Method of constructing and optimizing mountain flood rainfall station network in northwest inland region

Abstract

The present application relates to a method for constructing and optimizing a mountain flood rainfall station network in the northwest inland area, which comprises the following steps: multi-dimensional basic data collection and characteristic analysis; analyzing the regional hydro-meteorological and runoff response characteristics by collecting regional basic geographic, mountain flood disaster, hydro-meteorological and typical flood data; carrying out station network layout design suitable for runoff generation mechanism, combining with the basin area, terrain runoff generation characteristics, disaster risk and flood control demand to lay out monitoring sites, and completing site selection optimization; constructing a multi-scale data quality control system, building a progressive abnormal data identification mechanism, formulating differentiated data quality control rules, and establishing a dynamic updating mechanism for reference stations and a real-time correction mechanism for abnormal data; realizing dynamic optimization of the station network driven by both model and demand, dynamically adjusting the rainfall station network by combining the hydrological model calibration results, early warning indicators, site operation evaluation results and rainfall fusion model verification results.

Description

Technical Field

[0001] This invention relates to the field of flood disaster monitoring and early warning technology, specifically a method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region. Background Technology

[0002] The Northwest inland region has a complex topography, including mountains, glaciers, and deserts. The climate is arid with uneven spatial and temporal precipitation. Flash floods can be triggered by short-duration heavy rainfall, glacial meltwater, or both, and are characterized by their suddenness, rapid destructive power, and concentrated risk. Constructing a rain gauge network adapted to local characteristics is crucial for implementing a flash flood prevention system and enhancing disaster preparedness.

[0003] The current common model for rain gauge network deployment, data processing, and operation and maintenance is not suitable for the unique geographical and runoff generation conditions of Northwest China. Existing stations are mostly concentrated in oasis plains, while monitoring points in mountainous areas, upstream watersheds, and glacial meltwater areas are sparse, resulting in a large number of monitoring blind spots. The selection of station sites does not take into account topography and runoff characteristics, leading to insufficient representativeness of monitoring data. Furthermore, the lack of coordinated monitoring methods for rainfall and temperature in snowmelt areas makes it impossible to distinguish between different runoff sources.

[0004] Existing technologies still suffer from problems such as imperfect data quality control, rigid station network layout, and low system integration: traditional data screening methods are unable to guarantee the quality of monitoring data and have not established a dynamic update mechanism for benchmark stations; the station network is a static layout and cannot be adjusted in real time based on model simulation and early warning effects; monitoring, simulation, early warning, and optimization have not formed a closed loop, ultimately resulting in insufficient accuracy and short lead time for flash flood forecasting and early warning, making it difficult to meet the actual needs of regional flash flood defense. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for constructing and optimizing a mountain torrent and rainfall station network in the inland region of Northwest China, thus solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for constructing and optimizing a flash flood and rainfall station network in Northwest Inland China, comprising the following steps: S100: Multi-dimensional basic data collection and feature analysis; by collecting regional basic geography, flash flood disaster, hydro-meteorology and typical flood data, analyze the regional hydro-meteorology and runoff response characteristics; S200: Conduct station network layout design adapted to runoff generation mechanism, deploy monitoring stations in combination with watershed area, topographic runoff generation characteristics, disaster risk and flood control needs, and complete station site selection optimization; S300: Construct a multi-scale data quality control system, establish a progressive abnormal data identification mechanism, formulate differentiated data quality control rules, and establish a dynamic update mechanism for base stations and a real-time correction mechanism for abnormal data. S400: Enables dynamic optimization of the station network driven by both model and demand. It dynamically adjusts the rainfall station network by combining hydrological model calibration results, early warning indicators, station operation evaluation results, and rainfall fusion model verification results. S500: Complete the integrated interface development of the station network and monitoring platform, develop the hydrological simulation interface and configure data constraint rules to realize real-time data interaction between monitoring data and the flood and drought disaster monitoring and early warning platform.

[0007] Furthermore, in S100: the basic geographic data includes a 1:50,000 digital elevation model, digital line map, land use type map and soil texture type map, and the watershed area, slope, confluence path and river network distribution attributes of several key small watersheds are extracted based on the basic geographic data; The data on flash flood disasters includes information on villages requiring prevention and control, distribution of dangerous areas, flood control capacity parameters, and historical records of flash flood disasters.

[0008] Furthermore, S200 specifically includes: S201: Stations are deployed according to the catchment area; the catchment area is no more than 100 km². 2 One to two monitoring stations will be set up in the area, with a drainage area of ​​100 km². 2 ~300km 2 Two to three monitoring stations will be set up in the area, with a drainage area of ​​not less than 300 km². 2 No fewer than three monitoring stations shall be set up in the area, covering all key small watershed management units and areas prone to flash floods; S202: Deploy stations based on topography and runoff type; densify the deployment of monitoring stations in the upstream of mountain torrent gullies with a slope of not less than 15° and at key confluence nodes; deploy integrated monitoring stations for rainfall and temperature in glacier meltwater areas and mixed runoff areas of rainstorm and snowmelt to collect rainfall and temperature data simultaneously. S203: Deploy monitoring stations based on disaster risk and flood control needs; deploy monitoring stations within 5km to 10km upstream of historically prone flash floods and debris flows and dangerous areas; for villages with current flood control capacity not exceeding the 10-year flood standard, add monitoring stations upstream of their respective small watersheds. S204: Conduct site selection optimization; avoid interference areas such as depressions, steep slopes, and obstructions, and select locations with open views, convenient transportation, and stable power supply to deploy sites, with the distance between adjacent sites controlled between 3km and 10km.

[0009] Furthermore, S300 specifically includes: S301: Construct a progressive anomaly data identification mechanism; use the Hampel method to identify anomalous years of stations from the time dimension, combine the improved Grubbs criterion to screen benchmark stations from the spatial dimension, and use the surrounding station comparison method to identify anomalous data and anomalous time periods at the hour scale, forming a three-level data screening system that combines time dimension, spatial dimension, and radar verification. S302: Develop differentiated data quality control rules; unify the data step size for monitoring data with a frequency of 5 minutes to eliminate time series breakage issues; S303: Implement a dynamic update mechanism for benchmark stations; update the list of benchmark stations annually based on long-sequence rainfall observation data, retaining stations with stable monitoring status and qualified data quality, and removing stations with systematic data errors; S304: Establish a real-time data correction mechanism; for abnormal data obtained from screening, perform real-time correction by combining radar QPE data and monitoring data from surrounding base stations, and store the corrected data in the database.

[0010] 5. The method for constructing and optimizing a mountain torrent and rainfall station network in Northwest Inland Area according to claim 1, characterized in that: S400 specifically includes: S401: Optimize the station network based on the calibration results of the distributed hydrological model; select multiple flood processes to carry out simulation calculations, and for watersheds with a relative error of peak flow greater than 20% and a peak occurrence time error greater than 2 hours, add monitoring stations in the upstream of the watershed or in areas with high spatial heterogeneity of rainfall. S402: Optimize the station network layout by combining early warning indicators; verify the rationality of early warning indicators for each prevention and control village; and for areas where the deviation between early warning indicators and regional flood control capacity exceeds 20%, supplement the deployment of monitoring stations to calibrate the early warning threshold. S403: Optimize the station network based on the evaluation of station operation performance; collect and analyze monitoring data every six months to support the hit rate, false alarm rate and missed alarm rate of early warning work; set the qualified standards as follows: the hit rate of meteorological forecast and early warning is not less than 37%, and the hit rate of real-time monitoring and early warning is not less than 97%; for stations that fail the evaluation results for three consecutive times, relocation, equipment upgrade or station encryption operations shall be carried out. S404: Optimize the station network based on the verification results of the fusion model; when the deviation between the output data of the rainfall fusion model and the actual rainfall data exceeds two rainfall levels, supplement the monitoring stations in sparsely populated areas or areas with complex terrain.

[0011] Furthermore, the S500 specifically includes: developing several small watershed hydrological simulation interfaces, wherein the interfaces support HTTP protocol calls and the data interaction format adopts JSON format; An interface data constraint rule library is built to validate the input data, return error messages and terminate the calculation for invalid data; the monitoring data after quality control is pushed to the flood and drought disaster monitoring and early warning platform in real time, providing data support for the operation of distributed hydrological models and rainfall fusion models and the dynamic updating of early warning thresholds.

[0012] Furthermore, this method is implemented with the Xinjiang Uygur Autonomous Region as the core area, covering the mountain torrent disaster prevention and control areas including the Ili Kazakh Autonomous Prefecture, Urumqi City, Kashgar Prefecture, and Altay Prefecture, as well as all key small watershed management units.

[0013] Furthermore, after S100 is completed and before S200 is executed, a basic data preprocessing step is also included: which uses GIS technology to process basic geographic data, delineate small watershed boundaries and extract watershed attribute parameters. The flood control capacity data of the prevention and control villages were standardized, and invalid data with a missing rate of more than 10% were removed. Outliers and missing values ​​were removed from the rainfall and hydrological monitoring data, and the data were standardized to two time steps: 5 minutes and 1 hour. The frequency analysis method was used to calculate the design rainfall for 1 hour, 6 hours and 24 hours under the 5-year to 100-year return period standard for each region, and high-risk areas for rainfall were delineated.

[0014] Furthermore, the dynamic update mechanism for the base stations is executed monthly to collect statistical data quality, and the entire base station catalog is updated annually. The accuracy rate for correcting abnormal data is no less than 96%, and the accuracy rate for identifying abnormal data is no less than 98%.

[0015] Furthermore, after dynamic optimization of the station network, the proportion of flood events with a relative error of less than 20% in peak flow is no less than 86%, the average forecast period for flash floods is extended by 3 hours or more, and the station density in key small watersheds reaches 2-3 stations per 100km. 2 The monitoring coverage rate of key small watersheds has reached 100%.

[0016] Compared with the prior art, the technical solution of this application has the following beneficial effects: 1. This method for constructing and optimizing a flash flood and rainfall monitoring network in Northwest China's inland regions combines the complex topography of mountains, glaciers, and deserts with the characteristics of three types of runoff generation: rainfall, snowmelt, and mixed runoff. The network is planned according to watershed area, topographic slope, and disaster risk classification. Key monitoring points are densely deployed upstream of flash flood gullies, at confluence nodes, in glacial meltwater areas, and upstream of disaster-prone areas. This effectively addresses the problem of traditional monitoring networks being overly concentrated in oasis plains, mountainous areas, and snowmelt zones, leaving monitoring gaps. Simultaneously, integrated rainfall and temperature monitoring stations are specifically deployed to accurately distinguish the contributions of rainfall and glacial meltwater to runoff, fully adapting to local unique runoff mechanisms. Ultimately, this achieves 100% monitoring coverage of key small watersheds, significantly improving the completeness of flash flood monitoring across the entire region. 2. This method for constructing and optimizing a flash flood and rainfall monitoring network in Northwest China overcomes the shortcomings of traditional techniques, such as single-level data screening, fixed base stations, and the inability to promptly correct anomalies, by establishing a three-level anomaly data identification system that combines temporal, spatial, and radar verification. This is coupled with differentiated quality control rules, annual updates of base stations, and a real-time anomaly data correction mechanism. Strict quality control standards maintain a high accuracy rate in anomaly data identification and correction, significantly improving the quality of the data stored. This provides accurate and effective foundational data for hydrological model calibration, rainstorm analysis, and early warning threshold calculation, thereby enhancing the basic capabilities of flash flood forecasting and early warning from the data source.

[0017] 3. This method for constructing and optimizing mountain torrent and rainfall monitoring network in Northwest Inland Areas abandons the traditional static model after network construction. Instead, it uses four dimensions—hydrological model, early warning effect, station operation status, and rainfall fusion model—to drive dynamic adjustment of the network. Based on simulation errors, early warning deviations, and equipment operation problems, the method adjusts the station location, density, and equipment configuration in real time. After continuous optimization, the accuracy of flood simulation is significantly improved, the average early warning period for mountain torrents is extended by more than three hours, and the network layout can always match regional rainfall changes, disaster risk evolution, and flood control needs, achieving deep linkage between network construction and disaster prevention work. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the process framework of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1The method for constructing and optimizing the flash flood rainfall station network in the Northwest inland region in this embodiment mainly addresses the problems of the complex topography, uneven spatial and temporal distribution of precipitation, flash floods triggered by rainfall, glacial meltwater or a mixture of both in the Northwest inland region, and the blind layout, numerous monitoring blind spots, weak data quality control, static and fixed station network, and low system integration of traditional rainfall station networks. The implementation process adopts a complete approach, including multi-dimensional data analysis, hierarchical and zoned station network deployment, full-process data quality control, dynamic optimization driven by both models and requirements, and platform interface integration. This approach builds a mountain torrent and rainfall station network that is adapted to the geographical, climatic, and runoff characteristics of Northwest China. The entire method forms a closed-loop operation system from early planning, construction and deployment, data management and control, dynamic optimization to platform linkage.

[0021] The entire implementation process is divided into six stages: basic data collection and feature analysis, basic data preprocessing, website layout design, data quality control system construction, website dynamic optimization, interface development and platform integration. Each stage is implemented step by step.

[0022] First, we conducted multi-dimensional basic data collection and feature analysis. We comprehensively collected various basic data within the study area, using 1:50,000 digital elevation models, digital line maps, land use type maps, and soil texture type maps as basic geographic data. Based on this data, we extracted core attribute information such as watershed area, surface slope, confluence path, and river network distribution for 38 key small watersheds within the area. Simultaneously, we compiled data related to flash flood disasters, summarizing basic information on more than 3,395 prevention and control villages, the distribution range of disaster-prone areas, current flood control capacity parameters for various regions, and complete records of flash flood disasters over the years.

[0023] Simultaneously, long-term hydrological and meteorological data were collected, including hourly rainfall data from 2015 to 2025, water level and flow monitoring data from 269 hydrological stations in the region, hourly observation data from 184 meteorological stations, and data from no fewer than 52 typical rainstorm and flood events. Three types of flood samples were distinguished: rainstorm type, snowmelt type, and rainstorm and snowmelt mixed type. Based on all the collected data, the regional hydrological and meteorological change patterns and runoff response characteristics of different underlying surfaces were deeply analyzed. The spatial distribution characteristics of design rainstorms with different return periods were sorted out, and the linkage relationship between temperature, precipitation, and runoff in the glacier and snowmelt area was clarified, providing data and theoretical support for subsequent station network planning.

[0024] After the basic data collection was completed, basic data preprocessing was performed. GIS technology was used to process geographic data such as digital elevation models and digital line maps, accurately delineating the boundaries of each small watershed, improving the watershed attribute parameter archives, and standardizing the flood control capacity data of the prevention and control villages, removing invalid information with a missing data rate exceeding 10%. The raw rainfall and hydrological monitoring data were cleaned to remove obvious outliers and missing values, and the data time step was standardized to two standard formats: 5 minutes and 1 hour. Frequency analysis was used to calculate the design rainfall for 1 hour, 6 hours, and 24 hours under the 5-year, 10-year, 20-year, 50-year, and 100-year return period standards in the region. Based on the calculation results, high-risk areas for rainfall across the entire region were delineated, providing a basis for increased station density and focused monitoring.

[0025] Next, the station network layout design will be carried out to adapt to the runoff generation mechanism. This step will be implemented in stages according to four dimensions: watershed classification, topographic runoff generation, disaster risk, and site selection optimization. The stations will be classified and deployed according to the watershed area, with the watershed area not exceeding 100 km². 2 One or two monitoring stations are set up in small watersheds with a watershed area of ​​100 km². 2 Up to 300km 2 Two to three monitoring stations will be set up in the designated area, with a drainage area of ​​not less than 300 km². 2 No fewer than three monitoring stations will be set up in each area to ensure that all key small watershed management units and areas prone to flash floods are covered by the stations.

[0026] Based on the differences in terrain conditions and runoff types, monitoring stations are set up in a differentiated manner. Monitoring points are densified in the upstream of mountain torrent gullies with a surface slope of not less than 15° and at key nodes where watersheds converge. For glacier meltwater areas and mixed runoff areas caused by rainstorms and snowmelt, integrated monitoring stations for rainfall and temperature are specially set up to collect rainfall and temperature data simultaneously, so as to achieve comprehensive monitoring of the mixed runoff process.

[0027] In conjunction with disaster risk and flood control needs, additional monitoring stations will be deployed in areas prone to historical flash floods and debris flows, as well as within 5km to 10km upstream of disaster-prone areas. For villages with current flood control capacity not exceeding the 10-year flood standard, additional monitoring points will be added upstream of their respective small watersheds to improve the timeliness of early warning response. Finally, a unified site selection optimization will be carried out to proactively avoid areas that are prone to data distortion, such as depressions, steep slopes, and tall obstructions. Priority will be given to constructing stations in locations with open views, convenient transportation, and stable power supply. The distance between two adjacent stations will be strictly controlled between 3km and 10km to balance monitoring accuracy and construction and maintenance costs.

[0028] After the site deployment is completed, a multi-scale data quality control system is established, and data management rules and operating mechanisms for the entire process are set up. First, a progressive abnormal data identification mechanism is constructed, forming a three-level data screening system that combines time, space, and radar verification. The Hampel method is used to identify abnormal observation years of a single station from the time dimension, and the improved Grubbs criterion is used to screen regional benchmark stations from the spatial dimension. Then, abnormal data and abnormal time periods at the hourly scale are identified by comparing with surrounding stations.

[0029] Secondly, differentiated data quality control rules were formulated, and a unified data step size was adopted for 5-minute high-frequency monitoring data to eliminate the problem of data time sequence breaks. At the same time, rainfall thresholds and continuous value verification rules were set in combination with regional rainstorm characteristics to identify unreasonable monitoring data. A normalized dynamic update mechanism for benchmark stations was established, and the data quality of all stations in the region was statistically analyzed every month. The list of benchmark stations was updated once a year to retain stations with stable monitoring status and data quality that meet the standards, while removing stations with systematic data errors to ensure the reliability of the benchmark station system.

[0030] Establish a real-time correction mechanism for abnormal data. For abnormal data identified through screening, real-time correction is carried out by combining radar QPE data and synchronous monitoring data from surrounding qualified benchmark stations. After correction, compliant data is uniformly stored in the database. The entire quality control system requires that the accuracy rate of abnormal data identification is not less than 98%, and the accuracy rate of abnormal data correction is not less than 96%.

[0031] Subsequently, a dynamic optimization of the station network driven by both model and demand was initiated to achieve routine adjustments and iterative upgrades to the station network layout. Based on the calibration results of the distributed hydrological model, the station network layout was optimized. Simulation calculations were carried out on multiple typical flood processes. For watersheds with a relative error of more than 20% in peak flow and a peak occurrence time error of more than two hours, additional monitoring stations were added in the upstream areas of the watershed or in areas with strong spatial heterogeneity of rainfall.

[0032] The layout of the flood warning indicators was optimized by combining them with the actual flood control capacity of the area. The rationality of the warning indicators for each village was checked one by one. When the warning indicators deviated from the actual flood control capacity of the area by more than 20%, additional monitoring stations were set up. The warning threshold was calibrated by relying on refined monitoring data. The station network was optimized based on the actual operation effect of the stations. The hit rate, false alarm rate and missed alarm rate of the monitoring data of each station were counted every six months. The qualified standard was defined as a meteorological forecast warning hit rate of not less than 37% and a real-time monitoring warning hit rate of not less than 97%. For stations that failed to meet the standards for three consecutive assessments, relocation, equipment upgrade or densification of surrounding stations were carried out according to the actual situation.

[0033] Simultaneously, the layout was optimized based on the verification results of the rainfall fusion model. When the output data of the rainfall fusion model deviated from the actual rainfall data on site by more than two rainfall levels, monitoring points were added in sparsely populated areas and areas with complex terrain. After multiple rounds of dynamic optimization, it was required that the proportion of flood events with a relative error of less than 20% of the peak flow rate be no less than 86%, the density of key small watershed stations reach two to three per 100 square kilometers, and the monitoring coverage rate of key small watersheds reach 100%.

[0034] Finally, we carried out interface development and integrated the station network and platform. We developed hydrological simulation interfaces adapted to various small watersheds. All interfaces support HTTP protocol calls, and data interaction adopts JSON format. We built an interface data constraint rule library to perform validity checks on all input data. Once invalid data is identified, an error message is immediately returned and the calculation is terminated, ensuring the safety of model operation from the source. The monitoring data after multi-layer quality control is pushed to the flood and drought disaster monitoring and early warning platform in real time, realizing real-time data interaction between the rain gauge network, hydrological simulation model, rainfall fusion model, and early warning system. This provides continuous and reliable data support for model operation, dynamic updating of early warning thresholds, and the "four early warnings" of flash floods.

[0035] During the implementation of the entire methodology, daily operation and maintenance management will be carried out simultaneously. The equipment operation status, power supply and communication status of all monitoring stations will be inspected regularly. Data acquisition and transmission equipment will be maintained on time, and the list of reference stations and quality control rules will be reviewed regularly to ensure the long-term stable operation of the entire link of station network, data and platform.

[0036] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for constructing and optimizing a flash flood and rainfall monitoring network in the inland northwest region, characterized by: It includes the following steps: S100: Multi-dimensional basic data collection and feature analysis; by collecting regional basic geography, flash flood disaster, hydro-meteorology and typical flood data, analyze the regional hydro-meteorology and runoff response characteristics; S200: Conduct station network layout design adapted to runoff generation mechanism, deploy monitoring stations in combination with watershed area, topographic runoff generation characteristics, disaster risk and flood control needs, and complete station site selection optimization; S300: Construct a multi-scale data quality control system, establish a progressive abnormal data identification mechanism, formulate differentiated data quality control rules, and establish a dynamic update mechanism for base stations and a real-time correction mechanism for abnormal data. S400: Enables dynamic optimization of the station network driven by both model and demand. It dynamically adjusts the rainfall station network by combining hydrological model calibration results, early warning indicators, station operation evaluation results, and rainfall fusion model verification results. S500: Complete the integrated interface development of the station network and monitoring platform, develop the hydrological simulation interface and configure data constraint rules to realize real-time data interaction between monitoring data and the flood and drought disaster monitoring and early warning platform.

2. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: In S100: the basic geographic data includes a 1:50,000 digital elevation model, digital line map, land use type map and soil texture type map. Based on the basic geographic data, the watershed area, slope, confluence path and river network distribution attributes of several key small watersheds are extracted. The data on flash flood disasters includes information on villages requiring prevention and control, distribution of dangerous areas, flood control capacity parameters, and historical records of flash flood disasters.

3. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: Specifically, S200 includes: S201: Stations are deployed according to the catchment area; the catchment area is no more than 100 km². 2 One to two monitoring stations will be set up in the area, with a drainage area of ​​100 km². 2 ~300km 2 Two to three monitoring stations will be set up in the area, with a drainage area of ​​not less than 300 km². 2 No fewer than three monitoring stations shall be set up in the area, covering all key small watershed management units and areas prone to flash floods; S202: Deploy stations based on topography and runoff type; densify the deployment of monitoring stations in the upstream of mountain torrent gullies with a slope of not less than 15° and at key confluence nodes; deploy integrated monitoring stations for rainfall and temperature in glacier meltwater areas and mixed runoff areas of rainstorm and snowmelt to collect rainfall and temperature data simultaneously. S203: Deploy monitoring stations based on disaster risk and flood control needs; deploy monitoring stations within 5km to 10km upstream of historically prone flash floods and debris flows and dangerous areas; for villages with current flood control capacity not exceeding the 10-year flood standard, add monitoring stations upstream of their respective small watersheds. S204: Conduct site selection optimization; avoid interference areas such as depressions, steep slopes, and obstructions, and select locations with open views, convenient transportation, and stable power supply to deploy sites, with the distance between adjacent sites controlled between 3km and 10km.

4. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: Specifically, S300 includes: S301: Construct a progressive anomaly data identification mechanism; use the Hampel method to identify anomalous years of stations from the time dimension, combine the improved Grubbs criterion to screen benchmark stations from the spatial dimension, and use the surrounding station comparison method to identify anomalous data and anomalous time periods at the hour scale, forming a three-level data screening system that combines time dimension, spatial dimension, and radar verification. S302: Develop differentiated data quality control rules; unify the data step size for monitoring data with a frequency of 5 minutes to eliminate time series breakage issues; S303: Implement a dynamic update mechanism for benchmark stations; update the list of benchmark stations annually based on long-sequence rainfall observation data, retaining stations with stable monitoring status and qualified data quality, and removing stations with systematic data errors; S304: Establish a real-time data correction mechanism; for abnormal data obtained from screening, perform real-time correction by combining radar QPE data and monitoring data from surrounding base stations, and store the corrected data in the database.

5. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: Specifically, S400 includes: S401: Optimize the station network based on the calibration results of the distributed hydrological model; select multiple flood processes to carry out simulation calculations, and for watersheds with a relative error of peak flow greater than 20% and a peak occurrence time error greater than 2 hours, add monitoring stations in the upstream of the watershed or in areas with high spatial heterogeneity of rainfall. S402: Optimize the station network layout by combining early warning indicators; verify the rationality of early warning indicators for each prevention and control village; and for areas where the deviation between early warning indicators and regional flood control capacity exceeds 20%, supplement the deployment of monitoring stations to calibrate the early warning threshold. S403: Optimize the station network based on the evaluation of station operation performance; collect and analyze monitoring data every six months to support the hit rate, false alarm rate and missed alarm rate of early warning work; set the qualified standards as follows: the hit rate of meteorological forecast and early warning is not less than 37%, and the hit rate of real-time monitoring and early warning is not less than 97%; for stations that fail the evaluation results for three consecutive times, relocation, equipment upgrade or station encryption operations shall be carried out. S404: Optimize the station network based on the verification results of the fusion model; when the deviation between the output data of the rainfall fusion model and the actual rainfall data exceeds two rainfall levels, supplement the monitoring stations in sparsely populated areas or areas with complex terrain.

6. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: The S500 specifically includes: developing several small watershed hydrological simulation interfaces, which support HTTP protocol calls and use JSON format for data interaction; An interface data constraint rule library is built to validate the input data, return error messages and terminate the calculation for invalid data; the monitoring data after quality control is pushed to the flood and drought disaster monitoring and early warning platform in real time, providing data support for the operation of distributed hydrological models and rainfall fusion models and the dynamic updating of early warning thresholds.

7. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: This method is implemented with the Xinjiang Uygur Autonomous Region as the core area, covering the mountain torrent disaster prevention and control areas including the Ili Kazakh Autonomous Prefecture, Urumqi City, Kashgar Prefecture, and Altay Prefecture, as well as all key small watershed management units.

8. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: After S100 is completed and before S200 is executed, a basic data preprocessing step is also included: it uses GIS technology to process basic geographic data, delineate small watershed boundaries and extract watershed attribute parameters. The flood control capacity data of the prevention and control villages were standardized, and invalid data with a missing rate of more than 10% were removed. Outliers and missing values ​​were removed from the rainfall and hydrological monitoring data, and the data were standardized to two time steps: 5 minutes and 1 hour. The frequency analysis method was used to calculate the design rainfall for 1 hour, 6 hours and 24 hours under the 5-year to 100-year return period standard for each region, and high-risk areas for rainfall were delineated.

9. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: The dynamic update mechanism for the base stations is executed monthly to collect statistical data quality, and the entire base station catalog is updated once a year. The accuracy rate for correcting abnormal data is no less than 96%, and the accuracy rate for identifying abnormal data is no less than 98%.

10. The method for constructing and optimizing a mountain torrent and rainfall station network in the inland northwest region according to claim 1, characterized in that: After dynamic optimization of the station network, the proportion of flood events with a relative error of less than 20% in peak flow is no less than 86%, the average forecast period for flash floods is extended by 3 hours or more, and the station density in key small watersheds reaches 2-3 stations per 100km. 2 The monitoring coverage rate of key small watersheds has reached 100%.

Images