Single-gully debris flow disaster-causing intensity prediction method considering climatic change
Through sliding time window analysis and model coupling, the problem of ignoring the impact of climate change in debris flow simulation was solved, the dynamic prediction and assessment of debris flow disaster intensity was achieved, and the accuracy and adaptability of the prediction were improved.
Patent Information
- Application Number
- CN202510740121.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-05
- Publication Date
- 2025-09-19
AI Technical Summary
Existing debris flow simulation methods fail to effectively consider the impact of climate change on surface runoff evolution and debris flow triggering mechanisms, resulting in delayed prediction results and an inability to meet the current and future refined needs of disaster prevention and mitigation in mountainous areas.
A sliding time window-based method was used to analyze the rainfall extreme time series. A full-process prediction model was constructed by combining the surface runoff response with the debris flow dynamics process. Through the coupled simulation of the FSLAM and FLO-2D models, the debris flow disaster intensity prediction under the climate change scenario was achieved.
It can accurately reflect the impact of climate change on debris flow disasters, improve the accuracy and adaptability of debris flow disaster intensity prediction, and provide a quantitative assessment of future debris flow disasters.
Smart Images
Figure CN120671587A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of geological disaster prediction and disaster prevention engineering technology, specifically a method for predicting the disaster intensity of single-gully debris flow taking into account climate change. The method can cope with the prediction of the disaster intensity of single-gully debris flow facing climate change scenarios. By integrating climate scenario simulation, surface runoff dynamic response and debris flow dynamic process, a "climate change-runoff response-debris flow evolution" full-chain model is constructed to achieve the prediction and assessment of the disaster intensity of debris flow disasters under the background of extreme rainfall, providing technical support for disaster prevention decision-making under climate uncertainty. Background Art
[0002] Debris flows are a common geological disaster in mountainous areas. Characterized by suddenness, destructive power, and unpredictability, they pose a widespread threat to the safety of life, property, and infrastructure in mountainous areas. Debris flows are primarily driven by heavy rainfall, and their occurrence is closely linked to the evolution of surface runoff. In recent years, with the continued intensification of global climate change, extreme weather events have shown new characteristics: increased frequency, increased intensity, and uneven temporal and spatial distribution. The triggering mechanisms and evolution patterns of debris flow disasters are undergoing systematic changes. Numerous observations and studies have shown that the frequency of extreme precipitation events is increasing over time. In earlier periods (such as the mid-20th century), extreme precipitation events were relatively rare, but in recent years, the frequency of high-intensity precipitation has increased significantly, becoming a major meteorological feature. This trend reflects that, against the backdrop of overall warming and increasing instability in the climate system, extreme weather events, particularly short-duration torrential rains, are likely to become a regular trigger for debris flow disasters.
[0003] Traditional studies often assume that rainfall return periods remain constant over time, ignoring the potential impact of climate change on precipitation patterns. This results in models being less adaptable to future climate conditions. With the growing recognition that climate change can alter the intensity and frequency of extreme events beyond historical ranges, this assumption has become inadequate. Therefore, effective methods are needed to address these limitations and avoid errors in calculating the frequency or intensity of extreme events. Yang Can et al. proposed "A Dynamic Assessment Method for Regional Debris Flow Hazard Based on Spatiotemporal Probability." By integrating historical meteorological data with future forecasts, they constructed a dynamic probability model in the temporal dimension and a distributed probability model in the spatial dimension, enabling a dynamic assessment of debris flow hazard. However, this method still has certain limitations: it primarily considers the overall trends in long-term meteorological data and fails to account for time windows of varying lengths. Therefore, it cannot reveal the specific impacts of climate change on the evolution of extreme rainfall frequency and intensity over different periods of time. Furthermore, the rainfall input processing fails to incorporate analysis of the distribution of extreme rainfall extremes and simulation of events with varying return periods, resulting in an inadequate response to future extreme events.
[0004] Secondly, when predicting the disaster-causing intensity of geological disasters under extreme rainfall conditions, since the extreme rainfall amount only remains unchanged within a limited time period and will change at a certain future moment, the existing technology fails to determine the effective time period for prediction, and it is inappropriate to directly use the current extreme rainfall amount.
[0005] Therefore, the present invention proposes to introduce an analysis of extreme rainfall evolution based on a sliding time window, and then combine the coupled simulation of surface runoff response and debris flow dynamics process to construct a whole-process prediction method covering "future period climate prediction - surface runoff response - debris flow movement", so as to realize the dynamic calculation of the disaster-causing intensity of single-gully debris flow disasters that is more in line with the background of climate change. Summary of the Invention
[0006] In the context of frequent extreme climate events, traditional debris flow models based on rainfall data over the entire period are difficult to reflect the changing trend of future disaster risks, resulting in lagging prediction results and unable to meet the refined requirements of current and future mountain disaster prevention and mitigation. The present invention aims to solve the problem that the existing debris flow simulation methods generally ignore the impact of future climate change on surface runoff evolution and debris flow triggering mechanisms, and provides a method for predicting the disaster-causing intensity of single-gully debris flow considering climate change. This method realizes the simulation of single-gully debris flow under climate change scenarios by using a moving time window and stationary test to analyze the rainfall extreme value time series, and achieves the whole-process coupling simulation from climate drive to disaster response.
[0007] The technical solution adopted by the present invention to solve the above technical problems is as follows:
[0008] A method for predicting the disaster-causing intensity of single-gully debris flow considering climate change, the method comprising the following steps:
[0009] S1: Obtain the basic geographical and environmental data within the area to be studied, including digital elevation model DEM, land use and cover type LULC, soil type, rainfall records, soil physical properties, and on-site geological conditions;
[0010] S2: Obtain the rainfall extreme value time series of the area to be studied, with a total of n data. Divide it into 3 time periods according to the distance from the present time, and define them as the long-term, medium-term, and short-term respectively; the time span of each time period is the same, all being x, x < n, and the span of the start time nodes of adjacent two time periods is also the same, all being y, 2 ≤ y < n; Subsequently, use the sliding time window and the hypothesis testing method to determine the sequences of the three time periods with non-stationary characteristics as the rainfall extreme value time series in the long-term, medium-term, and short-term, and at the same time determine the time span x of each time period with non-stationary characteristics and the span y of the start time nodes of adjacent two time periods.
[0011] S3: Arrange the values of the long-term, medium-term, and short-term rainfall extreme value time series determined in step S2 in descending order, then use different extreme value distribution functions to fit the rainfall extreme value time series, apply the KS test to evaluate the goodness of fit, and screen out the optimal rainfall extreme value distribution function that can accurately reflect the probability characteristics of extreme rainfall in the study area;
[0012] S4: Using the optimal rainfall extreme value distribution function determined in step S3, the rainfall return period in different time periods is determined respectively; for the same return period, the extreme rainfall in three different periods can be obtained, and then the extreme rainfall in the three different periods is fitted to obtain a fitting formula, which is then extrapolated to a future time period, i.e., the next time period other than the long, medium, and near term, to obtain the extreme rainfall in the future time period; the first year of the future time period is 1+3y;
[0013] S5: Input the extreme rainfall in the future time period determined in step S4 into the FSLAM model, and output the surface runoff R of each grid cell under the extreme rainfall scenario;
[0014] S6: Use the surface runoff output of FSLAM as R, according to Q C = S·R·Z·D / Vc determines the peak runoff Q that can be achieved in the debris flow valley under surface runoff conditions C ; Use Q C The clean water flow is input into the FLO-2D model to simulate the debris flow movement process, and the flow velocity V, accumulation thickness h, and impact force F of each grid cell are obtained. Then, the debris flow disaster intensity I is obtained according to I = V·h·F, which is used to predict the debris flow disaster intensity.
[0015] Among them, Z is the debris flow blocking coefficient, S is the longitudinal slope coefficient, D is the loose coefficient of the deposit, and Vc is the volume concentration of the debris flow.
[0016] Furthermore, in step S2, the F test under the null hypothesis is used to determine the decay characteristics of the autocorrelation coefficient of the sequence. If the autocorrelation coefficient decays quickly to close to 0, it indicates that the sequence is a stationary time series, otherwise it is a non-stationary time series. If the sequence is a stationary time series, the span of the start time nodes of adjacent time periods is increased by 1, and the long-term, medium-term, and short-term are redefined, and the F test is repeated. The above operations are performed in a loop until the sequences of the three time periods with non-stationary characteristics are finally determined as the rainfall extreme value time series in the long-term, medium-term, and short-term.
[0017] Furthermore, the disaster intensity I is normalized to obtain the normalized disaster intensity I n , the value range is [0,1], and the normalized disaster intensity I nThe area to be studied is divided into three intensity levels: low, medium and high, to achieve a quantitative assessment of the intensity of debris flow disasters.
[0018] Furthermore, the on-site geological conditions include the abundance of loose deposits in the channel, the degree of blockage and the longitudinal slope.
[0019] The present invention also protects a computer-readable storage medium having a computer program stored thereon, wherein the program can implement the steps of the method when executed by a processor.
[0020] Compared with the prior art, the present invention has the following beneficial effects:
[0021] (1) It is possible to consider the non-stationary change characteristics of the rainfall sequence to effectively predict the rainfall extremes: Climate change is a known phenomenon, but the existing technology has not been able to effectively consider how extreme rainfall has changed in the past and in which time period the changes occurred. The present invention adapts to the changing trend of historical climate by moving the time window, gradually changing the time window from small to large, and performing a stationarity test on the characteristics of rainfall sequences in different historical periods, thereby determining rainfall time periods with different statistical characteristics. The span of the rainfall time period determines how many years in the future it is possible to analyze the rainfall sequence and the intensity of mudslide disasters. Subsequently, the rainfall extremes in different time periods are calculated and fitted, and then extrapolated to the rainfall extremes in future time periods. In addition, rainfall sequences in different regions may have different statistical characteristics, so the present invention considers different types of extreme value distribution functions to fit extreme value rainfall, thereby optimizing the distribution function.
[0022] (2) It can effectively combine the rainfall runoff generation and convergence process with the geological conditions of debris flow valleys: Extreme rainfall can generate surface runoff, but there is currently no effective technology that can simultaneously consider the changes in runoff values and the geological conditions of debris flow valleys to quantify the changes in debris flow disaster intensity. The present invention determines the surface runoff amplification effect under different rainfall conditions based on the specific geological conditions of the debris flow valley and its impact on surface runoff, thereby using FLO-2D simulation to achieve the prediction of debris flow disaster intensity. The present invention can achieve dynamic coupling of hydrodynamic changes and slope response mechanisms under climate change scenarios, and systematically reveal the influence of climate change on the triggering and movement of debris flow disasters. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 Climate change time scale segmentation chart;
[0024] Figure 2 100-year rainfall of four models based on a moving time window;
[0025] Figure 3Distribution map of debris flow disaster-causing intensity under different rainfall scenarios in the future period (from left to right are once-in-20-year, once-in-50-year, and once-in-100-year respectively). Detailed implementation manners
[0026] In order to make the purpose, technical solutions and advantages of the present invention clearer and more definite, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the specific embodiments described herein are only used to explain the technical principle of the present invention and do not constitute a limitation to the present invention.
[0027] A method for predicting debris flow disaster-causing intensity in a single gully considering climate change according to the present invention includes the following steps:
[0028] S1: Acquisition of basic data. Collect basic geographical and environmental data in the research area, including digital elevation model (DEM), land use and cover type (LULC), soil type, rainfall records, soil physical properties, on-site geological conditions (richness of loose debris in the gully, blockage degree, longitudinal slope drop), etc. These data serve as the basic input for subsequent hydrological modeling and debris flow simulation to ensure that the model can accurately reflect the topography, hydrological response and surface characteristics.
[0029] S2: Analysis of non-stationary characteristics of rainfall based on a sliding time window. Obtain the rainfall extreme value time series of the area to be studied, with a total of n data. Divide it into 3 time periods according to the distance from the current time, and define them as the long-term, medium-term, and short-term respectively; the time span of each time period is the same, all being x (x < n), and the span of the start time nodes of adjacent two time periods is also the same, all being y (2 ≤ y < n); Subsequently, use the sliding time window and the hypothesis testing method to determine the sequences with non-stationary characteristics in the three time periods as the rainfall extreme value time series in the long-term, medium-term, and short-term, and at the same time determine the time span x of each time period with non-stationary characteristics and the span y of the start time nodes of adjacent two time periods;
[0030] For example, if the long-term starts from the 1st year, then the medium-term starts from the 1 + y year, and the short-term starts from the 1 + 2y year. Specifically, use the F-test under the null hypothesis to determine the decay characteristics of the autocorrelation coefficient of the sequence. If the autocorrelation coefficient quickly decays to close to 0, it indicates that the above sequence is a stationary time series, otherwise it is a non-stationary time series. If the sequence is a stationary time series, then add 1 to the span of the start time nodes of adjacent time periods, redefine the long-term, medium-term, and short-term, and repeat the F-test. Perform the above operations in such a cycle until it is finally determined that the 3 sequences have non-stationary characteristics.
[0031] S3: Screening and Verification of Extreme Value Distribution Functions. The long-term, medium-term, and near-term rainfall extreme value time series identified above were statistically analyzed and arranged in descending order. Different extreme value distribution functions were then fitted to the rainfall extreme value time series. The Kolmogorov-Smirnov (KS) test was used to assess the goodness of fit. The optimal rainfall extreme value distribution function was selected by comparing the maximum difference between the empirical distribution function (ECDF) and the theoretical distribution function (CDF) of the observed data. This process used maximum likelihood estimation to optimize the model parameters, ensuring that the selected distribution accurately reflected the probabilistic characteristics of extreme rainfall in the study area.
[0032] S4: Determination of extreme rainfall values in future time periods. Using the optimal rainfall extreme value distribution function determined in S3, determine the rainfall return period for different time periods. For the same return period, the extreme rainfall values for three different periods can be obtained. Then, the extreme rainfall values for the three different periods are fitted to obtain a fitting formula. Subsequently, the formula is extrapolated to the future time period, that is, the next time period other than the long, medium, and short term, to obtain the extreme rainfall values in the future time period; the first year of the future time period is 1+3y;
[0033] For example, if x=10 and y=2, the long-term time period is 1-10 years, the medium-term time period is 3-13 years, and the near-term time period is 5-15 years. The next time period is 7-17 years. If calculations for the next time period are required, a sliding window is used, with the current medium-term time period as the long-term period, and the calculations are performed iteratively.
[0034] S5: Surface runoff simulation. The DEM, LULC, soil type, average pre-incident rainfall intensity (in mm / d) for the study area compiled from step S1, and the extreme rainfall data for the future time period obtained in step S4 are input into the FSLAM model to calculate surface runoff. The FSLAM model comprehensively considers topographic, hydrological, and soil characteristics and outputs the surface runoff volume R for each grid cell under an extreme rainfall scenario, providing the necessary runoff input data for subsequent debris flow simulations.
[0035] S6: Debris flow disaster intensity prediction. The surface runoff output of FSLAM is R, and according to Q C = S·R·Z·D / Vc determines the peak runoff Q that can be achieved in the debris flow valley under surface runoff conditions C ; Use Q C The clean water flow is input into the FLO-2D model to simulate the debris flow movement process, and the flow velocity V, accumulation thickness h, and impact force F of each grid cell are obtained. Then, the debris flow disaster intensity I is obtained according to I = V·h·F, which is used to predict the debris flow disaster intensity.
[0036] Among them, Z is the debris flow blocking coefficient, S is the longitudinal slope coefficient, D is the loose coefficient of the deposit, and Vc is the volume concentration of the debris flow.
[0037] Example 1
[0038] The present invention considers climate change to predict the intensity of single-channel debris flow disasters, considers debris flow risk simulation under future climate change scenarios, and comprehensively utilizes a three-stage model process of extreme rainfall distribution fitting, shallow landslide instability analysis, and debris flow dynamics simulation. By coupling the FSLAM and FLO-2D models, the whole process from climate drive to disaster response is simulated, including the following steps:
[0039] S1: Collect basic geographic and environmental data in the study area, including digital elevation model (DEM), land use and land cover type (LULC), soil type, rainfall records, soil physical properties, and on-site geological conditions (abundance of loose deposits in the channel, degree of blockage, and longitudinal slope).
[0040] S2: Analysis of non-stationary characteristics of rainfall based on a sliding time window. Obtain a total of n rainfall extreme value time series for the study area. Divide them into three time periods based on their distance from the present time, defined as the long term, medium term, and near term. Using a sliding time window and hypothesis testing, identify the three time periods with non-stationary characteristics as the rainfall extreme value time series in the long term, medium term, and near term.
[0041] This moving time window approach ensures that the analysis covers the entire historical reference period, providing a solid foundation for determining the optimal distribution function. By incorporating time windows and considering the potential impacts of climate change, the assessment of the evolution of extreme event frequencies and their associated risks is more reliable. Furthermore, rainfall extremes are estimated and analyzed using multiple extreme value distribution models.
[0042] S3: Extreme value distribution function screening and validation. To assess the suitability of extreme value distribution models for predicting extreme rainfall, the Kolmogorov-Smirnov (KS) test was applied as a statistical goodness-of-fit criterion. The KS test measures the maximum difference between the empirical cumulative distribution function of the observed data and the theoretical cumulative distribution function of the fitted model. This test provides a robust nonparametric method to assess the consistency between the data and the model, with the null hypothesis stating that the observed data follow a specified theoretical distribution. For each extreme value model, the KS test was performed by first fitting the distribution parameters to the observed rainfall data. Maximum likelihood estimation was used to optimize the parameters of each model to ensure a reliable representation of the data. The fitted cumulative distribution function (CDF) was then compared with the cumulative distribution function (ECDF) of the observed extreme rainfall events. The KS statistic (defined as the maximum vertical distance between the two CDFs) was calculated, and the corresponding p-value was obtained to determine the statistical significance of the fit.
[0043] The cumulative distribution function F of a random observation sample of size n n (X) is calculated as follows:
[0044]
[0045] Among them I [-∞,x] (X i ) is the indicator function, X i represents the i-th observation sample, x is the variable, and X is the observation sample.
[0046]
[0047] The formula for the KS test cumulative distribution function F(X) is as follows:
[0048] D n =SUP X |F n (X)-F(X)| (3)
[0049] Among them SUP X Is the upper limit of the distance; D n The smaller the value, the greater the extreme value sequence F n The more consistent (X) is with the hypothesized distribution F(X). The KS test verifies the null hypothesis that the data in vector x come from a standard normal distribution. If the test rejects the null hypothesis at the 5% significance level, the result is 1; if the null hypothesis is true, the result is 0.
[0050] The Kolmogorov-Smirnov (KS) test was used to compare the empirical cumulative distribution function (ECDF, F nThe maximum difference between the distribution function (X) and the theoretical distribution function (CDF, i.e., F(X)) is used to screen the optimal distribution function for extreme rainfall scenario generation.
[0051] S4: Determination of the extreme rainfall in the future time period. The optimal rainfall extreme value distribution function determined in S3 is used to determine the rainfall recurrence period. For the same recurrence period, the extreme rainfall in three different periods can be obtained, and then the extreme rainfall in the three different periods are fitted to obtain a fitting formula, and then extrapolated to the future time period (i.e., the next time period other than the long-term, medium-term, and short-term) to obtain the extreme rainfall in the future time period. The first year of the future time period is 1+3y, and the time span of each time period is the time window finally determined in S2. It should be noted that the time period span determined in S2 is x, so the present invention can predict the extreme rainfall in the future time period with a length of x years. When the time period to be predicted exceeds x years, the historical rainfall extreme time series in S2 needs to be redivided.
[0052] S5: Input the DEM, LULC, soil type, previous rainfall in the study area, and the extreme rainfall in the future time period determined in step S4 into the FSLAM runoff model to calculate the surface runoff R for each grid cell in the study area under the extreme rainfall scenario.
[0053] The main input data for FSLAM consists of five raster files and two text files. The raster files contain the following information: i) soil properties; ii) land use and land cover (LULC); iii) digital elevation model (DEM); iv) previous rainfall; and v) extreme rainfall.
[0054] S6: Debris flow intensity prediction. The FLO-2D model is used to simulate the movement characteristics of debris flows, where surface runoff is input in the form of peak runoff, which essentially refers to the maximum rate of water moving on the surface at any given moment during rainfall.
[0055] The surface runoff R was determined using the Runoff Module of FSLAM. The peak runoff was then determined using the following formula:
[0056] Q C =S·R·Z·D / Vc (4)
[0057] Where Q Cis the peak runoff input into FLO-2D; R is the surface runoff calculated by FSLAM; Z is the debris flow obstruction coefficient, with possible values of 1.5 (almost no channel obstruction), 2 (slight channel obstruction), 2.5 (moderate channel obstruction), and 3 (severe channel obstruction). S is the longitudinal slope coefficient, with possible values of 1 (average channel longitudinal slope between [0, 2‰)), 2 (average channel longitudinal slope between [2, 10‰)), 3 (average channel longitudinal slope between [10‰, 100‰)), and 4 (average channel longitudinal slope greater than 100‰). D is the looseness coefficient of the deposit, with possible values of 1 (dense deposit), 2 (relatively loose deposit), and 3 (relatively loose deposit). Vc is the volume concentration of the debris flow, which can be expressed as:
[0058] Vc=V S / (V S +V w ) (5)
[0059] Where V S is the total volume of solid matter in debris flow, V w is the total volume of water; Vc can be obtained through experiments or on-site investigations.
[0060] The above formula simultaneously considers the surface runoff caused by extreme rainfall conditions and the actual geological conditions of the debris flow valley. Using the improved calculated peak runoff as the input flow for the FLO-2D simulation can better reflect the influence of the on-site geological environment conditions. It is closer to the actual situation than traditional methods and helps to improve the accuracy of debris flow prediction.
[0061] Further, Q c The clear water flow data is input into the FLO-2D software. Parameters such as the Manning roughness, yield stress, viscosity coefficient, and laminar flow resistance coefficient are then set. The debris flow motion characteristics under these conditions are calculated. The velocity V, accumulation thickness h, and impact force F of each grid cell are output. The following formula is defined as the debris flow disaster intensity I:
[0062] I=V·h·F (6)
[0063] Normalize the disaster intensity of the entire area to obtain the normalized disaster intensity I n , the value range is [0,1], and the normalized disaster intensity I n The area to be studied is divided into three intensity levels: low, medium and high, to achieve a quantitative assessment of the intensity of debris flow disasters.
[0064] Example 2
[0065] This example takes a single-ditch debris flow in Jizhou District, Tianjin as the research object. Based on the SBAS-InSAR monitoring results and field investigation analysis, the study area showed obvious surface subsidence and accumulation characteristics from 2015 to 2024, and the subsidence in local areas was as high as 70 mm. Extreme rainfall has aggravated mountain erosion and loosened materials, the river channel is seriously blocked, the gullies are densely developed, and the terrain is steep. Taking into account the surface deformation characteristics and the impact of extreme climate, the study area has a high potential risk of debris flow disasters. The integrated processing of extreme rainfall analysis, surface runoff calculation and debris flow simulation that integrates future climate change scenarios is adopted to verify the effectiveness of the single-ditch debris flow disaster intensity prediction method considering climate change proposed in this invention.
[0066] S1: Collect basic geographic and environmental data in the study area, including digital elevation model (DEM), land use and land cover type (LULC), soil type, rainfall records, soil physical properties, and on-site geological conditions (abundance of loose deposits in the channel, degree of blockage, and longitudinal slope).
[0067] S2: The 48-hour rainfall extreme value time series from 1951 to 2017 is selected and divided into three different time periods: 1951-1997 (long term), 1961-2007 (medium term) and 1971-2017 (recent term) ( Figure 2 ), and the time span of each period is 46 years (i.e., x=46, y=10). Three moving time windows are used to analyze the data, and the occurrence of extreme precipitation events is as follows: in the long term, there are 2 times exceeding 200 mm and 5 times exceeding 150 mm; in the medium term, there are 2 times exceeding 200 mm and 6 times exceeding 150 mm; in the near term, there are 3 times exceeding 200 mm and 7 times exceeding 150 mm. The F test under the null hypothesis is used to calculate the decay characteristics of the autocorrelation coefficients of the above three series. The results show that the autocorrelation coefficient does not decay to 0 quickly, indicating that the above time series is a non-stationary time series. Therefore, it shows that the time period division at this time can prove the climate change in the study area, so the extreme rainfall prediction from 1981 to 2027 can be carried out. At this time, x=46 and y=10.
[0068] S3: Four models, namely exponential distribution, Gumbel distribution, Pearson III distribution (P-III), and Weibull distribution, are selected to fit the rainfall extreme time series.
[0069] The optimal distribution function was determined by the Kolmogorov-Smirnov test. The goodness of fit was assessed by comparing the cumulative distribution function (CDF) of the empirical data and the model at a significance level of 0.05. The results showed that the exponential distribution failed the KS test, as confirmed by a return value of 1, which rejected the null hypothesis at a significance level of 5%. In contrast, the Weibull distribution and the Pearson type III distribution performed well, with p-values of 0.740 and 0.785, respectively. n The values are 0.081 and 0.078 respectively.
[0070] Figure 2 The goodness-of-fit results of the empirical cumulative distribution function (CDF) and four distribution models were intuitively compared. The empirical cumulative distribution function (CDF) of the normalized data was highly consistent with the cumulative distribution functions of the Weibull distribution and the Pearson type III distribution, while the exponential distribution showed significant deviations. The measured data and fitting functions for the entire historical period were displayed. Among the models tested, the Pearson type III distribution had the best fitting curve, followed by the Weibull distribution. It is worth noting that the Pearson type III model is very close to the extreme rainfall event observed in 2016, highlighting its ability to effectively capture extreme precipitation patterns. In summary, the Pearson type III distribution was shown to be the most suitable for simulating extreme rainfall events in the study area, accurately characterizing historical precipitation patterns, and providing reliable predictions of extreme events. Ultimately, extreme rainfall predictions, based on different time windows (long-term, medium-term, near-term, and full-term) and return periods (P = 0.05, P = 0.02, and P = 0.01; P represents the probability of rainfall, which is the inverse of the return period; for example, the probability of a 20-year rainfall return period is 0.05), show a gradual increase in extreme rainfall over time. In the long-term, medium-term, and near-term periods, the rainfall amounts for a 20-year return period (P = 0.05) were 178.5 mm, 179.5 mm, and 183.1 mm, respectively; for a 50-year return period (P = 0.02), the rainfall amounts were 205.4 mm, 207.8 mm, and 213.3 mm, respectively; and for a 100-year return period (P = 0.01), the rainfall amounts were 224.8 mm, 228.3 mm, and 235.2 mm, respectively. The overall trend shows that under the background of future climate change, the intensity of extreme rainfall events will increase to a certain extent, which is particularly obvious in the recent stage, suggesting that the risk of mudslides and related geological disasters will increase in the future.
[0071] S4: Fitting the extreme rainfall values of three periods with a return period of 20 years (178.5 mm, 179.5 mm, and 183.1 mm) yields the following fitting formula:
[0072] Q=1.3A 2 -2.9A+180.1
[0073] Here, Q refers to the extreme rainfall, and A refers to the xth time period in descending order (A=1 for the long term; A=2 for the medium term; A=3 for the near term; and A=4 for the future). Therefore, the extreme rainfall for a 20-year return period in the future (i.e., 1981-2027) is 189.3 mm. Similarly, the extreme rainfall for a 50-year return period and a 100-year return period in the future is 221.9 mm and 245.5 mm, respectively.
[0074] S5: Based on the high-resolution DEM acquired by drones, LULC extracted from remote sensing images, soil parameters obtained from field surveys, and pre-effective infiltration calculated from historical meteorological data, combined with the extreme rainfall in the future time period determined in step S4, the FSLAM model is used to simulate the hydrological response under different extreme rainfall scenarios to determine surface runoff distribution. Simultaneously, the slope stability analysis module calculates the landslide instability probability for each grid cell, providing a basis for identifying debris flow initiation areas.
[0075] High-precision topographic and surface information for the study area was collected, including: (i) a digital elevation model (DEM) obtained through drone aerial surveys with a resolution of 3m×3m and centimeter-level accuracy; and (ii) land use and land cover (LULC) information extracted using PLANET images collected in June 2023 with a resolution of 3m×3m. Due to the lack of high-precision LULC data, the remote sensing images were classified using the maximum likelihood method, and the classification accuracy was tested using a confusion matrix. The overall classification accuracy reached 92.31% and the Kappa coefficient was 87.46%. Soil types and physical properties of different soil types (including cohesion, friction angle, density, hydraulic conductivity, and soil thickness) were obtained based on field surveys. The average initial rainfall in the study area was 0.35 mm / d. Subsequently, the rainfall extremes under different return periods obtained in S4 were input into the FSLAM model to calculate the surface runoff R under extreme rainfall conditions. The full-period scenario was 40.7 m / s, 10.8 m / s, and 11.7 m / s, respectively. 3 / s、49m 3 / s and 55m 3 / s.
[0076] S6: According to the field investigation, the blockage coefficient of the study area is 2 (slightly blocked channel), the longitudinal slope coefficient is 3 (average longitudinal slope of the channel is 32‰), the looseness coefficient of the sediment is 2 (relatively loose sediment), and the volume concentration of the debris flow is 0.5. Therefore, the peak runoff of the debris flow in the future time period calculated according to formula (4) is 976.8m 3 / s (once in 20 years), 1176m 3 / s(once in 50 years), 1320m 3 / s (100-year return period). The above peak runoff is input into the FLO-2D model as the clean water flow, and parameters such as Manning roughness, yield stress, viscosity coefficient, and laminar flow resistance coefficient are set according to the actual situation. FLO-2D simulates the flow velocity, accumulation thickness, and impact force of the debris flow during the movement process, and then calculates the debris flow disaster intensity and normalized disaster intensity map according to formula (6) ( Figure 3 ).
[0077] Any matters not described in the present invention are applicable to the prior art.
Claims
1. A method for predicting the intensity of debris flow disasters in a single gully considering climate change, characterized by: The method includes the following steps: S1: Obtain the basic geography and environmental data within the area to be studied, including digital elevation model (DEM), land use and cover type (LULC), soil type, rainfall records, soil physical properties, and on-site geological conditions; S2: Obtain the rainfall extreme value time series of the area to be studied, with a total of n data. Divide it into three time periods according to the distance from the current time, and define them as the long-term, medium-term, and short-term respectively; the time span of each time period is the same, both are x, where x < n, and the span of the start time nodes of adjacent two time periods is also the same, both are y, where 2 ≤ y < n; Subsequently, use the sliding time window and the hypothesis testing method to determine the sequences with non-stationary characteristics in the three time periods as the rainfall extreme value time series in the long-term, medium-term, and short-term, and at the same time determine the time span x of each time period and the span y of the start time nodes of adjacent two time periods when there are non-stationary characteristics; S3: Arrange the values of the rainfall extreme value time series in the long-term, medium-term, and short-term determined in step S2 in descending order, and then use different extreme value distribution functions to fit the rainfall extreme value time series, and apply the K-S test to evaluate the goodness of fit, and screen out the optimal rainfall extreme value distribution function that can accurately reflect the probability characteristics of extreme rainfall in the study area; S4: Use the optimal rainfall extreme value distribution function determined in step S3 to determine the rainfall recurrence periods in different time periods respectively; for the same recurrence period, three extreme rainfall amounts in different periods can be obtained, and then fit the three extreme rainfall amounts in different periods to obtain a fitting formula, and then extrapolate to the future time period, that is, the next time period except the long-term, medium-term, and short-term, to obtain the extreme rainfall amount in the future time period; the first year of the future time period is 1 + 3y; S5: Input the extreme rainfall amount in the future time period determined in step S4 into the FSLAM model, and output the surface runoff R of each grid cell under the extreme rainfall scenario; S6: Use the surface runoff output of FSLAM as R, according to Q C = S·R·Z·D / Vc determines the peak runoff Q that can be achieved in the debris flow valley under surface runoff conditions C ; Use Q C The clean water flow is input into the FLO-2D model to simulate the debris flow movement process, and the flow velocity V, accumulation thickness h, and impact force F of each grid cell are obtained. Then, the debris flow disaster intensity I is obtained according to I = V·h·F, which is used to predict the debris flow disaster intensity. Where, Z is the debris flow blockage coefficient, S is the longitudinal slope drop coefficient, D is the loose coefficient of the accumulation, and Vc is the volume concentration of the debris flow.
2. The method according to claim 1, characterized in that In step S2, use the F test under the null hypothesis to determine the decay characteristics of the autocorrelation coefficient of the sequence. If the autocorrelation coefficient decays rapidly to close to 0, it indicates that the sequence is a stationary time series, otherwise it is a non-stationary time series; if the sequence is a stationary time series, add 1 to the span of the start time nodes of adjacent time periods, redefine the long-term, medium-term, and short-term, and then repeat the F test; execute the above operations in a loop until the sequences with non-stationary characteristics in the three time periods are finally determined as the rainfall extreme value time series in the long-term, medium-term, and short-term.
3. The method according to claim 1, characterized in that The disaster intensity I is normalized to obtain the normalized disaster intensity I n , the value range is [0,1], and the normalized disaster intensity I n The area to be studied is divided into three intensity levels: low, medium and high, to achieve a quantitative assessment of the intensity of debris flow disasters.
4. The method according to claim 1, wherein The on-site geological conditions include the abundance, blockage degree, and longitudinal slope drop of the loose debris in the gully.
5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it can implement the steps of the method according to any one of claims 1-4.