A time-varying glacier degree-day factor calculation method suitable for plateau areas
By screening key driving factors from multi-source data and establishing a dynamic relationship model to calculate the glacier diurnal factor in plateau regions, the problem of low accuracy in glacier ablation simulation in plateau regions has been solved, enabling more efficient prediction of glacier ablation and runoff, and improving the scientific nature of water resource management and disaster early warning.
Patent Information
- Application Number
- CN202610538761.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-22
- Publication Date
- 2026-07-24
AI Technical Summary
Existing methods for determining the diurnal factor of glaciers in plateau regions suffer from insufficient spatial representation capabilities, inadequate representation of multi-physical process driving mechanisms, and a contradiction between data feasibility and model accuracy. This results in low accuracy of glacier ablation simulations and an inability to adapt to the complex terrain and climate changes in plateau regions.
By using multi-source remote sensing and reanalysis data, key driving factors were screened, and a dynamic relationship model between glacier day-to-day factor and net radiation, temperature, wind speed, etc. was established. The time-varying glacier day-to-day factor was calculated using linear or multiple regression models, and the albedo and solar radiation were combined to adapt to different underlying surface types.
It significantly improves the accuracy of glacier ablation simulation, reduces data acquisition costs, enhances the applicability and robustness of the model, and provides more accurate data support for water resource management and disaster early warning.
Smart Images

Figure CN122451630A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of cryosphere hydrological science and technology, and in particular relates to a method for calculating the time-varying glacier diurnal factor applicable to plateau regions. Background Technology
[0002] Glacier mass balance is a core indicator reflecting glacier changes and is of great significance for studying regional water resources, sea-level rise, and climate change. The degree-day factor (DDF) is a key parameter in glacier mass balance simulations, defined as the temperature at which a positive degree-day (PDD) is reached. The amount of glacial ice or snow melted during a given day is typically measured in millimeters of water equivalent per degree Celsius per day. Currently, methods for determining the glacial day-factor are mainly categorized as follows:
[0003] First, the empirical fixed value method: This method directly adopts empirical values reported in global or specific regional literature. However, the day factor is not a physical constant; it essentially reflects the contribution of various components of the glacier surface energy balance (such as net radiation, sensible heat flux, etc.) to melting. Therefore, this method completely ignores its spatiotemporal heterogeneity and has low accuracy.
[0004] Second, the direct calculation method based on mass balance observations: This method uses field-measured glacier mass balance data and synchronously observed temperature data to invert the degree-day factor. This method yields relatively accurate results, but it is highly dependent on expensive, arduous, and time-consuming field measurements, making it difficult to obtain large-scale, continuous time-series DDF values, resulting in poor reproducibility and scalability.
[0005] Third, parameterization methods based on energy balance theory: These methods estimate parameters by establishing statistical relationships between degree-day factors and readily available meteorological elements. While this method is an improvement over the fixed-value method, most of these parameterization schemes are based on data from specific glaciers or specific time periods, lacking broad applicability. When applied to glaciers of different sizes and in different watersheds, the estimation results exhibit significant uncertainty due to the vast differences in underlying surface characteristics and local climate conditions.
[0006] For high-altitude regions, the limitations of using the above methods to determine the diurnal factor of glaciers are even more pronounced:
[0007] (1) Insufficient spatial representation and simplification of mechanisms: The plateau has complex topography and dramatic spatial differentiation of climate and glacier types, with huge differences in DDF values from continental to maritime glaciers. Existing methods mostly use spatially fixed empirical values or local parameterization schemes based on a single meteorological variable (such as temperature). This highly simplified model fails to fully consider the spatial coordinated variation of multiple key physical factors (such as net radiation, precipitation, wind speed, etc.) that dominate glacier energy-mass exchange, resulting in DDF values that cannot truly reflect the dramatic environmental gradient within the plateau, poor spatial adaptability, and the introduction of systematic bias.
[0008] (2) Insufficient characterization of multi-physical process driving mechanisms: Glacier melting is driven by multiple components of energy balance. Existing methods, especially traditional single-factor daily models, have failed to effectively isolate and quantify the dynamic impact of net radiation, the most important direct energy source, on DDF. At the same time, they generally neglect the synergistic or antagonistic effects of factors such as precipitation phase (rain / snow), wind speed (affecting turbulent exchange), and humidity (affecting latent heat) on DDF by changing surface albedo and energy distribution. In plateau regions, factors such as net radiation are greatly affected by factors such as clouds and ice surface pollutants, resulting in significant spatiotemporal variations. Ignoring this synergistic physical driving force of multiple factors is the core of the model mechanism defects and accuracy bottlenecks.
[0009] (3) The contradiction between data feasibility and model accuracy is prominent: high-precision DDF calibration is extremely dependent on scarce and discrete field observations, which makes it difficult to support regional scale applications; while large-scale methods often result in low accuracy due to oversimplification of physical processes.
[0010] Therefore, there is currently a lack of a method for calculating the degree factor that can effectively integrate readily available multi-source remote sensing and reanalysis data, and on this basis construct a dynamic estimation framework for DDF with clear physical interpretation, so as to achieve the optimal balance between data feasibility and simulation accuracy. Summary of the Invention
[0011] The purpose of this invention is to provide a method for calculating the time-varying glacial day-degree factor applicable to plateau regions, so as to solve the above-mentioned technical problems.
[0012] To achieve the above objectives, the present invention provides the following technical solution:
[0013] This invention discloses a method for calculating the diurnal temperature variation factor of glaciers applicable to plateau regions, the method comprising the following steps:
[0014] Step 1: Multi-source data collection and processing: Obtain the measured glacier day-degree factor dataset of multiple glaciers in the study area during the study period. Each measured data point includes its geographic coordinates, observation period, and day-degree factor value. For the spatiotemporal location of each measured data point, simultaneously extract radiation factor, thermal factor, dynamic factor, precipitation factor, and surface attribute factor.
[0015] The radiation factors include downward shortwave radiation and downward longwave radiation; the thermal factors include air temperature; the dynamic factors include wind speed, specific humidity, and air pressure; the precipitation factors include precipitation amount; and the surface property factors include albedo.
[0016] Based on the measured data points and the downward shortwave radiation, downward longwave radiation, and albedo during the observation period, the net radiation was calculated as follows:
[0017] (1)
[0018] In the formula: R n Net radiation; S ↓ and L ↓ These represent downward shortwave radiation and downward longwave radiation, respectively; α is the albedo; ε is the emissivity of the glacier surface, taken as 0.98; σ is the Stefan-Boltzmann constant, with a value of [value missing]. ;T s This represents the absolute temperature of the glacier surface, taken as 0°C, or 273.15 K.
[0019] A dataset of meteorological driving factors consisting of net radiation, air temperature, wind speed, specific humidity, air pressure, and precipitation;
[0020] Step 2, Key Driving Factor Screening: Key driving factors are screened from the meteorological driving factor dataset. The specific process includes:
[0021] 1) Initial screening of correlation factors: Pearson correlation coefficients were calculated between net radiation, air temperature, wind speed, specific humidity, air pressure, precipitation and measured glacier day-time factors, respectively. Factors with a significance level of p<0.01 in the t-test were initially retained and denoted as the set. ;
[0022] 2) Selection of key driving factors: If the set Number of factors If the set contains factors that are key driving factors, then all factors in the set will be considered key driving factors; if the set contains factors that are key driving factors, then all factors in the set will be key driving factors. Number of factors Then for the set Principal component analysis was performed on the factors to identify key driving factors;
[0023] Step 3: Establishing a dynamic relationship model: Based on the key driving factors determined in Step 2, establish a statistical relationship model between them and the measured degree-day factors; with the measured degree-day factors as the dependent variable and the key driving factors as the independent variables, draw a scatter plot between the two, observe the function shape between the variables, and adaptively select the model form according to the number of driving factors and the relationship characteristics presented by the scatter plot.
[0024] Step 4: Calculation of time-varying glacier day-time factor: Based on the dynamic relationship model established in Step 3, the time-varying glacier day-time factor of the study area is calculated. That is, the combination data of key driving factors in the long-term series during the study period is used as input, and the time-varying glacier day-time factor of the study area is calculated according to the selected model form and fitting parameters.
[0025] Furthermore, the set described in step 2 The specific process of performing principal component analysis on the factors to determine the key driving factors includes:
[0026] (1) Data standardization: First, Z-score standardization is performed on each factor to obtain a standardized data matrix. , Given the sample size, the standardization formula is:
[0027] (2)
[0028] In the formula, Let represent the original observation value of the j-th driving factor on the i-th sample, i = 1, 2, …, n, j = 1, 2, …, m; This represents the standardized value of the j-th driving factor on the i-th sample after Z-score standardization. , The first Sample mean and standard deviation of each factor;
[0029] (2) Calculate the correlation coefficient matrix: the standardized data matrix Since the matrix has been centered (mean of each column is 0), its covariance matrix is the correlation coefficient matrix.
[0030] (3)
[0031] In the formula, for The transpose of the matrix; The covariance matrix of the standardized data is the correlation coefficient matrix, with diagonal elements of 1 and off-diagonal elements representing the Pearson correlation coefficients between each pair of factors.
[0032] (3) Eigenvalue decomposition: Solving for the correlation coefficient matrix Characteristic equation:
[0033] (4)
[0034] In the formula, for The identity matrix;
[0035] Solving the characteristic equation yields... nonnegative eigenvalues and the corresponding unit eigenvector Each feature vector Indicates the first Principal components and original The linear combination coefficients among the factors;
[0036] (4) Determine the number of principal components: Select according to the preset threshold. Principal components, The threshold setting is based on the Kaiser criterion: eigenvalues greater than 1 are retained. Principal components;
[0037] (5) Identify key driving factors: Before analysis The loading of each principal component is the product of the eigenvector and the square root of the eigenvalue. For each principal component, original factors with absolute loading values greater than a set threshold are selected as representative factors. The final set of retained original factors is the key driving factor. Furthermore, step 3, which describes adaptively selecting the model form based on the number of driving factors and the relationship characteristics presented by the scatter plot, specifically includes:
[0038] 1) If it is a single driving factor: select a linear model or a quadratic polynomial model, and choose the optimal form by comparing the deterministic coefficient R² and the root mean square error RMSE of the two.
[0039] Among them, the coefficient of determination R 2 The calculation formula is:
[0040] (5)
[0041] In the formula, For the first One true value; It is the first One predicted value; This represents the average of the true values; The number of samples; The closer the value is to 1, the higher the goodness of fit of the model;
[0042] The formula for calculating the root mean square error (RMSE) is:
[0043] (6)
[0044] The smaller the value, the smaller the prediction error of the model and the higher the accuracy;
[0045] 2) If there are multiple driving factors (≥2): then a multiple linear regression model is used.
[0046] The beneficial effects of the present invention are as follows:
[0047] 1) Significantly improves the accuracy of glacier ablation simulation and prediction: This invention establishes a physical relationship between the day-of-day factor and net radiation by coupling albedo and solar radiation, enabling dynamic calculation of the glacier day-of-day factor. It can accurately capture the significant differences in the day-of-day factor at the beginning, peak, and end of the ablation season due to changes in ice surface albedo, impurity content, and weather conditions (such as cloud cover), thereby greatly reducing errors in glacier mass balance and runoff simulation and improving simulation and prediction accuracy.
[0048] 2) Effectively solves the applicability problem in areas with complex underlying surfaces: By introducing key spatiotemporal variable parameters (albedo, solar radiation), this invention enables the model to adapt to different underlying surface types and their dynamic transformation processes, exhibiting good applicability and robustness.
[0049] 3) Reduced data acquisition threshold and cost for large-scale applications: This invention uses remote sensing data and reanalysis meteorological data as the main input drivers, avoiding the need to deploy expensive and difficult-to-maintain automatic weather stations on a large scale in the vast and treacherous plateau glacier areas. This makes it possible to calculate glacier day-degree factor and monitor ablation at the regional scale and over a long period of time, greatly improving the engineering application value and operational potential of the method.
[0050] 4) Provides more accurate data support for water resource management and disaster early warning: The time-varying glacier day factor calculation method provided by this invention can obtain more reliable glacier ablation and runoff prediction results, helping relevant departments to formulate more scientific water resource management strategies and disaster risk response measures, with significant social and economic benefits.
[0051] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description
[0052] Figure 1 This is a flowchart of the method described in this invention;
[0053] Figure 2 This is a graph showing the relationship between the glacier diurnal factor and net radiation, fitted using a linear model in Example 1.
[0054] Figure 3 The graph shows the relationship between the glacier diurnal factor and net radiation, which was fitted using a quadratic polynomial model in Example 1. Detailed Implementation
[0055] This invention discloses a method for calculating the diurnal temperature variation factor of time-varying glaciers applicable to plateau regions, such as... Figure 1 As shown, the method includes the following steps:
[0056] Step 1: Multi-source data collection and processing: Obtain a dataset of measured glacier day-degree factors for multiple glaciers within the study area during the study period (measured data originates from discontinuous field observation periods, dating back to 1940). Each measured data point should include its geographic coordinates, observation period, and day-degree factor value. For each measured data point's spatiotemporal location, simultaneously extract data on radiation factors, thermal factors, dynamic factors, precipitation factors, and surface attribute factors.
[0057] Among them, radiation factors include downward shortwave radiation and downward longwave radiation provided by the latest generation reanalysis dataset (ERA5) of the European Centre for Medium-Range Weather Forecasts or the China Regional High Spatiotemporal Resolution Surface Meteorological Driven Dataset (CMFD).
[0058] The thermal factor includes the air temperature provided by ERA5 or CMFD.
[0059] Dynamic factors include wind speed, specific humidity, and air pressure provided by ERA5 or CMFD.
[0060] Precipitation factors include: precipitation provided by ERA5 or CMFD.
[0061] Surface property factors include albedo provided by LandSat or MODIS.
[0062] Based on the measured data of downward shortwave radiation, downward longwave radiation, and albedo during the observation period, the net radiation is calculated as follows:
[0063] (1)
[0064] In the formula, R n Net radiation; S ↓ and L ↓ These represent downward shortwave radiation and downward longwave radiation, respectively; α is the albedo; ε is the emissivity of the glacier surface, taken as 0.98; σ is the Stefan-Boltzmann constant, with a value of [value missing]. ;T s The absolute temperature of the glacier surface (unit: Kelvin, K) is taken as 0°C (273.15 K).
[0065] The meteorological driving factor dataset consists of net radiation, air temperature, wind speed, specific humidity, air pressure, and precipitation.
[0066] Step 2, Key Driver Factor Screening: Identify the main driving factors, i.e., key driving factors, from the meteorological driving factor dataset that can effectively explain the variation of glacier day-degree factor. The specific process includes:
[0067] 1) Initial screening of correlation factors: Pearson correlation coefficients were calculated between net radiation, air temperature, wind speed, specific humidity, air pressure, precipitation and measured glacier day-temperature factors, respectively. Factors with high significance levels in t-tests (p<0.01) were initially retained and denoted as the set. .
[0068] 2) Selection of key driving factors: If the set Number of factors If all factors in the set are considered key driving factors, proceed directly to the next step; if Then, it is necessary to make a selection and determine the key driving factors.
[0069] If the number of initial screening factors exceeds 3 ( To avoid multicollinearity and further condense information, the set... Principal component analysis is performed on the factors to identify key driving factors. The specific process includes:
[0070] (1) Data standardization: First, Z-score standardization is performed on each factor to obtain a standardized data matrix. ( (where the sample size is 1), the standardization formula is:
[0071] (2)
[0072] In the formula, Let represent the original observation value of the j-th driving factor on the i-th sample, where i = 1, 2, …, n (sample number) and j = 1, 2, …, m (factor number). This represents the standardized value of the j-th driving factor on the i-th sample after Z-score standardization. , The first The sample mean and standard deviation of each factor.
[0073] (2) Calculate the correlation coefficient matrix: the standardized data matrix Since the matrix has been centered (mean of each column is 0), its covariance matrix is the correlation coefficient matrix:
[0074] (3)
[0075] In the formula, for The transpose matrix; due to Each column has been centered (mean is 0), therefore This is the covariance matrix of the standardized data, also known as the correlation coefficient matrix. Its diagonal elements are 1, and the off-diagonal elements are the Pearson correlation coefficients between each pair of factors.
[0076] (3) Eigenvalue decomposition: Solving for the correlation coefficient matrix Characteristic equation:
[0077] (4)
[0078] In the formula, for The identity matrix.
[0079] Solving the characteristic equation yields... nonnegative eigenvalues and the corresponding unit eigenvector Each feature vector Indicates the first Principal components and original The linear combination coefficients between the factors.
[0080] (4) Determine the number of principal components: Based on the preset threshold (according to the Kaiser criterion: eigenvalues greater than 1 are retained). Principal components) selection Principal components ( ).
[0081] (5) Identify key driving factors: Before analysis The loadings of each principal component (i.e., the product of the eigenvector and the square root of the eigenvalue) are calculated. For each principal component, original factors with absolute loading values greater than a set threshold (e.g., loading > 0.7) are selected as representative factors. The final set of retained original factors is the key driving factor. Step 3: Establishing a dynamic relationship model: Based on the key driving factors determined in Step 2, a statistical relationship model between them and the measured daily factors is established.
[0082] Regarding the choice of model form, using the measured daily degree factor obtained in step 1 as the dependent variable and the key driving factors determined in step 2 as the independent variables, a scatter plot is drawn between the two to observe the functional shape (e.g., linear, parabolic, etc.) of the variables. Based on the number of driving factors (one or more) and the relationship characteristics presented by the scatter plot, the model form is adaptively selected:
[0083] 1) If there is a single driving factor: a linear model or a quadratic polynomial model can be used. The optimal form is selected by comparing the coefficient of determination R² and the root mean square error RMSE of the two models.
[0084] Among them, the coefficient of determination R 2 Used to measure how well a regression model fits the observed data, it represents the proportion of variance explained by the model to the total variance, and is calculated using the following formula:
[0085] (5)
[0086] In the formula, For the first A true value, It is the first One predicted value, This represents the average of the true values. This represents the number of samples. The value of is usually between 0 and 1 (sometimes it may be negative, indicating that the model's prediction effect is extremely poor). The closer the value is to 1, the higher the goodness of fit of the model.
[0087] The root mean square error (RMSE) measures the deviation between predicted and actual values, reflecting the accuracy of the model's predictions. Its calculation formula is as follows:
[0088] (6)
[0089] The smaller the value, the smaller the prediction error of the model and the higher the accuracy.
[0090] 2) If there are multiple driving factors (≥2): then a multiple linear regression model is used.
[0091] Step 4: Calculation of Time-Varying Glacier Daily Degree Factor: Based on the dynamic relationship model established in Step 3, the time-varying glacier daily degree factor for the study area is calculated. This involves using long-term series data of key driving factors (from ERA5, CMFD, MODIS, or Landsat data products; the data period can be determined as needed, such as daily series data of ERA5 driving factors from January 1, 1940 to December 31, 2020) as input. Based on the selected model form (linear / quadratic polynomial / multivariate linear) and fitting parameters, the time-varying glacier daily degree factor is calculated pixel-by-pixel and time-by-time.
[0092] Example 1
[0093] This embodiment is an application example of the above method.
[0094] This embodiment takes the western glacier region of China (mainly the Qinghai-Tibet Plateau) as the study area, collects the measured geographic coordinates, observation time period, and degree-day factor value of glaciers in the study area, and extracts the downward longwave radiation, downward shortwave radiation, air temperature, wind speed, specific humidity, air pressure, precipitation, and LandSat albedo data for the observation time period corresponding to the CMFD grid where the measured glacier point is located. Since the LandSat albedo data started in 1984, for the Qiongtanlan Glacier, Cherkanbulak Glacier, Halong Glacier, Hailuogou Glacier, Dagongba Glacier, Xiaogongba Glacier, and Baishuihe No. 1 Glacier whose observation period of the glacier temperature factor value was earlier than 1984, the daily cumulative temperature value T1 within the observation period was obtained by combining CMFD temperature data, and the corresponding daily cumulative temperature value T2 for the observation period from 1984 to 2000 was found. The year with the smallest relative error between T1 and T2 was determined, and the LandSat albedo within the corresponding period of that year was used as the albedo of Qiongtanlan Glacier, Cherkanbulak Glacier, Halong Glacier, Hailuogou Glacier, Dagongba Glacier, Xiaogongba Glacier, and Baishuihe No. 1 Glacier. Combining the downward longwave radiation, downward shortwave radiation, and albedo data, the net radiation was calculated using formula (1), and the results are shown in Table 1.
[0095] Table 1. Comparison of Measured Data on Glacier Diurnal Factor and Net Radiation
[0096]
[0097] The radiation values in Table 1 are obtained from net radiation through unit conversion (1 W / m²). 2 =1J / (m 2 ·s)=1×24×60×60 J / m 2 = 1 × 24 × 60 × 60 × 10 -6 MJ / m 2 =0.0864 MJ / m 2 )).
[0098] By calculating the Pearson correlation coefficients between the glacier day-live factor and net radiation, precipitation, air pressure, air temperature, wind speed, and specific humidity, as shown in Table 2, it was found that the correlation coefficients between net radiation, precipitation, specific humidity, and the glacier day-live factor were 0.89, 0.73, and 0.74, respectively. Among them, the correlation between net radiation and the glacier day-live factor passed the 0.01 level significance t-test (p < 0.01), that is, net radiation was initially screened as a correlation factor, and net radiation was regarded as the key driving factor.
[0099] Table 2. Pearson correlation coefficients of various meteorological factors
[0100]
[0101] Linear and quadratic polynomial models were used to fit the relationship between the glacier diurnal factor and net radiation, respectively, and the results are as follows: Figure 2 , Figure 3 As shown in the figure, the red lines represent the results calculated using a linear model and a quadratic polynomial model, respectively. The relationships between the two models are: y = 1.2028x - 30.5039 and y = -0.1268x. 2 +9.6750 -170.7328, the blue hollow circles represent the measured data.
[0102] The coefficients of determination and root mean square errors (RMSEs) of the fitting results for the two models were calculated to evaluate the accuracy of the fitting results, as shown in Table 3. The results show that the quadratic polynomial model has a higher coefficient of determination than the linear model, and a smaller RMS error, indicating better performance. Therefore, the quadratic polynomial model (y = -0.1268x) is selected. 2 (+9.6750-170.7328), calculating the time-varying diurnal factor of glaciers in western China pixel by pixel and time period by time period.
[0103] Table 3. Accuracy of Fitting Results for the Two Models
[0104]
[0105] Finally, it should be noted that the above description is only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred arrangement, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.
Claims
1. A method for calculating the diurnal temperature variation factor of time-varying glaciers applicable to plateau regions, characterized in that, The method includes the following steps: Step 1: Multi-source data collection and processing: Obtain the measured glacier day-degree factor dataset of multiple glaciers in the study area during the study period. Each measured data point includes its geographic coordinates, observation period, and day-degree factor value. For the spatiotemporal location of each measured data point, simultaneously extract radiation factor, thermal factor, dynamic factor, precipitation factor, and surface attribute factor. The radiation factors include downward shortwave radiation and downward longwave radiation; the thermal factors include air temperature; the dynamic factors include wind speed, specific humidity, and air pressure; the precipitation factors include precipitation amount; and the surface property factors include albedo. Based on the measured data points and the downward shortwave radiation, downward longwave radiation, and albedo during the observation period, the net radiation was calculated as follows: (1) In the formula: R n Net radiation; S ↓ and L ↓ These represent downward shortwave radiation and downward longwave radiation, respectively; α is the albedo; ε is the emissivity of the glacier surface, taken as 0.98; σ is the Stefan-Boltzmann constant, with a value of [value missing]. ;T s This represents the absolute temperature of the glacier surface, taken as 0°C, or 273.15 K. A dataset of meteorological driving factors consisting of net radiation, air temperature, wind speed, specific humidity, air pressure, and precipitation; Step 2, Key Driving Factor Screening: Key driving factors are screened from the meteorological driving factor dataset. The specific process includes: 1) Initial screening of correlation factors: Pearson correlation coefficients were calculated between net radiation, air temperature, wind speed, specific humidity, air pressure, precipitation and measured glacier day-temperature factors, respectively. Factors with a significance level of p<0.01 in the t-test were initially retained and denoted as the set. ; 2) Selection of key driving factors: If the set Number of factors If the set contains factors that are key driving factors, then all factors in the set will be considered key driving factors; if the set contains factors that are key driving factors, then all factors in the set will be key driving factors. Number of factors Then for the set Principal component analysis was performed on the factors to identify key driving factors; Step 3: Establishing a dynamic relationship model: Based on the key driving factors determined in Step 2, establish a statistical relationship model between them and the measured degree-day factors; with the measured degree-day factors as the dependent variable and the key driving factors as the independent variables, draw a scatter plot between the two, observe the function shape between the variables, and adaptively select the model form according to the number of driving factors and the relationship characteristics presented by the scatter plot. Step 4: Calculation of time-varying glacier day-time factor: Based on the dynamic relationship model established in Step 3, the time-varying glacier day-time factor of the study area is calculated. That is, the combination data of key driving factors in the long-term series during the study period is used as input, and the time-varying glacier day-time factor of the study area is calculated according to the selected model form and fitting parameters.
2. The method for calculating the diurnal temperature variation factor of time-varying glaciers applicable to plateau regions as described in claim 1, characterized in that, The set described in step 2 The specific process of performing principal component analysis on the factors to determine the key driving factors includes: (1) Data standardization: First, Z-score standardization is performed on each factor to obtain a standardized data matrix. , Given the sample size, the standardization formula is: (2) In the formula, Let represent the original observation value of the j-th driving factor on the i-th sample, i = 1, 2, …, n, j = 1,2, …, m; This represents the standardized value of the j-th driving factor on the i-th sample after Z-score standardization. , The first Sample mean and standard deviation of each factor; (2) Calculate the correlation coefficient matrix: the standardized data matrix Since the matrix has been centered (mean of each column is 0), its covariance matrix is the correlation coefficient matrix. (3) In the formula, for The transpose of the matrix; The covariance matrix of the standardized data is the correlation coefficient matrix, with diagonal elements of 1 and off-diagonal elements representing the Pearson correlation coefficients between each pair of factors. (3) Eigenvalue decomposition: Solving for the correlation coefficient matrix Characteristic equation: (4) In the formula, for The identity matrix; Solving the characteristic equation yields... nonnegative eigenvalues and the corresponding unit eigenvector Each feature vector Indicates the first Principal components and original The linear combination coefficients among the factors; (4) Determine the number of principal components: Select according to the preset threshold. Principal components, The threshold setting is based on the Kaiser criterion: eigenvalues greater than 1 are retained. Principal components; (5) Identify key driving factors: Before analysis The loading of each principal component is the product of the eigenvector and the square root of the eigenvalue. For each principal component, the original factors whose absolute loading value is greater than a set threshold are selected as representative factors. The final set of original factors is the key driving factor.
3. The method for calculating the diurnal temperature variation factor of time-varying glaciers applicable to plateau regions as described in claim 1, characterized in that, Step 3, which describes adaptively selecting the model form based on the number of driving factors and the relationship characteristics presented in the scatter plot, specifically includes: 1) If it is a single driving factor: select a linear model or a quadratic polynomial model, and choose the optimal form by comparing the deterministic coefficient R² and the root mean square error RMSE of the two. Among them, the coefficient of determination R 2 The calculation formula is: (5) In the formula, For the first One true value; It is the first One predicted value; This represents the average of the true values; The number of samples; The closer the value is to 1, the higher the goodness of fit of the model; The formula for calculating the root mean square error (RMSE) is: (6) The smaller the value, the smaller the prediction error of the model and the higher the accuracy; 2) If there are multiple driving factors (≥2): then a multiple linear regression model is used.
