at least two administrative units, the flood resilience values ​​of the at least two administrative units are weighted and summed according to the area ratio of the sub-basin in each administrative unit to obtain the flood resilience characterization value of the sub-basin.

8. The method as described in claim 1, characterized in that, The spatial econometric model is constructed based on the spatial weight matrix, and the spatial econometric model is as follows: ； in, R represents the flood resilience characterization value of sub-basin i, R=( ) T , represents the column vector consisting of the flood resilience characterization values ​​of all sub-basins within the target watershed, and α represents the spatial transmission coefficient; Let X represent the spatial weight matrix, X represent the flood resilience influencing factor vector of the sub-basin, which is obtained by spatial mapping from the flood resilience influencing factor vector at the administrative unit scale; β represent the direct effect coefficient of the explanatory variable, θ represent the effect coefficient of the spatially lagged explanatory variable, and ɛ represent the random disturbance term.

9. The method according to any one of claims 1 to 8, characterized in that, Based on the model parameter estimation results of the target sub-basin and all sub-basins, scale transformation, effect decomposition, and action path identification are performed to characterize the impact of spatial transmission effects at the administrative unit scale, including: Based on the area proportion of the sub-basin within its administrative unit, the flood resilience characterization value at the sub-basin scale and its spatial transmission effect are weighted and summarized to achieve scale conversion from the sub-basin scale to the administrative unit scale, thereby obtaining the flood resilience impact effect characterization result of the administrative unit. Based on the model parameter estimation results of the target sub-basin, the effects of the flood control resilience impact characterization results are divided into direct effects, indirect effects, and total effects; Based on the spatial weight matrix and the model parameter estimation results of all sub-basins, the significant transmission direction and intensity of the flood control resilience influencing factor vector of each sub-basin between upstream and downstream sub-basins are identified, and key sub-basin nodes in the hydrological network structure are selected accordingly. Using the key sub-basin nodes as anchor points, the complete spatial path of the flood control resilience influencing factor vector of each sub-basin is output along the hydrological network structure, and all upstream related sub-basins that have an impact on the target sub-basin and the transmission links including the target sub-basin and the related sub-basins are extracted from it; Based on the aforementioned transmission link, the directionality, asymmetry, and attenuation law of the flood control resilience effect characterization results between the target sub-basin and related sub-basins are analyzed.

Images