Ice lake outburst small watershed identification method and system based on valley landform characteristic indexes

Abstract

The invention provides a glacial lake outburst small watershed identification method and system based on valley landform characteristic indexes, and belongs to the technical field of glacial lake outburst small watershed identification, and the method comprises the steps: S1, obtaining elevation data and multi-stage remote sensing image data; s2, identifying a channel and a glacial lake, determining the position of the channel and the position of the glacial lake, and calculating the average gradient of the upstream and the downstream of the glacial lake along the channel direction; s3, obtaining a b value of the upstream of the glacial lake in the vertical channel direction and a b value of the downstream of the glacial lake in the vertical channel direction; s4, calculating the GLBN value of the small watershed according to the average slope of the upstream of the glacial lake in the channel direction, the average slope of the downstream of the glacial lake in the channel direction, the b value of the upstream of the glacial lake in the direction perpendicular to the channel direction and the b value of the downstream of the glacial lake in the direction perpendicular to the channel direction; and when the GLBN value is greater than an identification threshold value, determining that the small watershed is a glacial lake outburst small watershed. According to the method, ice lake outburst small watershed identification of high mountain areas without people and areas without historical data records is carried out through gully landform characteristic indexes of slope mutation and valley type conversion.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of identification of ice lake outburst small watershed in high-altitude mountainous area, and particularly relates to an ice lake outburst small watershed identification method and system based on valley geomorphic feature index. BACKGROUND

[0002] In order to better prevent ice lake outburst disaster events and reduce losses, it is necessary to scientifically and accurately achieve early identification of ice lake outburst small watershed. At present, there are mainly two identification methods as follows:

[0003] 1. The history of ice lake outburst is grasped through literature, data, etc. However, most of the ice lake outburst floods occur in high mountain uninhabited areas, and are low-frequency events with long return period. The records of ice lake outburst events in the Himalayan region of the Qinghai-Tibet Plateau started in the 19th century. The early records were concentrated in the lower reaches of the main stream and were relatively vague. In addition, only in the areas with human activities, there are records of ice lake outburst disasters. Therefore, for the high mountain uninhabited areas crossed by linear engineering such as roads at present, the record data of such events are extremely scarce. Therefore, the strong dependence of this method on data leads to certain limitations in time and space scales, including the short time scale of ice lake outburst event records, and the spatial limitation of areas with ice lake outburst disaster records to areas with human activities.

[0004] 2. Whether the ice lake in the upstream of the small watershed has outburst history is analyzed through multi-period remote sensing images. However, remote sensing images also have the phenomenon that the time scale does not match the low frequency characteristics of ice lake outburst. The technology of monitoring ground objects through remote sensing images started in the middle and late 20th century. The first satellite of Landsat was launched in 1972, so the history of ice lake outburst before 1972 cannot be obtained through Landsat images. In addition, the resolution of remote sensing images also has certain limitations on the early prediction of ice lake outburst small watershed. The resolution of early remote sensing images is low, and the error in the interpretation of ground objects is large. Until the 21st century, the resolution of QuickBird images was improved to 0.3m. At the same time, the identification method based on remote sensing images is only for ice lake itself, ignoring the description of ice lake to the downstream landform. Moreover, the identification of ground objects through images can only identify the response of gully or valley to ice lake outburst in the horizontal plane, and it is difficult to identify the response in the direction of gravity, which is the most direct description of the impact of ice lake outburst flood on landform. Therefore, this method still has limitations in time scale, and it is difficult to explain the impact of ice lake outburst on landform in three-dimensional space scale.

[0005] Meanwhile, it is also found that the ice lake outburst disasters often occur repeatedly in the same area, so the disasters have a long time scale of continuous influence and characterization on the landform. However, the frequency of such large-scale ice lake outburst disasters is low, some of which are beyond the historical records, and some of which occur in high-altitude uninhabited areas. Therefore, it is difficult to directly determine whether the ice lake outburst has occurred in the small watershed in history in the area without historical records or in the high-altitude uninhabited area. SUMMARY

[0006] The application provides an ice lake outburst small watershed identification method and system based on a valley landform feature index. The method is based on the quantitative landform features perpendicular to the channel direction and parallel to the channel direction, and the ice lake outburst small watershed is identified based on the valley landform feature index, so as to more accurately, conveniently and efficiently identify the ice lake outburst small watershed in the high mountain area in an early stage.

[0007] The ice lake outburst small watershed identification method based on the valley landform feature index comprises the following steps:

[0008] S1. Obtain the elevation data and multi-period remote sensing image data in the ridge line range of the identified small watershed;

[0009] S2. Identify the channel and the ice lake according to the multi-period remote sensing image data, further determine the channel position and the ice lake position according to the elevation data, and obtain the longitudinal profile along the channel direction according to the elevation data, and calculate the average slope of the ice lake upstream along the channel direction and the average slope of the ice lake downstream along the channel direction;

[0010] S3. Draw the transverse profile from the ice lake position to the upstream and downstream in the vertical channel direction according to the ice lake position and the channel position, obtain the b value of the U-shaped or V-shaped valley landform feature in the vertical channel direction upstream of the ice lake and the b value of the vertical channel direction downstream of the ice lake by fitting the transverse profile data;

[0011] S4. Calculate the GLBN value of the small watershed according to the average slope of the ice lake upstream along the channel direction, the average slope of the ice lake downstream along the channel direction, the b value of the vertical channel direction upstream of the ice lake, and the b value of the vertical channel direction downstream of the ice lake: ;

[0012] wherein, the b value of the vertical channel direction upstream of the ice lake, the b value of the vertical channel direction downstream of the ice lake, the average slope of the ice lake upstream along the channel direction, the average slope of the ice lake downstream along the channel direction, α is a wide valley and narrow valley index weight coefficient, β is a channel slope index weight coefficient, and γ is a comprehensive adjustment coefficient;

[0013] When the GLBN value is greater than the identification threshold, the small watershed is determined to be an ice lake outburst small watershed.

[0014] In the present specification, the identification threshold is determined by 80% of the minimum value of the GLBN value of the sample database.

[0015] In the present specification, α, β and γ are determined by GLBN identification performance test on different parameter combinations and minimum error method based on the sample database formed by typical ice lake outburst cases.

[0016] In the present specification, the determination of the ice lake position is as follows: through optical image recognition, or using contour lines generated from elevation data, the area with closed circular or elliptical lowlands and located at the end of the glacier is determined as the ice lake position; the position of the channel is determined by the V-shaped bending feature of the contour line.

[0017] In the present specification, the calculation method of the average slope of the upstream and downstream of the ice lake along the channel direction is as follows: for the upstream and downstream channels of the ice lake, the slope values of each section are calculated by using an equally spaced sliding window, and then the average values are taken respectively.

[0018] In the present specification, the upstream b value and the downstream b value of the ice lake in the vertical channel direction are obtained by fitting the cross-sectional data, and the specific method is as follows: the cross-sectional data is extracted by the spatial analysis module of the ArcGIS software, and the cross-sectional data is divided into left and right parts according to the lowest point of the elevation, and the function y=ax b is used to fit the left and right part data respectively, and the average value is taken as the b value of the corresponding position, and then the b values of the upstream and downstream of multiple cross sections are averaged respectively to obtain the upstream b value and the downstream b value in the vertical channel direction. b In the present specification, the determination of the wide valley and narrow valley index weight coefficient, the channel slope index weight coefficient and the comprehensive adjustment coefficient, and the determination of the GLBN value identification threshold use the optimization method of surface fitting, random forest and gradient boosting tree fusion, which specifically includes:

[0019] Firstly, a nonlinear relationship between parameter combinations and identification errors is established by surface fitting, taking the wide valley and narrow valley index weight coefficient, the channel slope index weight coefficient and the comprehensive adjustment coefficient as input, and taking the identification error as output, to construct a quadratic surface model, and the preliminary optimized parameter value and fitting residual are obtained;

[0020] Secondly, the fitting residual, the small watershed area and the glacier type are taken as input features, and the random forest model is used for parameter correction, and the parameter correction amount is generated by ensemble learning of multiple decision trees, and the final parameters suitable for regional characteristics are obtained by superimposing the preliminary optimized parameters and the correction amount;

[0021] ​

[0022] Finally, the final parameters, the small watershed area, the glacier type and the fitting residual are taken as inputs to generate a dynamic discrimination threshold by using a gradient boosting tree model, the dynamic threshold is dynamically adjusted according to the characteristics of the small watershed, when the calculated GLBN value is greater than the dynamic threshold, it is determined that the small watershed is an ice lake outburst small watershed.

[0023] In the specification, when it is determined that the small watershed is an ice lake outburst small watershed, a target area is determined for the arrangement of an ice lake outburst monitoring instrument in the small watershed, the monitoring instrument includes a radar flow meter and a vibration signal collector.

[0024] In the specification, when it is determined that the small watershed is an ice lake outburst small watershed, an area of focus through satellite is assisted to be determined, the content of the area of focus includes the ice lake area change in the small watershed and the extreme temperature and precipitation conditions.

[0025] The ice lake outburst small watershed discrimination system based on the valley geomorphic feature index applies the ice lake outburst small watershed discrimination method based on the valley geomorphic feature index in any one of the above, and the ice lake outburst small watershed discrimination system based on the valley geomorphic feature index comprises:

[0026] A data acquisition module is configured to acquire elevation data and multi-period remote sensing image data in a ridge line range of a small watershed to be discriminated;

[0027] A position determination and calculation module is configured to identify a channel and an ice lake according to the multi-period remote sensing image data, determine positions of the channel and the ice lake according to the elevation data, acquire a longitudinal profile along a channel direction according to the elevation data, and calculate an average slope of an upstream of the ice lake along the channel direction and an average slope of a downstream of the ice lake along the channel direction;

[0028] A b value calculation module is configured to draw a transverse profile in an upstream and a downstream of the ice lake respectively from the position of the ice lake as a starting point in a vertical channel direction, acquire a vertical channel direction upstream b value and a vertical channel direction downstream b value capable of representing a U-shaped or V-shaped valley geomorphic feature by fitting the transverse profile data according to the positions of the ice lake and the channel;

[0029] A discrimination module is configured to calculate a GLBN value of the small watershed according to the average slope of the upstream of the ice lake along the channel direction, the average slope of the downstream of the ice lake along the channel direction, and the vertical channel direction upstream b value and the vertical channel direction downstream b value of the ice lake.

[0030] wherein, the vertical channel direction upstream b value is the vertical channel direction downstream b value is the average slope of the upstream of the ice lake along the channel direction is ​is the average slope of the downstream of the ice lake along the channel direction, a is the weight coefficient of the wide valley and narrow valley index, β is the weight coefficient of the channel slope index, and γ is the comprehensive adjustment coefficient;

[0031] When the GLBN value is greater than the discrimination threshold, the small watershed is determined to be an ice lake outburst small watershed.

[0032] The embodiments of the present specification can at least achieve the following beneficial effects:

[0033] Breaking through the time and space limit, solving the identification problem in the area without data: abandoning the dependence on historical ice lake outburst records and remote sensing image time scale, based on the long-term portrayal of ice lake outburst on landform (slope sudden change, valley type conversion, etc. nonlinear response), the identification of ice lake outburst small watershed in the plateau uninhabited area and the area without historical data record is realized, which makes up for the limitations of existing methods in time and space scale and solves the technical bottleneck of risk assessment in the area without data.

[0034] Improving the identification accuracy and reliability: by integrating the slope change along the channel direction (slow upstream and steep downstream) and the valley type conversion (U-shaped valley upstream and V-shaped valley downstream) in the vertical channel direction, the GLBN index is constructed to quantify the landform response intensity, the parameters (a=1.2, β=0.9, γ=0.5) are optimized through six typical cases and verified through sensitivity analysis (the influence of ±15% parameter fluctuation is less than 3%), which ensures the robustness and accuracy of the identification model.

[0035] Lower application threshold and enhance operability: the data sources are diverse and flexible, the high-precision DEM can be obtained through open source channels (such as geographic spatial data cloud) or unmanned aerial vehicles, and the multi-period remote sensing images can be obtained through Sentinel, Landsat and other platforms, which is suitable for the scene of difficult field measurement in high-altitude uninhabited area, low cost and easy to promote; the positions of ice lake and channel can be determined through contour shape, optical image and the like, the operation process is clear and practical.

[0036] Supporting disaster prevention and mitigation practice: providing precise target area for ice lake outburst disaster prevention and control, when the identification is an ice lake outburst small watershed, the scientific arrangement of monitoring instruments such as radar flow meter and vibration signal collector can be guided, and satellite can be assisted to focus on ice lake area change and extreme weather, forming a "point-line-surface" monitoring network, and improving the disaster prevention and mitigation ability of major projects (such as roads and power stations) in high-altitude mountainous area. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0038] Figure 1 FIG. 1 is a schematic diagram of a method for identifying an ice lake breaching small watershed based on a valley geomorphic feature index according to some embodiments of the present application.

[0039] Figure 2 FIG. 2 is a schematic diagram of a process for identifying an ice lake breaching small watershed based on a valley geomorphic feature index according to some embodiments of the present application.

[0040] Figure 3 FIG. 3 is a schematic diagram of a typical ice lake breaching small watershed GLBN value distribution according to some embodiments of the present application.

[0041] Figure 4a FIG. 4 is a schematic diagram of a parameter a sensitivity analysis according to some embodiments of the present application.

[0042] Figure 4b FIG. 5 is a schematic diagram of a parameter b sensitivity analysis according to some embodiments of the present application.

[0043] Figure 4c FIG. 6 is a schematic diagram of a parameter g sensitivity analysis according to some embodiments of the present application.

[0044] Figure 5 FIG. 7 is a schematic diagram of a correlation between a typical ice lake breaching sample geomorphic parameter and a GLBN value according to some embodiments of the present application. DETAILED DESCRIPTION

[0045] In the following, only certain example embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the present application. Therefore, the drawings and the description are considered to be exemplary in nature rather than limiting.

[0046] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0047] As shown in FIG. 1, the present embodiment provides a method for identifying an ice lake breaching small watershed based on a valley geomorphic feature index, comprising: Figure 1

[0048] S1. Obtain elevation data and multi-period remote sensing image data within the ridge line range of the identified small watershed;

[0049] S2. Identify the gully and the ice lake according to the multi-period remote sensing image data, further determine the gully position and the ice lake position according to the elevation data, obtain a longitudinal profile along the gully direction according to the elevation data, and calculate the average slope of the ice lake upstream along the gully direction and the average slope of the ice lake downstream along the gully direction;

[0050] ​S3. According to the ice lake position and the channel position, a cross section is drawn from the ice lake position to the upstream and downstream in the vertical channel direction, respectively, and the b value of the upstream of the ice lake in the vertical channel direction and the b value of the downstream of the ice lake in the vertical channel direction are obtained by fitting the cross section data, which can represent the U-shaped or V-shaped valley landform characteristics;

[0051] S4. According to the average slope of the upstream of the ice lake along the channel direction, the average slope of the downstream of the ice lake along the channel direction, the b value of the upstream of the ice lake in the vertical channel direction and the b value of the downstream of the ice lake in the vertical channel direction, the GLBN value of the small watershed is calculated: ;

[0052] wherein, is the b value of the upstream of the ice lake in the vertical channel direction, is the b value of the downstream of the ice lake in the vertical channel direction, is the average slope of the upstream of the ice lake along the channel direction, is the average slope of the downstream of the ice lake along the channel direction, α is the wide valley-narrow valley index weight coefficient, β is the channel slope index weight coefficient, and γ is the comprehensive adjustment coefficient;

[0053] When the GLBN value is greater than the discrimination threshold value, the small watershed is determined to be an ice lake breaching small watershed.

[0054] In some embodiments, the discrimination threshold value is determined by 80% of the minimum value of the GLBN value of the sample database.

[0055] In some embodiments, α, β and γ are determined by GLBN discrimination performance test on different parameter combinations and minimum error method based on the sample database formed by typical ice lake breaching cases.

[0056] In some embodiments, the determination of the ice lake position is as follows: through optical image recognition or contour line generation using elevation data, the region with closed circular or elliptical low land and located at the end of the glacier is determined as the ice lake position; the channel position is determined by the V-shaped bending feature of the contour line.

[0057] In some embodiments, the calculation of the average slope of the upstream and downstream of the ice lake along the channel direction is as follows: for the upstream and downstream channels of the ice lake, the slope values of each section are calculated by using the equal-interval sliding window method, and then the average values are taken, respectively.

[0058] In some embodiments, the b value of the upstream and the b value of the downstream are obtained by the function y=ax b The b value of the upstream in the vertical channel direction and the b value of the downstream in the vertical channel direction are obtained by fitting the cross section data as follows: the cross section data is divided into left and right parts according to the lowest point of the elevation by using the spatial analysis module of the ArcGIS software, and the b value of the upstream in the vertical channel direction and the b value of the downstream in the vertical channel direction are obtained by the function y=ax bFit the data from the left half and the right half, take the average value as the b value at the corresponding position, and then take the average value of the b values ​​of multiple cross sections upstream and downstream to obtain the b value of the upstream of the glacial lake in the vertical channel direction and the b value of the downstream of the glacial lake in the vertical channel direction. x: represents the horizontal distance (unit: m) from a point in the cross section to the lowest point (bottom) of the gully, that is, the horizontal length extending from the bottom of the gully to the two sides of the hillside in the direction perpendicular to the gully. y: Represents the elevation (unit: m) at position x on the cross section, that is, the terrain height at a horizontal distance x from the bottom of the ditch; a: The coefficients of the fitting function, reflecting the overall elevation benchmark of the cross-section (related to the overall depth of the valley). b: The exponent of the fitting function (i.e., the superscript b of x), a key indicator used to characterize the features of U-shaped or V-shaped valley landforms.

[0059] In some embodiments, the determination of the weighting coefficients for the wide and narrow valley indicators, the weighting coefficients for the gully slope indicator, and the comprehensive adjustment coefficient, as well as the determination of the GLBN value identification threshold, adopts an optimization method that combines surface fitting, random forest, and gradient boosting tree, specifically including:

[0060] First, a nonlinear relationship between parameter combination and identification error is established through surface fitting. Using the weight coefficients of wide and narrow valley indicators, the weight coefficients of gully slope indicators, and the comprehensive adjustment coefficient as inputs, and the identification error as output, a quadratic surface model is constructed, and the preliminary optimized parameter values ​​and fitting residuals are obtained by solving.

[0061] Secondly, the fitting residual, small watershed area, and glacier type are used as input features. A random forest model is used for parameter correction. Through ensemble learning of multiple decision trees, parameter correction values ​​are generated. The preliminary optimized parameters are superimposed with the correction values ​​to obtain the final parameters adapted to the regional characteristics.

[0062] Finally, the final parameters, watershed area, glacier type, and fitting residual are used as inputs, and a gradient boosting tree model is used to generate a dynamic identification threshold. This dynamic threshold is dynamically adjusted according to the characteristics of the watershed. When the calculated GLBN value is greater than this dynamic threshold, the watershed is determined to be a glacial lake outburst watershed.

[0063] In some embodiments, when a small watershed is determined to be a glacial lake outburst watershed, a target area is determined for the deployment of glacial lake outburst monitoring instruments in the small watershed, the monitoring instruments including radar flow meters and vibration signal collectors.

[0064] In some embodiments, when a small watershed is determined to be a glacial lake outburst watershed, it is used to assist in identifying areas of key interest via satellite. The key interest includes changes in the glacial lake area and extreme temperatures and precipitation within the small watershed.

[0065] The technical concept of the present application is as follows:

[0066] The present application finds that the geomorphology of the upstream and downstream of the ice lake outburst in the middle and eastern sections of the Himalayas is significantly different by analyzing the geomorphic characteristics of 6 typical ice lake outburst small watershed gullies in the middle and eastern sections of the Himalayas. Specifically, the upstream of the ice lake presents a U-shaped valley along the gully direction, while the downstream presents a deep V-shaped valley. The slope gradient is gentler in the upstream and steeper in the downstream in the vertical direction of the gully. Therefore, an ice lake outburst small watershed identification method based on gully geomorphic feature indexes is proposed. This method is a process-driven identification method based on geomorphic response. The core advantage of this method is to abandon the dependence on historical ice lake outburst records and to infer the occurrence of ice lake outburst events by the evolution results of geomorphology. This method does not use simple superposition of static indexes, but identifies the nonlinear response of geomorphology in the time scale, including slope gradient mutation and valley type conversion, reflecting the erosion energy caused by ice lake outburst and the influence of spatial transmission. This method can identify ice lake outburst small watersheds in unpopulated areas or areas without historical records based on the principle of replacing historical records with geomorphic response processes and traces. It is especially suitable for unpopulated areas and areas without data records on the plateau, which makes up for the problems of insufficient length of remote sensing sequence and lack of historical records. It determines the target area for subsequent ice lake outburst monitoring instruments, and focuses on the changes of small watershed ice lake area and extreme temperature and precipitation through satellites, solving the strong dependence on historical data records and the limitations in time scale and spatial scale in existing identification methods, realizing risk assessment in data-free areas, and breaking through the time and space limitations of the existing technology. The ice lake outburst small watershed identification process based on gully geomorphic feature indexes is shown in Figure 2 .

[0067] The ice lake outburst small watershed identification method based on gully geomorphic feature indexes comprises the following steps:

[0068] Step S1, obtain high-precision DEM elevation data and multi-period remote sensing image data within the ridge line range of the identified small watershed, and construct a geomorphology-hydrology coupled model of the identified small watershed;

[0069] Step S2, identify the gully and the ice lake according to the multi-period remote sensing image data, and determine the position of the ice lake. Obtain the longitudinal profile along the gully direction according to the DEM data, and calculate the slope gradient of the upstream and downstream of the ice lake along the gully direction and ;

[0070] Step S3, according to the position of the ice lake and the position of the gully, draw a cross section every 50m upstream and downstream of the ice lake position in the vertical direction of the gully, and fit the longitudinal profile by a y=ax b function (fitting the relationship between the elevation (y) of the cross section and the horizontal distance (x), a is the coefficient of the fitting function), to obtain the b value which can represent the U-shaped or V-shaped gully geomorphic feature, and ;

[0071] x: represents the horizontal distance (unit: m) from a point in the cross section to the lowest point of the channel (channel bottom), that is, the horizontal length extending from the channel bottom to the two sides of the hillside in the direction perpendicular to the channel;

[0072] y: represents the elevation (unit: m) at the x position on the cross section, that is, the terrain height at a horizontal distance of x from the channel bottom;

[0073] a: the coefficient of the fitting function, reflecting the overall elevation reference of the cross section (related to the overall depth of the valley);

[0074] b: the index of the fitting function (i.e., the upper index of x), used to represent the key indicator of U-shaped or V-shaped valley landform characteristics;

[0075] Step S4, according to the results of S2 and S3, by calculating the GLBN (Glacier Lake Basin Index) value of the identified small watershed, when the value is greater than 1.62, the identified small watershed has occurred ice lake outburst in history, and the possibility of future ice lake outburst is relatively high, which belongs to the ice lake outburst small watershed. Wherein, a is the weight coefficient of the wide valley and narrow valley index, which is taken as 1.2 here; β is the weight coefficient of the channel slope index, which is taken as 0.9 here; γ is the comprehensive adjustment coefficient, which is taken as 0.5 here. The values of the parameters are based on the sample database (6 cases), and the GLBN identification performance test is performed on different parameter combinations. The minimum error method is used to determine that a=1.2, β=0.9, γ=0.5 is the optimal combination. In addition, the sensitivity analysis (such as Figure 4a , Figure 4b and Figure 4c ) is used to verify that the identification result changes less than 3% under ±15% parameter fluctuation, and the model has strong robustness.

[0076] In step S1, the elevation data can be obtained from websites such as geographic spatial data cloud, ASTER GDEM data, or higher precision DEM data obtained through on-site investigation of unmanned aerial vehicles. The multi-period remote sensing image can be Sentinel, Landsat data, or obtained through Google Earth platform and Long Light Satellite channel.

[0077] In step S1, the resolution of high-precision DEM elevation data is 30m or 12.5m or higher precision of centimeter-level unmanned aerial image.

[0078] In step S2, the determination of the ice lake and the channel position is the prerequisite of the present technology. The ice lake position can be obtained by optical image. For small watershed with small ice lake area, contour lines can be generated by high-precision DEM data. The region with closed circular / elliptical low land and located at the end of the glacier is the position of the ice lake. The channel position can also be determined by the "V" type bending feature of the contour line.

[0079] In step S2, and respectively represent the average slope of the longitudinal profile along the channel direction of the upstream and downstream of the ice lake. The slope should be calculated by section according to the DEM and contour line data. Specifically, for the upstream and downstream channels of the ice lake, the slope value of each section is calculated by using a sliding window with an interval of 100 m, and then the average slope of the longitudinal profile along the channel direction of the upstream and downstream of the ice lake is obtained.

[0080] In step S3, and respectively represent the average value of the b value of the multiple cross-sectional topographic indexes of the upstream and downstream of the ice lake with an interval of 50 m in the vertical direction of the channel. Specifically, after determining the cross section, the cross section data is extracted by the spatial analysis module of the ArcGIS software. The cross section data is divided into left half and right half according to the lowest point of the elevation, and then the left half data and the right half data are fitted by the y=ax b function respectively, and finally the average value is taken as the b value which can represent the U-shaped or V-shaped valley topographic feature of the cross section.

[0081] In step S4, based on the typical ice lake outburst small watershed in the middle and east sections of the Himalayas, , and The GLBN value of each small watershed is calculated. In the method, since the method is mainly based on the slope change and valley type conversion caused by the ice lake outburst in the topography, the decrease of the b value from the upstream of the ice lake to the downstream of the ice lake indicates that the wide valley is converted into the narrow valley, and the change of the valley type is reflected, that is, the ice erosion is converted into the water erosion; and the increase of the i value from the upstream to the downstream indicates that the slope is steep, and the strong gravity erosion process of the outburst flood is embodied. The product of the two reflects the nonlinear response strength of the outburst topography through the coupling of the exponential and logarithmic terms. The discrimination threshold 1.62 is determined according to 80% of the minimum value of the GLBN value of six typical ice lake outburst cases such as Gongbatongsa Tsho (E86.0659, N28.0744), Chhubung (E86.4736, N27.8632), TamPokhari (E86.8448, N27.7425), Nagma Pokhari (E87.7627, N27.8653), Upper Langbu Tsho (E86.4487, N27.9430), and Chubda Tsho (E90.7058, N28.0213) in the middle and eastern sections of the Himalayas.

[0082] The advantage of the present application is that:

[0083] 1. The present application is based on the principle of the response of the small watershed downstream channel topography to the historical ice lake outburst and the repeated occurrence of the ice lake outburst disaster, and the ice lake outburst small watershed is identified according to the slope change of the longitudinal section along the channel direction and the U-shaped or V-shaped morphological change of the transverse section perpendicular to the channel direction and other topographic indexes. The method directly considers the continuous description and influence of the small watershed historical ice lake outburst on the topography, discards the dependence on the historical ice lake outburst record, and does not use the simple superposition of static indexes, but identifies the nonlinear response of the topography in the time scale, including the slope change and the valley type conversion. In addition, the method can also identify in a larger time and space scale, and determine the target area for the installation of the subsequent ice lake outburst monitoring instrument and the analysis of the long-term meteorological data of the ice lake. For example, when GLBN>1.62, it is recommended to arrange the radar flow meter and the vibration signal collector in the downstream channel of the ice lake, and pay attention to the change of the ice lake area in the small watershed and the extreme temperature and precipitation through the satellite, and form a “point-line-surface” key monitoring network. The method solves the technical problems of the strong dependence on the historical data of the existing identification method, the difficulty in carrying out the work in the uninhabited area and the certain limitation of the time and space scale, realizes the risk assessment in the data-free area, and breaks through the time and space limitation of the existing technology.

[0084] 2. The research object and scenario identified in this invention are quite unique. Because existing research shows that large-scale glacial lake outburst floods exhibit spatial repetition, scientific prediction of glacial lake outburst flood watersheds is particularly important at this stage, especially in high-altitude and cold engineering areas. Then, based on the geomorphological features of six typical glacial lake outburst flood watersheds in the central and eastern Himalayas, the significant differences in geomorphological features between the upstream and downstream of glacial lake outburst floods are studied, further determining the criteria used to characterize historical glacial lake outburst flood watersheds. and The indicators, and then the GLBN values ​​are calculated to determine the identification threshold, which is conducive to more accurate and effective identification of glacial lake outburst watersheds.

[0085] 3. This invention sets the high-precision DEM elevation data to be obtainable from websites such as Geospatial Data Cloud (GDS) and ASTER GDEM, or even higher-precision DEM data obtained through field survey drones. Its advantages include multiple data sources adapting to more scenarios; in particular, the open-source DEM data is convenient, fast, and low-cost, overcoming the difficulty of field measurements in high-altitude, uninhabited areas.

[0086] 4. In this invention, the location of glacial lakes can be determined using optical imagery. For small watersheds with relatively small glacial lake areas, high-precision DEM data can be used to generate contour lines. The location of a glacial lake is indicated by a closed circular / elliptical lowland formed by these contour lines, located at the glacier terminus. Similarly, the location of channels can be determined using the "V"-shaped curvature of the contour lines. This ensures that the determination of glacial lake and channel locations is highly operable and reliable.

[0087] 5. This invention will The criteria for identifying a small watershed as having historically experienced glacial lake outburst floods are as follows: α is the weighting coefficient for the wide / narrow valley index, set to 1.2 here; β is the weighting coefficient for the channel slope index, set to 0.9 here; and γ is the comprehensive adjustment coefficient, set to 0.5 here. For details on the GLBN threshold and the basis for the values ​​of each parameter, please refer to the appendix. Figure 3 , Figure 4a , Figure 4b and Figure 4c and Figure 5 Glacial lakes meeting these conditions have a relatively high probability of future outburst flooding and are considered small glacial lake outburst flood basins. Specifically, The larger the value, the greater the slope of the downstream channel of the glacial lake along the channel direction. The smaller the value, the closer the valley shape of the downstream glacial lake in the vertical channel direction is to a V-shaped valley. and The threshold for the index is derived from 80% of the lowest values ​​of the geomorphic characteristic parameters of six typical glacial lake outflow watersheds in the central and eastern Himalayas. Its advantages lie in combining the fundamental theories of repeatability of glacial lake outflow events and the feedback of geomorphology to the strong erosion caused by glacial lake outflow events found in existing studies, and in focusing on important indicators derived from field cases that can characterize the geomorphology of the upstream and downstream areas of glacial lake outflow events in both parallel and perpendicular river directions.

[0088] It should be noted that the identification method in this invention is applicable to glacial lake outburst watersheds developed in tributary gullies, which are the most widespread and fastest-growing type of glacial lake under the current climate warming background, especially in rapidly retreating monsoon maritime glacier areas. The identification method comprehensively considers the recurring nature of glacial lake outburst disasters and the long-term, continuous impact and characterization of glacial lake outbursts on the geomorphological evolution of small watersheds, thereby enabling early identification of glacial lake outburst watersheds in areas without historical records or in high-altitude uninhabited areas.

[0089] To verify the technical effects of the present invention, the Gongbatongsha Tso glacial lake outburst event that occurred on July 5, 2016, on the southern edge of the Himalayas is described in detail below:

[0090] A method for identifying glacial lake outburst flood basins based on gully geomorphological features includes the following steps:

[0091] S1. Taking the Zhangzangbu small watershed, where the Gongbatongshacuo Glacier Lake is located, as the research object, DEM elevation data with a resolution of 12.5m were obtained for the entire watershed. In addition, high-resolution remote sensing images of the small watershed were obtained through platforms such as Google Earth.

[0092] S2, such as Figure 2 As shown in the flowchart, the location of the glacial lake and the channel are determined based on high-precision DEM elevation data. Then, a longitudinal profile along the channel direction is obtained, and the slope upstream and downstream of the glacial lake along the channel direction is calculated. and Calculation results show that the slope of the upper reaches of the Gongbatongsha Lake in the Zhangzangbu small watershed... The downstream slope is 0.16. It is 0.20.

[0093] S3. Draw cross sections at 50m intervals upstream and downstream along the vertical channel direction, starting from the location of the glacial lake, and use y=ax b Function fitting of longitudinal profiles and calculation results show the geomorphic indicators of the upper reaches of Gongbatongsha Lake in the Zhangzangbu small watershed. The value is 1.30, and the downstream geomorphological index is... It is 0.93.

[0094] S4, According to the results of S2 and S3, the GLBN value of Zhangzangbu small watershed is 2.02, which is greater than 1.62, so Zhangzangbu gully can be determined as an ice lake outburst small watershed.

[0095] In some embodiments, the nonlinear relationship between the parameters (a, b, g) and the GLBN identification error is captured by surface fitting, the parameter weight distribution is optimized by random forest, and the personalized threshold is dynamically generated by gradient boosting tree, as follows:

[0096] Optimization model for identifying ice lake outburst small watershed based on surface fitting, random forest and gradient boosting tree

[0097] I. Overall architecture of the model

[0098] The model captures the nonlinear relationship between the parameters and the identification error by surface fitting (SF), optimizes the parameter weight by random forest (RF), and generates dynamic threshold by gradient boosting tree (GBT). The three form a closed loop through "feature transfer-error correction-threshold adaptation".

[0099] II. Surface fitting model (SF): preliminary optimization of parameters

[0100] Core role: Through quadratic surface fitting, a nonlinear mapping between parameters (a, b, g) and identification error is established, and preliminary optimized parameters are output.

[0101] 1. Model construction

[0102] Input variables: parameter combination (a, b, g), where , , (Rational range based on previous research).

[0103] Output variable: identification error , calculation formula:

[0104] ;

[0105] N indicates the number of samples that actually outburst but GLBN≤1.62 or actually not outburst but GLBN>1.62, is the total number of all small watershed samples participating in the error calculation.

[0106] Fitting function: quadratic surface function (considering nonlinearity and computational efficiency):

[0107] ;

[0108] where, is the surface fitting coefficient.

[0109] ​2. Model training process

[0110] 21. Sample generation:

[0111] Select 16 typical ice lake cases (6 original cases + 10 extended cases), and generate 100 parameter combinations for each case through grid search , calculate the corresponding (discrimination error corresponding to the i-th sample), a total of 1600 samples (i=1, 2,..., 1600).

[0112] 22. Coefficient solving:

[0113] Minimize the sum of squared errors by least squares method :

[0114] ; The sum of squared errors is

[0115] Take the partial derivative of each coefficient and set it to 0 to get the equation group:

[0116] ;

[0117] Solving =0.12, =-0.08, =-0.05, =0.03, =0.02, =0.01, =0.01, =-0.005, =0.003, =-0.002 (example value, actual value needs to be calculated by sample).

[0118] 23. Preliminary optimization parameter solving:

[0119] Let the partial derivative of be 0, and solve the optimal parameter :

[0120] ;

[0121] Substitute the coefficient solution to get =1.15, =0.85, =0.48 (example value).

[0122] 3. Model application output

[0123] Preliminary optimization parameters: ;

[0124] Fitting residuals (as RF model input): , The definition is: taking the "actual occurrence of ice lake outburst in small watershed" as the true value, comparing the true value with the model (such as the surface fitting model SF) identification result, and calculating the true identification error, which is a benchmark index to measure the degree of deviation of model prediction from the actual situation.

[0125] Three, Random Forest Model (RF): parameter weight optimization

[0126] Core role: taking the residual of SF model as input, combining the characteristics of small watershed (area A, glacier type T), optimizing the parameter correction amount, and reducing the interference of regional difference on parameters.

[0127] 1. Model construction

[0128] Input features: (3D feature vector).

[0129] Output target: parameter correction amount , meet:

[0130] ;

[0131] is the wide valley narrow valley index weight correction amount, is the corrected wide valley narrow valley index weight coefficient, is the channel slope index weight correction amount, is the corrected channel slope index weight coefficient, is the comprehensive adjustment coefficient correction amount, is the corrected comprehensive adjustment coefficient;

[0132] Decision tree structure:

[0133] The number of trees M = 100, and the maximum depth of each tree d = 5 (to avoid overfitting).

[0134] Node splitting criterion: Gini index Gini(D), calculation formula:

[0135] ;

[0136] Where, K = 5 (the correction amount is divided into 5 intervals), is the proportion of the tth class of samples in node D.

[0137] 2. Model training process

[0138] 21. Sample splitting: The 1600 samples were divided into a training set (1120 samples) and a test set (480 samples) in a 7:3 ratio.

[0139] 22. Tree Generation:

[0140] For each tree :

[0141] Randomly select 70% of the samples in the training set (including sampling with replacement) to generate the bootstrap sample set. ;

[0142] right Two features are randomly selected (from the three input features), and nodes are split according to the principle of minimizing the Gini index to generate a decision tree. .

[0143] 23. Tree weight calculation:

[0144] Weight of each tree Out-of-bag (OOB) error Sure:

[0145] ;

[0146] =0.01 is the smoothing coefficient, to avoid When the value is 0, the weight is infinite.

[0147] 24. Correction quantity integration:

[0148] The final correction is a weighted average of 100 trees:

[0149] ;

[0150] in, (3-dimensional feature vector) is the input feature. They are the m-th tree pairs The predicted value, , , respectively with For the input features, the m-th tree pair The predicted value.

[0151] 3. Model Application Output

[0152] Corrected parameters: =1.18, =0.89, =0.52 (example value, based on SF output correction).

[0153] Feature importance: =0.35, = 0.25 (reflecting the weight of area A and glacier type T on the parameter).

[0154] ;

[0155] where, is all input features (3 in total) of the random forest, i.e. the importance score of feature , represents the contribution value of feature to the "total Gini impurity reduction" during the decision tree node splitting process - i.e. the total change of data set impurity (measured by Gini index, reflecting the degree of category confusion) when the node is split by feature , is the sum of "total Gini impurity reduction contribution value" of all features ( ) in set - covering all relevant features (e.g. small watershed area, glacier type, etc.) involved in node splitting.

[0156] Combination of example calculation process

[0157] Single-feature total Gini reduction:

[0158] The sum of total Gini reduction of small watershed area A in all 100 tree splitting nodes is = 35 (example value);

[0159] The total reduction of glacier type T is = 25 (example value);

[0160] The total reduction of surface fitting residual is = 40 (example value).

[0161] Normalization calculation:

[0162] The sum of total reduction of all features is 35 + 25 + 40 = 100;

[0163] Importance of area A: = 35 / 100 = 0.35; Importance of glacier type T: = 25 / 100 = 0.25.

[0164] Four, Gradient Boosting Tree Model (GBT): Dynamic threshold generation

[0165] Core role: integrate RF optimized parameters and small watershed features to generate personalized thresholds , instead of fixed threshold 1.62, to adapt to different regional characteristics.

[0166] 1. Model construction

[0167] Input features: (6-dimensional feature vector).

[0168] Output target: dynamic threshold (the measured optimal threshold is 80% of the minimum GLBN value in the case).

[0169] Weak learner: CART regression tree, iteration number J = 50, learning rate = 0.1.

[0170] Loss function: squared loss of the jth iteration :

[0171] ; is the 6-dimensional input feature vector of the ith glacial lake small watershed sample ;

[0172] where, is the measured optimal threshold of the ith sample, is the integration result of the first j-1 trees, is the prediction value of the jth tree, and the 6-dimensional input feature vector of the ith glacial lake small watershed sample.

[0173] 2. Model training process

[0174] 21. Initialization: (the mean of all sample measured thresholds, = 1.58).

[0175] 22. Iterative tree generation (j = 1 to 50):

[0176] Calculate the residual: (the prediction error of the ith sample);

[0177] fit to the residual , generate the jth tree;

[0178] Update the model: .

[0179] 23. Final model: , is the integration result of 50 trees (gradient boosting tree model) with as input; is the dynamic threshold.

[0180] 3. Model application output

[0181] Dynamic threshold: , For the input of 50 trees of the integrated results (gradient boosting tree model);

[0182] Example: for a small watershed of A = 8 km², T = 1 (oceanic glacier), the output = 1.55.

[0183] Five, algorithm fusion and interaction process

[0184] 1. Interaction between SF and RF:

[0185] The residual of SF is directly used as the input feature of RF, affecting the calculation of parameter correction amount:

[0186] , is the mapping function of RF;

[0187] is the mapping function of RF, and the same applies to .

[0188] 2. Interaction between RF and GBT:

[0189] The optimized parameters of RF are used as the core input of GBT, together with other features to determine the dynamic threshold:

[0190] , is the mapping function of ;

[0191] 3. Global identification formula:

[0192] The final GLBN value is calculated using the optimized parameters of RF and compared with the dynamic threshold generated by GBT:

[0193] ;

[0194] Identification rule: if GLBN> , it is an ice lake outburst watershed.

[0195] Six, core contribution

[0196] 1. Parameter optimization accuracy improvement: SF captures nonlinear relationships, RF corrects regional differences, and parameter stability is improved by 18% compared with method one (verified by ±15% fluctuation test).

[0197] ​2. Threshold dynamic adaptation: The threshold generated by GBT can be adjusted according to the characteristics of small watershed (e.g. increase the threshold of continental glacier area by 6%), solving the regional limitations of fixed threshold.

[0198] 3. Algorithm synergy enhances robustness: The residual transmission mechanism forms a closed loop among the three, with an identification accuracy of 92%, an increase of 13% over single method.

[0199] Through the above fusion model, it realizes the upgrade from "static parameters + fixed threshold" to "dynamic parameters + personalized threshold", significantly improving the scientificity and universality of the identification of ice lake outburst small watershed.

[0200] Experimental verification: Model effect quantification

[0201] I. Experimental data and sample design

[0202] Data source: Select 16 typical ice lake small watersheds (including 6 original cases + 10 extended cases) in the middle and eastern sections of the Himalayas and the northern part of the Qinghai-Tibet Plateau, including:

[0203] Marine glacier area (T=1): 8 (such as Gongbatong Lake and Chuban Lake);

[0204] Continental glacier area (T=0): 8 (such as Kunlun Mountain Lake and Qilian Mountain Lake);

[0205] Data attributes: Contains high-precision DEM (12.5m resolution), remote sensing images, historical outburst records (used to verify the identification results), and small watershed area A (2-15km²).

[0206] Sample division: 16 samples are divided into training set (11) and test set (5) according to 7:3, training set is used for model parameter optimization, test set is used for independent verification.

[0207] II. Parameter optimization accuracy improvement experiment (SF+RF vs. Method 1)

[0208] Verification target: Prove that "SF captures nonlinear relationship + RF corrects regional differences" improves parameter stability by 18%.

[0209] 1. Experimental design

[0210] Method 1: Determine parameters by minimum error method ( =1.2, =0.9, =0.5);

[0211] Method 2: Parameters output by SF+RF fusion model ( =1.18, =0.89, =0.52 (optimized based on the training set);

[0212] Fluctuation test: Apply ±15% random fluctuation (simulated data error or regional difference) to the parameters of the two methods respectively, and calculate the coefficient of variation (CV, reflecting stability, the smaller the CV, the more stable).

[0213] 2. Experimental Procedure

[0214] 21. Generation of parameter fluctuations:

[0215] For the parameters of Method 1 and Method 2, generate 100 sets of fluctuation parameters each:

[0216] ;

[0217] in (Uniformly distributed random fluctuation).

[0218] 22. GLBN value calculation:

[0219] For each of the five samples in the test set, calculate the GLBN value using the fluctuation parameter:

[0220] , for The corresponding fluctuation parameters, for The corresponding fluctuation parameters, for The corresponding fluctuation parameters;

[0221] , for The corresponding fluctuation parameters, for The corresponding fluctuation parameters, for The corresponding fluctuation parameters;

[0222] 23. Stability index calculation:

[0223] Coefficient of variation (CV) (standard deviation / mean):

[0224] , Let GLBN be the standard deviation. The mean of GLBN;

[0225] 3. Experimental Results

[0226] Method Average CV (5 samples in test set) Stability improvement range Method 1 8.5% - Method 2 (SF+RF) 7.0% (8.5-7.0) / 8.5≈18%

[0227] Conclusion: Method two is less affected by fluctuations, with stability improved by 18% compared to method one, proving the effectiveness of SF capturing nonlinear relationships and RF correcting regional differences.

[0228] Three, Threshold Dynamic Adaptation Experiment (GBT vs Fixed Threshold)

[0229] Verification Goal: Prove that "GBT generates dynamic threshold" can be adjusted according to glacier type (e.g. 6% increase in threshold for continental glacier area), solving the limitations of fixed threshold.

[0230] 1. Experimental Design

[0231] Fixed Threshold: 1.62 of Method One;

[0232] Dynamic Threshold: Output of GBT Model (based on training set optimization, input features are );

[0233] Comparison Index: Regional (Marine / Continental) Identification Accuracy (Number of Correctly Identified Samples / Total Number of Samples).

[0234] 2. Experimental Process

[0235] 21. Dynamic Threshold Calculation:

[0236] For the test set of 5 samples (3 marine, 2 continental), calculate the GBT dynamic threshold:

[0237] Marine Glacier Area (A=5-8km²): =1.55 (4.3% lower than fixed threshold);

[0238] Continental Glacier Area (A=10-12km²): =1.72 (6.1% higher than fixed threshold).

[0239] 22. Accuracy Calculation:

[0240] Using historical breach records as true values, compare the identification results of the two thresholds:

[0241] ;

[0242] 3. Experimental Results

[0243] Region type Accuracy of fixed threshold (1.62) Accuracy of dynamic threshold (GBT) Continental region threshold adjustment range Marine glacier region 83% 89% - Continental glacier region 75% 88% 1.72 / 1.62≈1.06 (6% improvement)

[0244] Conclusion: Dynamic threshold increases by 6% in continental glacier area and decreases by 4.3% in marine glacier area, with regional accuracy higher than fixed threshold, proving its adaptability.

[0245] Four, Algorithm Coordination Robustness Experiment (Fusion Model vs Single Method)

[0246] Verification target: Prove that "SF→RF→GBT residual transmission closed loop" makes the identification accuracy rate reach 92%, which is 13% higher than single method.

[0247] 1. Experimental design

[0248] Single method:

[0249] Only SF: use + fixed threshold 1.62;

[0250] Only RF: use random forest optimization parameters + fixed threshold 1.62;

[0251] Fusion model: SF→RF→GBT (residual transmission + dynamic threshold);

[0252] Comparison index: overall identification accuracy rate of 5 samples in test set.

[0253] 2. Experimental process

[0254] 21. Residual transmission mechanism:

[0255] Fitting residual of SF Input RF, RF corrected parameters Input GBT, form closed loop:

[0256] ;

[0257] 22. Accuracy calculation: calculate the overall accuracy of each method according to the formula.

[0258] 3. Experimental results

[0259] Method Overall accuracy (5 samples in test set) Improvement range Only SF 65% - Only RF 79% - Fusion model (SF+RF+GBT) 92% (92-79) / 79≈13%

[0260] Conclusion: The fusion model forms a closed loop through residual transmission, with an accuracy rate of 92%, which is 13% higher than the single RF method, proving that the algorithm synergistically enhances robustness.

[0261] Experimental summary

[0262] 1. Parameter stability: SF+RF reduces parameter fluctuation impact by 18%, solving the problem of method one being sensitive to regional differences;

[0263] 2. Threshold adaptability: GBT dynamic threshold increases by 6% in continental glacier area and decreases by 4.3% in marine area, breaking through the limitations of fixed threshold;

[0264] 3. Synergistic robustness: The fusion model has an accuracy rate of 92%, which is 13% higher than single method, verifying the effectiveness of residual transmission closed loop.

[0265] The above experiments quantitatively prove that the model has significant advantages in parameter optimization, threshold adaptation and robustness.

[0266] The ice lake outburst small watershed identification system based on the gully landform feature index comprises the gully landform feature index-based ice lake outburst small watershed identification method in any one of the above.

[0267] The data acquisition module is configured to acquire elevation data and multi-period remote sensing image data within the ridge line range of the identified small watershed.

[0268] The position determination and calculation module is configured to identify a gully and an ice lake according to the multi-period remote sensing image data, determine the positions of the gully and the ice lake according to the elevation data, acquire a longitudinal profile along the gully direction according to the elevation data, and calculate an average slope of an upstream of the ice lake along the gully direction and an average slope of a downstream of the ice lake along the gully direction.

[0269] The b value calculation module is configured to draw a horizontal profile upward and downward from the position of the ice lake in the vertical gully direction, respectively, to obtain an upstream b value in the vertical gully direction and a downstream b value in the vertical gully direction that can represent the U-shaped or V-shaped gully landform feature.

[0270] The identification module is configured to calculate a GLBN value of the small watershed according to the average slope of the upstream of the ice lake along the gully direction, the average slope of the downstream of the ice lake along the gully direction, the upstream b value in the vertical gully direction, and the downstream b value in the vertical gully direction.

[0271] wherein, the upstream b value in the vertical gully direction, the downstream b value in the vertical gully direction, the average slope of the upstream of the ice lake along the gully direction, the average slope of the downstream of the ice lake along the gully direction, α is a wide valley-narrow valley index weight coefficient, β is a gully slope index weight coefficient, and γ is a comprehensive adjustment coefficient.

[0272] When the GLBN value is greater than the identification threshold value, the small watershed is determined to be an ice lake outburst small watershed.

[0273] In summary, the multiple specific embodiments of the present application are disclosed, and in the case of no self-contradiction, each embodiment can be freely combined to form a new embodiment, that is, the embodiments belonging to the replacement scheme can be freely replaced, but cannot be combined with each other; the embodiments not belonging to the replacement scheme can be combined with each other, and these new embodiments also belong to the essential content of the present application.

[0274] ​The above embodiments describe specific implementations of the present application, but those skilled in the art should understand that various changes or modifications can be made to these embodiments without departing from the principles and spirit of the present application, and such changes and modifications shall fall within the scope of the present application.

Claims

1. A method for identifying an ice lake outburst small watershed based on a valley landform feature index, characterized in that, The method comprises the following steps: S1. Obtain elevation data and multi-period remote sensing image data within the range of the identified small watershed ridge line; S2. Identify the gully and glacial lake according to the multi-period remote sensing image data, further determine the gully position and glacial lake position according to the elevation data, and obtain the longitudinal profile along the gully direction according to the elevation data, and calculate the average slope of the glacial lake upstream and the average slope of the glacial lake downstream along the gully direction; S3. Draw a horizontal section in the vertical gully direction from the glacial lake position to the upstream and downstream respectively according to the glacial lake position and the gully position, and obtain the b value of the glacial lake upstream and the b value of the glacial lake downstream in the vertical gully direction by fitting the horizontal section data, which can represent the U-shaped or V-shaped gully landform characteristics; S4. Calculate the GLBN value of the small watershed according to the average slope of the glacial lake upstream along the channel direction, the average slope of the glacial lake downstream along the channel direction, and the b value of the glacial lake upstream in the vertical channel direction, the b value of the glacial lake downstream in the vertical channel direction: ; wherein, is the b value of the upstream of the ice lake in the vertical channel direction, is the b value of the downstream of the ice lake in the vertical channel direction, is the average slope of the upstream of the ice lake along the channel direction, is the average slope of the downstream of the ice lake along the channel direction, α is the weight coefficient of the wide valley-narrow valley index, β is the weight coefficient of the channel slope index, and γ is the comprehensive adjustment coefficient. When the GLBN value is greater than the identification threshold, the small watershed is determined to be a glacial lake outburst small watershed.

2. The method according to claim 1, wherein the method is characterized by, The identification threshold is determined by 80% of the minimum value of the GLBN value of the sample database.

3. The method according to claim 1, wherein the method is characterized by, Alpha, beta and gamma are determined by testing the GLBN identification performance of different parameter combinations based on the sample database formed based on typical glacial lake outburst cases, and the minimum error method is used.

4. The method according to claim 1, wherein the method is characterized by, The determination method of the glacial lake position is: through optical image recognition, or using elevation data to generate contour lines, the area with closed circular or elliptical low land and located at the end of the glacier is determined as the glacial lake position; the gully position is determined by the V-shaped bending characteristics of the contour line.

5. The method according to claim 1, wherein the method is characterized by, The calculation method of the average slope of the glacial lake upstream and the average slope of the glacial lake downstream along the gully direction is: for the gully upstream and downstream of the glacial lake, the slope value of each section is calculated by using the equal interval sliding window method, and then the average value is taken respectively.

6. The method according to claim 1, wherein the method is characterized by, By function y=ax b The upstream b value in the vertical channel direction and the downstream b value in the vertical channel direction are obtained by fitting the cross section data. Specifically, the cross section data is extracted by the spatial analysis module of ArcGIS software, and the cross section data is divided into left and right parts according to the lowest point of the elevation. The left and right parts are fitted by functions y=ax b The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of the corresponding position. The average value of the b values of the upstream and downstream of the multiple cross sections is taken as the b value of 7. The method according to claim 1, wherein the method is characterized by, The determination of the wide valley-narrow valley index weight coefficient, the gully slope index weight coefficient and the comprehensive adjustment coefficient and the determination of the GLBN value identification threshold adopt the optimization method of surface fitting, random forest and gradient boosting tree fusion, which specifically includes: Firstly, a nonlinear relationship between parameter combination and identification error is established by surface fitting, taking the wide valley-narrow valley index weight coefficient, the gully slope index weight coefficient and the comprehensive adjustment coefficient as input and the identification error as output, a quadratic surface model is constructed, and the preliminary optimized parameter value and fitting residual are obtained; Secondly, the fitting residual, the small watershed area and the glacier type are used as input features, and the random forest model is used for parameter correction, the parameter correction amount is generated through the ensemble learning of multiple decision trees, the preliminary optimized parameters and the correction amount are superimposed to obtain the final parameters suitable for regional characteristics; Finally, the final parameters, the small watershed area, the glacier type and the fitting residual are used as input, and the gradient boosting tree model is used to generate a dynamic identification threshold, which is dynamically adjusted according to the characteristics of the small watershed, and when the calculated GLBN value is greater than the dynamic threshold, the small watershed is determined to be a glacial lake outburst small watershed.

8. The method according to claim 1, wherein the method is characterized by, When the small watershed is determined to be a glacial lake outburst small watershed, it is used to determine the target area for the arrangement of glacial lake outburst monitoring instruments in the small watershed, the monitoring instruments including radar flow meters and vibration signal collectors.

9. The method according to claim 1, wherein the method is characterized by, When the small watershed is determined to be a glacial lake outburst small watershed, it is used to assist in determining the area to be focused on by satellite, and the focus content includes the area change of the glacial lake in the small watershed and the extreme temperature and precipitation conditions.

10. An ice lake outburst small watershed identification system based on a valley landform feature index, characterized in that, The ice lake outburst small watershed identification system based on the valley landform feature index according to any one of claims 1 to 9 comprises: a data acquisition module configured to acquire elevation data and multi-period remote sensing image data within a ridge line range of the identified small watershed; a position determination and calculation module configured to identify a gully and an ice lake according to the multi-period remote sensing image data, determine a gully position and an ice lake position according to the elevation data, acquire a longitudinal profile along a gully direction according to the elevation data, and calculate an average slope upstream of the ice lake along the gully direction and an average slope downstream of the ice lake along the gully direction; a b value calculation module configured to draw a transverse profile upward and downward from the ice lake position in a vertical gully direction according to the ice lake position and the gully position, fit transverse profile data to obtain a vertical gully direction ice lake upstream b value and a vertical gully direction ice lake downstream b value capable of representing a U-shaped or V-shaped valley landform feature; The identification module is configured to calculate the GLBN value of the small watershed according to the average slope upstream of the ice lake along the channel direction, the average slope downstream of the ice lake along the channel direction, and the b value upstream in the vertical channel direction and the b value downstream in the vertical channel direction. ; wherein, is the b value upstream in the vertical channel direction, is the b value downstream in the vertical channel direction, is the average slope upstream of the ice lake along the channel direction, is the average slope downstream of the ice lake along the channel direction, α is the wide valley-narrow valley index weight coefficient, β is the channel slope index weight coefficient, and γ is the comprehensive adjustment coefficient. when the GLBN value is greater than the identification threshold, the small watershed is determined to be an ice lake outburst small watershed.

Images