A Multi-Scale Attribution Identification Method for Baseflow Changes in Rivers in High-Altitude Cold Regions

Abstract

This application discloses a multi-scale attribution identification method for baseflow changes in rivers in high-altitude and cold regions. By constructing a trained soil and moisture assessment tool model to simulate key hydrological parameters and determining the elasticity coefficient based on a hydrothermal coupling correction model, the multi-scale attribution of baseflow changes can be accurately identified. This method has the advantages of overcoming the limitations of traditional hydrothermal balance models in high-altitude and cold regions, improving the accuracy and reliability of baseflow change attribution, and providing a scientific basis for sustainable water resource management.

Description

Technical Field

[0001] This application relates to the field of hydrology and water resources technology, and more specifically, to a multi-scale attribution identification method for baseflow changes in rivers in high-altitude and cold regions. Background Technology

[0002] Hydrological processes in high-altitude, cold regions exhibit significant unique characteristics, posing fundamental limitations to traditional hydrothermal balance models in simulating baseflow changes. Snowfall constitutes a prominent proportion of annual precipitation in these environments, and its phase transition characteristics lead to a delayed release effect, disrupting the synchronization between actual evapotranspiration and precipitation. The secondary water supply mechanism formed by seasonal snowmelt significantly alters the temporal structure of the hydrothermal balance, making the traditional Budyko-Fu model unable to accurately capture the complex runoff generation and confluence mechanisms in cold regions. When applied to high-altitude areas, this model fails to consider the phase characteristics of snowfall, the lag effect of snowmelt, and the dynamic changes in seasonal water storage, resulting in hydrothermal balance simulations that deviate from actual hydrological processes. Particularly in typical high-altitude, cold basins such as the Yangtze River source region, traditional methods struggle to quantify the combined impact of precipitation phase transitions, snowmelt processes, and water storage fluctuations on baseflow, leading to systematic biases in the attribution analysis of baseflow changes. Existing technologies lack an attribution framework for baseflow changes across multiple timescales in cold regions, and cannot effectively analyze the driving mechanisms of climate factors and underlying surface factors on baseflow, which seriously restricts the scientific nature of sustainable water resource management and ecological protection decisions.

[0003] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0004] This application provides a multi-scale attribution identification method for baseflow changes in rivers in high-altitude and cold regions. It has the advantages of overcoming the limitations of traditional water-heat balance models in high-altitude and cold regions, improving the accuracy and reliability of baseflow change attribution, and providing a scientific basis for sustainable water resource management.

[0005] This application provides a multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions, including: Based on the research data of the study area, a soil and moisture assessment tool model was constructed, and the parameters of the soil and moisture assessment tool model were optimized and calibrated using the measured cross-sectional runoff data of the study area to obtain the trained soil and moisture assessment tool model. The research data included digital elevation model data, soil data, land use remote sensing data and meteorological and hydrological data. The trained soil and moisture assessment tool model was used for hydrological simulation in plateau and cold regions. The snowfall, snowmelt, and actual evapotranspiration in the study area were simulated using a trained soil and moisture assessment tool model at different time scales, and the changes in snowfall ratio and water storage at different time scales were determined; the snowfall ratio is the ratio of snowfall to total precipitation. The potential evapotranspiration and base flow of the study area at different time scales were obtained; the potential evapotranspiration was determined based on meteorological data of the study area, and the base flow was determined based on the base flow segmentation results of measured runoff data at the outlet section of the study area. A hydrothermal coupling correction model corresponding to the plateau cold region and the functional relationship between base flow and runoff were constructed. The hydrothermal coupling correction model introduced changes in snowfall ratio, snowmelt amount and water storage as correction parameters. The precipitation in the hydrothermal coupling correction model is the effective precipitation obtained after correction. Based on the hydrothermal coupling correction model and functional relationship, the elasticity coefficient of the base flow rate with respect to multiple influencing factors is determined. The influencing factors include precipitation, snowfall ratio, snowmelt, potential evapotranspiration and changes in water storage. The baseflow time series is divided into a baseline period and a change period based on the abrupt change points of the baseflow time series. Based on the elasticity coefficient and the change of each influencing factor between the baseline period and the change period, the contribution of each influencing factor to the baseflow change is determined, so as to output multi-timescale attribution identification.

[0006] The multi-scale attribution identification method for baseflow changes in rivers in high-altitude and cold regions provided in this application simulation method for key hydrological parameters by constructing a trained soil and moisture assessment tool model and determining the elasticity coefficient based on a hydrothermal coupling correction model. This method accurately identifies the multi-scale attribution of baseflow changes and overcomes the limitations of traditional hydrothermal balance models in high-altitude and cold regions. It also improves the accuracy and reliability of baseflow change attribution and provides a scientific basis for sustainable water resource management. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic diagram of a multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions provided in the embodiments of this application; Figure 2 This is a schematic diagram of the Budyko-Fu model provided in the embodiments of this application; Figure 3 These are the seven time scales provided in the embodiments of this application. A diagram illustrating the value; Figures 4a to 4f The base current provided in the embodiments of this application A schematic diagram of the elasticity coefficients of the six influencing factors; Figure 5 The base current provided in the embodiments of this application A schematic diagram at 7 time scales; Figure 6 This is a schematic diagram of the statistic U provided in the embodiments of this application across 7 time scales; Figures 7a to 7g This is a schematic diagram of the contribution rate of each influencing factor to the base current change provided in the embodiments of this application across seven time scales. Detailed Implementation

[0009] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0010] This application provides a multi-scale attribution identification method for baseflow changes in rivers in high-altitude and cold regions. The following will describe this application in detail with reference to embodiments.

[0011] The traditional Budyko-Fu model (the traditional model corresponding to the hydrothermal coupling correction model mentioned below) is difficult to accurately capture the delayed release effect of snowfall when simulating runoff processes and hydrothermal balance in high-altitude cold regions. Furthermore, the secondary water supply formed by snowmelt significantly alters the timing of hydrothermal balance, making the model unable to adapt to the complex hydrological processes and seasonal climate change disturbances in cold regions.

[0012] To address this, this application proposes a multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions, such as... Figure 1 As shown, the multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions provided in this application includes the following steps: S101: Based on the research data of the study area, a soil and moisture assessment tool model is constructed, and the parameters of the soil and moisture assessment tool model are optimized and calibrated using the measured cross-sectional runoff data of the study area to obtain the trained soil and moisture assessment tool model; the research data includes digital elevation model data, soil data, land use remote sensing data and meteorological and hydrological data, and the trained soil and moisture assessment tool model is used for hydrological simulation in plateau and cold regions.

[0013] S102: The snowfall, snowmelt, and actual evapotranspiration in the study area at different time scales are simulated using the trained soil and moisture assessment tool model, and the changes in snowfall ratio and water storage at different time scales are determined; the snowfall ratio is the ratio of snowfall to total precipitation.

[0014] S103: Obtain the potential evapotranspiration and base flow of the study area at different time scales; the potential evapotranspiration is determined based on the meteorological data of the study area, and the base flow is determined based on the base flow segmentation results of the measured runoff data at the outlet section of the study area.

[0015] S104: Construct a hydrothermal coupling correction model corresponding to the plateau cold region and a functional relationship between base flow and runoff; the hydrothermal coupling correction model introduces the snowfall ratio, snowmelt amount and water storage change as correction parameters, and the precipitation in the hydrothermal coupling correction model is the corrected effective precipitation.

[0016] S105: Based on the hydrothermal coupling correction model and the functional relationship, determine the elasticity coefficient of the base flow rate with respect to multiple influencing factors, including precipitation, snowfall ratio, snowmelt, potential evapotranspiration, and changes in water storage.

[0017] S106: Based on the abrupt change points of the baseflow time series, the baseflow time series is divided into a baseline period and a change period. Based on the elasticity coefficient and the change of each influencing factor between the baseline period and the change period, the contribution degree and contribution rate of each influencing factor to the baseflow change are determined to output multi-timescale attribution identification.

[0018] For ease of understanding, the following explains some key terms in this embodiment: The Soil and Moisture Assessment Tool (SWAT) model is a semi-distributed hydrological model used to simulate hydrological processes, mass migration, and crop growth within a watershed. This model can divide the watershed into multiple hydrological response units, thereby enabling refined simulation of hydrological responses in different regions.

[0019] The snowfall ratio is the ratio of snowfall to total precipitation. It reflects the composition of precipitation phases and is of great significance for understanding hydrological processes in cold regions.

[0020] The hydrothermal coupling correction model is a model that is based on the traditional hydrothermal balance model (i.e., the traditional Budyko-Fu model), but incorporates parameters specific to cold regions, such as snowfall ratio, snowmelt amount, and water storage changes, to make corrections. Its aim is to more accurately describe the hydrothermal balance process in cold regions.

[0021] The elasticity coefficient refers to the sensitivity of the base flow to changes in a certain influencing factor. It quantifies the relative change in the base flow when the influencing factor changes by a unit.

[0022] The abrupt change point in a baseflow time series refers to the moment when the statistical characteristics (such as mean and variance) of the baseflow time series change significantly, and it is used to divide different hydrological periods.

[0023] This embodiment first constructs a soil and moisture assessment tool model based on research data from the study area, and then optimizes and calibrates the model using measured cross-sectional runoff data from the study area, resulting in a trained soil and moisture assessment tool model. The research data includes digital elevation model data, soil data, land use remote sensing data, and meteorological and hydrological data. The trained soil and moisture assessment tool model is used for hydrological simulation in high-altitude and cold regions. In one implementation, research data can be manually input, and model parameters can be adjusted based on experience. Preliminary model calibration can be performed by visually comparing the trends of simulated and measured runoff. This method is simple to operate, but may suffer from strong subjectivity and limited accuracy.

[0024] Secondly, the trained soil and moisture assessment tool model simulated the snowfall, snowmelt, and actual evapotranspiration in the study area at different time scales, and determined the changes in snowfall ratio and water storage at different time scales. The snowfall ratio is the ratio of snowfall to total precipitation. In one implementation, snowfall and snowmelt can be estimated using simple empirical formulas based on historical meteorological data. For example, precipitation is considered snowfall when the temperature is below a certain preset threshold; snow melts when the temperature is above another preset threshold. Actual evapotranspiration can be estimated using empirical formulas based on temperature. Changes in water storage can be estimated by statistical averaging of historical observation data or by simple linear interpolation methods.

[0025] Next, the potential evapotranspiration and baseflow of the study area at different time scales are obtained. Potential evapotranspiration is determined based on meteorological data of the study area, while baseflow is determined based on the baseflow segmentation results of measured runoff data at the outlet section of the study area. In one implementation, potential evapotranspiration can be calculated using simple empirical formulas, such as the Thornthwaite method or Hargreaves method based on daily average temperature, which only require temperature data for estimation. Baseflow can be estimated using graphical segmentation methods, for example, by drawing straight lines or curves along the runoff process line to separate baseflow from surface runoff.

[0026] Then, a hydrothermal coupling correction model corresponding to the plateau cold region and the functional relationship between base flow and runoff are constructed. The hydrothermal coupling correction model introduces changes in snowfall ratio, snowmelt amount, and water storage as correction parameters, and the precipitation in the model is the corrected effective precipitation. In one implementation, the hydrothermal coupling correction model can be based on the traditional Budyko-Fu framework and introduce a simple linear correction term. For example, changes in snowfall ratio, snowmelt amount, and water storage are added as independent linear factors to the precipitation term of the traditional model to form the corrected effective precipitation. The functional relationship between base flow and runoff can be constructed using a simple linear regression model.

[0027] Furthermore, based on the hydrothermal coupling correction model and functional relationships, the elasticity coefficients of the base flow rate with respect to multiple influencing factors are determined. These factors include precipitation, snowfall ratio, snowmelt, potential evapotranspiration, and changes in water storage. In one implementation, the elasticity coefficients can be determined using sensitivity analysis methods. For example, by gradually changing the small values ​​of each influencing factor and observing the corresponding changes in the base flow rate, the elasticity coefficients of each factor can be calculated. This method can be implemented using numerical perturbation or regression analysis.

[0028] Finally, based on the abrupt change points in the baseflow time series, the baseflow time series is divided into a baseline period and a change period. Based on the elasticity coefficients and the changes in each influencing factor between the baseline and change periods, the contribution degree and contribution rate of each influencing factor to the baseflow change are determined to output multi-timescale attribution identification. In one implementation, the abrupt change points in the baseflow time series can be identified through visual inspection or a simple moving average method. For example, by observing the long-term trend changes in the baseflow series or calculating the differences in average values ​​across different time periods, significant change points can be determined. Once the abrupt change points are identified, the time series can be divided into a baseline period and a change period. The contribution degree and contribution rate of each influencing factor to the baseflow change can be calculated by linearly superimposing the elasticity coefficients and the average changes of each factor over different periods.

[0029] This method effectively solves the problems of delayed snowfall release and misaligned snowmelt timing in traditional models for simulating water and heat balance in high-altitude and cold regions by constructing a hydrological model adapted to the characteristics of high-altitude and cold regions and introducing key correction parameters such as changes in snowfall ratio, snowmelt amount, and water storage. Therefore, it can more accurately quantify the driving mechanisms of various climate and underlying surface factors on baseflow changes, providing a scientific basis for water resource management and ecological protection in high-altitude and cold regions.

[0030] In some embodiments of this application, measured cross-sectional runoff data of the study area are used to optimize and calibrate the parameters of the soil and water assessment tool model to obtain a trained soil and water assessment tool model. Specifically, this includes: inputting standardized measured cross-sectional runoff data into the soil and water assessment tool model to divide the hydrological response units of the study area and construct a database; using historical measured cross-sectional runoff data of the outlet section of the study area, using a preset optimization algorithm to adjust the key hydrological parameters in the soil and water assessment tool model; and adjusting the coefficient of determination and Nash efficiency coefficient of the soil and water assessment tool model by comparing simulated runoff and measured runoff until the coefficient of determination and the Nash efficiency coefficient meet a preset accuracy threshold.

[0031] The standardized measured cross-sectional runoff data aims to eliminate differences between data of different dimensions or magnitudes, reduce the impact of data noise and outliers on model training, and ensure that the data has a uniform scale when input into the model, thereby improving the model's convergence speed and the stability of parameter optimization. For example, min-max normalization can be used to linearly scale the data to a fixed interval, such as [0, 1]; or Z-score normalization can be used to convert the data into a distribution with a mean of 0 and a standard deviation of 1, which is suitable for cases where the data distribution is approximately normal. These data are input into the soil and moisture assessment tool model through the model's built-in data interface or API, importing the data into the model in a specific format; or by writing scripts or programs to read the data and convert it into an internal data structure that the model can recognize, and then loading it into the model. Dividing the study area into hydrological response units refers to grouping areas with similar topographic, soil, land use, and vegetation characteristics together to simplify the simulation of hydrological processes while preserving the heterogeneity within the watershed. This can be achieved automatically through overlay analysis and threshold setting using Geographic Information System (GIS) tools, combined with Digital Elevation Models (DEMs), soil type maps, and land use maps; or through semi-automatic or manual methods, interactively defining and dividing the data within the GIS environment based on expert knowledge and watershed characteristics. Simultaneously, database construction aims to systematically store and manage all input and output data required for model operation, ensuring data integrity, consistency, and traceability, facilitating model configuration, operation, result analysis, and subsequent maintenance. This can be achieved using a Relational Database Management System (RDBMS), such as PostgreSQL, MySQL, or SQL Server, with table structures designed to store different types of data; or using a NoSQL database, such as MongoDB or Cassandra, storing data in document, key-value pair, or column family format.

[0032] Historical measured runoff data from the outlet section of the study area are used as the "true values" for model parameter calibration. These values ​​are compared with the runoff simulated by the model to guide the direction and magnitude of parameter adjustments, ensuring that the model can accurately reproduce historical hydrological processes. This data can be obtained from long-term continuous runoff observation records from hydrological stations, meteorological bureaus, or relevant research institutions; or through remote sensing technology or hydrological model inversion methods. A pre-defined optimization algorithm can automatically search the model parameter space to find the parameter combination that minimizes the error between the model simulation results and the measured data, avoiding the subjectivity and inefficiency of manual parameter tuning. For example, heuristic optimization algorithms such as the SCE-UA (Shuffled Complex Evolution-University of Arizona) algorithm can be used; or swarm intelligence optimization algorithms such as Particle Swarm Optimization (PSO) or Genetic Algorithm (GA) can be used. Adjusting the key hydrological parameters in the soil and moisture assessment tool model refers to systematically adjusting parameters in the model that have a significant impact on the simulation results of hydrological processes, such as soil hydraulic conductivity, groundwater recharge coefficient, evapotranspiration parameters, and snow melt parameters, in order to improve the simulation accuracy of the model.

[0033] Comparing simulated and measured runoff is a core step in evaluating model performance and guiding parameter adjustments, visually demonstrating the model's ability to reproduce actual hydrological processes. This can be done through visual inspection by plotting time series graphs, scatter plots, or cumulative frequency curves of simulated and measured runoff; or by calculating various error indices between simulated and measured values, such as root mean square error (RMSE) and mean absolute error (MAE). Adjusting the coefficient of determination and Nash efficiency coefficient of the soil and moisture assessment tool model refers to optimizing parameters to better fit the model's simulation results to the measured data, thereby improving the values ​​of these two indices. This can be achieved by including the coefficient of determination and Nash efficiency coefficient in the objective function of the optimization algorithm, allowing the algorithm to directly aim at maximizing these indices when searching for parameters; or by recalculating these coefficients after each parameter adjustment and judging the effect of the parameter adjustment based on their changing trends. The model is considered ready for use in subsequent hydrological simulations and analyses when the coefficient of determination and Nash efficiency coefficient meet preset accuracy thresholds, such as a coefficient of determination greater than 0.7 and a Nash efficiency coefficient greater than 0.6, indicating that the model has sufficient accuracy and reliability.

[0034] Through the aforementioned technical solution, this application introduces a systematic parameter optimization and calibration process to ensure that the soil and moisture assessment tool model can accurately capture the complex hydrological response characteristics of high-altitude cold regions. First, the measured cross-sectional runoff data is standardized and, combined with the division of hydrological response units and database construction, provides the model with high-quality, structured input data and refined spatial representation. Second, using historical measured runoff data, a preset optimization algorithm is used to automatically and efficiently adjust key hydrological parameters in the model, avoiding biases caused by subjective experience. Finally, based on the comparison between simulated and measured runoff, the coefficient of determination and Nash efficiency coefficient are continuously adjusted until a preset accuracy threshold is reached, forming a closed-loop accuracy control mechanism. This systematic optimization and calibration process effectively solves the problem of insufficient parameter adjustment in traditional models, which fails to accurately capture the complex hydrological response characteristics of high-altitude cold regions. After this optimization and calibration process, the trained soil and moisture assessment tool model can more accurately simulate hydrological processes in high-altitude cold regions, especially in key aspects such as snowfall, snowmelt, and actual evapotranspiration. This provides highly reliable basic data for simulating snowfall, snowmelt, and actual evapotranspiration at different time scales, as well as determining changes in snowfall ratio and water storage. This significantly improves the accuracy and reliability of the multi-scale attribution identification method for baseflow changes in rivers throughout the plateau and cold regions. It enables the hydrothermal coupling correction model and the functional relationship between baseflow and runoff to be established on a more solid data foundation, ultimately improving the accuracy and scientific rigor of baseflow attribution analysis.

[0035] In some embodiments of this application, the steps of simulating snowfall, snowmelt, and actual evapotranspiration in the study area at different time scales using the trained soil and moisture assessment tool model specifically include: First, based on the difference between the daily maximum temperature and the critical snowmelt temperature, combined with the snow cover rate, the daily snowmelt equivalent is calculated. The daily maximum temperature can be obtained from measured data from meteorological stations within the study area or through regional climate model simulation. The critical snowmelt temperature can be set to an empirical value of 0℃ or slightly above 0℃ based on the physical characteristics of the snow cover and local climate conditions. Snow cover rate can be extracted from satellite remote sensing imagery data (e.g., snow cover products acquired using remote sensing platforms such as MODIS or Landsat) or estimated through interpolation from ground observation data. The snowmelt equivalent can be calculated using the temperature index method, for example, by multiplying the difference between the daily maximum temperature and the critical snowmelt temperature by a snowmelt coefficient and combining it with the snow cover rate, or by using a more complex energy balance model that comprehensively considers factors such as net radiation, sensible heat, and latent heat.

[0036] Secondly, the snow surface temperature for the current day is recursively calculated based on the previous day's snow cover temperature, the current day's average air temperature, and the snow cover lag coefficient. The previous day's snow cover temperature is the snow surface temperature calculated in the previous simulation time step, and the current day's average air temperature can be obtained from meteorological observation data or forecast data of the study area. The snow cover lag coefficient is a parameter characterizing the thermal inertia of snow cover, reflecting the speed at which snow cover responds to changes in air temperature. Its value can be determined through model calibration or by referring to empirical values ​​for similar snow cover types in high-altitude and cold regions. The recursive calculation of the current day's snow surface temperature can be performed using a weighted average method, such as weighting the previous day's snow cover temperature and the current day's average air temperature according to the snow cover lag coefficient, or by using a physical model that considers the heat conduction process within the snow layer to more accurately simulate the dynamic changes in snow cover temperature.

[0037] Next, based on the date's position within the year's sequence, the daily snowmelt coefficient is dynamically determined between preset maximum and minimum snowmelt coefficients. The date's position within the year's sequence typically refers to Julian Day, i.e., which day of the year. Preset maximum and minimum snowmelt coefficients can be obtained by analyzing historical snowmelt observation data of the study area or through model calibration. The dynamic determination of the daily snowmelt coefficient can employ linear interpolation, interpolating between the maximum and minimum snowmelt coefficients based on Julian Day's position within the year to reflect the seasonal variation in snowmelt efficiency; or a periodic function (such as a sine function) can be used to simulate its seasonal variation.

[0038] Subsequently, based on the snowmelt equivalent, snow surface temperature, and snowmelt coefficient, the trained soil and moisture assessment tool model simulates and outputs a continuous snowmelt sequence and changes in snow cover status. The trained soil and moisture assessment tool model can, for example, be a SWAT model that has been optimized and calibrated for the study area, and it includes a detailed snowmelt module. This model can comprehensively consider the above input parameters, simulate the daily snowmelt amount, and dynamically update snow cover status parameters such as snow depth and snow water equivalent, thereby providing a continuous and physically consistent snowmelt process.

[0039] Finally, based on the snowmelt sequence and changes in snow cover, the snowfall, snowmelt, and actual evapotranspiration in the study area at different time scales are determined. This is typically achieved by summing or averaging the daily snowfall, snowmelt, and actual evapotranspiration output from the model to obtain corresponding values ​​at different time scales such as months, seasons, and years, to meet the needs of subsequent analysis.

[0040] Through the aforementioned technical solutions, this application effectively addresses the shortcomings of traditional methods in capturing daily snowmelt dynamics by refining the simulation of snow accumulation and melting processes in high-altitude and cold regions. Specifically, by considering the difference between the daily maximum temperature and the critical snowmelt temperature, as well as snow cover rate, the snowmelt equivalent is calculated, making the snowmelt potential assessment more accurate. By introducing the previous day's snow temperature and snow lag coefficient to recursively calculate the snow surface temperature, the thermal inertia of snow is effectively reflected, avoiding simulation errors caused by sudden temperature changes. By dynamically adjusting the snowmelt coefficient according to the date of the year, the model can adapt to the seasonal changes in snowmelt rates in high-altitude and cold regions. These improvements work together to enable the trained soil and moisture assessment tool model to output continuous and physically consistent snowmelt amount sequences and snow cover state changes, thereby more accurately determining snowfall, snowmelt, and actual evapotranspiration at different time scales. This provides more reliable and accurate key hydrological parameters for the subsequent construction of hydrothermal coupling correction models, significantly improving the accuracy and credibility of baseflow attribution analysis, and ensuring a deep understanding and effective management of complex hydrological processes in high-altitude and cold regions.

[0041] In some of the embodiments described above in this application, potential evapotranspiration is obtained and input into the hydrothermal coupling correction model. However, if the calculation of potential evapotranspiration is inaccurate, especially if the integration of specific meteorological factors in high-altitude and cold regions, such as net radiation, soil heat flux, and water vapor pressure, is neglected, it will lead to distortion of energy balance and water vapor transport simulation, thereby affecting the accuracy of the hydrothermal coupling correction model and the reliability of the baseflow attribution results.

[0042] In some embodiments of this application, the method for obtaining the potential evapotranspiration and base flux of the study area at different time scales includes: obtaining net radiation and soil heat flux data; determining saturated vapor pressure and actual vapor pressure based on air temperature data; obtaining the slope of the vapor pressure curve and the ecliptic constant; and determining the potential evapotranspiration of the study area at different time scales by integrating the net radiation, soil heat flux, saturated vapor pressure and actual vapor pressure, the slope of the vapor pressure curve, and the ecliptic constant based on wind speed data.

[0043] Obtaining net solar radiation and soil heat flux data is fundamental to calculating potential evapotranspiration. Net solar radiation refers to the difference between shortwave solar radiation reaching and being absorbed by the Earth's surface and longwave solar radiation emitted outward from the surface. It is a key component of the Earth's surface energy balance and directly drives the evapotranspiration process. Soil heat flux refers to the amount of heat passing through a unit area of ​​soil per unit time, reflecting heat exchange within the soil. Especially in high-altitude and cold regions, the freeze-thaw process significantly affects its value. These data can be directly measured using net radiometers and soil heat flux plates equipped at ground-based radiation observation stations. Alternatively, they can be estimated using satellite remote sensing data combined with surface energy balance models, or calculated using meteorological data from weather stations combined with empirical formulas.

[0044] Determining the saturated vapor pressure and actual vapor pressure based on air temperature data is a crucial step in quantifying the air vapor pressure difference. Saturated vapor pressure refers to the maximum vapor pressure at which air reaches saturation at a given temperature; it increases with increasing temperature. Actual vapor pressure refers to the pressure exerted by the actual water vapor present in the air. The difference between the two reflects the dryness and evaporation potential of the air. Saturated vapor pressure is typically calculated from air temperature data using the Clausius-Clapeyron equation or its simplified form. Actual vapor pressure can be determined by multiplying the dew point temperature or relative humidity by the saturated vapor pressure. Alternatively, it can be calculated using hygroscopic charts or empirical formulas based on dry-bulb and wet-bulb temperatures measured by dry-bulb and wet-bulb thermometers.

[0045] Obtaining the slope of the vapor pressure curve and the ecliptic constant are crucial input parameters for physical models such as the Penman-Monteith model. The slope of the vapor pressure curve represents the rate of change of saturated vapor pressure with temperature, a key parameter in the energy term, reflecting the influence of temperature on evaporation potential. The ecliptic constant, related to atmospheric pressure and latent heat, is used to convert energy units to vapor pressure units, reflecting the resistance of air to water vapor transport. The slope of the vapor pressure curve can be calculated from the derivative of saturated vapor pressure with respect to temperature, and its value varies with temperature. The ecliptic constant is typically calculated based on local atmospheric pressure and the latent heat of vaporization of water; atmospheric pressure can be corrected for altitude. Alternatively, empirical or typical values ​​can be obtained from standard meteorological manuals or hydrological model parameter tables, based on the average temperature and altitude of the study area.

[0046] Based on wind speed data, the potential evapotranspiration of the study area at different time scales is determined by integrating net radiation, soil heat flux, saturated vapor pressure and actual vapor pressure, the slope of the vapor pressure curve, and the ecliptic constant. Wind speed data reflects the intensity of air convection and is a key factor affecting the rate of water vapor transport from the Earth's surface to the atmosphere. By integrating all the above parameters, physical models such as the Penman-Monteith model can be used to accurately calculate the potential evapotranspiration by comprehensively considering both energy balance and aerodynamic factors. Specifically, the obtained net radiation, soil heat flux, saturated vapor pressure, actual vapor pressure, vapor pressure curve slope, ecliptic constant, and wind speed data are substituted into the Penman-Monteith formula for calculation. This formula can comprehensively consider the influence of energy supply, vapor pressure difference, and aerodynamic drag on evapotranspiration. When data is limited, the simplified form recommended by FAO Penman-Monteith can also be used, or empirical coefficients and local calibration can be combined to estimate these parameters.

[0047] Through the aforementioned technical solutions, this application can accurately obtain the potential evapotranspiration in high-altitude and cold regions at different time scales. Specifically, by acquiring net radiation and soil heat flux data, the surface energy input and heat exchange characteristics of high-altitude and cold regions can be accurately quantified, overcoming the problem of energy estimation bias under snow and ice cover. Determining the saturated vapor pressure and actual vapor pressure based on temperature data can accurately capture the impact of temperature fluctuations on vapor pressure dynamics in high-altitude and cold regions, reflecting the constraint of air humidity on evapotranspiration potential. Furthermore, obtaining the slope of the vapor pressure curve and the ecliptic constant ensures that the potential evapotranspiration calculation conforms to the physical mechanism of the Penman-Monteith equation, improving the model's applicability in high-altitude and cold regions. Finally, based on wind speed data, by integrating all key meteorological parameters, multi-factor collaborative calculation is achieved, comprehensively considering energy balance and aerodynamic effects, thereby outputting reliable potential evapotranspiration under complex meteorological conditions in high-altitude and cold regions. This accurate calculation of potential evapotranspiration provides high-quality input for the aforementioned hydrothermal coupling correction model, significantly improving the model's accuracy in simulating hydrothermal balance processes in cold regions. This, in turn, enhances the scientific validity and reliability of the baseflow attribution analysis, avoiding distortions in energy balance and water vapor transport simulations caused by inaccurate calculations of potential evapotranspiration.

[0048] In some embodiments of this application, the potential evapotranspiration and baseflow of the study area at different time scales are obtained by: obtaining the maximum baseflow index and the recession constant; using the maximum baseflow index, the recession constant, and the baseflow of the previous time step as recursive inputs; weighting the total runoff of the current time step with the recursive inputs to obtain the baseflow of the current time step; and recursively separating the baseflow from the surface runoff of the entire runoff sequence according to the time series to obtain the baseflow of the study area at different time scales.

[0049] Obtaining the maximum baseflow index and the recession constant are crucial steps in the baseflow separation process. The maximum baseflow index represents the largest proportion of the total runoff to the baseflow, setting a reasonable upper limit for baseflow estimation and ensuring the physical validity of the baseflow separation results. This parameter can be determined through long-term observation and statistical analysis of the watershed's hydrological characteristics, for example, by analyzing historical dry season runoff data or using watershed geological and soil information for empirical estimation. The recession constant reflects the response rate and attenuation characteristics of the watershed's groundwater system to changes in runoff, controlling the rate at which baseflow is released from groundwater reserves. This constant is typically obtained by analyzing runoff recession curves, or it can be calibrated through hydrological models or empirically set based on the watershed's hydrogeological characteristics.

[0050] Using the maximum baseflow index, the recession constant, and the baseflow rate from the previous time step as recursive inputs, this method aims to leverage the temporal continuity of the runoff process. The recursive input mechanism ensures that the baseflow calculation at the current time step fully considers the baseflow state of the previous time step, thus maintaining the smoothness and logical consistency of baseflow changes over time. This approach avoids computational jumps or unreasonable results that might result from treating each time step in isolation, ensuring the dynamic coherence of the baseflow sequence.

[0051] The core of achieving dynamic baseflow separation lies in calculating the baseflow by weighting the total runoff at the current time step with the recursive input. This weighting mechanism allows the system to dynamically adjust the baseflow estimate based on the current total runoff and historical baseflow conditions. For example, when runoff increases rapidly, the weights may be more biased towards the total runoff to reflect the rapid response of surface runoff; while when runoff decreases, the weights may be more biased towards the baseflow and recession constant of the previous time step to simulate the slow release of groundwater. This weighting method effectively integrates real-time hydrological data with the inherent hydrological response characteristics of the watershed, enabling baseflow estimation to adapt to the rapid fluctuations and seasonal changes in runoff in high-altitude and cold regions.

[0052] By recursively analyzing the time series, the baseflow and surface runoff of the entire runoff sequence are separated, yielding baseflow at different time scales for the study area. This step ensures the systematic nature and completeness of the baseflow segmentation process. By starting from the beginning of the runoff sequence and applying the above recursive and weighted calculation method step by step, a continuous and complete baseflow time series can be generated. This recursive approach guarantees the consistency of the entire segmentation process and allows baseflow data to be aggregated to different time scales (such as daily, monthly, quarterly, and annually) as needed, providing accurate and reliable baseflow data for subsequent multi-timescale attribution analysis.

[0053] Through the above technical solutions, this application provides a baseflow segmentation method based on recursive filtering. By introducing key hydrological parameters such as the maximum baseflow index and the recession constant, and utilizing a recursive calculation mechanism, it effectively solves the accuracy problem of baseflow separation in dynamically changing runoff sequences in plateau and cold regions. This method leverages the continuity of time series, incorporating historical baseflow information into the current calculation, overcoming the temporal disconnect problem caused by isolated time steps in traditional methods. By weighting the total runoff at the current time step with the recursive input, it dynamically fuses real-time runoff data with historical baseflow status, enabling baseflow estimation to adapt to the rapid fluctuations and seasonal changes in runoff in plateau and cold regions. The separation of baseflow from surface runoff throughout the entire runoff sequence is completed recursively according to the time series, ensuring the consistency and integrity of the segmentation process and generating reliable baseflow data at multiple time scales. This provides more accurate and reliable baseflow data input for the multi-scale attribution identification method of baseflow changes in rivers in high-altitude and cold regions, thereby significantly improving the accuracy and scientific nature of the entire attribution identification method. Especially in the complex and dynamic environment of high-altitude and cold regions, it can more realistically reflect the actual changes in baseflow and provide a solid data foundation for water resource management and ecological protection.

[0054] In some embodiments of this application, the method for determining the changes in snowfall ratio and water storage at different time scales includes: for each time scale, obtaining the total precipitation, actual evapotranspiration, and total runoff during that period; subtracting the actual evapotranspiration from the total precipitation, and then subtracting the total runoff to obtain a calculation result; using the calculation result as the net change in water storage during that period; and determining the changes in water storage at different time scales based on the net change in water storage during that period.

[0055] Specifically, for each time scale, the total precipitation, actual evapotranspiration, and total runoff are obtained to provide basic data for subsequent water balance calculations. Total precipitation represents the water input to the basin, actual evapotranspiration represents the water output, and total runoff represents the water output. Obtaining this data at different time scales (e.g., monthly, seasonal, annual) is to analyze the dynamic changes in water storage during these specific time periods. This data can be obtained in various ways. For example, it can be obtained using meteorological station networks, satellite remote sensing data (such as TRMM, GPM, and other precipitation products), hydrological station runoff observation data, and actual evapotranspiration simulated by hydrological models based on soil and moisture assessment tools. Alternatively, it can be aggregated by summing or averaging the raw daily-scale data to the required monthly, seasonal, or annual time scales. For example, daily precipitation can be summed to obtain the monthly total precipitation, daily actual evapotranspiration can be summed to obtain the monthly actual evapotranspiration, and daily runoff can be summed to obtain the monthly total runoff.

[0056] Based on this, the total precipitation is subtracted from the actual evapotranspiration, and then the total runoff is subtracted to obtain the calculation result. This step is a core calculation based on the water balance principle, that is, precipitation minus actual evapotranspiration minus total runoff equals the change in water storage. By performing this subtraction operation, the net change in water storage within the watershed over a specific time period can be directly quantified. This calculation can be performed in a Geographic Information System (GIS) platform using a raster calculator or vector data processing tool to aggregate and calculate data from different spatial units; alternatively, it can be performed using a script written in a programming language (such as Python, R, Matlab) to read pre-processed time-series data and perform time-step arithmetic operations to obtain the calculation result.

[0057] Subsequently, the calculation result is used as the net change in watershed storage during that period. This step clarifies the physical meaning of the values ​​calculated in the previous step; that is, the calculation result directly represents the increase or decrease in the total amount of water in the watershed (including soil water, groundwater, lake water, snow, etc.) within a specific time period. Positive values ​​indicate an increase in water storage, and negative values ​​indicate a decrease in water storage. This net change can be stored as a new data field or variable in a database or data file and associated with the corresponding time scale and watershed unit; alternatively, this net change can be directly used as the input to the water storage change parameters in the subsequent hydrothermal coupling correction model to ensure that the model can accurately reflect the dynamics of watershed water storage.

[0058] Finally, based on the net change in watershed storage during this period, the changes in water storage at different time scales are determined. This step formally confirms the calculated net change as the water storage change parameter used for model analysis at different time scales. It emphasizes that the water storage change is specific to a particular time period and is a key input for multi-scale analysis. Specifically, the net change calculated for each time scale (e.g., each month, each quarter) can be directly assigned to the corresponding water storage change parameter as a correction parameter in the hydrothermal coupling correction model; alternatively, these net changes arranged in time series can be constructed into a complete water storage change time series for subsequent elasticity coefficient calculation and attribution analysis.

[0059] Through the aforementioned technical solution, this application clearly and quantitatively calculates the water storage change parameter in the multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions. Specifically, by accurately acquiring key hydrological elements such as total precipitation, actual evapotranspiration, and total runoff, and strictly adhering to the water balance principle in the calculation, it avoids the accumulation of errors caused by data ambiguity or inaccurate calculations in traditional methods. The calculation result is directly defined as the net change in watershed water storage, ensuring that the physical meaning of this parameter is clear and traceable, thereby eliminating the ambiguity in the calculation of water storage changes. This enables the hydrothermal coupling correction model to obtain more accurate input parameters, significantly improving the model's simulation accuracy for complex hydrological processes in high-altitude cold regions and its ability to describe hydrothermal balance. Given that changes in water storage are an important correction parameter in the hydrothermal coupling correction model, their accurate determination directly enhances the reliability of the calculation of the elasticity coefficients of factors such as precipitation, snowfall ratio, snowmelt, and potential evapotranspiration, thereby improving the accuracy and scientific nature of multi-scale attribution identification of baseflow changes. This provides a more solid data foundation and decision support for water resource management and ecological protection in plateau and cold regions.

[0060] In some embodiments of this application, the step of obtaining effective precipitation includes: calculating the amount of water in the non-snowfall portion of the total precipitation, i.e., multiplying the total precipitation by (1 minus the snowfall ratio); obtaining the effective precipitation based on the non-snowfall water, the concurrent snowmelt amount, and the concurrent water storage change; adding the snowfall water to the concurrent snowmelt amount, subtracting the concurrent water storage change, and finally obtaining the effective precipitation.

[0061] The calculation of the non-snowfall portion of total precipitation, i.e., total precipitation multiplied by (1 minus the snowfall ratio), aims to accurately distinguish between liquid and solid precipitation to determine the amount of liquid water immediately participating in the hydrological cycle at a specific time scale. This is particularly crucial for high-altitude and cold regions because the delayed release effect of snowfall means that not all precipitation is immediately available. This step can be achieved by using real-time meteorological observation data combined with a preset temperature threshold method to distinguish precipitation phases. For example, when the temperature is above 0°C, precipitation is considered rain; when the temperature is below 0°C, precipitation is considered snow. The snowfall ratio is then estimated based on historical statistical data or real-time observation data. Alternatively, remote sensing data (such as snow cover and snow depth) can be combined with meteorological models to dynamically estimate snowfall and total precipitation, and then calculate the snowfall ratio. The non-snowfall portion of the water is the total precipitation minus the snowfall.

[0062] The effective precipitation is derived from the non-snowfall water volume, concurrent snowmelt volume, and concurrent changes in water storage. This step aims to integrate various water input and storage changes to derive the "effective precipitation" that truly contributes to the hydrological system. It acknowledges that not all precipitation is immediately available, and that snowmelt and storage changes play a crucial role. This step is implemented based on the water balance principle, comprehensively considering the immediately available liquid precipitation, the water released by snowmelt, and the net change in water storage within the watershed. Alternatively, a hydrological model can be established, using non-snowfall water volume as a direct input and snowmelt volume as an additional water source input, while simultaneously monitoring or simulating changes in soil moisture content and groundwater storage within the watershed, thereby calculating the effective water volume actually involved in runoff formation and evapotranspiration over a specific time scale.

[0063] The effective precipitation is obtained by adding snowfall to the concurrent snowmelt and subtracting the change in water storage during the same period. This step provides a specific method for calculating effective precipitation, emphasizing the contribution of snowmelt and its adjustment to changes in water storage. It corrects for the delayed release of water from snow accumulation and the actual dynamic changes in watershed water storage. The specific calculation process can be expressed as a formula, for example: Effective Precipitation = (Total Precipitation * (1 - Snowfall Ratio)) + Snowmelt - Change in Water Storage. The snowfall ratio, snowmelt, and change in water storage must all be obtained through the aforementioned steps or model simulation. Alternatively, these parameters can be continuously adjusted in the hydrological model using iterative optimization methods to match the simulated runoff with the measured runoff, thereby obtaining a more accurate effective precipitation.

[0064] The above technical solution achieves several key benefits. First, by calculating the non-snowfall portion of total precipitation, it accurately distinguishes between liquid and solid precipitation, preventing the misclassification of snowfall as immediately available water and ensuring the accuracy of non-snowfall water volume, thus providing a foundation for subsequent integration. Second, by comprehensively considering non-snowfall water volume, concurrent snowmelt, and concurrent changes in water storage, it effectively captures the characteristics of delayed snowfall release and seasonal water storage fluctuations in high-altitude cold regions, ensuring that effective precipitation truly reflects the actual water supply timeline. Finally, by adding snowfall water volume to concurrent snowmelt and subtracting concurrent changes in water storage, it quantifies the contribution of snowmelt and adjusts the impact of water storage, eliminating model input bias caused by neglecting snowmelt or changes in water storage. This ensures the accuracy of the hydrothermal balance simulation, thereby improving the reliability of baseflow attribution and enabling the multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions to more accurately reflect actual hydrological processes.

[0065] In some embodiments of this application, the method for dividing the basestream time series into a base period and a change period based on the abrupt change points of the basestream time series specifically includes: for a basestream time series of length n, calculating the order statistic corresponding to each possible segmentation point; finding the time point where the absolute value of the order statistic reaches its maximum; comparing the absolute value of the maximum order statistic with the critical value at a given significance level to determine whether the time point is a statistically significant abrupt change point; and dividing the entire basestream time series into two stages, the base period and the change period, based on the abrupt change points.

[0066] For a basestream time series of length n, the order statistic for each possible split point is calculated to quantify the degree of change at each potential split point. The order statistic is a nonparametric statistic that does not depend on a specific data distribution and can effectively detect structural changes in a time series. For example, the Pettitt test statistic U_t,n can be used, which reflects the change in the mean of the series by comparing the rank sums of the two segments before and after the split point; alternatively, the Mann-Whitney U test can be used to calculate the difference in rank sums between the subsequences before and after each possible split point. Calculating these statistics provides objective data support for subsequent abrupt change point identification, avoiding the arbitrariness of pre-determined split points.

[0067] Finding the time point where the absolute value of the order statistic reaches its maximum is crucial for identifying the most significant changes in the time series. After calculating the order statistics for all possible breakpoints, the point with the largest absolute value is selected as the most likely abrupt change. This step focuses the analysis on the time point that best reflects the change in the series' properties, effectively eliminating the risk of bias from subjective selection and ensuring that the identified abrupt change has the greatest statistical significance.

[0068] Comparing the absolute value of the largest order statistic with a critical value at a given significance level to determine whether the time point is a statistically significant abrupt change is a crucial step in verifying the reliability of the identified abrupt change. After determining the time point with the largest absolute value of the order statistic, it needs to be compared with a critical value at a pre-set significance level (e.g., 0.05 or 0.01). The critical value can be obtained by consulting statistical tables, Monte Carlo simulations, or asymptotic distribution theory. If the absolute value of the largest order statistic exceeds the critical value, the time point is considered a statistically significant abrupt change, meaning the change is not due to random fluctuations but rather a substantial alteration in the intrinsic structure of the sequence. This step utilizes the rigor of statistical testing to ensure the scientific validity and reliability of the partitioning results.

[0069] Dividing the entire baseflow time series into two phases—the baseline period and the change period—based on the aforementioned abrupt change points provides a clear time periodization for subsequent attribution analysis. Once a statistically significant abrupt change point is identified, the series data before that point is defined as the baseline period, while the series data after that point is defined as the change period. This statistically validated division method provides a solid and objective time periodization basis for subsequent calculations of the contribution of each influencing factor to baseflow changes, making the results of the attribution analysis more convincing.

[0070] Through the aforementioned technical solution, this application can objectively and accurately identify abrupt changes in baseflow time series, avoiding errors that may arise from subjective judgment. By introducing a statistical testing mechanism, the division between the baseline period and the change period is ensured to be statistically significant, thus providing a reliable time-based basis for subsequent calculations of the contribution degree and contribution rate of each influencing factor to baseflow changes based on elasticity coefficients. This significantly improves the scientific rigor and accuracy of the multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions, making the quantitative analysis of the driving mechanisms of climate and underlying surface factors more precise, and thus providing more solid data support for water resource management and ecological protection.

[0071] In some embodiments of this application, the foregoing method further includes the following steps: Snow observation data aims to reflect the current actual state of snow cover in the study area. Acquisition methods may include, but are not limited to: obtaining snow cover extent, snow depth, or snow water equivalent products through satellite remote sensing technology; obtaining real-time snow depth, snow water equivalent, and temperature data through ground sensor networks, such as automatic weather stations or snow depth sensors; or obtaining high-precision snow depth data using UAV-borne lidar scanning. Short-term weather forecast data is used to predict meteorological elements over a future period. For example, it can be derived from temperature, precipitation (and its phase), and wind speed data output from numerical weather prediction models (such as WRF and GFS); or obtained from short-term forecast products of regional climate models (RCM); or it can be output from meteorological prediction models based on machine learning or deep learning. Regular access to this data (e.g., daily or weekly) allows for timely acquisition of the latest environmental information, effectively avoiding the lag problem caused by relying solely on historical data.

[0072] Based on the real-time snow cover observation data and the short-term weather forecast data, the snowfall ratio and snowmelt parameters used for short-term prediction are dynamically corrected. This dynamic correction process aims to adjust key parameters in the model according to real-time information to better adapt to the dynamic changes in snow cover in high-altitude and cold regions. Specifically, based on real-time snow cover observation data and combined with temperature information from short-term weather forecasts, simplified physical models or empirical relationships (e.g., calibration based on the temperature index method combined with real-time snow depth or snow water equivalent data) can be used to estimate the current initial values ​​of snowfall ratio and snowmelt in real time. Then, different weights are assigned to the parameter values ​​statistically obtained from the soil and moisture assessment tool model and the current initial values, and these are fused to obtain the corrected snowfall ratio and snowmelt parameter values. Alternatively, data assimilation techniques can be used to integrate real-time observation data into the model's state variables, thereby indirectly correcting the relevant parameters. Furthermore, machine learning models can be used to train or fine-tune the parameters using real-time data. The snowfall ratio, which is the ratio of snowfall to total precipitation, reflects the phase characteristics of precipitation; the snowmelt amount represents the amount of water formed by melting snow, reflecting the snow ablation process. By dynamically correcting these parameters, their accuracy in short-term forecasting can be significantly improved, making them closer to actual dynamic changes.

[0073] The corrected snowfall ratio and snowmelt parameters, along with precipitation and potential evapotranspiration data from short-term weather forecasts, are input into the hydrothermal coupling correction model to output a predicted baseflow value for the near future. The hydrothermal coupling correction model refers to the model constructed using the aforementioned method, which incorporates changes in snowfall ratio, snowmelt, and water storage as correction parameters. By using the snowfall ratio and snowmelt parameters corrected with real-time data, along with precipitation and potential evapotranspiration data from short-term weather forecasts, as input, the hydrothermal coupling correction model is driven to perform calculations, thereby obtaining a predicted baseflow value for the near future. This step ensures that the prediction model can utilize the latest and most accurate parameter information, thereby generating more reliable short-term baseflow prediction results to address the non-stationarity of hydrological processes in high-altitude and cold regions.

[0074] Based on multiple meteorological scenarios or parameter perturbations, baseflow forecasts are generated, and the uncertainty of the forecasts is expressed in the form of forecast intervals. These multiple meteorological scenarios may include, but are not limited to: using ensemble forecast products to obtain meteorological data from different forecast members as scenario inputs; or constructing different meteorological scenarios such as optimistic, neutral, and pessimistic based on historical extreme events or statistical distributions. The parameter perturbations refer to small-range changes to model parameters to assess their impact on the forecast results. For example, Monte Carlo simulations can be performed on the corrected snowfall ratio and snowmelt parameters to introduce random perturbations; or sensitivity analysis can be performed on other uncertain parameters in the hydrothermal coupling correction model. By generating a series of baseflow forecasts and expressing them in the form of forecast intervals (e.g., by calculating statistics such as mean, standard deviation, and percentiles of different scenario or perturbation results to determine the intervals), rather than a single deterministic value, the uncertainty of the forecasts can be quantified, providing water resource managers with more comprehensive and reliable decision support.

[0075] Through the above technical solution, this application effectively solves the parameter reliability problem caused by dynamic changes in snow cover in short-term baseflow forecasting in high-altitude and cold regions. Regularly accessing real-time snow cover observation data and short-term weather forecast data enables the immediate capture of snow cover status and future weather trends, avoiding the lag caused by relying solely on historical data and providing real-time basis for parameter correction. Based on this, the snowfall ratio and snowmelt parameters are dynamically corrected. By fusing real-time observations with model statistics, key parameters are directly adjusted to reflect actual dynamic changes, significantly improving the accuracy of parameters in short-term forecasting. The corrected reliable parameters drive a hydrothermal coupling correction model, ensuring that the forecast results better reflect the non-stationarity of hydrological processes in high-altitude and cold regions. Furthermore, by generating forecast intervals based on multiple meteorological scenarios or parameter perturbations, potential variability is quantified, providing the necessary reliability range for water resource management, rather than a single forecast value. This allows this method to transform a static historical attribution model into a dynamically adaptable forecasting tool, significantly improving the accuracy and reliability of short-term baseflow forecasting, thereby better supporting refined water resource scheduling and management decisions.

[0076] In some embodiments of this application, a method for dynamically correcting the snowfall ratio and snowmelt amount parameter values ​​used for short-term prediction includes: estimating the current initial values ​​of snowfall ratio and snowmelt amount in real time based on the real-time snow cover observation data; assigning different weights to the parameter values ​​obtained statistically based on the soil and moisture assessment tool model and the current initial values, and fusing them to obtain corrected snowfall ratio and snowmelt amount parameter values.

[0077] The initial values ​​of snowfall ratio and snowmelt amount are estimated in real time based on the real-time snow cover observation data. This aims to capture the immediate dynamic changes in snow cover conditions and provide a current, realistic starting point for subsequent parameter correction. This is particularly crucial for high-altitude cold regions, where the accumulation and melting of snow are significantly affected by factors such as temperature and precipitation, and are highly uncertain. In practice, various methods can be employed. For example, the snowmelt coefficient can be dynamically calibrated based on the temperature index method, combined with real-time data on snow cover extent, snow depth, or snow water equivalent, to estimate the initial value of the current snowmelt amount. Simultaneously, the ratio of snowfall to total precipitation in the current period, i.e., the initial value of the snowfall ratio, can be estimated through real-time precipitation phase observation or by combining temperature threshold judgment. Another approach is to utilize a machine learning model trained on historical real-time observation data (such as satellite remote sensing data and ground sensor data) and using current real-time snow cover observation data as input to directly output the initial values ​​of the snowfall ratio and snowmelt amount.

[0078] This paper assigns different weights to the parameter values ​​obtained from the soil and moisture assessment tool model and the current initial values, and then fuses them to obtain corrected snowfall ratio and snowmelt amount parameter values. This aims to effectively integrate the advantages of information from different sources and overcome the limitations of a single data source. The parameter values ​​obtained from the soil and moisture assessment tool model typically represent long-term, statistically significant averages or trends, exhibiting good stability and reliability; while the current initial values ​​reflect immediate, local conditions, possessing high timeliness. By assigning different weights, adjustments can be made flexibly according to actual needs and data characteristics. For example, the weights can be dynamically adjusted based on the quality, coverage, and timeliness of real-time observation data. When the real-time data is of high quality and fully reflects the current situation, it is given a higher weight; conversely, it relies more on the model statistical parameters. Furthermore, a time decay weighting method can be used, where for short-term predictions, the current initial value is given a higher weight, and as the prediction time scale increases, the weight of the current initial value gradually decreases, while the weight of the model statistical parameters gradually increases. Another approach is to use data assimilation techniques such as Kalman filtering to take the parameter values ​​obtained from the soil and moisture assessment tool model as prior information and the current initial value as observation information, and then fuse them through optimal estimation theory to obtain the corrected parameter values.

[0079] Through the above technical solution, this application effectively solves the problem of decreased reliability of key parameters (snowmelt amount and snowfall ratio) caused by the uncertainty of snow cover dynamics in short-term baseflow prediction in high-altitude cold regions. Specifically, by estimating the current initial values ​​of snowfall ratio and snowmelt amount in real time based on real-time snow cover observation data, the correction process ensures that it can respond promptly to the immediate state of snow cover, avoiding bias caused by the lag of historical data, and making the parameters more timely and accurate. On this basis, different weights are assigned to the parameter values ​​obtained statistically based on the soil and moisture assessment tool model and the current initial values, and then fused, realizing an intelligent balance between long-term statistical regularities and short-term real-time changes. This differentiated weight allocation mechanism can dynamically adjust the contribution of the two types of parameters according to the reliability and timeliness of real-time data and the needs of the prediction time scale, thereby avoiding the accuracy loss that may be caused by simple averaging or fixed weight fusion, especially when the dynamic changes of snow cover in high-altitude cold regions are nonlinear or sudden, it can more accurately capture their impact. Ultimately, the corrected snowfall ratio and snowmelt parameters obtained through this optimized fusion can be more accurately input into the hydrothermal coupling correction model, significantly improving the accuracy and reliability of baseflow prediction in the short term and providing a more refined and reliable scientific basis for water resource scheduling and management decisions in plateau and cold regions.

[0080] The following example will provide a more detailed explanation of the above technical solution: In high-altitude and cold environments, snowfall accounts for a significantly higher proportion of annual precipitation than in other watersheds. This unique precipitation phase characteristic leads to a delayed release effect of snow accumulation, disrupting the synchronicity between actual evapotranspiration (ET) and precipitation (P). Secondary water supply from snowmelt at seasonal scales significantly alters the hydrothermal balance timeline. Therefore, the traditional Budyko-Fu model is ill-suited to the complex hydrological processes and seasonal climate change disturbances in cold regions. To overcome this limitation, this method couples the snowmelt runoff mechanism to the SWAT model (i.e., a soil and moisture assessment tool model), fully considering the impact of snowfall-snowmelt-runoff on the runoff generation and collection mechanisms in high-altitude and cold regions, and introducing the snowfall ratio (…). =Snowfall / Total Precipitation) and snowmelt (M) are used as improved parameters, and water storage changes are considered on a seasonal scale ( To mitigate the impact of [the pandemic], a seasonal-scale hydrothermal coupling correction model for cold regions was constructed. This model [is used in conjunction with other models]. The value corrects the disturbance effect of precipitation phase on water balance, and will effectively reduce precipitation ( To replace precipitation P in the original model and reflect the actual water supply characteristics after snowmelt lag, a functional relationship B=f(R) between baseflow (B) and runoff (R) is constructed, and baseflow is calculated using the chain rule. By analyzing the elasticity coefficients of each driving factor, the attribution of baseflow can be ultimately achieved. This model can better reflect the complex process of snowfall-melt-runoff in cold regions, establishing a new analytical framework for the study of hydrological responses in cold regions under climate change.

[0081] The improved Budyko-Fu elasticity coefficient method for attributing base current variations provided in this embodiment includes the following steps: Step a: Acquire DEM (Digital Elevation Model) data, soil data, land use remote sensing data, and meteorological and hydrological data for the study area. Perform SWAT model standardization on the data. Optimize and calibrate the model parameters using measured cross-sectional runoff data, and determine the coefficients (R²). 2 The optimal SWAT model for cold regions was obtained through evaluation using the temperature index method and the Nash coefficient (NSE). The SWAT model uses the temperature index method to estimate the snow accumulation and melting process, and controls the influence of the previous day's air temperature on the current day's snow temperature through a lag coefficient. The specific implementation process is shown in Formula 1. (1) In the formula: This represents the snowmelt equivalent t (mm). Snow melting factor (mm / d°C) Indicates the snow cover rate of the HRU unit. Indicates the surface temperature of the snow (°C). This indicates the simulated daily maximum temperature. This indicates the critical temperature for snow melting (°C).

[0082] The formula for calculating the surface temperature of snow is shown in Formula 2: (2) In the formula: This represents the simulated daily snow cover temperature (°C). This indicates the previous day's snow cover temperature (°C). Indicates the snow accumulation lag coefficient. This indicates the daily average temperature (°C). Snowmelt coefficient The calculation formula is shown in Formula 3: (3) In the formula The daily snowmelt coefficient (mm / d°C) This indicates the maximum snow melting rate (mm / d°C). The minimum snow melting coefficient (mm / d·°C) Refers to Julian Day (d) within the current year.

[0083] Step b: Using the SWAT model to simulate the output results, calculate the snowfall ratio at multiple time scales (spring, summer, autumn, winter, warm, cold, and annually). The series of processes including snow melting (M) and evapotranspiration (E).

[0084] Step c: The Penman-Monteith (PM) method is used to calculate the potential evapotranspiration sequences (PET) for multiple timescales in the study area during spring, summer, autumn, winter, warm, cold, and throughout the year. The PM calculation formula is shown in Formula 4. (4) In the formula: PET is the potential evaporation rate (mm / d). Rn is the slope of the water vapor pressure curve (kPa / ℃), and Rn is the net radiation (MJ / (m2)). d) ), G is the soil heat flux ( MJ / (m2) d) ), γ is the ecliptic constant (kPa / ℃), T is the daily average temperature at 2 meters (℃), u2 is the wind speed at 2 meters (m / s), e s e is the saturated vapor pressure (kPa). a This represents the actual water vapor pressure (kPa).

[0085] Step d: The Eckhardt two-parameter digital filtering method is applied to segment the baseflow of the outlet section of the study area, and the baseflow processes at multiple time scales (spring, summer, autumn, winter, warm, cold, and year-round) are estimated. The Eckhardt calculation formula is shown in Formula 5. (5) In the formula: The recession constant, This represents the maximum percentage of baseflow to total runoff over a long period. and These are the total runoff and baseflow at time step i, respectively.

[0086] Step e: Apply hydrological statistical analysis methods to construct the functional relationship B=f(R) of baseflow (B) and runoff (R) at multiple time scales for spring, summer, autumn, winter, warm, cold and all year round.

[0087] Step f, as follows Figure 2 As shown, an improved Budyko-Fu model for high-altitude and cold regions is constructed. The calculation principle of this model is shown in Equation 6. (6) Step g: Using the evapotranspiration (E) and measured precipitation (P) and runoff (R) obtained in step b above, the change in water storage is calculated through the water balance equation. The specific formula is shown in Formula 7: (7) Step h involves simultaneously improving the Budyko-Fu equations using the water balance equations to obtain the final expression for the runoff R, as shown in Equation 8: (8) Step i: Obtain M and fs through actual measurements of R and P and calculations. PET, through inversion using formula (8), yields multi-timescale data. Value, specifically The diagram illustrates the values ​​at seven different time scales (including spring, summer, autumn, winter, warm, cold, and year-round). Figure 3 As shown.

[0088] Step j, based on the functional relationship between baseflow and runoff Other parameters , , , , , As independent variables, the fundamental current can be obtained using the chain rule. The elasticity coefficients for each influencing factor are shown in Formula 9: (9) In the formula: is the elasticity coefficient of the baseflow to the influencing factor x, which indicates the sensitivity of the baseflow to various climatic factors; x represents a certain climatic factor that affects the change of the baseflow.

[0089] Assumption Then the base current The formulas for calculating the elasticity coefficients of each influencing factor are shown in Formulas 10 to 15, respectively, for the base current. The elasticity coefficients of each influencing factor at seven different time scales correspond as follows: Figures 4a to 4f As shown.

[0090] (10) (11) (12) (13) (14) (15) Step k involves applying the Pettitt test to identify abrupt changes in the base current across multiple scales (spring, summer, autumn, winter, warm, cold, and year-round), thus separating the baseline and variation periods of R. This method effectively detects abrupt changes in time series data and can analyze the statistical significance of these changes.

[0091] Using statistics To test the sequence Two subsamples and , The calculation method is shown in Formula 16: (16) In the formula It is a sign function. The Pettitt test's null hypothesis is that the original sequence has no abrupt change points. If the statistic is satisfied at time t... Then time t is the mutation point to be detected. If the conditions shown in Formula 17 are met at the corresponding significance level α, it is considered that a sudden change occurs at time t at the significance level α.

[0092] (17) In this step, the correspondence of base current B at 7 different time scales is as follows: Figure 5 As shown, the correspondence of statistic U at 7 different time scales is as follows: Figure 6 As shown.

[0093] Step 1, set the base period and the change period. , , , , , Substituting into the following formula, multi-scale attribution identification of each driving factor can be achieved using R. Based on the Petitt mutation analysis results, the base current sequence is divided into a baseline period and a change period. The base current change in dB in both stages can be interpreted as... , , , , , Changes in base current resulting from the change (dfs, dP, dM, dPET, d d The contribution of each driving factor to the baseflow change is calculated as shown in Formula 18: (18) In the formula: dx represents two periods , , , , , The change value, The baseflow change is characterized by the amount of change caused by each variable, and the contribution rate of each driving factor to the baseflow change can be calculated as shown in Formula 19: (19) In the formula: The calculated value represents the base current change. for , , , , The contribution rate of factors such as the base current variation.

[0094] After performing this step, the corresponding relationships of the contribution rates of different influencing factors to baseflow changes at seven different time scales are as follows: Figures 7a to 7g As shown.

[0095] The above provides a detailed description of a multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions, as provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions, characterized in that, include: Based on the research data of the study area, a soil and moisture assessment tool model was constructed, and the parameters of the soil and moisture assessment tool model were optimized and calibrated using the measured cross-sectional runoff data of the study area to obtain the trained soil and moisture assessment tool model; the research data included digital elevation model data, soil data, land use remote sensing data and meteorological and hydrological data, and the trained soil and moisture assessment tool model was used for hydrological simulation in plateau and cold regions. The trained soil and moisture assessment tool model was used to simulate the snowfall, snowmelt, and actual evapotranspiration in the study area at different time scales, and to determine the changes in snowfall ratio and water storage at different time scales; the snowfall ratio is the ratio of snowfall to total precipitation. The potential evapotranspiration and base flow of the study area at different time scales are obtained; the potential evapotranspiration is determined based on meteorological data of the study area, and the base flow is determined based on the base flow segmentation results of measured runoff data at the outlet section of the study area. A hydrothermal coupling correction model corresponding to the plateau cold region and a functional relationship between base flow and runoff are constructed. The hydrothermal coupling correction model introduces the snowfall ratio, snowmelt amount and water storage change as correction parameters. The precipitation in the hydrothermal coupling correction model is the corrected effective precipitation. Based on the aforementioned hydrothermal coupling correction model and the aforementioned functional relationship, the elasticity coefficients of the base flow rate with respect to multiple influencing factors are determined. These influencing factors include precipitation, snowfall ratio, snowmelt, potential evapotranspiration, and changes in water storage. The baseflow time series is divided into a baseline period and a change period based on the abrupt change points of the baseflow time series. Based on the elasticity coefficient and the change of each influencing factor between the baseline period and the change period, the contribution degree and contribution rate of each influencing factor to the baseflow change are determined to output multi-timescale attribution identification.

2. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in claim 1, characterized in that, The process involves optimizing and calibrating the soil and moisture assessment tool model using measured cross-sectional runoff data from the study area to obtain a trained soil and moisture assessment tool model, including: The standardized measured cross-sectional runoff data are input into the soil and moisture assessment tool model to divide the hydrological response units of the study area and construct the database. Using historical measured cross-sectional runoff data from the outlet section of the study area, the key hydrological parameters in the soil and moisture assessment tool model were adjusted using a preset optimization algorithm. By comparing simulated runoff with measured runoff, the coefficient of determination and Nash efficiency coefficient of the soil and moisture assessment tool model are adjusted until the coefficient of determination and the Nash efficiency coefficient meet the preset accuracy threshold.

3. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in claim 1, characterized in that, The snowfall, snowmelt, and actual evapotranspiration in the study area at different time scales were simulated using the trained soil and moisture assessment tool model, including: Based on the difference between the daily maximum temperature and the critical temperature for snow melting, and combined with the snow cover rate, the daily snow melting equivalent is calculated. Based on the previous day's snow temperature, the current day's average air temperature, and the snow lag coefficient, the surface temperature of the snow cover for the current day is calculated recursively. Based on the date's position within the year's sequence, the snow melting coefficient for each day is dynamically determined between preset maximum and minimum snow melting coefficients. Based on the snowmelt equivalent, snow surface temperature and snowmelt coefficient, the trained soil and moisture assessment tool model simulates and outputs a continuous snowmelt sequence and snow cover state changes. Based on the snowmelt sequence and changes in snow cover, the snowfall, snowmelt, and actual evapotranspiration in the study area at different time scales were determined.

4. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in claim 1, characterized in that, The acquisition of the potential evapotranspiration and base flux of the study area at different time scales includes: Obtain data on net radiation and soil heat flux; Determine the saturated vapor pressure and the actual vapor pressure based on temperature data; Obtain the slope of the water vapor pressure curve and the constant of the wet / dry gauge; Based on wind speed data, the potential evapotranspiration of the study area at different time scales is determined by integrating the net radiation, soil heat flux, saturated vapor pressure and actual vapor pressure, the slope of the vapor pressure curve and the ecliptic constant.

5. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in claim 1, characterized in that, The acquisition of the potential evapotranspiration and base flux of the study area at different time scales includes: Obtain the maximum value of the baseflow index and the recession constant; The maximum value of the baseflow index, the drainage constant, and the baseflow rate of the previous time step are used as recursive inputs; The total runoff at the current time step is weighted and calculated with the recursive input to obtain the base runoff at the current time step; By recursively extrapolating the time series, the baseflow and surface runoff of the entire runoff sequence are separated to obtain the baseflow of the study area at different time scales.

6. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in claim 1, characterized in that, Determining the changes in snowfall ratio and water storage at different time scales includes: For each time scale, obtain the total precipitation, actual evapotranspiration and total runoff for that period. The total precipitation is subtracted from the actual evapotranspiration, and then the total runoff is subtracted to obtain the calculation result; The calculation results are taken as the net change in water storage in the basin during this period. Based on the net change in water storage in the basin during that period, the changes in water storage at different time scales are determined.

7. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in claim 1, characterized in that, The steps for obtaining effective precipitation data include: Calculate the amount of water in the non-snowfall portion of the total precipitation, which is the total precipitation multiplied by (1 minus the snowfall ratio). The effective precipitation is obtained based on the non-snowfall water volume, the concurrent snowmelt volume, and the concurrent water storage change. The effective precipitation is then obtained by adding the snowfall water volume to the concurrent snowmelt volume and subtracting the concurrent water storage change.

8. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in claim 1, characterized in that, Based on the abrupt change points in the base current time series, the base current time series is divided into a baseline period and a change period, including: For a basic current time series of length n, calculate the order statistic corresponding to each possible split point; Find the time point at which the absolute value of the order statistic reaches its maximum; Compare the absolute value of the maximum order statistic with the critical value at a given significance level to determine whether the time point is a statistically significant mutation point; Based on the aforementioned mutation points, the entire base current time series is divided into two stages: the baseline period and the change period.

9. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in any one of claims 1 to 8, characterized in that, After outputting multi-timescale attribution recognition, it also includes: Regularly access real-time snow cover observation data and short-term weather forecast data for the study area; Based on the real-time snow cover observation data and the short-term weather forecast data, the snowfall ratio and snowmelt parameters used for short-term forecasting are dynamically corrected. The corrected snowfall ratio and snowmelt parameters, along with precipitation and potential evapotranspiration data from short-term weather forecasts, are input into the hydrothermal coupling correction model, which outputs the baseflow prediction value for the near future. Based on various meteorological scenarios or parameter disturbances, baseflow prediction results are generated, and the uncertainty of the prediction is expressed in the form of prediction intervals.

10. The multi-scale attribution identification method for baseflow changes in rivers in high-altitude cold regions as described in claim 9, characterized in that, The dynamic correction of the snowfall ratio and snowmelt parameters used for short-term forecasting based on the real-time snow cover observation data and the short-term weather forecast data includes: Based on the real-time snow cover observation data, the current initial values ​​of snowfall ratio and snowmelt amount are estimated in real time. Different weights are assigned to the parameter values ​​obtained statistically from the soil and moisture assessment tool model and the current initial values, and then the values ​​are fused to obtain the corrected snowfall ratio and snowmelt amount parameter values.

Images