Rainfall-induced landslide deformation stage identification and early warning method and system

Abstract

The application discloses a rainfall landslide deformation stage identification and early warning method and system, obtains a cumulative displacement sequence; calculates an improved tangent angle of each monitoring moment; obtains a displacement rate sequence, sums all positive instantaneous accelerations in a sliding window to obtain sliding window cumulative acceleration; and takes the improved tangent angle and the sliding window cumulative acceleration as inputs, sequentially generates a final warning level through three layers of criteria, namely, basic grading, grade correction and persistence smoothing. The application takes the improved tangent angle calculated from the original displacement sequence as a main criterion, retains complete response capability to reactivation events, takes the sliding window cumulative acceleration as an auxiliary criterion, avoids the smoothing problem of late events of a trend fitting method, reduces the high sensitivity of a single criterion through mutual verification of the two criteria, supports rolling update of positive acceleration accumulation quantile values, avoids long-term drag of historical cumulative quantities, can adapt to magnitude differences of different measuring points, and significantly improves parameter transferability.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring and early warning technology, and is particularly applicable to methods and systems for identifying and warning of deformation stages of rainfall-induced landslides based on GNSS automated monitoring data. Background Technology

[0002] Rainfall-induced landslides exhibit typical data characteristics in engineering practice, such as a generally step-like evolution in displacement sequences. During periods of concentrated rainfall, the displacement rate rapidly increases, forming a step-like rise, followed by a plateau phase lasting several months, and may experience a secondary acceleration under subsequent heavy rainfall. Existing early warning methods for rainfall-induced landslide data sequences include the fixed rate threshold method, the single tangent angle criterion method, and the all-time cumulative acceleration method. These methods have the following main shortcomings in engineering implementation: (1) The fixed rate threshold method directly uses the daily displacement rate exceeding the preset threshold as the basis for upgrading. It is sensitive to acceleration events, but occasionally exceeding the threshold during the plateau period will trigger an early warning. The false alarm rate is generally high (20%–30% in actual measurements). Early warning fatigue and mismatch of emergency resources are likely to occur at the engineering site.

[0003] (2) The single tangent angle criterion method classifies the level based on the degree of deviation of the instantaneous rate from the average rate of the whole process. The physical meaning is clear, but the single-point threshold crossing is upgraded, which is sensitive to monitoring noise. Frequent level jumps will occur near the boundary between the plateau period and the acceleration period, and the engineering readability is poor.

[0004] (3) The displacement master curve trend fitting method fits the displacement master curve with an analytical function (such as an S-shaped function) and then classifies it. It works well for displacement sequences with single-peak evolution; however, for composite sequences of "long plateau + later reactivation", it is difficult to use a single function to characterize two independent acceleration processes at the same time, and the later events are easily "flattened" by the fitted curve, resulting in missed reports.

[0005] (4) The full-time cumulative acceleration method uses cumulative acceleration as an auxiliary criterion. The full-time cumulative form produces a lasting "memory effect" on the early cumulative amount. The plateau period is prone to jump due to the superposition of historical tails. At the same time, most existing methods use fixed numerical thresholds, which are difficult to adapt to the magnitude differences between different measurement points and have poor parameter transferability.

[0006] In summary, existing technical solutions have significant shortcomings in terms of early warning stability, parameter transferability, and engineering interpretability for step displacement sequences. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for identifying and warning of deformation stages in rainfall-induced landslides, which addresses the problems of insufficient early warning stability, poor parameter transferability, and weak engineering interpretability of step displacement sequences.

[0008] To achieve the above objectives, the rainfall-induced landslide deformation stage identification and early warning system of the present invention includes the following steps: S1, the raw monitoring data is aligned into an equally spaced time series according to a unified time format to obtain the cumulative displacement series; S2, calculate each monitoring time. t i Improved tangent angle β i ; S3. Obtain the displacement rate sequence based on the cumulative displacement sequence, and smooth the displacement rate sequence using Savitzky-Golay filtering; calculate the displacement rate sequence at each monitoring time point after smoothing. t i instantaneous acceleration ; in length of w The cumulative acceleration of the sliding window is obtained by summing all positive instantaneous accelerations within the sliding window. ; S4, to improve the tangent angle and sliding window cumulative acceleration Using the input as the basis, the final warning level is generated through three layers of criteria: basic classification, level correction, and continuous smoothing. L i (3) The warning levels include safety level, attention level, warning level, alert level, and alarm level.

[0009] Furthermore, in step S1, multiple GNSS measuring points are set up at different locations of the landslide body, with a sampling rate of no less than 1 time / day, and the cumulative displacement of each measuring point is continuously collected to obtain the cumulative displacement sequence of each measuring point.

[0010] Furthermore, each monitoring time mentioned in step S2 t i Improved tangent angle The For monitoring time t i The instantaneous displacement rate; For monitoring time t i and the start time t 0 The average rate over the entire process; , ε This is the preset minimum rate threshold.

[0011] Furthermore, the basic grading described in step S4 is based on the improved tangent angle. β i Determine the basic early warning level Li (1) The level correction is based on the cumulative acceleration of the sliding window. Peak rate relative to basic warning level L i (1) Make corrections to obtain the correction level. L i (2) The continuous smoothing process generates the final warning level. L i (3) Determine using the following formula: , m↑ To upgrade the required number of continuous monitoring moments, m↓ The number of continuous monitoring moments required for downgrading, and m↓>m↑ c represents the monitoring time. t i Correction level L i (2) ; This represents the final warning level at the previous monitoring time. The correction level is for monitoring time k.

[0012] Furthermore, the peak rate is the monitoring time. t i instantaneous displacement rate Greater than the 80th percentile of the instantaneous displacement rate of the previous 30 days q 80 v The product with the coefficient r, i.e. q 80 v × r .

[0013] Furthermore, monitoring time t i If the number of valid samples is less than 30, the quantile value is calculated using existing historical samples, or the rank correction is temporarily not enabled.

[0014] Furthermore, when the basic early warning level L i (1) Below the alert level and the cumulative acceleration of the sliding window Monitoring time t i 80th percentile of instantaneous displacement rate in the first 30 days q 80 v At that time, the correction level L i (2) Basic warning levelL i (1) The previous warning level; when the basic warning level L i (1) Equal to alert level, cumulative acceleration of sliding window Monitoring time t i 80th percentile of instantaneous displacement rate in the first 30 days q 80 v And monitoring time t i instantaneous displacement rate v i ≥ q 80 v × r Correction level L i (2) Basic warning level L i (1) The previous warning level; when the basic warning level L i (1) Equal to alert level, but monitoring time t i instantaneous displacement rate v i < q 80 v × r Maintain basic early warning level L i (1) Unchanged, i.e., the correction level remains unchanged. L i (2) Basic warning level L i (1) .

[0015] Furthermore, the final warning level The system uses color-coded displays to indicate the time sequence of events, and pushes warning information to preset receiving terminals for on-site audible and visual alarms.

[0016] The present invention also provides a system for identifying and warning of deformation stages of rainfall-induced landslides, used to implement the aforementioned method for identifying and warning of deformation stages of rainfall-induced landslides, specifically including a monitoring system, a data processing system, and an early warning system; The monitoring system includes a GNSS displacement acquisition module and a data transmission module, used to acquire raw monitoring data and transmit the raw monitoring data to the data processing system. The data processing system preprocesses the raw monitoring data, calculates the improved tangent angle, smooths the displacement rate, calculates the acceleration, and calculates the cumulative acceleration of the sliding window. The early warning system generates the final early warning level through three layers of criteria: basic classification, level correction, and continuous smoothing. L i (3) .

[0017] Furthermore, it also includes a warning information output and rainfall coupling verification module; the warning information output displays color codes and time sequences corresponding to the warning level through a display device, pushes the warning information to a preset receiving terminal through a communication network, and provides audible and visual alarms through an on-site alarm device; The rainfall coupling verification module collects rainfall data of the monitoring area through the monitoring system, preprocesses the rainfall data through the data processing system, and then performs left concatenation according to the time alignment of the cumulative displacement sequence. The physical rationality of the early warning result is externally verified by calculating the Pearson correlation coefficient and significance level.

[0018] The advantages of this invention are: 1. This invention no longer performs analytical function trend fitting on the displacement master curve, but directly calculates the improved tangent angle using the original displacement sequence as the primary criterion, preserving the complete response capability to reactivation events; simultaneously, it uses the sliding window cumulative acceleration as an auxiliary criterion to provide short-memory evidence of whether acceleration continues in the current stage. This architecture structurally avoids the smoothing problem of later events inherent in trend fitting methods, while reducing the high sensitivity of a single criterion through the mutual verification of the two criteria.

[0019] 2. This invention employs a sliding window approach instead of accumulating positive acceleration over the entire time period, supporting rolling updates of the accumulated quantile values ​​for positive acceleration. This avoids the long-term drag of historical accumulation, which could lead to erroneous cancellation of deceleration periods. The dynamically updated quantile values ​​can adapt to the magnitude differences at different measurement points, significantly improving parameter portability.

[0020] 3. The number of continuous monitoring times required for the upgrade of this invention m ↑ The number of continuous monitoring times required for downgrading m ↓ They are all different, and m ↓ > m ↑ This asymmetric design implements an engineering early warning logic that provides a fast response to acceleration events and strict requirements for degradation judgment, ensuring a rapid response when new acceleration events occur; it also maintains good robustness during occasional rate drops. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method for identifying and warning of deformation stages in rainfall-induced landslides according to the present invention.

[0022] Figure 2 This is a block diagram of the rain-induced landslide deformation stage identification and early warning system of the present invention.

[0023] Figure 3 This is a graph showing the original data curves of the monitoring area in Example 2.

[0024] Figure 4 This is a data curve generated by applying the method of the present invention to the original data of the monitoring area in Example 2. Detailed Implementation

[0025] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0026] Example 1 like Figure 1 As shown, the method for identifying and warning of deformation stages in rainfall-induced landslides according to the present invention includes the following steps: S1. Multiple high-precision GNSS measuring points are deployed at different locations on the landslide body (such as the upper unloading zone, the middle main deformation body, and the lower accumulation zone). A GNSS receiver compatible with three constellations (GPS+BDS+GLONASS) is preferably used, with a horizontal displacement accuracy better than ±2 mm and a vertical displacement accuracy better than ±5 mm, and a sampling rate of no less than 1 time / day. The cumulative displacement sequence of each measuring point is continuously collected, and the raw monitoring data is aligned to an equally spaced time series according to a unified time format to obtain the cumulative displacement sequence. The collected monitoring data is uploaded for calculation via wireless (4G / 5G) or wired network. In some embodiments, local storage can be provided to cope with communication interruptions, or InSAR, surface displacement gauges, inclinometers, etc., can be used to obtain the cumulative displacement sequence of each measuring point.

[0027] S2, calculate each monitoring time. t i Improved tangent angle β i The monitoring time mentioned above. t i Improved tangent angle The aforementioned For monitoring time t i The instantaneous displacement rate. That is... in, , For monitoring time t i , t i-1 The corresponding cumulative displacement value.

[0028] For monitoring time t i and the start time t 0 The average rate over the entire process. That is, , ε This is the preset minimum rate threshold. Minimum rate threshold ε Preferred 1×10⁻³ mm / d .

[0029] Improved tangent angle β i This reflects the degree of deviation of the instantaneous deformation rate at a certain moment from the average level of the entire process: gradual deformation stage , Smaller; Acceleration phase , It will rise rapidly.

[0030] S3. Obtain the displacement rate sequence based on the cumulative displacement sequence, and smooth the displacement rate sequence using Savitzky-Golay filtering; calculate the displacement rate sequence at each monitoring time point after smoothing. t i instantaneous acceleration ; in length of w The cumulative acceleration of the sliding window is obtained by summing all positive instantaneous accelerations within the sliding window. .

[0031] In some embodiments, median filtering can also be used to smooth the displacement rate sequence. The Savitzky-Golay filter window length can be selected as an odd number between 5 and 11, and the order is 2. In this invention, a window of 7 and an order of 2 are preferred for the best smoothing effect, which can suppress high-frequency noise on the order of ±5 mm / d while preserving the true peak value of 100+ mm / d in the later stage more completely.

[0032] Let the smoothed displacement rate sequence be... vismooth Then acceleration The calculation is as follows: In length wWithin the sliding window, only positive values ​​of acceleration are taken and summed to obtain the cumulative acceleration of the sliding window. ; in, w The sliding window length is [3, 5, 7, 10, 14] days, preferably [3, 5, 7, 10, 14]. w = 7 days. This indicator emphasizes the continuous accumulation of acceleration effects within the recent time window, effectively reflecting whether the landslide is in a state of continuous intensification at the current stage. Compared with the cumulative acceleration over the entire time period, the sliding window form only focuses on the acceleration amount in the current stage, structurally avoiding the level jump caused by the "residual heat" of historical accumulation continuing to generate during the plateau period.

[0033] S4, to improve the tangent angle and sliding window cumulative acceleration Using the input as the basis, the final warning level is generated through three layers of criteria: basic classification, level correction, and continuous smoothing. L i (3) The warning levels include safety level, attention level, warning level, alert level, and alarm level.

[0034] First, the basic grading is based on the improved tangent angle. β i Determine the basic early warning level L i (1) The specific standards are shown in Table 1 below.

[0035] Table 1 The 42° to 48° range serves as a buffer zone (corresponding to the ATTENTION level). The width of this buffer zone is based on the theoretical median of the improved tangent angle, approximately 45°, with a 3° tolerance on both sides. This is to mitigate frequent level jumps caused by fluctuations in the improved tangent angle around 45°. Local adjustments can also be made during engineering implementation based on the noise level at the measurement points.

[0036] Secondly, based on the cumulative acceleration of the sliding window Peak rate relative to basic warning level L i (1) Make corrections to obtain the correction level. L i (2) .

[0037] Peak rate is the monitoring time t i instantaneous displacement rate Greater than the 80th percentile of the instantaneous displacement rate of the previous 30 daysq 80 v the product with the coefficient r, that is q 80 v × r . When the number of valid samples before the monitoring time t i is less than 30, the existing historical samples are used to calculate the quantile value, or the level correction is not temporarily enabled.

[0038] When the basic warning level L i (1) is lower than the alert level ( L i (1) <ALERT) and the sliding window cumulative acceleration is greater than or equal to the 80% quantile of the instantaneous displacement rate in the 30 days before the monitoring time t i q 80 v ( Σa i (w) ≥ q 80 v ), the correction level L i (2) is the previous warning level of the basic warning level, that is L i (1) L i (2) = L i (1) + 1, the determined basic warning level is raised by one level.

[0039] When the basic warning level L i (1) is equal to the alert level ( L i (1) = ALERT), the sliding window cumulative acceleration is greater than or equal to the 80% quantile of the instantaneous displacement rate in the 30 days before the monitoring time t i q 80 v ( Σa i (w) ≥ q 80 ​​​v And monitoring time t i instantaneous displacement rate v i ≥ q 80 v × r Correction level L i (2) Basic warning level L i (1) The previous warning level; that is L i (1) = ALERT, L i (2) =ALARM alert level.

[0040] When the basic warning level L i (1) Equal to alert level ( L i (1) = ALERT), but monitoring time t i instantaneous displacement rate v i < q 80 v × r Maintain basic early warning level L i (1) Unchanged, i.e., the correction level remains unchanged. L i (2) Basic warning level L i (1) .

[0041] The physical meaning of the criterion is: when the tangent angle is already high, and the cumulative acceleration of the sliding window also reaches a historical high ( q 80 v Furthermore, when the current displacement rate shows a peak that significantly exceeds the recent background, the three types of evidence jointly point to a high level of risk, thus allowing the warning to jump directly to the highest level.

[0042] Furthermore, the final warning level L i (3) Determine using the following formula: m↑To upgrade the required number of continuous monitoring moments, m↓ The number of continuous monitoring moments required for downgrading, and m↑ > m ↓ ; such as in the direction of upgrade m↑ = 2, in the downgrade direction m↓ = 3; This enables the method of the present invention to respond more quickly when a new acceleration event occurs; it has better robustness to occasional rate drops, avoiding premature degradation before the acceleration event has ended. The warning levels are arranged from low to high as follows: safety level, attention level, warning level, alert level, and alarm level.

[0043] Compared with landslide early warning methods based on a single threshold, this invention can continuously and regularly transform isolated spikes caused by noise into "instantaneous events" that can be swallowed up by rules, resulting in a more continuous early warning sequence and better engineering readability.

[0044] In the above formula, c represents the monitoring time. t i Correction level L i (2) ; This represents the final warning level at the previous monitoring time. The correction level is for monitoring time k.

[0045] In some embodiments, a regional-level early warning can be triggered when multiple GNSS monitoring points in the monitoring area simultaneously reach the alert level. Using a deep learning-based displacement trend prediction module, the method of this invention can be applied based on predicted displacement as an auxiliary means of early warning.

[0046] This invention provides a system for identifying and warning of deformation stages in rainfall-induced landslides, used to implement the aforementioned method for identifying and warning of deformation stages in rainfall-induced landslides, specifically as follows: Figure 2 As shown, it includes a monitoring system, a data processing system, and an early warning system.

[0047] The monitoring system includes a GNSS displacement acquisition module and a data transmission module, used to acquire raw monitoring data and transmit the raw monitoring data to the data processing system.

[0048] The data processing system preprocesses the raw monitoring data, calculates the improved tangent angle, smooths the displacement rate, calculates the acceleration, and calculates the cumulative acceleration of the sliding window.

[0049] The early warning system generates the final early warning level through three layers of criteria: basic classification, level correction, and continuous smoothing. L i (3) .

[0050] In some embodiments, the rainfall-induced landslide deformation stage identification and early warning system further includes an early warning information output and a rainfall coupling verification module.

[0051] The warning information output is displayed on a display device with color codes corresponding to the warning level (e.g., green, blue, yellow, orange, and red for safety level, attention level, warning level, alert level, and alarm level, respectively) and time sequence. The warning information is pushed to preset receiving terminals (e.g., the manager's mobile phone or the dispatch center's large screen) through a communication network, and is also triggered by sound and light alarms through on-site alarm devices.

[0052] The rainfall coupling verification module collects rainfall data of the monitoring area through the monitoring system, preprocesses the rainfall data through the data processing system, and then performs left concatenation according to the time alignment of the cumulative displacement sequence. The physical rationality of the early warning result is externally verified by calculating the Pearson correlation coefficient and significance level.

[0053] In some embodiments, rainfall data is synchronously acquired via rain gauges or weather stations, with the rainfall sequence aligned with the GNSS measurement point time. A tipping bucket rain gauge with a resolution of at least 0.2 mm is preferred. The rainfall data is left-joined and aligned to the GNSS time series, with zeros added to confirm dates without rainfall; missing dates are interpolated, removed, or marked as missing. The Pearson correlation coefficient and its significance level are calculated for the rainfall and displacement rate sequences within the monitoring period at multiple lag days (e.g., 0, 1, 2, 3, 5, 7, 10, 14 days). When rainfall and displacement rate reach a significant correlation within a lag range of 3 to 7 days (e.g., p-value less than 0.05 at α = 0.05), a valid rainfall-driven evidence signal is output as an auxiliary basis for the final warning level upgrade in this invention. This mechanism provides external physical verification independent of displacement monitoring, further reducing the false alarm rate and making the warning conclusions more physically interpretable.

[0054] Example 2 This invention selects a 386-day diurnal monitoring sequence of a GNSS measuring point for a rainfall-induced landslide as the verification object. The cumulative displacement of this measuring point increased from 3.97 mm to 3152.83 mm, generally experiencing three stages: rapid growth in the early stage (approximately 64 days), a plateau and gradual change in the middle stage (approximately 8 months), and a re-acceleration in the later stage (approximately 49+9 days). The two acceleration periods highly correspond to the concentrated rainfall periods, belonging to a typical "long plateau + reactivation" composite displacement sequence, which can comprehensively test the early warning performance of this system under different deformation stages. The coupled response relationship between the cumulative displacement, displacement rate and rainfall process of the GNSS measuring point is shown in Figure 3.

[0055] The comprehensive early warning output results for the landslide deformation stage using the method described in this invention on the data sequence of the measuring points are as follows: Figure 4 As shown. Figure 4 (a) is the cumulative displacement time series curve; Figure 4 (b) is the improved tangent angle calculated based on the original displacement sequence. It can be seen that both acceleration periods reached the warning level of 80° or above, and the long platform in the middle section remained at a low level. Figure 4 (c) represents the cumulative acceleration of the sliding window, which shows a significant peak only during the two acceleration periods (maximum value of approximately 290), and remains close to zero during the plateau period; Figure 4 (d) is the final warning level color band after smoothing by three-layer criteria: both acceleration periods are stably identified as alert or alarm level (orange-red), and the middle plateau period maintains a stable safe level (green).

[0056] The period when the daily displacement rate is greater than 3 mm / d is defined as the acceleration period and used as the evaluation benchmark (Ground-Truth). The method described in this invention is compared with four conventional methods listed in the background art, and the quantitative data are shown in Table 2.

[0057] Table 2 As can be seen from the table, compared with the single tangent angle criterion method, the F1 index of this invention has been improved from 0.165 to 0.738, an improvement of 347%; the acceleration period coverage has been improved from 9.0% to 59.0%; and the false negative rate has been reduced from 91.0% to 41.0%.

[0058] Compared with the fixed rate threshold method, the false alarm rate of the present invention is reduced from 29.2% to 4.5%, a reduction of 84.6%, which significantly reduces false alarms during the plateau period while maintaining the event recognition capability.

[0059] Compared with the displacement master curve trend fitting method, the number of plateau period jumps in this invention is reduced from 8 to 3, which verifies the advantage of the sliding window form over the full-time cumulative form in suppressing the "memory effect".

[0060] Compared with the all-time cumulative acceleration method, the number of plateau jumps in the final model was reduced from 9 to 3, a decrease of 67%, which verifies the engineering value of the persistence constraint.

[0061] The final warning distribution of this invention is as follows: 259 days for the safety level, 11 days for the attention level, 43 days for the warning level, 62 days for the alert level, and 11 days for the alarm level. Both acceleration periods were stably captured, and the middle 7-month plateau period maintained a stable safety level, which is in line with engineering expectations.

[0062] Furthermore, this invention is applicable to days w∈[5, 14]. m ↑Within a reasonable parameter range of [2, 3], the F1 score of the present invention showed changes of less than 10%. Savitzky-Golay smoothing improved the F1 score by 20.2% and 13.2% compared to no smoothing and median filtering, respectively. This indicates that the present invention exhibits good robustness to key parameters, and that the selection of each key parameter has a clear physical or statistical meaning, allowing for localized adjustments based on the monitoring frequency and noise level of different measurement points.

[0063] according to Figure 3 Statistical results of the coupled response diagram of cumulative displacement, displacement rate and rainfall process at GNSS measuring points show that the correlation between daily rainfall and displacement rate reaches its peak after a 3-day lag, with a Pearson correlation coefficient of r=0.229 (significance level p<0.001). The correlation is further lower after a 7-day lag, with r=0.222 (p<0.001), reflecting that the main timescale for rainwater infiltration into the deeper slope through joint fissures and triggering deformation in the study area is 3–7 days. Considering the effects of rainfall accumulation and attenuation, the effective rainfall amount is significantly positively correlated with the improved tangent angle, with r=0.690 (p<0.001), verifying the physical rationality of using the improved tangent angle as the main criterion in this invention from an independent meteorological perspective.

[0064] This invention is applicable to landslides that accelerate significantly during the rainy season, such as steep rock slopes at dam sites of water conservancy projects, mountain highways, deep cuts and high embankment slopes along railways, steep slopes in open-pit mines, reservoir bank landslides and reservoir drawdown zones, slopes in urban built-up areas, and dangerous rock masses in abandoned mining areas. These landslides involve large engineering impact areas and high costs associated with false alarms. There are many monitoring points, requiring a unified early warning framework with strong parameter transferability. The coupling effect between rainfall and reservoir water level is significant, necessitating landslide early warning with an external verification mechanism.

Claims

1. A method for identifying and warning of deformation stages in rainfall-induced landslides, characterized in that, Includes the following steps: S1, the raw monitoring data is aligned into an equally spaced time series according to a unified time format to obtain the cumulative displacement series; S2, calculate each monitoring time. t i Improved tangent angle β i ; S3. Obtain the displacement rate sequence based on the cumulative displacement sequence, and smooth the displacement rate sequence using Savitzky-Golay filtering; calculate the displacement rate sequence at each monitoring time point after smoothing. t i instantaneous acceleration ; in length of w The cumulative acceleration of the sliding window is obtained by summing all positive instantaneous accelerations within the sliding window. ; S4, to improve the tangent angle and sliding window cumulative acceleration Using the input as the basis, the final warning level is generated through three layers of criteria: basic classification, level correction, and continuous smoothing. L i (3) The warning levels include safety level, attention level, warning level, alert level, and alarm level.

2. The method for identifying and warning of deformation stages in rainfall-induced landslides according to claim 1, characterized in that: In step S1, multiple GNSS measuring points are set up at different locations of the landslide body, with a sampling rate of no less than 1 time / day, and the cumulative displacement of each measuring point is continuously collected to obtain the cumulative displacement sequence of each measuring point.

3. The method for identifying and warning of deformation stages in rainfall-induced landslides according to claim 1, characterized in that: Each monitoring moment mentioned in step S2 t i Improved tangent angle The For monitoring time t i The instantaneous displacement rate; For monitoring time t i and the start time t 0 The average rate over the entire process; , ε This is the preset minimum rate threshold.

4. The method for identifying and warning of deformation stages in rainfall-induced landslides according to claim 1, characterized in that: The basic grading described in step S4 is based on the improved tangent angle. β i Determine the basic early warning level L i (1) ; The level correction is based on the cumulative acceleration of the sliding window. Peak rate relative to basic warning level L i (1) Make corrections to obtain the correction level. L i (2) The continuous smoothing process generates the final warning level. L i (3) Determine using the following formula: , m↑ To upgrade the required number of continuous monitoring moments, m↓ The number of continuous monitoring moments required for downgrading, and m↓ > m↑ c represents the monitoring time. t i Correction level L i (2) ; This represents the final warning level at the previous monitoring time. The correction level is for monitoring time k.

5. The method for identifying and warning of deformation stages in rainfall-induced landslides according to claim 4, characterized in that: The peak rate is the monitoring time. t i instantaneous displacement rate Greater than the 80th percentile of the instantaneous displacement rate of the previous 30 days q 80 v The product with the coefficient r, i.e. q 80 v × r .

6. The method for identifying and warning of deformation stages in rainfall-induced landslides according to claim 5, characterized in that: Monitoring time t i If the number of valid samples is less than 30, the quantile value is calculated using existing historical samples, or the rank correction is temporarily not enabled.

7. The method for identifying and warning of deformation stages in rainfall-induced landslides according to claim 5, characterized in that: When the basic warning level L i (1) Below the alert level and the cumulative acceleration of the sliding window Monitoring time t i 80th percentile of instantaneous displacement rate in the first 30 days q 80 v At that time, the correction level L i (2) Basic warning level L i (1) The previous warning level; When the basic warning level L i (1) Equal to alert level, cumulative acceleration of sliding window Monitoring time t i 80th percentile of instantaneous displacement rate in the first 30 days q 80 v And monitoring time t i instantaneous displacement rate v i ≥ q 80 v × r Correction level L i (2) Basic warning level L i (1) The previous warning level; When the basic warning level L i (1) Equal to alert level, but monitoring time t i instantaneous displacement rate v i < q 80 v × r Maintain basic early warning level L i (1) Unchanged, i.e., the correction level remains unchanged. L i (2) Basic warning level L i (1) .

8. The method for identifying and warning of deformation stages in rainfall-induced landslides according to claim 1, characterized in that: Final warning level The system uses color-coded displays to indicate the time sequence of events, and pushes warning information to preset receiving terminals for on-site audible and visual alarms.

9. A system for identifying and warning of deformation stages in rainfall-induced landslides, characterized in that: The method for identifying and warning of deformation stages of rainfall-induced landslides as described in any one of claims 1-8 includes a monitoring system, a data processing system, and an early warning system; The monitoring system includes a GNSS displacement acquisition module and a data transmission module, used to acquire raw monitoring data and transmit the raw monitoring data to the data processing system. The data processing system preprocesses the raw monitoring data, calculates the improved tangent angle, smooths the displacement rate, calculates the acceleration, and calculates the cumulative acceleration of the sliding window. The early warning system generates the final early warning level through three layers of criteria: basic classification, level correction, and continuous smoothing. L i (3) .

10. The rainfall-induced landslide deformation stage identification and early warning system according to claim 9, characterized in that: It also includes a warning information output and rainfall coupling verification module; the warning information output displays color codes and time series corresponding to the warning level through a display device, pushes the warning information to a preset receiving terminal through a communication network, and provides audible and visual alarms through an on-site alarm device; The rainfall coupling verification module collects rainfall data of the monitoring area through the monitoring system, preprocesses the rainfall data through the data processing system, and then performs left concatenation according to the time alignment of the cumulative displacement sequence. The physical rationality of the early warning result is externally verified by calculating the Pearson correlation coefficient and significance level.

Images