Snow depth inversion method, device and equipment based on multi-source data, medium and product

Abstract

The invention discloses an accumulated snow depth inversion method, device and equipment based on multi-source data, a medium and a product, and relates to the field of accumulated snow depth inversion. The method comprises the steps of obtaining information data of a target research area; preprocessing the information data, and constructing an accumulated snow depth inversion feature set based on the preprocessed information data, including an accumulated snow depth index SDI; constructing a plurality of training feature sets based on a plurality of features contained in the snow depth inversion feature set; taking the plurality of training feature sets as training data sets, and respectively inputting the training data sets into a plurality of initial snow depth inversion models for training to obtain a plurality of snow depth inversion models; the initial snow depth inversion model is determined based on a machine learning method; performing model selection in the snow depth inversion model based on an inversion precision result, and determining a target model; and carrying out inversion on the snow depth of the target research area based on the target model. According to the method, the high-resolution snow depth can be inverted in the shallow snow area.

Description

Technical Field

[0001] This application relates to the field of snow depth inversion, and in particular to a method, apparatus, equipment, medium and product for snow depth inversion based on multi-source data. Background Technology

[0002] Snow depth is a crucial factor in agricultural management. It influences soil moisture storage and spring sowing timing, making it a key variable in agricultural operations. Traditional meteorological stations provide sparse snow depth data, which is insufficient for large-area snow depth monitoring. At the regional scale, passive microwave remote sensing can penetrate snow cover and effectively monitor snow depth, but its spatial resolution is low (10-25 km), failing to capture the fine details of surface snow depth. Therefore, there is an urgent need to explore a remote sensing method for snow depth detection that combines high resolution and large-scale coverage.

[0003] In recent years, scholars have developed various methods to improve the spatial resolution of snow depth retrieval. Among them, machine learning methods based on active microwave remote sensing data have shown great potential, thanks to the fine-scale features provided by active microwave remote sensing imagery and the feature extraction capabilities of machine learning methods. However, most existing studies are limited to high-altitude areas, and their effectiveness is poor in plains areas with shallow snow depth. Large-area snow depth retrieval usually assumes uniform snow distribution with consistent density and grain size. However, snow grain size has a significant impact on microwave scattering; generally, larger snow grains scatter more microwave radiation than smaller ones. In practical applications, the spatiotemporal heterogeneity of snow characteristics significantly limits the accuracy of snow depth retrieval based on large-area or long-term microwave data. Therefore, how to retrieve high-resolution snow depth in shallow snow areas is crucial. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, equipment, medium, and product for snow depth inversion based on multi-source data, which can invert high-resolution snow depth in areas with shallow snow.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] Firstly, this application provides a method for snow depth inversion based on multi-source data, including:

[0007] Obtain information and data about the target study area;

[0008] The information data is preprocessed, and a snow depth inversion feature set is constructed based on the preprocessed information data; the snow depth inversion feature set includes: Snow Depth Index (SDI);

[0009] Multiple training feature sets are constructed based on the multiple features contained in the snow depth inversion feature set;

[0010] The multiple training feature sets are used as training datasets and input into various initial snow depth inversion models for training, resulting in multiple snow depth inversion models; the initial snow depth inversion models are determined based on machine learning methods.

[0011] The target model is determined by selecting a model from among the multiple snow depth inversion models based on the inversion accuracy results; the inversion accuracy results include: root mean square error and coefficient of determination.

[0012] The snow depth in the target study area is inverted based on the target model.

[0013] In one embodiment, the information data includes: snow depth data, satellite remote sensing data, and auxiliary data; wherein, the snow depth data includes observed snow depth from meteorological stations and snow depth data from manual field observations; the satellite remote sensing data includes active microwave remote sensing data and optical remote sensing data; and the auxiliary data includes snow cover products, coarse spatial resolution snow depth products, topographic data, and latitude and longitude data.

[0014] In one embodiment, the information data is preprocessed, and a snow depth inversion feature set is constructed based on the preprocessed information data, specifically including:

[0015] The information data is preprocessed to obtain preprocessed information data; wherein, the snow cover product, the coarse spatial resolution snow depth product, and the terrain data are resampled to the spatial resolution of the satellite remote sensing data using bilinear interpolation; the active microwave remote sensing data are subjected to orbit file application, GRD boundary noise removal, thermal noise removal, application of radiometric calibration values, terrain correction, and mean synthesis; the optical remote sensing data are subjected to atmospheric correction, cloud masking, and mean synthesis.

[0016] A snow depth inversion feature set is constructed based on the preprocessed information data.

[0017] In one embodiment, the snow depth inversion feature set further includes basic geographic features, active microwave remote sensing features, optical remote sensing features, and differential remote sensing features; the differential remote sensing features include: differential active microwave remote sensing features and differential optical remote sensing features;

[0018] The basic geographic features include longitude and latitude calculated from latitude and longitude data, altitude, slope and aspect calculated from topographic data, coarse-resolution snow depth obtained from coarse spatial resolution snow depth products, and cumulative snow days and cumulative snow volume calculated from snow cover products.

[0019] The active microwave remote sensing features include each band and band ratio of the active microwave remote sensing data; the optical remote sensing features include each band of the optical remote sensing data; the differential remote sensing features are the difference between the active microwave remote sensing data and the optical remote sensing data before and after the snowfall.

[0020] The Snow Depth Index (SDI) is calculated as follows:

[0021]

[0022] Where, diff is the differential processing of data before and after snowfall; NIR is the near-infrared band; SWIR2 is the short-wave infrared band; VH is the polarization band for vertical transmission and horizontal reception in synthetic aperture radar; and VHmin is the minimum value of the VH band.

[0023] In one embodiment, the multiple training feature sets include eight feature sets;

[0024] The first feature set is constructed from the basic geographic features; the second feature set is constructed from active microwave remote sensing features; the third feature set is constructed from optical remote sensing features; the fourth feature set is constructed from differential active microwave remote sensing features; the fifth feature set is constructed from differential optical remote sensing features; the sixth feature set is constructed from the Snow Depth Index (SDI); the seventh feature set is constructed from basic geographic features, active microwave remote sensing features, optical remote sensing features, differential active microwave remote sensing features, and differential optical remote sensing features; the eighth feature set is constructed from basic geographic features, active microwave remote sensing features, optical remote sensing features, differential active microwave remote sensing features, differential optical remote sensing features, and the Snow Depth Index (SDI). Among the first six feature sets, which are composed of features of a single category, the sixth feature set constructed from the Snow Depth Index (SDI) has the fewest features and the best prediction accuracy. After adding the Snow Depth Index (SDI) to the seventh feature set, the eighth feature set has the best prediction accuracy.

[0025] In one embodiment, the snow depth inversion model includes RF, SVM, and XGBoost; the target model is an SVM model that uses the eighth feature set as input features.

[0026] Secondly, this application provides a snow depth inversion device based on multi-source data, comprising:

[0027] The information and data acquisition module is used to acquire information and data about the target study area.

[0028] The processing module is used to preprocess the information data and construct a snow depth inversion feature set based on the preprocessed information data; the snow depth inversion feature set includes: snow depth index (SDI);

[0029] A construction module is used to construct multiple training feature sets based on multiple features contained in the snow depth inversion feature set;

[0030] The inversion training module is used to take the multiple training feature sets as training datasets and input them into multiple initial snow depth inversion models for training, thereby obtaining multiple snow depth inversion models; the initial snow depth inversion models are determined based on machine learning methods.

[0031] The determination module is used to select a target model from multiple snow depth inversion models based on inversion accuracy results; the inversion accuracy results include: root mean square error and coefficient of determination.

[0032] The inversion module is used to invert the snow depth of the target study area based on the target model.

[0033] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the snow depth inversion method based on multi-source data described above.

[0034] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the snow depth inversion method based on multi-source data described above.

[0035] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the snow depth inversion method based on multi-source data described above.

[0036] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0037] This application provides a method, apparatus, device, medium, and product for snow depth inversion based on multi-source data. The method involves acquiring information data of the target study area; preprocessing the information data; and constructing a snow depth inversion feature set based on the preprocessed information data. The snow depth inversion feature set includes the Snow Depth Index (SDI). Multiple training feature sets are constructed based on the features contained in the snow depth inversion feature set. These multiple training feature sets are used as training datasets and input into various initial snow depth inversion models for training, resulting in multiple snow depth inversion models. The initial snow depth inversion models are determined using machine learning methods. A target model is determined based on the inversion accuracy results among the snow depth inversion models. The snow depth of the target study area is then inverted based on the target model. This application, by preprocessing the information data and then constructing a training dataset combined with the inversion accuracy results for model selection, can invert high-resolution snow depth in shallow snow areas. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0039] Figure 1 This is a flowchart of a snow depth inversion method based on multi-source data;

[0040] Figure 2 This is a schematic diagram of the technical process for high spatial resolution snow depth inversion based on multi-source data;

[0041] Figure 3 A graph showing the feature importance results;

[0042] Figure 4 This is a graph showing the results of snow depth inversion.

[0043] Figure 5 This is a schematic diagram of the snow depth inversion system in practical applications;

[0044] Figure 6 This is a structural diagram of a snow depth inversion device based on multi-source data;

[0045] Figure 7 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0046] 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.

[0047] Previous studies have shown that visible / infrared remote sensing data can provide rich information on surface snow characteristics and evidence of snow presence. Therefore, constructing a snow depth index by integrating active microwave and optical remote sensing data and exploring the spatiotemporal heterogeneity of snow properties can effectively improve the accuracy of snow depth inversion.

[0048] This application combines active microwave remote sensing data with optical remote sensing data to construct a novel snow depth index feature, which, when added to a machine learning model, exhibits good snow depth inversion accuracy.

[0049] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] In one exemplary embodiment, such as Figure 1 As shown, a method for snow depth inversion based on multi-source data is provided, including:

[0051] Step 100: Obtain information data for the target study area.

[0052] The information data includes: snow depth data, satellite remote sensing data, and auxiliary data; among which, snow depth data includes snow depth observed at meteorological stations and snow depth observed manually in the field; satellite remote sensing data includes active microwave remote sensing data and optical remote sensing data; auxiliary data includes snow cover products, coarse spatial resolution snow depth products, topographic data, and latitude and longitude data.

[0053] Step 200: Preprocess the information data and construct a snow depth inversion feature set based on the preprocessed information data. The snow depth inversion feature set includes the Snow Depth Index (SDI).

[0054] The information data is preprocessed, and a snow depth inversion feature set is constructed based on the preprocessed information data, specifically including:

[0055] The information data is preprocessed to obtain preprocessed information data. Among them, the snow cover product, coarse spatial resolution snow depth product, and topographic data are resampled to the spatial resolution of satellite remote sensing data using bilinear interpolation. The active microwave remote sensing data are processed by applying orbit files, removing GRD boundary noise, removing thermal noise, applying radiometric calibration values, topographic correction, and mean synthesis. The optical remote sensing data are processed by atmospheric correction, cloud masking, and mean synthesis.

[0056] A snow depth inversion feature set is constructed based on the preprocessed information data.

[0057] In one embodiment, the snow depth inversion feature set further includes basic geographic features, active microwave remote sensing features, optical remote sensing features, and differential remote sensing features; the differential remote sensing features include differential active microwave remote sensing features and differential optical remote sensing features.

[0058] The basic geographic features include longitude and latitude calculated from latitude and longitude data, altitude, slope and aspect calculated from topographic data, coarse-resolution snow depth obtained from coarse spatial resolution snow depth products, and cumulative snow days and cumulative snow volume calculated from snow cover products.

[0059] Active microwave remote sensing features include each band and band ratio of active microwave remote sensing data; optical remote sensing features include each band of optical remote sensing data; differential remote sensing features are the differences between active microwave remote sensing data and optical remote sensing data before and after snowfall.

[0060] The cumulative number of snow days is calculated as follows: the cumulative number of days from the earliest snowfall date to the date when active microwave remote sensing data passes through the study area; the cumulative snow cover is calculated as follows: the cumulative snow cover from the earliest snowfall date to the date when active microwave remote sensing data passes through the study area.

[0061] The differential remote sensing features are calculated as follows: the average values ​​of the pre-snow active microwave remote sensing data and optical remote sensing data from one month before the snowfall to the earliest snowfall date are combined to form the pre-snow baseline image; the average values ​​of the post-snow active microwave remote sensing data and optical remote sensing data are combined according to the transit dates of the active microwave remote sensing data in the study area over fifteen days to form the post-snow feature image; the differential features are obtained by subtracting the pre-snow baseline image from the post-snow feature image.

[0062] The Snow Depth Index (SDI) is calculated as follows:

[0063]

[0064] Where, diff is the differential processing of data before and after snowfall; NIR is the near-infrared band; SWIR2 is the short-wave infrared band; VH is the polarization band for vertical transmission and horizontal reception in synthetic aperture radar; and VHmin is the minimum value of the VH band.

[0065] Step 300: Construct multiple training feature sets based on the multiple features contained in the snow depth inversion feature set.

[0066] The training feature sets include eight feature sets: the first feature set is constructed from basic geographic features; the second feature set is constructed from active microwave remote sensing features; the third feature set is constructed from optical remote sensing features; the fourth feature set is constructed from differential active microwave remote sensing features; the fifth feature set is constructed from differential optical remote sensing features; the sixth feature set is constructed from the Snow Depth Index (SDI); the seventh feature set is constructed from basic geographic features, active microwave remote sensing features, optical remote sensing features, differential active microwave remote sensing features, differential optical remote sensing features, and differential optical remote sensing features; the eighth feature set is constructed from basic geographic features, active microwave remote sensing features, optical remote sensing features, differential active microwave remote sensing features, differential optical remote sensing features, and the Snow Depth Index (SDI). Among the first six feature sets, which consist of features of a single category, the sixth feature set, constructed from the Snow Depth Index (SDI), has the fewest features and the best prediction accuracy. After adding the Snow Depth Index (SDI) to the seventh feature set, the eighth feature set has the best prediction accuracy. That is, among all feature sets, the eighth feature set has the best prediction accuracy.

[0067] Step 400: Multiple training feature sets are used as training datasets and input into various initial snow depth inversion models for training, resulting in multiple snow depth inversion models. The initial snow depth inversion models are determined based on machine learning methods. These models include RF, SVM, and XGBoost algorithms.

[0068] Step 500: Select a target model from multiple snow depth inversion models based on the inversion accuracy results. The inversion accuracy results include root mean square error and coefficient of determination. The target model is an SVM model using the eighth feature set as input features.

[0069] Step 600: Invert the snow depth of the target study area based on the target model.

[0070] like Figure 2 As shown, a high spatial resolution snow depth inversion method based on multi-source data includes:

[0071] Step 1: Acquire snow depth data, satellite remote sensing data, and auxiliary data. Snow depth data includes snow depth observed at meteorological stations and snow depth observed in the field. Satellite remote sensing data includes active microwave remote sensing data (Sentinel-1 microwave remote sensing data) and optical remote sensing data (Sentinel-2 optical remote sensing data). Auxiliary data includes snow cover products, coarse spatial resolution snow depth products, topographic data, and latitude and longitude data, namely, MOD10A1 snow cover products, coarse spatial resolution daily snow depth products from EAR5-Land reanalysis climate data, SRTM3 DEM topographic data, and latitude and longitude data.

[0072] Step 2: Preprocess the acquired data, extract features based on the preprocessed data, construct the snow depth index feature and construct the inversion feature set.

[0073] Data preprocessing includes: resampling snow cover products, coarse spatial resolution snow depth products, and topographic data to the spatial resolution of satellite remote sensing data using bilinear interpolation; applying orbit files, removing GRD boundary noise, removing thermal noise, applying radiometric calibration values, topographic correction, and mean synthesis to active microwave remote sensing data; and performing atmospheric correction, cloud masking, and mean synthesis on optical remote sensing data.

[0074] Specifically, bilinear interpolation was used to resample the MOD10A1 snow cover product, the coarse spatial resolution daily snow depth product in the EAR5-Land reanalysis climate data, and the SRTM3 DEM topographic data to the spatial resolution of the satellite remote sensing data; orbit files, GRD boundary noise removal, thermal noise removal, application of radiometric calibration values, topographic correction, and mean synthesis were performed on the Sentinel-1 microwave remote sensing data; atmospheric correction, cloud masking, and mean synthesis were performed on the Sentinel-2 optical remote sensing data.

[0075] The snow depth index is calculated as follows: Where diff represents differential processing of data before and after snowfall, NIR is the near-infrared band, SWIR2 is the short-wave infrared band, VH is the polarization band for vertical transmission and horizontal reception in synthetic aperture radar, and VHmin is the minimum value of the VH band.

[0076] Basic geographic features include longitude and latitude calculated from latitude and longitude data, elevation, slope and aspect calculated from topographic data, coarse-resolution snow depth obtained from EAR5-Land reanalysis climate data, and cumulative snow cover days and cumulative snow cover calculated from MOD10A1 snow cover products; active microwave remote sensing features include VV, VH and VV / VH from Sentienl-1 data; optical remote sensing features include various bands from Sentienl-2 data; differential remote sensing features are the differences between active microwave remote sensing features and optical remote sensing features before and after snowfall.

[0077] The cumulative number of snow days is calculated as follows: the cumulative number of days from the earliest snowfall date to the date the Sentienl-1 satellite passes over the study area; the cumulative snow cover is calculated as follows: the cumulative snow cover from the earliest snowfall date to the date the Sentienl-1 satellite passes over the study area.

[0078] In one embodiment, the differential remote sensing features are calculated as follows: the Sentienl-1 and Sentienl-2 data before the snowfall are averaged and synthesized from one month before the snowfall to the earliest snowfall date to serve as the pre-snowfall baseline image; the Sentienl-1 and Sentienl-2 data after the snowfall are averaged and synthesized over fifteen days according to the transit date of the Sentienl-1 data within the study area to serve as the post-snowfall feature image; the differential features are obtained by subtracting the pre-snowfall baseline image from the post-snowfall feature image.

[0079] Step 3: Input the training datasets obtained from multiple training feature sets composed of different features into various machine learning-based snow depth inversion models for training, and obtain various snow depth inversion models.

[0080] The training dataset, composed of multiple training feature sets with different features, includes: constructing eight training feature sets; wherein, the first feature set is constructed from the aforementioned basic geographic features; the second feature set is constructed from active microwave remote sensing features; the third feature set is constructed from optical remote sensing features; the fourth feature set is constructed from differential active microwave remote sensing features; the fifth feature set is constructed from differential optical remote sensing features; the sixth feature set is constructed from snow depth index features (SDI); and the seventh feature set is constructed from basic geographic features. The first set of features is constructed from basic geographic features, active microwave remote sensing features, optical remote sensing features, differential active microwave remote sensing features, and differential optical remote sensing features. The second set of features is constructed from basic geographic features, active microwave remote sensing features, optical remote sensing features, differential active microwave remote sensing features, differential optical remote sensing features, and snow depth index features. As shown in Table 1, among the first six feature sets composed of single-category features, the sixth feature set constructed from snow depth index features has the fewest number of features and the best prediction accuracy. After adding snow depth index features to the seventh feature set, the eighth feature set has the best prediction accuracy. Figure 3 The feature importance of each individual feature is calculated using the random forest algorithm, where the snow depth index (SDI) is the most important feature in the training model. Figure 3 The vertical axis represents each individual feature, and the horizontal axis represents the importance of the corresponding feature.

[0081] Table 1 Prediction Results

[0082]

[0083] Step 4: Verify the accuracy of the various trained snow depth inversion models, as shown in Table 1. Based on the inversion accuracy results, select the model with the best accuracy as the target model and perform snow depth inversion throughout the entire study area. The inversion accuracy results include: root mean square error (RMSE) and coefficient of determination (R).

[0084] The accuracy evaluation results of snow depth inversion models based on different feature combinations and machine learning algorithms are presented. The SVM using the eighth feature combination achieves the best snow depth inversion accuracy and significantly outperforms the other algorithms.

[0085] The snow depth inversion results for the study area based on the target model are as follows: Figure 4As shown, 10 days after the first snowfall, the snow depth distribution exhibited significant spatial heterogeneity. The northernmost county in XX city generally had deeper snow depths, averaging approximately 14-16 cm, with some areas (such as the northeast) reaching over 18 cm. In contrast, the snow depth in XX county in XX city was slightly shallower, averaging between 12-14 cm, while the southernmost district in XX city had the lowest snow depth, averaging only 10-12 cm. This north-south difference may be related to the spatial distribution of latitudinal gradient and snowfall amount. Forty days after the first snowfall, the overall snow depth decreased. In XX county, the snow depth decreased to 12-14 cm; in XX county, it decreased to 10-12 cm; and in XX district, it decreased to 8-10 cm. This change was influenced by snow compression, localized snowmelt, and wind action, resulting in a less uniform spatial distribution of snow depth. Ten days after another snowfall, the snow depth increased again. The new round of snowfall replenished the snow cover, while the lower temperatures likely slowed the snowmelt rate, allowing snow to accumulate. Temporally, the snow depth in the study area underwent a dynamic process of "increase-decrease-re-increase" during the snow season. Spatially, the snow depth exhibited a "deeper in the north, shallower in the south" distribution pattern, with the deepest snow in County XX and the shallowest in District XX. Furthermore, significant heterogeneity existed at local scales due to surface conditions (such as ridged farmland and straw mulch). Based on the model's performance, this application effectively captures the spatiotemporal variations in snow depth, providing reliable data support for agricultural management in plain areas.

[0086] This application provides high spatial resolution snow depth data for shallow snow areas, providing necessary data support for extensive and timely monitoring of snow depth.

[0087] like Figure 5 As shown, the snow depth inversion system in practical applications includes:

[0088] The data acquisition unit acquires snow depth data, satellite remote sensing data, and auxiliary data. The snow depth data includes snow depth observed at meteorological stations and snow depth observed in the field. The satellite remote sensing data includes Sentinel-1 microwave remote sensing data and Sentinel-2 optical remote sensing data. The auxiliary data includes MOD10A1 snow cover product, coarse spatial resolution daily snow depth product from EAR5-Land reanalysis climate data, SRTM3 DEM topographic data, and latitude and longitude data.

[0089] The data preprocessing unit resamples MOD10A1 snow cover products, coarse spatial resolution daily snow depth products from EAR5-Land reanalysis climate data, and SRTM3 DEM topographic data to the spatial resolution of satellite remote sensing data using bilinear interpolation; it applies orbit files, removes GRD boundary noise, removes thermal noise, applies radiometric calibration values, performs topographic correction, and synthesizes means for Sentinel-1 microwave remote sensing data; and it performs atmospheric correction, cloud masking, and means synthesis for Sentinel-2 optical remote sensing data.

[0090] The feature extraction unit is used to construct a set of features for snow depth inversion, including basic geographic features, active microwave remote sensing features, optical remote sensing features, differential active microwave remote sensing features, differential optical remote sensing features, and snow depth index features.

[0091] The snow depth index is calculated as follows: Where *diff* represents the differential processing of pre-snow and post-snow data, *NIR* is the near-infrared band, *SWIR2* is the shortwave infrared band, *VH* is the polarization band for vertical transmission and horizontal reception in synthetic aperture radar, and *VHmin* is the minimum value of the *VH* band. Basic geographic features include longitude and latitude calculated from latitude and longitude data, elevation, slope, and aspect calculated from topographic data, coarse-resolution snow depth obtained from coarse spatial resolution snow depth products, and cumulative snow days and cumulative snow cover calculated from snow cover products. Active microwave remote sensing features include each band and band ratio of the active microwave remote sensing data; optical remote sensing features include each band of the optical remote sensing data; and differential remote sensing features are the differences between the pre-snow and post-snow active microwave remote sensing data and the optical remote sensing data.

[0092] The cumulative number of snow days is calculated as follows: the cumulative number of days from the earliest snowfall date to the date when active microwave remote sensing data passes through the study area; the cumulative snow cover is calculated as follows: the cumulative snow cover from the earliest snowfall date to the date when active microwave remote sensing data passes through the study area.

[0093] The differential remote sensing features are calculated as follows: the average values ​​of the pre-snow active microwave remote sensing data and optical remote sensing data from one month before the snowfall to the earliest snowfall date are combined to form the pre-snow baseline image; the average values ​​of the post-snow active microwave remote sensing data and optical remote sensing data are combined according to the transit dates of the active microwave remote sensing data in the study area over fifteen days to form the post-snow feature image; the differential features are obtained by subtracting the pre-snow baseline image from the post-snow feature image.

[0094] The model training unit is used to train different snow depth inversion models. The input features include eight sets of training features, and the training models include three machine learning algorithms.

[0095] The accuracy evaluation unit evaluates the accuracy of the trained snow depth inversion models and selects the model with the best accuracy as the target model for snow depth inversion.

[0096] In one exemplary embodiment, such as Figure 6 As shown, a snow depth inversion device based on multi-source data is provided, comprising:

[0097] The information and data acquisition module is used to acquire information and data about the target study area.

[0098] The processing module is used to preprocess the information data and construct a snow depth inversion feature set based on the preprocessed information data. The snow depth inversion feature set includes the Snow Depth Index (SDI).

[0099] The module is used to construct multiple training feature sets based on multiple features contained in the snow depth inversion feature set.

[0100] The inversion training module is used to take multiple training feature sets as training datasets and input them into multiple initial snow depth inversion models for training, thereby obtaining multiple snow depth inversion models; the initial snow depth inversion models are determined based on machine learning methods.

[0101] The determination module is used to select a target model from multiple snow depth inversion models based on the inversion accuracy results; the inversion accuracy results include: root mean square error and coefficient of determination.

[0102] The inversion module is used to invert the snow depth of the target study area based on the target model.

[0103] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 7As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores snow depth inversion data based on multi-source data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a snow depth inversion method based on multi-source data.

[0104] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0105] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0106] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0107] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0108] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0109] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0110] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0111] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] This document uses specific examples 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 methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for snow depth inversion based on multi-source data, characterized in that, The method comprises the following steps: acquiring information data of a target study area; preprocessing the information data and constructing a snow depth inversion feature set based on the preprocessed information data; the snow depth inversion feature set comprises a snow depth index SDI; a plurality of training feature sets are constructed based on a plurality of features contained in the snow depth inversion feature set; the plurality of training feature sets are used as training data sets and are input into a plurality of initial snow depth inversion models for training to obtain a plurality of snow depth inversion models; the initial snow depth inversion model is determined based on a machine learning method; a target model is determined by selecting a model based on inversion accuracy results in the plurality of snow depth inversion models; the inversion accuracy results comprise a root mean square error and a determination coefficient; the target model is used to invert the snow depth of the target study area.

2. The multi-source data based snow depth inversion method according to claim 1, characterized in that, The information data comprises snow depth data, satellite remote sensing data and auxiliary data; the snow depth data comprises observed snow depth at weather stations and manually observed snow depth in the field; the satellite remote sensing data comprises active microwave remote sensing data and optical remote sensing data; the auxiliary data comprises a snow cover product, a coarse spatial resolution snow depth product, terrain data and latitude and longitude data.

3. The multi-source data based snow depth inversion method according to claim 2, characterized in that, The information data is preprocessed and a snow depth inversion feature set is constructed based on the preprocessed information data, specifically comprising: The information data is preprocessed to obtain preprocessed information data; the snow cover product, the coarse spatial resolution snow depth product and the terrain data are resampled to the spatial resolution of the satellite remote sensing data by using a bilinear interpolation method; the active microwave remote sensing data is processed by applying an orbit file, removing GRD boundary noise, removing thermal noise, applying a radiation calibration value, terrain correction and mean synthesis; the optical remote sensing data is processed by atmospheric correction, cloud mask processing and mean synthesis; a snow depth inversion feature set is constructed based on the preprocessed information data.

4. The multi-source data based snow depth inversion method according to claim 1, characterized in that, The snow depth inversion feature set further comprises basic geographic features, active microwave remote sensing features, optical remote sensing features and differential remote sensing features; The differential remote sensing features comprise differential active microwave remote sensing features and differential optical remote sensing features; The basic geographic features comprise longitude and latitude calculated from the latitude and longitude data, elevation, slope and aspect calculated from the terrain data, coarse resolution snow depth obtained from the coarse spatial resolution snow depth product, and cumulative snow days and cumulative snow amount calculated from the snow cover product; The active microwave remote sensing features comprise each band and band ratio of the active microwave remote sensing data; the optical remote sensing features comprise each band of the optical remote sensing data; the differential remote sensing features are the difference between the active microwave remote sensing data and the optical remote sensing data before and after snowfall; The calculation method of the snow depth index SDI is: Wherein, diff is the difference between the data before and after the snow; NIR is the near-infrared band; SWIR2 is the short-wave infrared band; VH is the polarization band of the vertical emission and horizontal reception in the synthetic aperture radar; VHmin is the minimum value of the VH band.

5. The multi-source data based snow depth inversion method according to claim 4, characterized in that, The plurality of training feature sets includes eight feature sets; Wherein, the first feature set is constructed from the basic geographic features; the second feature set is constructed from the active microwave remote sensing features; the third feature set is constructed from the optical remote sensing features; the fourth feature set is constructed from the differential active microwave remote sensing features; the fifth feature set is constructed from the differential optical remote sensing features; the sixth feature set is constructed from the snow depth index SDI; the seventh feature set is constructed from the basic geographic features, the active microwave remote sensing features, the optical remote sensing features, the differential active microwave remote sensing features, and the differential optical remote sensing features; the eighth feature set is constructed from the basic geographic features, the active microwave remote sensing features, the optical remote sensing features, the differential active microwave remote sensing features, the differential optical remote sensing features, and the snow depth index SDI; among the first six feature sets each composed of a single type of features, the sixth feature set constructed from the snow depth index SDI has the least number of features and the best prediction accuracy; after adding the snow depth index SDI to the seventh feature set, the eighth feature set has the best prediction accuracy.

6. The multi-source data based snow depth inversion method according to claim 5, characterized in that, The snow depth inversion model includes RF, SVM, and XGBoost; the target model is an SVM model using the eighth feature set as the input features.

7. A multi-source data based snow depth inversion device, characterized in that, It comprises: an information data acquisition module configured to acquire information data of a target study area; a processing module configured to pre-process the information data and construct a snow depth inversion feature set based on the pre-processed information data; the snow depth inversion feature set includes a snow depth index SDI; a construction module configured to construct a plurality of training feature sets based on a plurality of features included in the snow depth inversion feature set; an inversion training module configured to input the plurality of training feature sets as training data sets into a plurality of initial snow depth inversion models for training to obtain a plurality of snow depth inversion models; the initial snow depth inversion model is determined based on a machine learning method; a determination module configured to select a target model based on inversion accuracy results among the plurality of snow depth inversion models; the inversion accuracy results include root mean square error and determination coefficient; an inversion module configured to invert the snow depth of the target study area based on the target model.

8. A computer device comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that the processor executes the computer program to implement the multi-source data-based snow depth inversion method of any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the multi-source data-based snow depth inversion method of any one of claims 1-6.

10. A computer program product comprising a computer program, characterized in that, The computer program, when executed by the processor, implements the multi-source data based snow depth inversion method of any one of claims 1-6.

Images