A disaster-affected population assessment method based on dynamic correction of disaster reporting population data

Abstract

This invention discloses a method for assessing the affected population based on dynamically corrected disaster-reported population data. The method includes the following steps: Step 1, collecting basic data of the study area; Step 2, initializing the population density field; Step 3, constructing a spatial adjacency graph; Step 4, determining the dynamic flood inundation range; Step 5, comparing the reported data to correct population errors. This method, even with limited reported data, dynamically optimizes local population density estimation by combining the spatiotemporal patterns of flood evolution, while ensuring total population conservation, non-negativity, and spatial smoothness. It achieves dynamic updates of population density at the grid scale within administrative regions, thereby obtaining a more accurate and realistic spatial distribution of the population and the number of affected people. This provides a basis for decision support such as population relocation and precise rescue, demonstrating strong practicality.

Description

Technical Field

[0001] This invention belongs to the field of urban emergency management and population spatial distribution modeling technology, and in particular relates to a disaster-affected population assessment method based on dynamic correction of disaster-reporting population data. Background Technology

[0002] my country faces a vast and widely distributed area threatened by floods, particularly in densely populated and affluent plains regions. Floods in these areas have a profound impact on local socio-economic development. In recent years, under the dual pressures of global climate change and intense human activity, the likelihood and uncertainty of extreme rainstorms and floods have increased, especially in developing countries. Due to relatively lagging infrastructure development and weak disaster prevention capabilities, excessive flooding can exacerbate infrastructure damage, leading to urban disruptions such as road, internet, power, and water outages. This can isolate affected areas for a period, creating "information islands" and hindering rescue and reconstruction efforts. Under the current principle of "people first, life first," ensuring the safety of the people is the primary task of flood control and emergency response. Accurately estimating the affected population is crucial for developing rescue and dispatch plans, resource allocation, and risk warnings.

[0003] Traditional methods typically assume a uniform population distribution within residential areas, which differs significantly from the actual population, leading to substantial discrepancies in the final assessment of the affected population. Existing technologies have attempted to construct prior population distribution data using auxiliary data such as nighttime light, mobile phone signaling, or building density. However, these approaches face challenges including data acquisition difficulties, update delays, and privacy compliance issues. For instance, mobile phone signaling is difficult to use in disaster-stricken "information silos," while data such as nighttime light and building density struggle to adapt to the real-time correction requirements of sudden disaster scenarios. Furthermore, in actual flood disaster emergency response, the sparse nature of reported data (i.e., far fewer observations than the parameters to be estimated) leads to severe underdeterminism in direct inversion, lacking effective regularization mechanisms and prone to overfitting or oscillations. Additionally, many methods fail to consider the total population conservation constraint, resulting in population drift after multiple updates and compromising physical plausibility. In summary, effectively utilizing locally reported data and proposing a fast, dynamically corrected, and accurate method for estimating the affected population is a pressing technical challenge. Summary of the Invention

[0004] The purpose of this invention is to provide a method for assessing the disaster-affected population based on dynamically corrected disaster-reporting population data, in order to solve the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] This invention discloses a method for assessing the disaster-affected population based on dynamically corrected disaster-reporting population data. The method includes the following steps:

[0007] Step 1: Collect basic data for the study area: Collect basic data for the study area, including land use and population data, real-time rainfall and engineering data, geographic information data, and flood control and drainage data; the land use and population data includes residential areas, land use layers, administrative boundary layers at different levels, and population data for townships and streets; the geographic information data includes topography, river systems, and DEM data; the flood control and drainage data includes data on flood control and drainage projects and structures.

[0008] Step 2: Initialize the population density field: First, the study area is divided into grids to form a flood simulation grid. Then, using spatial GIS technology, the residential map layer is overlaid with the grid to obtain the spatial overlay relationship between the residential area and the grid.

[0009] Initially, assuming the population is uniformly distributed within the residential areas of the study region, the initial population density of the study region is:

[0010]

[0011] In the formula: ρ (0) Population density of the study area at the initial moment, unit: people / km 2 ;P total A represents the total population of the study area, in persons; total The total residential area of ​​the study region is expressed in km². 2 Among them, A total =∑ i ∑ j A i,j , i is the residential unit index, i∈{1,…,I}; j is the flood simulation grid index, j∈{1,…,J}; A i,j The intersection area of ​​residential cell i and flood simulation grid j, in km². 2 When the two have no intersection, A i,j =0;

[0012] At the initial moment, the population density ρ of the intersection region (i,j) is... i,j The population density ρ of the study area is the same, that is: for all A i,j >0, the initial population density of the intersection region (i,j) is:

[0013]

[0014] In the formula: The population density of the intersection region (i,j) at the initial time, in people / km. 2 ;

[0015] At the initial moment, the population N of the intersection region (i,j) is... i,j for:

[0016]

[0017] In the formula: The population of the intersection region (i,j) at the initial time, in people;

[0018] Step 3: Construct a spatial adjacency graph: Construct an adjacency graph based on intersection regions; use all intersection regions (i,j) as the node set V:

[0019] V={(i,j)∣A i,j >0} (4) If two overlapping areas within the same residential unit share a boundary in space, they are considered adjacent;

[0020] Let l((i,j),(i,j′)) be the length of the common edge of the two intersection regions, and Perimeter(i,j) be the geometric perimeter of the intersection region (i,j), i.e., the total length of the boundary. Define the weights of nodes (i,j) and (i,j′) as follows:

[0021]

[0022] This yields a symmetric sparse adjacency matrix W = [w uv In the diagram, nodes u and v represent intersection regions; one node represents one intersection region (i,j), so there are |V| intersection regions. W represents the adjacency weight of any two nodes u and v; if these two nodes are not adjacent, the weight is w. uv =0;

[0023] Define the degree matrix and the graph Laplacian matrix as follows:

[0024] D uu =∑ v w uv (6)

[0025] L = DW (7)

[0026] In the formula: D is a diagonal matrix; the elements of D are... uu L represents the total connectivity strength between node u = (i,j) and all its adjacent regions; L is used to measure the difference between each node and its neighboring regions.

[0027] Step 4: Determine the dynamic flood inundation range: Construct a flood numerical simulation model to obtain the inundation state of the flood simulation grid at different times; that is, based on real-time rainfall, river water level or flow, and reservoir dam scheduling data, output the inundation state of each grid on the discrete time series (i,j):

[0028]

[0029] In the formula: t sim The output time for the flood simulation is: when a certain grid j is flooded, all intersection regions (i,j) falling on that grid are considered flooded;

[0030] Step 5: Compare and correct population error based on reported data: Calculate and correct population error based on the received reported data, specifically including the following steps:

[0031] Step 5-1: Receive and parse the reported data:

[0032] Receive reported data and represent each reported record as a triple: (a, t) report ,P real (a)); where a represents the reporting administrative unit, including villages, subdistricts, and townships; t report Indicates the reporting time; P real (a) represents the actual number of disaster-affected people in the administrative unit; the set of residential units covered by administrative unit a is defined as follows:

[0033] Step 5-2, Time Alignment and Submerged Mesh Determination:

[0034] According to t report Back to the closest flood simulation output time t sim , t sim Get and report time t report The most recent simulation output time; obtain the grid state G(t) of the flooded area at the corresponding time. sim ,j)∈{0,1}, statistics up to t sim Is grid j submerged?

[0035] J flood (t sim )={j∣G(t sim ,j)=1} (9)

[0036] In the formula: J flood (t sim ) for t sim A collection of grids that are constantly being submerged;

[0037] Step 5-3: Calculate the estimated current affected population:

[0038] Based on the current population density estimate ρ i,j The estimated affected population for administrative unit a is calculated as follows:

[0039]

[0040] In the formula: P est (a,t sim ) for t sim The estimated disaster-affected population of administrative unit A at any given time, in persons;

[0041] Step 5-4: Calculate population error and update the judgment:

[0042] Population error calculation:

[0043]

[0044] The calculated population error value e(a) is compared with the preset threshold ε. If e(a) ≤ ε, it means that the estimated disaster-affected population P of the reported administrative unit a is calculated. est (a,t sim If the accuracy requirement is met, no update is needed; if e(a) > ε, then the population density should be updated.

[0045] Furthermore, the population data of townships and subdistricts mentioned in step 1 are obtained by querying the local statistical yearbook or the data from the Seventh National Population Census; if the data cannot be obtained through the local statistical yearbook or the data from the Seventh National Population Census, then a standard grid population product is used.

[0046] Furthermore, the process of dividing the study area into a flood simulation grid in step 2 involves selecting water-blocking features that are above ground level from linear layers such as highways, railways, and embankments, and dividing the study area into triangular grids or unstructured irregular grids that conform to the terrain and fit the features.

[0047] Furthermore, the specific process for updating the population density described in step 5-4 is as follows:

[0048] 1) Global optimization of incremental field:

[0049] Define the global mask:

[0050] m i,j =1{i∈I(a)}·G(t) sim ,j) (12)

[0051] Then the incremental field Δρ i,j The optimization problem is:

[0052]

[0053] The first term constrains the matching of errors through masking; the second term constrains the population density increment in unaffected areas to remain unchanged; and the third term is a regularization term that ensures spatial smoothness across the entire domain through graph Laplace regularization. Here, λ>0 is the regularization coefficient, and the larger the value, the smoother the spatial distribution of population density increment.

[0054] 2) Global Update:

[0055] ρ i,j ←ρ i,j +αΔρ i,j ,α=α0·min(1,Δt / τ) (14)

[0056] Where α0∈(0,1], τ>0 is the reference time scale, and Δt=t report -t last , where t last The timestamp of the last update was displayed; α represents the degree of trust in the "newly reported data": the more frequent the reporting (i.e., the smaller Δt), the smaller α; the longer the reporting interval (i.e., the larger Δt), the larger α.

[0057] 3) Population conservation correction:

[0058] Population conservation correction includes two methods:

[0059] Method 1: Truncation + Global Proportion Conservation:

[0060]

[0061] Where s is the population conservation scaling factor, which keeps the total population constant across the entire region; ε0 is the non-negative lower bound, with a value of 1e. -6 ;

[0062] Method 2: Based on KL divergence:

[0063] First, we must also ensure that the population density is non-negative:

[0064] ρ i,j ←max{ρ i,j The optimization objective is clearly defined as follows:

[0065]

[0066] in:

[0067]

[0068] Where ρ′ i,j The problem to be solved is the total population P. total Conserved density, π i,j with π′ i,jThe population density is normalized based on area and total population, and is used as the input for KL divergence;

[0069] The optimal solution ρ′ of the optimization objective i,j for:

[0070]

[0071] Therefore, population conservation correction is achieved by solving the following problem:

[0072]

[0073] And let ρ i,j ←ρ′ i,j ;

[0074] Finally, record the state. After the update is complete, record the last update time of the intersection region (i,j) involved in this update, and set t. last ←t report .

[0075] The beneficial effects of this invention are: the method described in this invention can dynamically optimize the estimation of local population density by combining the spatiotemporal pattern of flood evolution with only a small amount of reported data, while ensuring the conservation, non-negativity and spatial smoothness of the total population. This enables dynamic updates of population density at the grid scale within the administrative region, thereby obtaining a more accurate and realistic spatial distribution of the population and the number of affected people. This provides a basis for decision support such as population relocation and precise rescue, and is highly practical.

[0076] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0077] Figure 1 This is a schematic diagram of the method flow described in this invention;

[0078] Figure 2 A schematic diagram of a portion of the grid in the study area;

[0079] Figure 3 This is a schematic diagram of the overlay of a resident map layer and a grid.

[0080] Figure 4 This is a schematic diagram of the grid in the study area of ​​Example 1. Detailed Implementation

[0081] This invention discloses a method for assessing the disaster-affected population based on dynamically corrected disaster-reporting population data, such as... Figure 1 As shown, the method includes the following steps:

[0082] Step 1: Collect basic data for the study area. This includes land use and population data, real-time rainfall and engineering data, geographic information data, and flood control and drainage data. Land use and population data includes residential areas, land use layers, administrative boundary layers at different levels (townships, streets, villages, etc.), and population data for townships and streets. Geographic information data includes the latest topography, river systems, and DEM data. Flood control and drainage data includes data on dikes and other flood control and drainage projects and structures.

[0083] Township and street population data can generally be obtained by consulting local statistical yearbooks or by consulting the data from the Seventh National Population Census. When the statistical yearbooks of townships and streets in the study area do not contain detailed population information, or when statistical yearbooks or the Seventh National Population Census data are difficult to obtain, standard grid population products can be used, such as the 1km spatial resolution kilometer grid dataset of China's population spatial distribution published by the Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences.

[0084] Step 2: Initialize the population density field: First, the study area is meshed to form a flood simulation grid. Specifically, water-blocking features above ground level are selected from linear layers such as highways, railways, and embankments, and the study area is divided into triangular grids or unstructured irregular grids that conform to the terrain and fit the features. Figure 2 The diagram shows a partial grid of the study area. The red bolded areas represent roads with elevations above ground level, which act as water barriers; therefore, these roads are designated as inner boundaries for grid partitioning. Then, using spatial GIS technology, the residential map layer is overlaid with the grid to obtain the spatial overlay relationship between the residential areas and the grid, as shown below. Figure 3 As shown, the purple areas represent residential areas, and each purple area is considered a residential unit.

[0085] Initially, assuming the population is uniformly distributed within the residential areas of the study region, the initial population density of the study region is:

[0086]

[0087] In the formula: ρ (0) Population density of the study area at the initial moment, unit: people / km 2 ;P total A represents the total population of the study area, in persons; total The total residential area of ​​the study region is expressed in km². 2 Among them, A total =∑ i ∑ j A i,j , i is the residential unit index, i∈{1,…,I}; j is the flood simulation grid index, j∈{1,…,J}; Ai,j The intersection area of ​​residential cell i and flood simulation grid j, in km². 2 When the two have no intersection, A i,j =0.

[0088] At the initial moment, the population density ρ of the intersection region (i,j) is... i,j The population density ρ of the study area is the same, that is: for all A i,j >0, the initial population density of the intersection region (i,j) is:

[0089]

[0090] In the formula: The population density of the intersection region (i,j) at the initial time, in people / km. 2 .

[0091] At the initial moment, the population N of the intersection region (i,j) is... i,j for:

[0092]

[0093] In the formula: This represents the population of the intersection region (i,j) at the initial time, in people.

[0094] To avoid loss of generality, in the subsequent conservation correction, the "population conservation region" is defined as the entire study area or any set of sub-regions. This method assumes global conservation by default, meaning that the total population remains constant (∑) throughout the entire study area. i,j ρ i,j A i,j =P total This method is also applicable if it is necessary to divide sub-regions according to administrative boundaries or other boundaries for conservation in specific applications.

[0095] Step 3: Construct a spatial adjacency graph: During the dynamic adjustment of population density, updating only a single intersection region (i,j) may cause unreasonable abrupt changes in the density of its spatially adjacent regions. To introduce spatial linkage, suppress local oscillations, and maintain smoothness during the update, an adjacency graph based on the intersection region is constructed.

[0096] Let all intersecting regions (i,j) be the node set V:

[0097] V={(i,j)∣A i,j >0} (4)

[0098] If two overlapping areas within the same residential unit share a boundary segment (not just corner contact), they are considered adjacent. Figure 3As shown in the figure, the blue and yellow boundary grids are adjacent.

[0099] Let l((i,j),(i,j′)) be the length of the common edge of the two intersection regions, and Perimeter(i,j) be the geometric perimeter (i.e., the total length of the boundary) of the intersection region (i,j). Define the weights of nodes (i,j) and (i,j′) as follows:

[0100]

[0101] This yields a symmetric sparse adjacency matrix W = [w uv ], where nodes u and v represent intersection regions respectively. One node represents one intersection region (i,j), so there are a total of |V| intersection regions. The adjacency weight of any two nodes u and v is represented by W (two-dimensional matrix) (if the two regions are adjacent, they must share an edge, and the weight is calculated using the length of this edge, i.e., formula (5)); if the two nodes are not adjacent, the weight is w uv =0.

[0102] Further define the degree matrix and the graph Laplacian matrix:

[0103] D uu =∑ v w uv (6)

[0104] L = DW (7)

[0105] In the formula: D is a diagonal matrix; the elements of D are... uu Let D represent the total connectivity strength between node u = (i,j) and all its adjacent intersection regions. If an intersection region shares a boundary with many surrounding regions, or if the common edge is very long, then D... uu The value is relatively large; if there is only slight contact with a few areas, then D uu The value is relatively small. L is used to measure the difference between each node and its neighboring regions, thus ensuring the spatial smoothness of the update field.

[0106] Step 4: Determine the dynamic flood inundation range: Construct a flood numerical simulation model to obtain the inundation state of the flood simulation grid at different times. Specifically, based on real-time rainfall, river water level (flow), reservoir dam scheduling, and other data, output the inundation state of each grid on the discrete time series (i,j):

[0107]

[0108] In the formula: t sim The output time for the flood simulation is: when a certain grid j is flooded, all intersection regions (i,j) falling on that grid are considered to be flooded.

[0109] Step 5: Compare reported data to correct population error: Calculate and update the population error based on the received reported data, specifically including the following steps:

[0110] Step 5-1: Receive and parse the reported data:

[0111] Receive reported data and represent each reported record as a triple: (a, t) report ,P real (a)). Where 'a' represents the reporting administrative unit (e.g., a village, street, township, etc.), and 't' represents the reporting administrative unit. report Indicates the reporting time, P real (a) represents the actual number of disaster-affected people in the administrative unit. The set of residential units covered by administrative unit a is defined as follows:

[0112] Step 5-2, Time Alignment and Submerged Mesh Determination:

[0113] According to t report Back to the closest flood simulation output time t sim , t sim Get and report time t report The most recent simulation output time; obtain the grid state G(t) of the flooded area at the corresponding time. sim ,j)∈{0,1}, statistics up to t sim Is grid j submerged?

[0114] J flood (t sim )={j∣G(t sim ,j)=1} (9)

[0115] In the formula: J flood (t sim ) for t sim A collection of grids that are constantly being submerged.

[0116] Step 5-3: Calculate the estimated current affected population:

[0117] Based on the current population density estimate ρ i,j The estimated affected population for administrative unit a is calculated as follows:

[0118]

[0119] In the formula: P est (a,t sim ) for t sim The estimated number of people affected by the disaster in administrative unit a at any given time, in persons.

[0120] Step 5-4: Calculate population error and update the judgment:

[0121] First, calculate the population error:

[0122]

[0123] The calculated population error value e(a) is compared with the preset threshold ε. If e(a) ≤ ε, it means that the estimated disaster-affected population P of the reported administrative unit a is calculated. est (a,t sim If the accuracy requirement is met, no update is needed; if e(a) > ε, then the population density should be updated.

[0124] The specific process for updating population density is as follows:

[0125] 1) Global optimization of incremental field:

[0126] Define the global mask:

[0127] m i,j =1{i∈I(a)}·G(t) sim ,j) (12)

[0128] Then the incremental field Δρ i,j The optimization problem is:

[0129]

[0130] The first term constrains the matching of errors through masking; the second term constrains the population density increment in unaffected areas to remain unchanged; and the third term is a regularization term that ensures spatial smoothness across the entire domain through graph Laplacian regularization. Here, λ>0 is the regularization coefficient; the larger the value, the smoother the spatial distribution of the population density increment.

[0131] 2) Global Update:

[0132] ρ i,j ←ρ i,j +αΔρ i,j ,α=α0·min(1,Δt / τ) (14)

[0133] Where α0∈(0,1], τ>0 is the reference time scale (e.g., 6 hours), and Δt=t report -t last , where t last This represents the timestamp of the last update. α represents the level of trust in the "newly reported data": the more frequent the reporting (smaller Δt), the smaller α (smaller iterations, avoiding oscillations); the longer the reporting interval (larger Δt), the larger α (allowing for larger adjustments, accelerating convergence). Furthermore, since the regularization term applies globally, flooded regions typically generate larger Δρ, while unflooded regions receive smaller adjustments due to spatial smoothness constraints, thus ensuring global consistency.

[0134] 3) Population conservation correction:

[0135] Population conservation correction includes two methods:

[0136] Method 1: Truncation + Global Proportion Conservation:

[0137]

[0138] Here, s is the population conservation scaling factor, which maintains the total population constant across the entire region. Without loss of generality, this constraint can also be extended to grouped conservation. ε0 is a non-negative lower bound, taking a value of 1e. -6 .

[0139] The advantages of this method are that it is simple to implement, has low computational cost, can quickly satisfy non-negativity and total quantity conservation, and is suitable for real-time application scenarios with high-frequency reporting and low latency.

[0140] Method 2: Based on KL divergence (Kullback-Leibler Divergence):

[0141] First, we must also ensure that the population density is non-negative:

[0142] ρ i,j ←max[ρ i,j The optimization objective is clearly defined as follows:

[0143]

[0144] in:

[0145]

[0146] Where ρ′ i,j The problem to be solved is the total population P. total Conserved density, π i,j with π′ i,j The population density is normalized based on area and total population, and is used as the input for KL divergence.

[0147] The optimal solution ρ′ of the optimization objective (Equation 17) i,j for:

[0148]

[0149] Therefore, population conservation correction can be achieved by solving the following problem:

[0150]

[0151] And let ρ i,j ←ρ′ i,j .

[0152] The advantage of this method is that it achieves minimal changes under constraints, prioritizes the stability of high-density areas, and allocates corrections to low-density areas, thereby avoiding the "one-size-fits-all" scaling in Method 1. The results are smoother, more natural, and more stable, making it suitable as a final step for final convergence or high-quality output.

[0153] Finally, record the state. After the update is complete, record the last update time of the intersection region (i,j) involved in this update, and set t. last ←t report This is for subsequent adaptive calculation of the learning rate.

[0154] Example 1

[0155] This embodiment is an application example of the above method.

[0156] 1. Basic information about the study area:

[0157] The study area was divided into a 3×3 grid, with each grid measuring 1 km. 2 ,like Figure 4 As shown in the image, only the purple grid contains residential areas, and the residential areas within the grid are fully covered.

[0158] Total population: 12,000;

[0159] Initial population density: 12000 / 4 = 3000 people / km² 2 .

[0160] 2. Spatial adjacency diagram:

[0161] Nodes: A, B, C, D; Adjacency relationships (strictly determined by shared edges) are shown in Table 1.

[0162] Table 1

[0163] Node pairs Are they adjacent? Common side length weight w A–B Yes (top and bottom) 1km 1 / 4＝0.25 B–C Yes (left and right) 1km 0.25 B–D Yes (top and bottom) 1km 0.25 A–C No (diagonal) 0 0 A–D No (alternate one line vertically) 0 0 C–D No (diagonal) 0 0

[0164] 3. Dynamic correction process for population density:

[0165] Set the population error threshold ε to 10%;

[0166] (1) First report (t=6h)

[0167] Submerged grid: B, C;

[0168] Reported number of people affected by the disaster: 7,000;

[0169] Initial estimate of the affected population: 3000 * 2 = 6000 people;

[0170] Error: e = 1000 / 7000 = 14.28%;

[0171] Since e > 10%, the population density needs to be corrected. The specific correction and update process is as follows:

[0172] Incremental field optimization:

[0173] Objective: Increase population density in regions B and C, while simultaneously influencing population density in regions A and D through Laplace smoothing.

[0174] After solving:

[0175] Δρ_A = +8 (due to adjacent B);

[0176] Δρ_B = +333 (It is necessary to reduce the population error while keeping the relative changes with A, C, and D as small as possible);

[0177] Δρ_C = +658 (It is necessary to reduce the population error while keeping the relative change with B not too large);

[0178] Δρ_D = +8 (due to adjacent B);

[0179] The updated density is:

[0180] A:3008

[0181] B:3333

[0182] C:3658

[0183] D:3008

[0184] Total population = 13007 > 12000;

[0185] Population conservation correction (using method one):

[0186] Scaling factor: α = 12000 / 13007 ≈ 0.92;

[0187] The corrected population density is as follows:

[0188] A:2775

[0189] B:3075

[0190] C:3375

[0191] D:2775

[0192] Total population = 12,000.

[0193] (2) Second report (t=12h)

[0194] New flooded grid: D;

[0195] Current flooded grid: B, C, D;

[0196] Reported number of people affected by the disaster: 9,200;

[0197] Current estimate: 3075 + 3375 + 2775 = 9225 people;

[0198] Error: e = -25 / 9200 = 0.27%, e < 10%; therefore, no update is required.

Claims

1. A method for assessing the disaster-affected population based on dynamically corrected disaster-reported population data, characterized in that, The method includes the following steps: Step 1: Collect basic data for the study area: Collect basic data for the study area, including land use and population data, real-time rainfall and engineering data, geographic information data, and flood control and drainage data; the land use and population data includes residential areas, land use layers, administrative boundary layers at different levels, and population data for townships and streets; the geographic information data includes topography, river systems, and DEM data; the flood control and drainage data includes data on flood control and drainage projects and structures. Step 2: Initialize the population density field: First, the study area is divided into grids to form a flood simulation grid. Then, using spatial GIS technology, the residential map layer is overlaid with the grid to obtain the spatial overlay relationship between the residential area and the grid. Initially, assuming the population is uniformly distributed within the residential areas of the study region, the initial population density of the study region is: （1） In the formula: Population density of the study area at the initial moment, in people / km²; The total population of the study area, in persons; The total residential area of ​​the study region is expressed in km²; among which, , where i is the resident unit index. ;j is the flood simulation grid index, ; For residential units With flood simulation grid The area of ​​their intersection, in km²; when they have no intersection. ; At the initial moment, the intersection region population density With the population density of the study area Same, that is: for all The initial intersection region The population density is: （2） In the formula: The intersection region at the initial time Population density, in people / km²; At the initial moment, the intersection region The population is: （3） In the formula: The intersection region at the initial time Population number, unit: people; Step 3: Construct a spatial adjacency graph: Construct an adjacency graph based on intersection regions; using all intersection regions... As a set of nodes : （4） If two overlapping areas within the same residential unit share a boundary in space, they are considered adjacent. remember Let be the length of the common edge of the two intersecting regions. Intersection area The geometric perimeter is the total length of the boundary, defining the nodes. and The weights are: （5） This yields a symmetric sparse adjacency matrix. , where nodes Each node represents an intersection region; a node represents an intersection region. There are |V| intersection regions, where W represents the adjacency weight of any two nodes u and v; if the two nodes are not adjacent, the weight is... = 0; Define the degree matrix and the graph Laplacian matrix as follows: （6） （7） In the formula: It is a diagonal matrix; elements Represents a node The total connectivity strength with all its adjacent intersecting regions; Used to measure the difference between each node and its neighboring regions; Step 4: Determine the dynamic flood inundation range: Construct a flood numerical simulation model to obtain the inundation state of the flood simulation grid at different times; that is, based on real-time rainfall, river water level or flow, and reservoir dam scheduling data, in discrete time series... Output the flooding status of each grid: （8） In the formula: Output time for flood simulation; when a certain grid When flooded, all intersecting areas falling on this grid All are considered to be flooded; Step 5: Compare and correct population error based on reported data: Calculate and correct population error based on the received reported data, specifically including the following steps: Step 5-1: Receive and parse the reported data: Receive reported data and represent each reported record as a triple: ;in, The administrative unit to which the report is submitted includes villages, subdistricts, and townships. Indicates the reporting time; This represents the actual number of people affected by the disaster in the administrative unit; it defines the administrative unit. The covered residential units are ; Step 5-2, Time Alignment and Submerged Mesh Determination: according to Back to the closest flood simulation output time , Time of collection and reporting The most recent simulation output time; obtain the grid state of the flooded area at the corresponding time. ∈{0,1}, statistics up to Is grid j submerged? （9） In the formula: for A collection of grids that are constantly being submerged; Step 5-3: Calculate the estimated current affected population: Based on current population density estimates Calculate and report administrative units The estimated number of people affected by the disaster is: （10） In the formula: for Time Administrative Unit The estimated number of people affected by the disaster, in persons; Step 5-4: Calculate population error and update the judgment: Population error calculation: （11） The calculated population error value With preset threshold In comparison, if This indicates that the administrative unit that was reported has been calculated. The estimated number of people affected by the disaster If the accuracy requirements are met, no update is needed; Then the population density will be updated.

2. The method for assessing the disaster-affected population based on dynamically corrected disaster-reporting population data according to claim 1, characterized in that, The population data of townships and subdistricts mentioned in step 1 can be obtained by querying the local statistical yearbook or the data from the Seventh National Population Census; if the data cannot be obtained through the local statistical yearbook or the data from the Seventh National Population Census, then the standard grid population product is used.

3. The method for assessing the disaster-affected population based on dynamically corrected disaster-reporting population data according to claim 1, characterized in that, Step 2, which describes the process of dividing the study area into a grid to form a flood simulation grid, involves selecting water-blocking features that are above ground level from linear layers such as highways, railways, and embankments, and dividing the study area into triangular grids that conform to the terrain and fit the features, or unstructured irregular grids.

4. The method for assessing the disaster-affected population based on dynamically corrected disaster-reporting population data according to claim 1, characterized in that, The specific process for updating population density as described in step 5-4 is as follows: 1) Incremental field global optimization: Define the global mask: （12） Then the incremental field The optimization problem is: （13） The first term constrains error matching through masking; the second term constrains population density increments in unaffected areas to remain unchanged; and the third term is a regularization term that ensures spatial smoothness across the entire domain through graph Laplacian regularization. This is the regularization coefficient; the larger the value, the smoother the spatial distribution of population density increment. 2) Global Update: （14） in, , For reference time scale, ,in Indicates the timestamp of the last update execution; This indicates the level of trust in "newly reported data": the more frequent the reporting, the higher the level of trust. Small, The smaller the value, the longer the reporting interval. big, The larger; 3) Population conservation correction: Population conservation correction includes two methods: Method 1: Truncation + Global Proportion Conservation: （15） in, It is a population conservation scaling factor, which means keeping the total population constant across the entire region; It is a non-negative lower bound, and takes the value of ; Method 2: Based on KL divergence: First, ensure that the population density is non-negative. （16） The optimization objective is clearly defined as follows: （17） in: （18） in The problem to be solved is the total population. Conserved density, and The population density is normalized based on area and total population, and is used as the input for KL divergence; Optimal solution of the optimization objective for: （19） Therefore, population conservation correction is achieved by solving the following problem: （20） And order ; Finally, the status is recorded, and after the update is complete, the intersection area involved in this operation is recorded. The last update time, and .

Images