A method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion

Abstract

This paper provides a method for constructing a snowmelt dataset for the Yangtze River source region based on multi-source data fusion. The method involves collecting snowwater equivalent data, snowmelt data, ground station data, potential evapotranspiration data, and hydrological station flow data for the study area; spatially matching ground stations and calculating snowmelt data; fusing the spatiotemporal characteristics of multi-source snowmelt data; evaluating the effectiveness of snowmelt data fusion; calculating site snowmelt using a temperature coefficient algorithm, and utilizing deep learning and other methods to achieve efficient fusion of spatiotemporal snowmelt information. This method improves the spatiotemporal resolution of snowmelt data, more accurately capturing the spatial distribution and temporal dynamics of snowmelt in the Yangtze River source region, which is of great significance for dry season runoff prediction, flood warning, and ecological protection.

Description

Technical Field

[0001] The present invention relates to the field of hydrological and meteorological data applications, and in particular to a method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion. Background Art

[0002] Snowmelt is a crucial source of spring water resources in many regions and is crucial for water resource management. Snowmelt in the Yangtze River source region influences runoff in the Jinsha River basin, significantly impacting hydropower generation and ecosystems. While snowmelt monitoring is crucial, its complexity and variability mean limited observational tools and challenges remain, particularly in the Yangtze River source region at high altitudes or with complex terrain.

[0003] Currently, mainstream snow cover data include satellite observation data, ground-based observation data, and reanalysis data. Satellite observation data provides wide spatial coverage and can monitor the distribution and changes of snow cover in real time, but is greatly affected by weather conditions, and cloud cover can lead to missing observation data. Data from ground-based observation stations provide accurate snow depth data, but no snowmelt data. Reanalysis data is generated by combining numerical weather forecast models with multiple observation data, providing long-term series of snow cover information, but its accuracy depends on the accuracy of the model and the quality of the input data. In short, existing snow cover data cannot accurately reflect the snow melt process. Satellite observation and reanalysis data each have limitations and cannot provide comprehensive and accurate snowmelt information alone. Therefore, a more comprehensive approach is urgently needed to improve the accuracy of snowmelt data in the source region of the Yangtze River. Summary of the Invention

[0004] The purpose of the present invention is to address the deficiencies of the above-mentioned existing technologies and provide a method for constructing a snowmelt dataset in the source area of ​​the Yangtze River based on multi-source data fusion, effectively fuse data of different temporal and spatial resolutions, extract effective information from multi-source data through deep learning methods, and improve the accuracy of snowmelt data.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] The present invention provides a method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion, comprising the following steps:

[0007] S1. Collect data in the study area: collect snow water equivalent data, snowmelt data, ground station data, potential evapotranspiration data and hydrological station flow data;

[0008] The snow water equivalent data is daily snow water equivalent product data;

[0009] The snowmelt data is the reanalysis of daily snowmelt data;

[0010] The ground station data are daily-scale measured snow depth data, temperature data, and snowfall data;

[0011] The potential evapotranspiration data is a monthly potential evapotranspiration dataset at 1 km in the source area of ​​the Yangtze River.

[0012] The flow data of the hydrological stations are measured flow data on a daily basis;

[0013] S2. Data pre-processing: This includes ground station spatial matching and snowmelt data calculation. Ground station spatial matching involves matching uniform resolution snow cover data and potential evapotranspiration data with ground station data using geographic coordinates to generate sample data sequences and time series sample data.

[0014] The snowmelt data is calculated by using satellite remote sensing snow water equivalent data and ground station data to calculate the satellite remote sensing snowmelt sequence and the station snowmelt sequence respectively;

[0015] S3. Spatiotemporal feature fusion of multi-source snowmelt data: including spatial feature fusion of snowmelt data and time series feature fusion of snowmelt data;

[0016] The spatial feature fusion of snowmelt data refers to fusing the ground-based and satellite remote sensing snowmelt datasets in S2 using a two-layer long-short-term memory network model and then restoring them using activation functions and an anti-normalization method to form fused monthly snowmelt data.

[0017] The snowmelt data time series feature fusion refers to further fusing the reanalyzed daily snowmelt data in S1 with the fused monthly snowmelt data in S2 using an adaptive instance normalization method to ultimately obtain daily snowmelt data;

[0018] S4. Evaluation of snowmelt data fusion effect: Use snowmelt observation alternative data and accuracy assessment indicators to evaluate the accuracy of the fused monthly snowmelt data.

[0019] Furthermore, in S2, the spatial matching of ground stations is the daily snow water equivalent product data. Taking the grid center point as the reference, the grid value closest to the longitude and latitude of the meteorological station is selected as the snow water equivalent data corresponding to the meteorological station, which is used to calculate the ground station snowmelt data based on remote sensing data.

[0020] Furthermore, in S2, the extraction of the time series sample data is specifically as follows: based on the time resolution of the reanalyzed daily snowmelt data, extracting the grid daily snowmelt time series sample data closest to the longitude and latitude of the meteorological station;

[0021] If a month has 31 days, it forms a 31×1 time series.

[0022] Furthermore, the snowmelt data in S2 is calculated as follows:

[0023] The monthly snowmelt data are calculated using a temperature index model based on the relationship between snowmelt and temperature. Monthly snowmelt data based on ground station data and remote sensing data are generated. The specific formula is:

[0024] ;

[0025] in, For the Monthly site snowmelt data; is the temperature coefficient, in millimeters per degree Celsius per day, which is an empirical coefficient relating snowmelt rate to air temperature; is the cumulative positive air temperature in degrees Celsius; For the The snow cover data of the ground stations in July can use the daily snow water equivalent data, which is the total amount of snowmelt;

[0026] When using snowfall data from ground stations, it is necessary to substitute Ground station snow cover data for the month :

[0027] ;

[0028] in, is the snowfall at the ground site, in millimeters; is the snow sublimation at the ground site, in millimeters;

[0029] The formula for snow sublimation at ground stations is:

[0030] ;

[0031] in, To obtain the minimum function, the smaller value between the snowmelt amount calculated by the function and the total amount currently used for snowmelt is taken as the final snowmelt amount; is the total amount of snow used for sublimation; The potential evapotranspiration is generated by matching the monthly potential evapotranspiration dataset of the Yangtze River source area at 1 km with the ground station data through the geographic point coordinates.

[0032] Furthermore, in S3, the specific process of using a two-layer long short-term memory network model to fuse ground station and remote sensing snowmelt sample data is as follows: first, the monthly snowmelt sequence samples based on the station data are input into the memory neural unit of the long short-term memory network model with multiple hidden layers for cyclic connection; the output feature values ​​and the monthly snowmelt sequence samples based on remote sensing data are input into the second-layer long short-term memory network model; finally, the activation function and denormalization are used to output the fused feature variables as the fused monthly snowmelt data. The specific formula is:

[0033] ;

[0034] in, For the Site, Monthly scale snowmelt data after monthly fusion; For the Ground stations, Monthly snowmelt eigenvector; is the standard deviation function; is the arithmetic mean function.

[0035] Furthermore, in S3, the fusion of snowmelt data time series features is specifically as follows: the daily snowmelt data is reanalyzed using the adaptive normalization method and further integrated with the monthly snowmelt data after fusion using the long short-term memory network model, and finally the daily snowmelt data is obtained. The specific formula is:

[0036] ;

[0037] in, For the Site, Month daily snowmelt data; For the Site, Month Snowmelt reanalysis data for the day.

[0038] Furthermore, using the Pearson correlation coefficient Conduct reliability assessment on the fused daily snowmelt data;

[0039] Due to the lack of snowmelt observation data, alternative data will be used to assist in proving the feasibility of snowmelt data. Therefore, the snowmelt observation alternative products in S4 include ground station snow depth change data and hydrological station measured flow data; the ground station snow depth change is obtained by subtracting the snow depth of the day from the snow depth of the previous day as the previous day's snowmelt observation alternative data; the hydrological station measured flow data is the dry season flow data of the nearest hydrological station as the dry season snowmelt observation alternative data.

[0040] The beneficial effects of this invention include: This dataset construction method effectively utilizes remote sensing data, reanalysis data, and site data, calculates site snowmelt using a temperature coefficient algorithm, and utilizes deep learning and other methods to achieve efficient integration of snowmelt spatiotemporal information. This overcomes the drawbacks of a single data source, further improves the spatiotemporal resolution of snowmelt data, and more accurately captures the spatial distribution and temporal dynamics of snowmelt in the Yangtze River source region, which is of great significance for dry season runoff prediction, flood warning, and ecological protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flowchart of a method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion;

[0042] Figure 2 is a flow chart of an embodiment;

[0043] Figure 3 It is the daily snowmelt data of a certain year and month in the study area. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0045] See also Figure 1 and Figure 2 A method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion includes the following steps:

[0046] S1. Collect data in the study area: collect snow water equivalent data, snowmelt data, ground station data, potential evapotranspiration data and hydrological station flow data;

[0047] The snow water equivalent data is daily snow water equivalent product data with a spatial resolution of 0.25°;

[0048] Among them, the daily snow water equivalent product data is the satellite remote sensing snow water equivalent product data developed by the Japan Aerospace Exploration Agency;

[0049] The snowmelt data are reanalyzed daily snowmelt data with a spatial resolution of 0.25°;

[0050] Snowmelt data are daily snowmelt data analyzed from the fifth-generation full-climate reanalysis product provided by the European Centre for Medium-Range Weather Forecasts.

[0051] The ground station data are daily-scale measured snow depth data, temperature data, and snowfall data;

[0052] The potential evapotranspiration data is a monthly potential evapotranspiration dataset at 1 km in the source area of ​​the Yangtze River.

[0053] The flow data at the hydrological stations are measured flow data on a daily scale;

[0054] S2. Data pre-processing:

[0055] Including ground station spatial matching and snowmelt data calculation; the ground station spatial matching is to match the snow cover data and potential evapotranspiration data of uniform resolution with the ground station data through geographic point coordinates to generate sample data series and time series sample data;

[0056] The snowmelt data is calculated by using satellite remote sensing snow water equivalent data and ground station data to calculate the satellite remote sensing snowmelt sequence and the station snowmelt sequence respectively;

[0057] S3. Spatiotemporal feature fusion of multi-source snowmelt data: including spatial feature fusion of snowmelt data and time series feature fusion of snowmelt data;

[0058] The spatial feature fusion of snowmelt data refers to fusing the snowmelt dataset based on ground stations and satellite remote sensing in S2 using a two-layer long short-term memory network model and then restoring it using an activation function and an anti-normalization method to form the fused monthly snowmelt data;

[0059] The snowmelt data time series feature fusion refers to further fusing the ERA5 reanalysis daily snowmelt data in S1 with the fused monthly snowmelt data in S2 using an adaptive instance normalization method, ultimately obtaining daily snowmelt data.

[0060] S4. Evaluation of snowmelt data fusion effect: Use snowmelt observation alternative data and accuracy assessment indicators to evaluate the accuracy of the fused snowmelt data.

[0061] In S2, the spatial matching of ground stations is the daily snow water equivalent data. Taking the grid center point as the reference, the grid value closest to the longitude and latitude of the meteorological station is selected as the snow water equivalent data corresponding to the meteorological station, which is used to calculate the station snowmelt data based on remote sensing data.

[0062] In S2, the extraction of time series sample data is specifically as follows: based on the time resolution of the reanalyzed daily snowmelt data, extracting the grid daily snowmelt time series sample data closest to the longitude and latitude of the meteorological station;

[0063] If a month has 31 days, it forms a 31×1 time series.

[0064] The snowmelt data calculation in S2 is specifically as follows:

[0065] The monthly snowmelt data are calculated using a temperature index model based on the relationship between snowmelt and temperature. Monthly snowmelt data based on station data and remote sensing data are generated. The specific formula is:

[0066] ;

[0067] in, For the Monthly site snowmelt data; is the temperature coefficient, in millimeters per degree Celsius per day, which is an empirical coefficient relating snowmelt rate to air temperature; is the cumulative positive air temperature in degrees Celsius; For the The snow cover data of the ground stations in July can use the daily snow water equivalent data, which is the total amount of snowmelt;

[0068] When using the snowfall data of the site, it is necessary to substitute the Ground station snow cover data for the month :

[0069] ;

[0070] in, is the snowfall at the ground site, in millimeters; is the snow sublimation at the ground site, in millimeters;

[0071] The formula for snow sublimation at ground stations is:

[0072] ;

[0073] in, To obtain the minimum function, the smaller value between the snowmelt amount calculated by the function and the total amount currently used for snowmelt is taken as the final snowmelt amount; is the total amount of snow used for sublimation; is the potential evapotranspiration, the empirical coefficient Query according to Table 1;

[0074] Potential evapotranspiration is a sample data generated by matching the 1km monthly potential evapotranspiration dataset of the Yangtze River source region with ground station data through geographic point coordinates.

[0075] Table 1 Empirical parameters for estimating snow sublimation in different climate zones and snow types value

[0076]

[0077] In S3, the specific process of using a two-layer long-short-term memory network model to fuse site and remote sensing snowmelt sample data is as follows: first, the monthly snowmelt sequence samples based on site data are input into the memory neural unit of the long-short-term memory network model with multiple hidden layers, and then the output feature values ​​and the monthly snowmelt sequence samples based on remote sensing data are input into the second layer of the long-short-term memory network model. The specific formula of the memory neural unit of the long-short-term memory model is:

[0078] ;

[0079] Among them, the superscript 1 is the first-layer long short-term memory network model; For the Gate of Forgetfulness; is the activation function; is the weight matrix; and are the hidden states of the current site and the previous site respectively. The data of each unit in the first layer model is In the second-layer long short-term memory network model, the remote sensing snowmelt sample data and the output of the first-layer model are used as the input of the second-layer model; is a bias term; Input data for site snowmelt samples; is the hyperbolic tangent function; is the cell status of the current site;

[0080] The data features of the site and remote sensing will be integrated using the structure of the long short-term memory network model. The specific formula is as follows:

[0081] ;

[0082] Among them, the superscript 2 refers to the second-layer long short-term memory network model; Input for remote sensing snowmelt sample data.

[0083] The activation function and denormalization are used to output the fused feature variables as fused monthly snowmelt data. The specific formula is:

[0084] ;

[0085] in, For the Site, Monthly scale snowmelt data after monthly fusion; For the Site, Monthly snowmelt eigenvector; is the standard deviation function; is the arithmetic mean function.

[0086] In S3, the fusion of snowmelt data time series features is specifically as follows: the daily snowmelt data is reanalyzed using the adaptive normalization method and further integrated with the monthly snowmelt data after fusion using the long short-term memory network model, and finally the daily snowmelt data is obtained. The specific formula is:

[0087] ;

[0088] in, For the Site, Month daily snowmelt data; For the Site, Month Snowmelt reanalysis data for the day.

[0089] Due to the lack of snowmelt observation data, alternative data will be used to assist in proving the feasibility of snowmelt data. Therefore, the snowmelt observation alternative products in S4 include site snow depth change data and hydrological site measured flow data; the site snow depth change is the subtraction of the snow depth on the current day from the snow depth on the previous day as the previous day's snowmelt observation alternative data; the hydrological site measured flow data is the dry season flow data of the nearest hydrological station as the dry season snowmelt observation alternative data.

[0090] Among them, the Pearson correlation coefficient The specific formula is:

[0091] ;

[0092] in, is the number of meteorological stations in the study area; The snow depth change data measured at the ground station or the dry season flow data at the nearest hydrological station, the average value is ; is the fused daily snowmelt data product, with an average value of .

[0093] After the data is fused, the accuracy of the snowmelt data fusion method of multi-source data in this embodiment is evaluated, where the daily snowmelt data of a certain year and month in the study area is as follows: Figure 3 As shown. Figure 3 It can be seen that the fused snowmelt data obtained based on the multi-source snowmelt data fusion method proposed in this invention has a relatively consistent change pattern with the snow depth change data measured at ground stations in the study area, and is closer to the station observation results than the reanalysis data.

[0094] Furthermore, the accuracy assessment results based on the Pearson correlation coefficient are shown in Table 2. As shown in Table 2, the correlation coefficients between the fused snowmelt data obtained by the multi-source snowmelt data fusion method proposed in this invention and the measured snow depth change data at ground stations in the study area and the measured flow at nearby hydrological stations are higher than those of the reanalysis data products, reaching 0.8 and 0.68, respectively. This shows that the multi-source snowmelt data fusion method proposed in this invention can obtain highly accurate snowmelt data, fully considering the multiple data characteristics of ground stations, satellite remote sensing, and reanalysis data, thereby improving the accuracy of the snowmelt data obtained after multi-source snowmelt data fusion.

[0095] Table 2 Accuracy evaluation of multi-source snowmelt data fusion results in a certain study area in the example

[0096]

[0097] The above-described embodiments merely illustrate the implementation methods of the present invention. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion, characterized in that: The following steps are included: S1. Collect data in the study area: collect snow water equivalent data, snowmelt data, ground station data, potential evapotranspiration data and hydrological station flow data; The snow water equivalent data is daily snow water equivalent data; The snowmelt data is the reanalysis of daily snowmelt data; The ground station data are daily-scale measured snow depth data, temperature data, and snowfall data; The potential evapotranspiration data is a monthly potential evapotranspiration dataset at 1 km in the source area of ​​the Yangtze River. The flow data of the hydrological stations are measured flow data on a daily basis; S2. Data pre-processing: This includes ground station spatial matching and snowmelt data calculation. Ground station spatial matching involves matching uniform resolution snow cover data and potential evapotranspiration data with ground station data using geographic coordinates to generate sample data sequences and time series sample data. The snowmelt data is calculated by using snow water equivalent data and ground station data to calculate the satellite remote sensing snowmelt sequence and the site snowmelt sequence respectively; S3. Spatiotemporal feature fusion of multi-source snowmelt data: including spatial feature fusion of snowmelt data and time series feature fusion of snowmelt data; The spatial feature fusion of snowmelt data refers to fusing the satellite remote sensing snowmelt sequence and the site snowmelt sequence in S2 using a two-layer long short-term memory network model and then restoring them using an activation function and an anti-normalization method to form fused monthly snowmelt data; The snowmelt data time series feature fusion refers to further fusing the reanalyzed daily snowmelt data in S1 with the fused monthly snowmelt data in S3 using an adaptive instance normalization method to ultimately obtain daily snowmelt data; S4. Evaluation of snowmelt data fusion effect: Use snowmelt observation alternative data and accuracy assessment indicators to evaluate the accuracy of the fused daily snowmelt data.

2. The method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion according to claim 1, characterized in that: In S2, the spatial matching of ground stations is the daily snow water equivalent product data. Taking the grid center point as the reference, the grid value closest to the longitude and latitude of the meteorological station is selected as the snow water equivalent data corresponding to the meteorological station, which is used to calculate the ground station snowmelt data based on remote sensing data.

3. The method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion according to claim 2, characterized in that: In S2, the extraction of time series sample data is specifically as follows: based on the time resolution of the reanalyzed daily snowmelt data, extracting the grid daily snowmelt time series sample data closest to the longitude and latitude of the meteorological station; If a month has 31 days, it forms a 31×1 time series.

4. The method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion according to claim 3 is characterized in that: The snowmelt data calculation in S2 is specifically as follows: The monthly snowmelt data are calculated using a temperature index model based on the relationship between snowmelt and temperature. Monthly snowmelt data based on ground station data and remote sensing data are generated. The specific formula is: ； in, For the Monthly site snowmelt data; For the The monthly temperature coefficient, in millimeters per degree Celsius per day, is an empirical coefficient relating snowmelt rate to air temperature; For the The cumulative positive air temperature for the month, in degrees Celsius; For the The snow cover data of the ground stations in July can use the daily snow water equivalent data, which is the total amount of snowmelt; When using snowfall data from ground stations, it is necessary to substitute Ground station snow cover data for the month : ； in, For the Monthly snowfall at ground stations, in millimeters; For the Snow sublimation at ground stations for the month, in millimeters; The formula for snow sublimation at ground stations is: ； in, To obtain the minimum function, the smaller value between the snowmelt amount calculated by the function and the total amount currently used for snowmelt is taken as the final snowmelt amount; is the total amount of snow used for sublimation; The potential evapotranspiration is generated by matching the monthly potential evapotranspiration dataset of the Yangtze River source area at 1 km with the ground station data through the geographic point coordinates.

5. The method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion according to claim 4 is characterized in that: In S3, the specific process of using a two-layer long short-term memory network model to fuse ground station and remote sensing snowmelt sample data is as follows: first, the monthly snowmelt sequence samples based on the station data are input into the memory neural unit of the long short-term memory network model with multiple hidden layers for cyclic connection; the output feature values ​​and the monthly snowmelt sequence samples based on remote sensing data are input into the second layer of the long short-term memory network model; finally, the activation function and denormalization are used to output the fused feature variables as the fused monthly snowmelt data. The specific formula is: ； in, For the Site, Monthly scale snowmelt data after monthly fusion; For the Ground stations, Monthly snowmelt eigenvector; is the standard deviation function; is the arithmetic mean function.

6. The method for constructing a snowmelt dataset in the source region of the Yangtze River based on multi-source data fusion according to claim 5 is characterized in that: In S3, the fusion of snowmelt data time series features is specifically as follows: the daily snowmelt data is reanalyzed using the adaptive normalization method and further integrated with the monthly snowmelt data after fusion using the long short-term memory network model, and finally the daily snowmelt data is obtained. The specific formula is: ； in, For the Site, Month daily snowmelt data; For the Site, Month Snowmelt reanalysis data for the day.

7. The method for constructing a Yangtze River source region snowmelt dataset based on multi-source data fusion according to claim 6, characterized in that: Using the Pearson correlation coefficient Conduct reliability assessment on the fused daily snowmelt data; Due to the lack of snowmelt observation data, alternative data will be used to assist in proving the feasibility of snowmelt data. Therefore, the snowmelt observation alternative products in S4 include ground station snow depth change data and hydrological station measured flow data; the ground station snow depth change is obtained by subtracting the snow depth of the day from the snow depth of the previous day as the previous day's snowmelt observation alternative data; the hydrological station measured flow data is the dry season flow data of the nearest hydrological station as the dry season snowmelt observation alternative data.

Images