Method for estimating avalanche hazard parameters in different recurrence periods
By drawing empirical frequency curves and fitting the theoretical frequency curves in combination with RAMMS avalanche simulation, the accuracy and efficiency of avalanche hazard parameter estimation are solved, and accurate prediction and scientific decision-making support of avalanche hazard parameters in different recurrence periods are achieved.
Patent Information
- Application Number
- CN202510328402.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-19
- Publication Date
- 2025-07-08
AI Technical Summary
The prior art is inaccurate in estimating avalanche hazard parameters, especially in areas where climate change is obvious or lack of observation data, and traditional methods require more parameter settings and take time to make full use of the data of each avalanche event.
By obtaining monthly precipitation data from the research area, drawing empirical frequency curves and fitting theoretical frequency curves, combining RAMMS avalanche simulation, avalanche hazard parameters are calculated in different recurrence periods, reducing parameter dependence, and improving evaluation efficiency.
Accurate prediction of avalanche hazard parameters during different recurrence periods is achieved, forecast uncertainty is taken into account, scientific decision-making basis is provided, parameter setting is simplified, and evaluation errors are reduced.
Smart Images
Figure CN120277884A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of natural disaster assessment, and particularly to a method for estimating avalanche hazard parameters with different return periods. Background Art
[0002] Estimating avalanche hazard parameters with different return periods is an important part of avalanche risk management and disaster prevention and mitigation, and is of great significance for ensuring personal safety and infrastructure protection. The estimation of avalanche hazard parameters mainly depends on multiple factors such as snow depth, snowfall amount, and meteorological conditions. The changes of these factors directly affect the occurrence and the degree of harm of avalanches.
[0003] Currently, the estimation methods of avalanches mainly include empirical formulas and model simulations. Empirical formulas are usually based on historical observation data and establish relationships with meteorological conditions (such as snowfall amount, air temperature, etc.) through statistical analysis. This method is effective in areas with rich data, but its accuracy is often limited in regions with obvious climate change or lack of observation data. In contrast, model simulations (such as SNOWPACK and CROCUS models) can dynamically simulate the processes of snow layer accumulation, compaction, and ablation, and combine high-resolution meteorological and terrain data to provide more accurate estimates of snow depth. However, the calculation of the model is complex and depends on high-quality input data.
[0004] Although a large number of studies have been dedicated to the estimation of avalanche hazard parameters, current forecasting methods mostly rely on traditional empirical models, often emphasizing the prediction of peaks, while paying insufficient attention to the quantitative forecasting of the overall avalanche hazard. In addition, traditional methods usually require more parameter settings, the parameter calibration process is time-consuming, and have high requirements for historical snowfall and avalanche data, and do not fully utilize the data of each avalanche event.
[0005] Therefore, there is an urgent need to propose a simple and efficient new method from the perspective of data mining to systematically mine existing snowfall and avalanche data. By combining with an avalanche model with fewer parameters, accurate prediction of avalanche hazard parameters with different return periods can be achieved. At the same time, considering the uncertainty of avalanche forecasting, comparing with historical avalanches, and giving the performance of the prediction results in history, so as to provide a scientific basis for relevant decisions. Summary of the Invention
[0006] The object of the present invention is to overcome the above deficiencies and provide a method for estimating avalanche hazard parameters with different return periods, so as to systematically mine existing snowfall and avalanche data. By combining with an avalanche model with fewer parameters, accurate prediction of avalanche hazard parameters with different return periods can be achieved. At the same time, considering the uncertainty of avalanche forecasting, comparing with historical avalanches, and giving the performance of the prediction results in history, so as to provide a scientific basis for relevant decisions.
[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for estimating avalanche hazard parameters with different return periods, comprising:
[0008] S1. Obtain the monthly precipitation data of the study area for n years, sum them up according to the main precipitation months that affect avalanches, and obtain the annual precipitation data that affects avalanches for n years;
[0009] S2. Draw an empirical frequency curve based on the annual precipitation data affecting avalanches in n years, and obtain a theoretical frequency curve by fitting the empirical frequency curve;
[0010] S3, obtaining corresponding precipitation data according to the theoretical frequency curve corresponding to different recurrence periods, and converting the data into an avalanche thickness index;
[0011] S4. Combine the obtained avalanche thickness index with RAMMS avalanche simulation to deduce the relevant parameters of avalanche hazards with different return periods, thereby improving the accuracy of avalanche hazard assessment.
[0012] Furthermore, in S1, the annual precipitation data that affects avalanches is considered to be the annual precipitation data that affects avalanches by summing up the main precipitation months that affect avalanches in the study area.
[0013] Furthermore, the step of drawing an empirical frequency curve according to the annual precipitation data affecting avalanches in n years in S2 specifically includes the following steps:
[0014] S21. Arrange the annual precipitation data affecting avalanches in n years in descending order;
[0015] S22. Calculate the empirical frequency of each annual precipitation data;
[0016] S23. Using the empirical frequency as the horizontal coordinate and the annual precipitation data as the vertical coordinate, an empirical frequency curve is plotted.
[0017] Furthermore, the empirical frequency in S22 is calculated by the following formula:
[0018] Pj = j / (n+1);
[0019] Among them, j is the arrangement number, n is the number of years, and Pj is the empirical frequency.
[0020] Furthermore, obtaining the theoretical frequency curve according to the empirical frequency curve fitting in S2 specifically includes the following steps:
[0021] S24, comparing the hydrological frequency curve with the empirical frequency curve, and selecting the frequency curve type with the most linear similarity;
[0022] S25. Determine the curve parameters by using the maximum likelihood method to obtain a theoretical frequency curve that best fits the empirical frequency curve.
[0023] Furthermore, the hydrological frequency curve is a P-III curve.
[0024] Furthermore, the S25 specifically includes the following steps:
[0025] Define the likelihood function: According to the P-III distribution, construct the probability density function of this distribution, and use it as the distribution model of precipitation. For each observed value of precipitation data, calculate the likelihood value of this value under the current model parameters;
[0026] Maximum likelihood estimation: By constructing the likelihood function L(θ), where θ represents the parameters of the distribution model, which includes the coefficient of variation Cv and the coefficient of skewness Cs; take the logarithm of the likelihood function to obtain the log-likelihood function l(θ), and then transform it into the negative logarithm form -l(θ) for minimization in numerical optimization;
[0027] Numerical optimization: Use the numerical optimization algorithm to search in the parameter space to make the log-likelihood function reach the maximum value; reasonable initial values need to be set during the optimization process, and the parameters Cv and Cs are restricted within a specific range;
[0028] Fitting and verification: After obtaining the optimal parameters Cv and Cs, substitute them into the P-III distribution equation, draw the theoretical frequency curve, and compare this curve with the empirical frequency curve to verify the fitting effect and accuracy.
[0029] Furthermore, during the numerical optimization process, Cv and Cs are limited between 0 and 0.6.
[0030] Further, the S3 specifically includes the following steps:
[0031] S31. Obtain the corresponding precipitation data in the theoretical frequency curve according to the required recurrence period;
[0032] S32. Obtain the avalanche thickness corresponding to the precipitation data where an avalanche occurred in a certain year;
[0033] S33. According to the precipitation data of different recurrence periods corresponding to the precipitation data of the known avalanche thickness in proportion, obtain the avalanche thickness indexes of different recurrence periods respectively.
[0034] Furthermore, the avalanche thickness corresponding to the precipitation data where an avalanche occurred in a certain year is calculated by the following formula:
[0035]
[0036] Where: c: the frictional resistance between the snow cover and the hillside; γ: the density of the snow cover, unit: g / cm 3 ; Internal friction coefficient of snow cover; α: Hillside slope angle, unit is °.
[0037] Advantages of the present invention:
[0038] 1. By combining the measured precipitation data with the RAMMS avalanche simulation model, the present invention improves the efficiency of avalanche risk assessment. This method reduces the dependence on subjective assumption parameters by fitting the empirical frequency curve and the theoretical frequency curve, thereby reducing potential errors in the assessment process. At the same time, the data mining technology is used to simplify the parameter setting and improve the operation convenience of the model.
[0039] 2. The present invention systematically mines the existing snowfall and avalanche data. By combining with the avalanche model with fewer parameters, it can accurately predict the avalanche hazard parameters of different return periods. At the same time, considering the uncertainty of avalanche forecasting, comparing with historical avalanches, and giving the performance of the prediction results in history, so as to provide a scientific basis for relevant decisions. Description of the drawings
[0040] Figure 1 It is a schematic flow chart of a method for estimating avalanche hazard parameters of different return periods;
[0041] Figure 2 It is the selected area of the specific implementation mode of the present invention;
[0042] Figure 3 It is the theoretical frequency curve after fitting in the specific implementation mode of the present invention;
[0043] Figure 4 It is the combined RAMMS accumulation simulation result diagram in the specific implementation mode of the present invention. Specific implementation mode
[0044] The present invention will be further described in detail below with reference to the drawings and specific embodiments.
[0045] Embodiment 1: This embodiment provides a method for estimating avalanche hazard parameters of different return periods, including:
[0046] S1. Obtain the monthly precipitation data of the research area for n years, and sum according to the main precipitation months affecting avalanches to obtain the annual precipitation data affecting avalanches in n years; the annual precipitation data affecting avalanches is obtained by summing the main precipitation months affecting avalanche occurrence in the research area, and it is considered as the annual precipitation data affecting avalanches.
[0047] S2. Draw an empirical frequency curve according to the annual precipitation data of the research area for n years, and obtain a theoretical frequency curve by fitting the empirical frequency curve;
[0048] S21. Arrange the annual precipitation data affecting avalanches in n years in descending order;
[0049] S22. Calculate the empirical frequency of each annual precipitation data;
[0050] S23. Using the empirical frequency as the abscissa and the annual precipitation data as the ordinate, plot to obtain the empirical frequency curve.
[0051] Furthermore, the empirical frequency in S22 is calculated by the following formula:
[0052] Pj = j / (n + 1);
[0053] where j is the permutation serial number, n is the number of years, and Pj is the empirical frequency.
[0054] S24. Compare the hydrological frequency curve with the empirical frequency curve, and screen out the type of frequency curve with the most similar linearity; the hydrological frequency curve is the P-III curve.
[0055] S25. Determine the curve parameters by the maximum likelihood method to obtain the theoretical frequency curve that best fits the empirical frequency curve. Specifically, it includes the following steps:
[0056] Define the likelihood function: According to the P-III distribution (Pearson Type III distribution), construct the probability density function of this distribution as the distribution model of precipitation. For each observed value of precipitation data, calculate the likelihood value of this value under the current model parameters;
[0057] Maximum likelihood estimation: By constructing the likelihood function L(θ), where θ represents the parameters of the distribution model (including the coefficient of variation Cv and the coefficient of skewness Cs); take the logarithm of the likelihood function to obtain the log-likelihood function l(θ), and then convert it to the negative logarithm form -e(θ) for minimization in numerical optimization;
[0058] Numerical optimization: Use a numerical optimization algorithm (such as L-BFGS-B) to search in the parameter space to make the log-likelihood function reach the maximum value; reasonable initial values need to be set during the optimization process, and the parameters are restricted within a specific range (such as Cv and Cs are between 0 and 0.6);
[0059] Fitting and verification: After obtaining the optimal parameters (Cv, Cs), substitute them into the P-III distribution equation, plot the theoretical frequency curve, and compare this curve with the empirical frequency curve to verify the fitting effect and accuracy.
[0060] S3. Obtain the corresponding precipitation data according to the theoretical frequency curve corresponding to different return periods, and convert it into the avalanche thickness index;
[0061] S31. Obtain the corresponding precipitation data in the theoretical frequency curve according to the required return period;
[0062] S32. Obtain the avalanche thickness corresponding to the precipitation data with known avalanches in a certain year;
[0063] S33. According to the precipitation data with different return periods corresponding to the precipitation data of the known avalanche thickness in proportion, obtain the avalanche thickness indexes with different return periods respectively.
[0064] The avalanche thickness corresponding to the precipitation data with avalanches in a certain year is calculated by the following formula:
[0065]
[0066] where: c: the frictional resistance between the snow cover and the hillside; γ: the density of the snow cover, in g / cm 3 ; the internal friction coefficient of the snow; α: the hillside slope angle, in °.
[0067] S4. Combine the obtained avalanche thickness indexes with the RAMMS avalanche simulation to deduce the relevant parameters of avalanche hazards with different return periods, so as to improve the accuracy of avalanche hazard assessment.
[0068] Example 2: This example takes the avalanche that occurred on February 9, 2024 in the third bid section of the construction of National Highway G219 as an example to illustrate the application of this method. As Figure 2 shown, the third bid section of the construction of National Highway G219 is located at the border of Zayu County. This example can be implemented according to the following steps:
[0069] Step 1: Establish a data set;
[0070] Collect the monthly precipitation data of Zayu Meteorological Station from 1969 to 2024. Considering that the snowfall in this area generally occurs from September to December of the previous year and from January to April of the current year, we classify the precipitation from September to December of the previous year and the winter and spring precipitation from January to April of the current year as the precipitation related to snowfall in the current year. For example, the precipitation related to snowfall in the winter and spring of 2024 is the sum of the precipitation from September to December of 2023 and from January to April of 2024.
[0071] Step 2: Obtain the theoretical frequency curve;
[0072] Sort the data, calculate the empirical frequency of each annual precipitation data according to the empirical frequency calculation formula, use the empirical frequency as the abscissa and the annual precipitation data as the ordinate to plot the empirical frequency curve. The calculation formula is as follows:
[0073] Pj = j / (n + 1);
[0074] where j is the ranking number, n is the number of years, and Pj is the empirical frequency.
[0075] Select the P-III distribution (Pearson Type III distribution) as the theoretical frequency model, and obtain the optimal theoretical frequency curve through maximum likelihood estimation. The specific steps are as follows:
[0076] Define the likelihood function: According to the P-III distribution, construct the probability density function of this distribution, and use it as the distribution model of precipitation. For each observed value of precipitation data, calculate the likelihood value of this value under the current model parameters.
[0077] Maximum likelihood estimation: By constructing the likelihood function L(θ), where θ represents the parameters of the distribution model (including the coefficient of variation Cv and the coefficient of skewness Cs). Take the logarithm of the likelihood function to obtain the log-likelihood function e(θ), and then transform it into the negative logarithm form -e(θ) for minimizing the solution during numerical optimization.
[0078] Numerical optimization: Use a numerical optimization algorithm (such as L-BFGS-B) to search in the parameter space to make the log-likelihood function reach the maximum value. During the optimization process, reasonable initial values need to be set, and the parameters are restricted within a specific range (such as Cv and Cs are between 0 and 0.6).
[0079] Fitting and verification: After obtaining the optimal parameters (Cv, Cs), substitute them into the P-III distribution equation, draw the theoretical frequency curve, and compare this curve with the empirical frequency curve to verify the fitting effect and accuracy.
[0080] Figure 3 The theoretical frequency curve obtained by fitting with the parameters Cv = 0.27 and Cs = 0.27 obtained through maximum likelihood estimation.
[0081] Step 3: Determine the avalanche thickness and recurrence period;
[0082] According to the fitted theoretical frequency curve, the precipitation in 2024 is 440.3 mm, the corresponding frequency is 30%, and the recurrence period is 3.33 years. Considering the characteristics of the comprehensive climate in 2024 resulting in a slightly larger avalanche scale, the avalanche in 2024 is comprehensively determined to be once in five years; the snow cover thickness in the avalanche source area in 2024 is obtained according to the avalanche fracture depth calculation formula. The calculation formula is as follows:
[0083]
[0084] Where: c: The frictional resistance between the snow cover and the hillside; γ: The density of the snow cover (g / cm3); The internal friction coefficient of the snow; α: The hillside slope angle (°).
[0085] According to on-site observations and referring to "Avalanches and Governance in the Tianshan Mountains of China" and "Avalanches and Their Prevention and Control", the density γ of the snow cover is taken as 0.1 g / cm3 , the internal friction coefficient of snow cover takes a value of 0.22, and the friction resistance c between the snow cover and the hillside takes a value of 10 g / cm 2 , the average angle of the hillside is 32.21°, and the snow cover thickness in 2024 is calculated to be 2.88 m.
[0086] Considering the disaster prevention standards of national highways, it is necessary to calculate the avalanche scales with return periods of 20 years, 50 years, and 100 years. Based on the snowfall with a return period of 5 years in 2024, on the P-III type curve from 1969 to 2024, the precipitation amounts for a return period of 20 years (5%), 50 years (2%), and 100 years (1%) can be read out respectively as 600, 655, and 695 mm. Since the snow density takes the same value, the snow cover thicknesses under the snowfall conditions with return periods of 20 years, 50 years, and 100 years in the study area can be inferred according to the precipitation ratio, and then the relevant parameters of avalanche hazards can be simulated and calculated.
[0087] Step 4: Comparison between the simulation results and the actual situation;
[0088] It can be seen from the theoretical frequency curve that the return period in 2024 is once in 5 years. Among the parameters selected for the model, the friction coefficient will be selected according to the scale of once in 5 years, and the calculated snow cover thickness and the selected friction coefficient will be used as the input values of RAMMS. Figure 4 is the model output result. The red line is the on-site avalanche deposit line. It can be seen that the avalanche deposit range is basically simulated correctly, indicating that the parameter selection by this method is applicable in the simulation of this avalanche process. Table 1 shows the maximum values of avalanche simulation parameters, which can provide reference values for on-site avalanche prevention and control.
[0089] Table 1 Maximum values of avalanche simulation parameters
[0090]
[0091] The above embodiments are only the preferred technical solutions of the present invention and should not be regarded as limitations to the present invention. The protection scope of the present invention should be the technical solutions recorded in the claims, including the equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, the equivalent replacement improvements within this scope are also within the protection scope of the present invention.
Claims
1. A method for estimating avalanche hazard parameters for different return periods, characterized in that: Including: S1. Obtain the monthly precipitation data of the study area for n years, sum according to the main precipitation months affecting avalanches, and obtain the annual precipitation data affecting avalanches for n years; S2. Draw an empirical frequency curve based on the annual precipitation data affecting avalanches for n years, and obtain a theoretical frequency curve by fitting the empirical frequency curve; S3. Obtain the corresponding precipitation data according to the theoretical frequency curve corresponding to different return periods, and convert it into an avalanche thickness index; S4. Combine the obtained avalanche thickness index with the RAMMS avalanche simulation to calculate the relevant parameters of avalanche hazards for different return periods, thereby improving the accuracy of avalanche hazard assessment.
2. The method for estimating avalanche hazard parameters with different return periods according to claim 1, characterized in that: In S1, the annual precipitation data affecting avalanches is obtained by summing the main precipitation months affecting avalanche occurrences in the study area, and it is considered as the annual precipitation data affecting avalanches.
3. The method for estimating avalanche hazard parameters with different return periods according to claim 1, characterized in that: The specific steps for drawing the empirical frequency curve according to the annual precipitation data affecting avalanches for n years in S2 are as follows: S21. Arrange the annual precipitation data affecting avalanches for n years in descending order; S22. Calculate the empirical frequency of each annual precipitation data; S23. Take the empirical frequency as the abscissa and the annual precipitation data as the ordinate, and draw an empirical frequency curve.
4. The method for estimating avalanche hazard parameters with different return periods according to claim 3, characterized in that: The empirical frequency in S22 is calculated by the following formula: Pj = j / (n + 1); where j is the ranking number, n is the number of years, and Pj is the empirical frequency.
5. The method for estimating avalanche hazard parameters for different return periods according to claim 3, characterized in that: The specific steps for obtaining the theoretical frequency curve by fitting according to the empirical frequency curve in S2 are as follows: S24. Compare the hydrological frequency curve with the empirical frequency curve, and select the type of frequency curve with the most similar linearity; S25. Determine the curve parameters by the maximum likelihood method to obtain the theoretical frequency curve that fits the empirical frequency curve best.
6. The method for estimating avalanche hazard parameters with different return periods according to claim 5, characterized in that: The hydrological frequency curve is the P-III curve.
7. The method for estimating avalanche hazard parameters with different return periods according to claim 5, characterized in that: S25 specifically includes the following steps: Define the likelihood function: According to the P-III type distribution, construct the probability density function of this distribution as the distribution model of precipitation, and calculate the likelihood value of each observed value of precipitation data under the current model parameters; Maximum likelihood estimation: Construct the likelihood function L(θ), where θ represents the parameters of the distribution model, including the coefficient of variation Cv and the coefficient of skewness Cs; take the logarithm of the likelihood function to obtain the log-likelihood function l(θ), and then convert it into the negative logarithm form -e(θ) for minimization in numerical optimization; Numerical optimization: Use a numerical optimization algorithm to search in the parameter space to make the log-likelihood function reach the maximum value; reasonable initial values need to be set during the optimization process, and the parameters Cv and Cs are restricted within a specific range; Fitting and verification: After obtaining the optimal parameters Cv and Cs, substitute them into the P-III type distribution equation, draw the theoretical frequency curve, and compare this curve with the empirical frequency curve to verify the fitting effect and accuracy.
8. The method for estimating avalanche hazard parameters with different return periods according to claim 7, characterized in that: During the numerical optimization process, Cv and Cs are restricted between 0 and 0.
6.
9. The method for estimating avalanche hazard parameters with different return periods according to claim 1, characterized in that: The specific steps of S3 are as follows: S31. Obtain the corresponding precipitation data in the theoretical frequency curve according to the required return period; S32. Obtain the avalanche thickness corresponding to the precipitation data of a certain year when an avalanche is known to occur; S33. Corresponding the precipitation data with different return periods to the precipitation data of the known avalanche thickness according to a proportion, and respectively obtaining the avalanche thickness indexes with different return periods.
10. The method for estimating avalanche hazard parameters with different return periods according to claim 9, characterized in that: Obtaining the avalanche thickness corresponding to the precipitation data of an avalanche occurring in a certain year is calculated by the following formula: Where: c: frictional resistance between the snow cover and the hillside; γ: density of the snow cover, unit: g / cm 3 ; Coefficient of internal friction of the snow; α: hillside slope angle, unit: °.