A forward modeling method, device, medium and product for snow albedo

Abstract

The present application discloses a forward modeling method, device, medium and product for snow albedo, relating to the technical field of deep learning, including: obtaining forward modeling parameters of the target snow; the forward modeling parameters include: snow physical parameters, normalized difference snow index and observation geometry data, the snow physical parameters include: snow grain size and snow depth, and the observation geometry data includes: solar zenith angle, observation zenith angle, solar azimuth angle and observation azimuth angle of MODIS; inputting the forward modeling parameters of the target snow into the snow albedo forward modeling model to obtain the forward modeling value of the snow albedo of the target snow; the snow albedo forward modeling model is obtained by training the Transformer network using a training data set; the training data set includes forward modeling parameters and true values of snow albedo at multiple target locations, and the snow albedo includes: visible light band snow albedo and near-infrared band snow albedo. The present application improves the forward modeling accuracy of snow albedo.

Description

Technical Field

[0001] The present application relates to the field of deep learning technology, and particularly to a forward inversion method, device, medium and product for snow reflectivity. Background Art

[0002] Monitoring and researching snow resources not only helps to deeply understand complex environmental problems such as hydrology, meteorology and climate change in the Earth system, but also provides important scientific support and decision-making references for related fields. It has important significance for target detection, tracking and recognition technologies and has always been a research focus in related fields. As an important parameter in the optical properties of snow, snow reflectivity is an important indicator for measuring the surface energy budget, and the research on snow reflectivity has important significance. The value of snow reflectivity is affected by factors including snow grain size, snow depth, pollutants in snow, terrain and solar zenith angle.

[0003] With the development of detection technology means, using satellite observation data and combining with a snow radiation transfer model to forward invert snow reflectivity has become one of the important ways to study snow resources in the Qinghai-Tibet Plateau region. Satellite observation data has advantages such as convenient acquisition, high spectral and temporal resolution, wide detection range and low cost. The snow radiation transfer model can forward invert the reflectivity of the snow surface by inputting weather data, snow parameters and terrain information, and can be calculated at multiple time and space scales.

[0004] Currently, the traditional satellite data-based snow reflectivity forward inversion method has high computational complexity, difficult parameter acquisition, and the application of the model is restricted by some special problems and the calculation difficulty of eigenvalues. Moreover, the forward inversion effects of the traditional radiation transfer model in each band are different, and the forward inversion accuracy of individual bands is not high. Summary of the Invention

[0005] The purpose of the present application is to provide a forward inversion method, device, medium and product for snow reflectivity, which improves the forward inversion accuracy of snow reflectivity.

[0006] To achieve the above purpose, the present application provides the following solutions:

[0007] In a first aspect, the present application provides a forward inversion method for snow reflectivity, including:

[0008] Obtaining forward inversion parameters of a target snow; the forward inversion parameters include: snow physical parameters, normalized difference snow index and observation geometry data, the snow physical parameters include: snow grain size and snow depth, and the observation geometry data includes: solar zenith angle, observation zenith angle, solar azimuth angle and observation azimuth angle of MODIS;

[0009] Input the forward modeling parameters of the target snow cover into the snow cover reflectance forward modeling model to obtain the forward modeling value of the snow cover reflectance of the target snow cover; the snow cover reflectance forward modeling model is obtained by training a Transformer network using a training dataset; the training dataset includes the forward modeling parameters and the true values of the snow cover reflectance at multiple target locations, and the snow cover reflectance includes: visible light band snow cover reflectance and near-infrared band snow cover reflectance.

[0010] Optionally, the training process of the snow cover reflectance forward modeling model includes:

[0011] Obtain the training dataset;

[0012] Construct the Transformer network;

[0013] Use the forward modeling parameters of the snow cover at each target location in the training dataset as the input and the corresponding true value of the snow cover reflectance as the output to train the Transformer network to obtain the snow cover reflectance forward modeling model.

[0014] Optionally, obtaining the training dataset includes:

[0015] Obtain MODIS data at multiple locations; the MODIS data includes: surface reflectance data and observation geometry data; the surface reflectance data includes: visible light band reflectance data and near-infrared band reflectance data;

[0016] Calculate the normalized difference snow index (NDSI) for the corresponding location based on the visible light band reflectance data at each location;

[0017] Perform snow cover identification based on the NDSI at each location to determine the snow cover situation at the corresponding location; the snow cover situation is either snow cover exists or no snow cover exists;

[0018] Screen each location based on the surface reflectance data at the locations where snow cover exists, and determine the screened locations as target locations;

[0019] Determine the surface reflectance data of each target location as the true value of the snow cover reflectance of the corresponding target location;

[0020] Invert the snow grain size of the corresponding target location based on the visible light band reflectance data of each target location and the ART model;

[0021] Invert the snow depth of the corresponding target location based on the surface reflectance data of each target location and the linear regression model;

[0022] Determine the true values of the snow particle size, snow depth, and snow reflectivity of the snow at each target location as the training dataset.

[0023] Optionally, calculate the normalized difference snow index (NDSI) for each location based on the visible light band reflectivity data at each location, including:

[0024] Calculate the normalized difference snow index for each location according to the reflectivity data of the fourth band and the short-wave infrared band reflectivity data in the visible light band reflectivity data at each location.

[0025] Optionally, perform snow cover identification based on the normalized difference snow index of each location to determine the snow cover situation at the corresponding location, including:

[0026] Determine that there is snow cover at the locations where the normalized difference snow index is greater than 0.4.

[0027] Optionally, screen each location based on the surface reflectivity data of each location where the snow cover situation is determined to be snow cover, and determine the screened locations as target locations, including:

[0028] Determine the screening rule based on the surface reflectivity data of each location; the screening rule is that the reflectivity data of the first band, the third band, and the fourth band are all greater than 0.7 and the reflectivity data of the second band is greater than 0.11; the wavelength range corresponding to the reflectivity data of the first band is 620nm - 670nm, the wavelength range corresponding to the reflectivity data of the second band is 841nm - 876nm, the wavelength range corresponding to the reflectivity data of the third band is 459nm - 479nm, and the wavelength range corresponding to the reflectivity data of the fourth band is 620 - 670nm;

[0029] Determine the locations where the surface reflectivity data meets the screening rule as target locations.

[0030] Optionally, the Transformer network includes: an encoder module and a decoder module;

[0031] The encoder module includes: 6 encoders connected in sequence; each encoder includes: a multi-headed self-attention mechanism and a feed-forward neural network;

[0032] The decoder module includes: 6 decoders connected in sequence; each decoder includes: a masked multi-headed attention mechanism, a multi-headed attention mechanism, and a feed-forward neural network.

[0033] In a second aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the snow reflectivity forward modeling method described in any one of the above.

[0034] In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the snow reflectance forward modeling method described in any one of the above is implemented.

[0035] In a fourth aspect, the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, the snow reflectance forward modeling method described in any one of the above is implemented.

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

[0037] The present application discloses a snow reflectance forward modeling method, device, medium and product, which comprehensively utilizes various MODIS product data, fully considers various factors affecting snow reflectance such as snow physical parameters and observation geometry, and the most important one is the snow grain size factor. Considering factors from multiple angles and in all directions enables the snow reflectance forward modeling model to be applicable to a variety of different situations and has good universality; it is proposed to use a Transformer network to construct a snow reflectance forward modeling model, which requires fewer parameters and simple parameter forward modeling, and can be applicable to the forward modeling of a large number of reflectance data, overcoming the difficulties of complex traditional methods and large computational amounts; it is proposed to use a Transformer network to construct a snow reflectance forward modeling model, which fully approximates the complex non-linear relationship between key snow physical parameters and snow reflectance through a large amount of training data, and solves the problem that the traditional radiative transfer calculation model is not universal for individual bands or the low accuracy of reflectance forward modeling caused by physical parameter errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0039] Figure 1 It is a schematic flowchart of the snow reflectance forward modeling method provided in Embodiment 1 of the present application;

[0040] Figure 2 It is a schematic diagram of the architecture of the snow reflectance forward modeling method;

[0041] Figure 3 It is a schematic diagram of the DS-Encoder structure;

[0042] Figure 4 It is a schematic diagram of the DS-Decoder structure;

[0043] Figure 5Schematic diagram of the forward modeling result for the first band;

[0044] Figure 6 Schematic diagram of the forward modeling result for the third band;

[0045] Figure 7 Schematic diagram of the structure of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0046] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0047] The purpose of the present application is to provide a method, device, medium and product for forward modeling of snow albedo, aiming to improve the forward modeling accuracy of snow albedo.

[0048] To make the above objects, features and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the drawings and specific implementation manners.

[0049] In an exemplary embodiment, as Figure 1 and Figure 2 shown, a method for forward modeling of snow albedo is provided, including:

[0050] Step 1: Obtain the forward modeling parameters of the target snow; the forward modeling parameters include: snow physical parameters, normalized difference snow index, and observation geometry data. The snow physical parameters include: snow grain size and snow depth. The observation geometry data includes: solar zenith angle, observation zenith angle, solar azimuth angle, and observation azimuth angle of MODIS.

[0051] Step 2: Input the forward modeling parameters of the target snow into the snow albedo forward modeling model to obtain the forward modeling value of the snow albedo of the target snow; the snow albedo forward modeling model is obtained by training the Transformer network using a training data set; the training data set includes the forward modeling parameters and the true values of the snow albedo of the snow at multiple target locations. The snow albedo includes: visible light band snow albedo and near-infrared band snow albedo.

[0052] Specifically, by comprehensively utilizing various MODIS product data and fully considering various factors affecting snow reflectivity, such as snow physical parameters and observation geometry, among which the most important is the snow grain size factor, considering factors from multiple angles and in all directions enables the forward snow reflectivity model to be applicable to various different situations and have good universality; it is proposed to use the Transformer network to construct the forward snow reflectivity model, which requires fewer parameters and has a simple parameter forward calculation, and can be applied to the forward calculation of a large number of reflectivity data, overcoming the difficulties of complex traditional methods and large computational amounts; it is proposed to use the Transformer network to construct the forward snow reflectivity model, which fully approximates the complex non-linear relationship between key snow physical parameters and snow reflectivity through a large amount of training data, and solves the problem that the traditional radiative transfer calculation model is not universal for individual bands or the low accuracy of reflectivity forward calculation caused by physical parameter errors.

[0053] As an alternative implementation, the training process of the forward snow reflectivity model includes:

[0054] Step 21: Obtain a training data set.

[0055] As an alternative implementation, Step 21 includes:

[0056] Step 211: Obtain MODIS data at multiple locations; the MODIS data includes: surface reflectivity data and observation geometry data; the surface reflectivity data includes: visible light band reflectivity data and near-infrared band reflectivity data.

[0057] Step 212: Calculate the normalized snow cover index corresponding to each location based on the visible light band reflectivity data at each location.

[0058] As an alternative implementation, Step 212 includes:

[0059] Calculate the normalized snow cover index corresponding to each location according to the reflectivity data of the fourth band and the short-wave infrared band in the visible light band reflectivity data at each location.

[0060] Step 213: Conduct snow cover identification based on the normalized snow cover index at each location to determine the snow cover situation at the corresponding location; the snow cover situation is either snow cover exists or no snow cover exists.

[0061] As an alternative implementation, Step 213 includes:

[0062] Determine that the snow cover situation at the location where the normalized snow cover index is greater than 0.4 is snow cover exists.

[0063] Step 214: Screen each location based on the surface reflectivity data at each location where the snow cover situation is snow cover exists, and determine the screened locations as target locations.

[0064] As an alternative implementation, step 214 includes:

[0065] Step 2141: Determine a screening rule based on the surface reflectance data at each location; the screening rule is that the first-band reflectance data, the third-band reflectance data, and the fourth-band reflectance data are all greater than 0.7 and the second-band reflectance data is greater than 0.11; the wavelength range corresponding to the first-band reflectance data is 620 nm to 670 nm, the wavelength range corresponding to the second-band reflectance data is 841 nm to 876 nm, the wavelength range corresponding to the third-band reflectance data is 459 nm to 479 nm, and the wavelength range corresponding to the fourth-band reflectance data is 620 to 670 nm.

[0066] Specifically, when the first-band reflectance data, the third-band reflectance data, and the fourth-band reflectance data are all greater than 0.7, it is ensured that the snow cover is in a full-coverage state; when the second-band reflectance data is greater than 0.11, the influence of water bodies can be removed.

[0067] Step 2142: Determine each location where the surface reflectance data satisfies the screening rule as the target location.

[0068] Step 215: Determine the surface reflectance data at each target location as the true value of the snow reflectance at the corresponding target location.

[0069] Step 216: Invert based on the visible light band reflectance data and the ART model at each target location to obtain the snow grain size at the corresponding target location.

[0070] Specifically, using the single-band inversion formula based on the ART theory, invert based on the visible light band reflectance data and the ART model at each target location to obtain the snow grain size at the corresponding target location. The single-band inversion formula based on the ART theory is:

[0071]

[0072] Where is the snow grain size; γ(λ) is the absorption coefficient with respect to the wavelength λ; λ is the wavelength; C is the snow grain size shape and asymmetry parameter; is with respect to μ s 、μ v and is the angular function of the viewing and illumination geometry of; μ s 、μ v and are all observation geometry parameters and can be calculated and obtained according to the observation geometry data; R is the band reflectance value of the wavelength band where the wavelength is located; is with respect to μ s 、μv and The reflectivity distribution function of the semi-infinite space snow layer

[0073] Actually, the wavelengths λ used in the snow forward model are generally 1050nm and 1240nm. These wavelengths are sensitive to the absorption caused by snow particle size, and 1240nm is the central wavelength of the fifth band of MODIS. Therefore, this application uses the reflectivity data of the fifth band as the wavelength parameter for single-band inversion of snow particle size.

[0074] Step 217: Invert based on the surface reflectivity data of each target location and the linear regression model to obtain the snow depth corresponding to each target location.

[0075] Specifically, the formula of the linear regression model is:

[0076] Y = 10.513 + 207.15X1 - 312.502X2 + 97.058X3 + 1.105X4 + 29.554X5.

[0077] Among them, Y is the snow depth; X1 is the reflectivity data of the third band; X2 is the reflectivity data of the fourth band; X3 is the reflectivity data of the first band; X4 is the reflectivity data of the second band; X5 is the reflectivity data of the fifth band. The wavelength range corresponding to the reflectivity data of the fifth band is 1230nm - 1250nm. The first band, the third band, the fourth band, and the fifth band all belong to the visible light band, and the second band belongs to the near-infrared band.

[0078] Step 218: Determine the true values of the snow particle size, snow depth, and snow reflectivity of the snow at each target location as the training data set.

[0079] Step 22: Construct a Transformer network.

[0080] As an optional implementation, the Transformer network includes: an encoder module and a decoder module.

[0081] The encoder module includes: 6 encoders (DS-Encoder) connected in sequence; the encoder includes: a multi-headed self-attention mechanism and a feed-forward neural network.

[0082] The decoder module includes: 6 decoders (DS-Decoder) connected in sequence; the decoder includes: a masked multi-headed attention mechanism, a multi-headed attention mechanism, and a feed-forward neural network.

[0083] Step 23: Use the forward parameters of the snow at each target location in the training data set as the input and the true value of the corresponding snow reflectivity as the output to train the Transformer network to obtain a snow reflectivity forward model.

[0084] Specifically, as Figure 3 shown, the specific structure of the DS-Encoder is as Figure 2 shown. Each layer includes two sub-layers. The first sub-layer is the Multi-Head Self-Attention mechanism, which is used to calculate the attention of the input.

[0085] The second sub-layer is the Feed Forward neural network layer, which usually consists of two consecutive fully connected layers and an activation function. This structure is the standard setting of the Transformer, but in the practice of processing multi-feature and large-batch data, it is necessary to extract features more deeply and effectively. In this application, the Feed Forward neural network structure in the Transformer model is changed to a Pyramid Pooling module. The improved Pyramid Pooling module can capture object information at different scales and provide multi-scale feature representations. The steps of its feature fusion are as follows:

[0086] 1) Convolve the input sequence with a Convolutional Neural Networks (CNN) to obtain an input feature map.

[0087] 2) Perform Pyramid Pooling operations of three different scales, 1×1, 3×3, and 5×5, on the input feature map to extract three groups of features with different ratios at different scales.

[0088] 3) Convolve the features of different scales with a Depthwise Separable Convolution layer respectively. This convolution layer is a lightweight convolution operation. Since the pyramid has a three-layer structure and three ratios of features need to be fused, in order to maintain the weight of the global features, the dimension of the context representation needs to be reduced to 1 / 3 of the original, and three groups of features with different scales are extracted and output.

[0089] 4) Upsample the three groups of features by bilinear interpolation to restore them to the length and width of the input.

[0090] 5) Concatenate the features of different scales through a concatenation operation to obtain the global feature of the Pyramid Pooling, thereby obtaining a comprehensive multi-scale feature representation.

[0091] 6) Place the Recurrent Neural Network, namely the Long Short-Term Memory (LSTM), at the end of the Feed Forward neural network layer, which can allow the model to model the temporal dependencies in the sequence, better capture the long-term dependencies of the sequence, and overcome some limitations of the Transformer itself.

[0092] As shown Figure 4 in the figure, the DS-Decoder includes three sub-layers. The first sub-layer uses a masked multi-headed attention mechanism to prevent the model from seeing the data to be predicted and causing data leakage. The second sub-layer is a multi-head attention mechanism that performs attention calculations on the input of the DS-Encoder. The third sub-layer is a feed-forward neural network layer that is the same as the DS-Encoder, and a Linear layer and a Sigmoid layer are used after its output to predict the prediction probability of the forward value of the corresponding snow reflectivity.

[0093] During training, the mean squared error loss function is selected as the loss function of the Transformer network, and the Adam optimizer is selected as the optimizer.

[0094] As an alternative implementation, after step 23, it further includes:

[0095] Step 24: Use the root mean square error and the mean absolute percentage error to evaluate the performance of the snow reflectivity forward model.

[0096] Specifically, use the test set to evaluate the performance of the snow reflectivity forward model using the root mean square error and the mean absolute percentage error. The test set includes the true values of the snow reflectivity of the snow at multiple test locations.

[0097] Specifically, the calculation formula for the root mean square error is:

[0098]

[0099] where RMSE is the root mean square error; n is the number of test locations in the test set; is the forward value of the snow reflectivity of the snow at the i-th test location in the test set; y i is the true value of the snow reflectivity of the snow at the i-th test location in the test set.

[0100] The calculation formula for the mean absolute percentage error (MAPE) is:

[0101]

[0102] where MAPE is the mean absolute percentage error.

[0103] The ranges of both the root mean square error and the mean absolute percentage error are [0, +∞). When the forward value and the true value are exactly the same, it is equal to 0, that is, perfect forward calculation; the worse the forward calculation effect, the larger the above two error values.

[0104] To verify the actual effect of the present invention, the method of the present invention is applied to a part of the Qinghai-Tibet Plateau region, and the forward modeling results shown in Figure 5 and Figure 6 are obtained, verifying the effectiveness of the method of the present invention.

[0105] The present application also provides an application scenario that applies the above snow albedo forward modeling method. Specifically: The snow albedo forward modeling method provided in this embodiment can be applied to the evaluation of regional snow cover. Obtain the forward modeling parameters of the snow in the target area. Input the forward modeling parameters into the snow albedo forward modeling model to obtain the forward modeling value of the snow albedo in the target area; evaluate the regional snow cover based on the forward modeling value of the snow albedo in the target area.

[0106] In an exemplary embodiment, a computer device is provided, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the computer program to implement the snow albedo forward modeling method in Embodiment 1.

[0107] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the snow albedo forward modeling method in Embodiment 1.

[0108] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, it implements the snow albedo forward modeling method in Embodiment 1.

[0109] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal, and its internal structure diagram can be as shown in Figure 7 . The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store video tag processing data. The input / output interface of the computer device is used for the processor to exchange information with external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements a video tag processing method.

[0110] Those skilled in the art can understand that Figure 7 The structure shown in Figure 7 is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0111] 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 for analysis, stored data, displayed data, 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 relevant data need to comply with relevant regulations.

[0112] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. 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), magnetoresistive 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 be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0113] In each of the embodiments provided in the present application, the database involved may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., without limitation. In each of the embodiments provided in the present application, the processor involved may be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without limitation.

[0114] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, 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, it should be considered as the scope described in this specification.

[0115] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A snow reflectivity forward modeling method, characterized in that: The method comprises: Obtaining forward modeling parameters of target snow; the forward modeling parameters include: snow physical parameters, normalized snow index and observation geometry data, the snow physical parameters include: snow particle size and snow depth, and the observation geometry data include: MODIS solar zenith angle, observation zenith angle, solar azimuth angle and observation azimuth angle; The forward modeling parameters of the target snow are input into the snow reflectivity forward model to obtain the forward modeling value of the snow reflectivity of the target snow; the snow reflectivity forward model is obtained by training the Transformer network using a training data set; the training data set includes the forward modeling parameters of snow at multiple target locations and the true value of snow reflectivity, and the snow reflectivity includes: snow reflectivity in the visible light band and snow reflectivity in the near infrared band; The training process of the snow reflectivity forward model includes: Obtaining the training data set; Constructing the Transformer network; Taking the forward modeling parameters of snow at each target location in the training data set as input and the corresponding true value of snow reflectivity as output, the Transformer network is trained to obtain the snow reflectivity forward model; Acquiring the training data set includes: Obtaining MODIS data at multiple locations; the MODIS data includes: surface reflectance data and observation geometry data; the surface reflectance data includes: visible light band reflectance data and near infrared band reflectance data; The normalized snow cover index at the corresponding location is calculated based on the visible light band reflectance data at each location; Performing snow accumulation identification based on the normalized snow accumulation index of each location to determine the snow accumulation condition at the corresponding location; the snow accumulation condition is the presence or absence of snow accumulation; Screening each location based on the surface reflectance data of each location where snow is present, and determining each screened location as a target location; The surface reflectivity data of each target location is determined as the true value of the snow reflectivity of the corresponding target location; Based on the visible light band reflectance data of each target location and the ART model, the snow particle size at the corresponding target location is obtained; Based on the surface reflectivity data of each target location and the linear regression model, the snow depth at the corresponding target location is obtained; The true values ​​of snow particle size, snow depth and snow reflectivity of snow at each target location are determined as the training data set.

2. The snow reflectivity forward modeling method according to claim 1, characterized in that: The normalized snow cover index at the corresponding location is calculated based on the visible light band reflectance data at each location, including: The normalized snow cover index at the corresponding location is calculated based on the fourth band reflectance data and the shortwave infrared band reflectance data in the visible light band reflectance data at each location.

3. The snow reflectivity forward modeling method according to claim 1, characterized in that: Snow accumulation is identified based on the normalized snow accumulation index at each location to determine the snow accumulation conditions at the corresponding location, including: Snow accumulation conditions at locations where the normalized snow accumulation index is greater than 0.4 are determined to be present.

4. The snow reflectivity forward modeling method according to claim 1, characterized in that: Based on the snow accumulation condition, the surface reflectance data of each location where snow exists are screened, and each screened location is determined as a target location, including: Determine a screening rule based on the surface reflectivity data of each location; the screening rule is that the reflectivity data of the first band, the reflectivity data of the third band and the reflectivity data of the fourth band are all greater than 0.7 and the reflectivity data of the second band is greater than 0.11; the wavelength range corresponding to the reflectivity data of the first band is 620nm~670nm, the wavelength range corresponding to the reflectivity data of the second band is 841nm~876nm, the wavelength range corresponding to the reflectivity data of the third band is 459nm~479nm, and the wavelength range corresponding to the reflectivity data of the fourth band is 620~670nm; Each position whose surface reflectance data satisfies the screening rule is determined as a target position.

5. The snow reflectivity forward modeling method according to claim 1, characterized in that: The Transformer network includes: an encoder module and a decoder module; The encoder module includes: 6 encoders connected in sequence; the encoder includes: a multi-cast self-attention mechanism and a feedforward neural network; The decoder module includes: 6 decoders connected in sequence; the decoder includes: a masked multi-attention mechanism, a multi-attention mechanism and a feedforward neural network.

6. A computer device comprising: 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 reflectivity forward modeling method according to any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the snow reflectivity forward modeling method described in any one of claims 1 to 5 is implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the snow reflectivity forward modeling method described in any one of claims 1 to 5 is implemented.

Images