Plateau region temperature and humidity profile inversion method and system based on airborne temperature and humidity profile instrument
By combining localized datasets and various optimization techniques, a high-precision temperature and humidity profile inversion model suitable for the Qinghai-Tibet Plateau was constructed, solving the accuracy and applicability issues of general algorithms in special environments and achieving efficient inversion in plateau regions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CMA METEOROLOGICAL OBSERVATION CENT
- Filing Date
- 2026-03-05
- Publication Date
- 2026-04-24
AI Technical Summary
Existing general inversion algorithms suffer from low accuracy and large systematic biases when applied in special environments such as the Qinghai-Tibet Plateau, making it difficult to meet the accuracy requirements of meteorological operations and scientific research.
By combining a pre-trained general atmospheric temperature and humidity profile inversion model with a localized synchronous observation small sample dataset, and through a phased freeze-thaw fine-tuning strategy, an SNR/sensitivity adaptive weighted loss function, physical consistency regularization constraints and joint estimation of surface emissivity, a high-precision and robust localized inversion model is constructed.
It significantly improves inversion accuracy, suppresses drift in inversion results, enhances the model's generalization ability to complex surfaces, and solves the application challenges of remote sensing algorithms in areas with sparse data and special environments.
Smart Images
Figure CN121920239A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of atmospheric remote sensing technology, and in particular to a method and system for inverting temperature and humidity profiles in plateau regions based on an airborne temperature and humidity profiler. Background Technology
[0002] Atmospheric temperature and humidity profiles are key physical quantities characterizing atmospheric thermodynamics and the vertical structure of water vapor. They directly affect the radiation balance of the land-atmosphere system and the evolution of weather processes. Therefore, obtaining continuous and accurate temperature and humidity profile data is crucial for improving the accuracy of weather forecasts, strengthening disaster early warning capabilities, and deepening the understanding of atmospheric physical processes.
[0003] Currently, atmospheric temperature and humidity profile detection methods mainly fall into two categories: in-situ detection and remote sensing. In-situ detection, represented by weather balloons, can provide accurate data with high vertical resolution, but its spatiotemporal resolution is low and its cost is high, making it difficult to meet the high timeliness requirements of modern meteorological operations. Although satellite remote sensing has a wide coverage, it is easily interfered with in cloudy and rainy weather, and its detection accuracy for the lower atmosphere, such as the boundary layer, is limited. Ground-based microwave radiometers, as an important remote sensing device, can provide continuous observations with high temporal resolution, and their inversion algorithms have been extensively studied, but their detection range is limited to the airspace above fixed stations. These detection methods all face severe challenges in the complex terrain and sparsely populated Qinghai-Tibet Plateau region, resulting in a particularly scarce vertical profile data in this area. This severely restricts the research on the occurrence and development mechanisms of plateau weather systems and the improvement of forecasting capabilities.
[0004] To overcome the limitations of the aforementioned detection methods, airborne microwave radiometers (or airborne temperature and humidity profilers), as a mobile and flexible remote sensing platform, can achieve high-resolution detection of any region and have become an important tool for filling data gaps in atmospheric profile detection. However, the quality of data produced by airborne temperature and humidity profilers is highly dependent on the performance of their inversion algorithms. Currently, most mainstream inversion algorithms are based on statistical models such as neural networks. These models are typically trained under global or regional average atmospheric conditions, aiming to establish a universal "brightness temperature-profile" inversion relationship.
[0005] When this general algorithm is directly applied to the unique environment of the Qinghai-Tibet Plateau, its applicability faces severe challenges, as preliminary application studies have shown. This is mainly due to the significant differences between the unique physical environment of the plateau and the global average, causing the assumptions of the pre-trained model to no longer be fully applicable. Specific challenges include: 1) Complex underlying surface and emissivity uncertainties: The plateau has diverse surface types, with huge and dynamic differences in microwave emissivity, which can severely pollute the inversion signal as background noise; 2) Unique atmospheric background: The plateau has a high average altitude, low air pressure, and thin air, which significantly weakens the pressure broadening effect of atmospheric absorption lines and changes the shape of the weighting function; 3) Extremely low water vapor content: The plateau atmosphere is usually extremely dry with very low absolute humidity, resulting in very weak signals from water vapor absorption channels and low signal-to-noise ratios, posing fundamental difficulties for humidity inversion and even causing channel saturation problems; 4) Representativeness bias of the training dataset: In the global training dataset, the proportion of samples that can reflect the unique atmospheric stratification characteristics of the plateau (such as strong near-surface radiation inversion in winter and deep convective boundary layers in the afternoon in summer) is severely insufficient, causing the general model to fail to fully learn the physical laws of the plateau region.
[0006] In summary, directly applying general inversion algorithms trained on global data leads to severe systematic biases (such as systematic cold bias) and large random errors in high-altitude regions, making it difficult to meet the accuracy requirements of meteorological operations and scientific research. Therefore, there is an urgent need for a technical solution that can efficiently adapt general inversion algorithms to specific regions such as high-altitude areas to overcome the limitations of existing algorithms and significantly improve the inversion accuracy of airborne temperature and humidity profilers in special environments. Summary of the Invention
[0007] This disclosure provides a method and system for inverting temperature and humidity profiles in plateau regions based on an airborne temperature and humidity profiler, which solves the technical problems of low accuracy and large systematic deviations caused by the unique environment when general inversion algorithms are applied to special areas such as plateaus.
[0008] According to a first aspect of this disclosure, a method for inverting temperature and humidity profiles in plateau regions based on an airborne temperature and humidity profiler is provided. The method includes: Brightness temperature data for plateau regions were obtained using an airborne temperature and humidity profiler. The brightness temperature data of the plateau region is input into a pre-trained plateau region atmospheric temperature and humidity profile inversion model to obtain the corresponding atmospheric temperature and humidity profile. The pre-trained atmospheric temperature and humidity profile inversion model for plateau regions is obtained through the following steps: Brightness temperature data of the plateau region was obtained using an airborne temperature and humidity profiler, and atmospheric temperature and humidity profile data of the corresponding region were obtained simultaneously using a drop-in radiosonde. The plateau region brightness temperature data and the atmospheric temperature and humidity profile data were spatiotemporally matched to construct a synchronous matching dataset. Obtain a pre-trained general atmospheric temperature and humidity profile inversion model; The general atmospheric temperature and humidity profile inversion model is trained using the synchronous matching dataset to obtain a pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
[0009] In addition to the aspects and any possible implementations described above, a further implementation is provided in which training the general atmospheric temperature and humidity profile inversion model using the synchronous matching dataset includes: The general atmospheric temperature and humidity profile inversion model is based on a neural network and has been trained to parameter convergence on a global atmospheric temperature and humidity profile dataset. Using the network parameters of the general atmospheric temperature and humidity profile inversion model as initial parameters, the model is iteratively fine-tuned using the synchronous matching dataset until the independent evaluation index meets the preset performance threshold, thus obtaining the pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
[0010] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the fine-tuning employs a phased fine-tuning strategy, including: Freeze the network parameters from the input layer to the hidden layer, update only the output layer parameters, and train using the first preset learning rate; Unfreeze the network parameters of the input layer and hidden layer, and fine-tune the entire network with a second preset learning rate; wherein the second preset learning rate is less than the first preset learning rate.
[0011] As described above and in any possible implementation, a further implementation is provided in which the fine-tuning training employs a joint loss function, which is: in, For joint losses, The loss is an adaptive hierarchical weighted loss based on SNR / sensitivity. For physical consistency regularization, For joint estimation of surface emissivity and prior constraints, These are the weight parameters.
[0012] As described above and in any possible implementation, a further implementation is provided in which the SNR / sensitivity adaptive hierarchical weighted loss is: Among them, weight For the first Error weights for each height layer For the first Each height level corresponds to the signal-to-noise ratio of the microwave channel. For the first Each height level corresponds to the signal-to-noise ratio of the microwave channel. For the first Sensitivity factors at each height level, For the first Sensitivity factors at each height level, For hyperparameters, This represents the total number of height layers in the temperature and humidity profile. For the first Predicted temperature and humidity values for each altitude level For the first True values of temperature and humidity at each altitude level.
[0013] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the physical consistency regularization term is: in, This represents the total number of microwave channels. For the forward model of radiative transfer in the channel Brightness temperature simulation, For surface emissivity, Indicates in the channel Actual observed brightness temperature This is the prediction vector for the temperature and humidity profile.
[0014] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the joint estimation of surface emissivity and prior constraints are: in, It is the prior emissivity obtained from the surface type library. μ It is the constraint strength. It represents the surface emissivity.
[0015] According to a second aspect of this disclosure, a system for inverting temperature and humidity profiles in high-altitude areas based on an airborne temperature and humidity profiler is provided. The system includes: The acquisition module is used to acquire brightness temperature data for plateau regions based on an airborne temperature and humidity profiler. The inversion module is used to input the brightness temperature data of the plateau region into a pre-trained plateau region atmospheric temperature and humidity profile inversion model to obtain the corresponding atmospheric temperature and humidity profile. The pre-trained atmospheric temperature and humidity profile inversion model for plateau regions is obtained through the following steps: Brightness temperature data of the plateau region was obtained using an airborne temperature and humidity profiler, and atmospheric temperature and humidity profile data of the corresponding region were obtained simultaneously using a drop-in radiosonde. The plateau region brightness temperature data and the atmospheric temperature and humidity profile data were spatiotemporally matched to construct a synchronous matching dataset. Obtain a pre-trained general atmospheric temperature and humidity profile inversion model; The general atmospheric temperature and humidity profile inversion model is trained using the synchronous matching dataset to obtain a pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
[0016] According to a third aspect of this disclosure, an electronic device is provided. The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.
[0017] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the methods according to the first and / or second aspects of this disclosure.
[0018] This disclosure is based on a pre-trained general atmospheric temperature and humidity profile inversion model, combined with a localized synchronous observation small sample dataset, and introduces key techniques such as a staged freeze-thaw fine-tuning strategy, an SNR / sensitivity adaptive weighted loss function, physical consistency regularization constraints, and joint estimation of surface emissivity. It constructs a high-precision, robust, and localized inversion model applicable to the target plateau region, thereby significantly improving the inversion accuracy. Physical constraints suppress the drift of inversion results, and joint emissivity estimation enhances the model's generalization ability to complex surfaces. This provides an effective technical system for solving the problem of localized application of remote sensing algorithms in areas with sparse data and special environments.
[0019] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description
[0020] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. The drawings are provided for a better understanding of the invention and are not intended to limit the scope of this disclosure. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein: Figure 1 A flowchart is shown for a method for inverting temperature and humidity profiles in high-altitude areas based on an airborne temperature and humidity profiler, according to an embodiment of the present disclosure. Figure 2 A flowchart illustrating the acquisition of a pre-trained atmospheric temperature and humidity profile inversion model for plateau regions according to an embodiment of the present disclosure is shown. Figure 3 A block diagram of a plateau temperature and humidity profile inversion system based on an airborne temperature and humidity profiler, according to an embodiment of the present disclosure, is shown. Figure 4 A block diagram of an exemplary electronic device capable of implementing embodiments of the present disclosure is shown. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.
[0022] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.
[0023] This disclosure is based on a pre-trained general atmospheric temperature and humidity profile inversion model, combined with a localized synchronous observation small sample dataset, and introduces key techniques such as a staged freeze-thaw fine-tuning strategy, an SNR / sensitivity adaptive weighted loss function, physical consistency regularization constraints, and joint estimation of surface emissivity. It constructs a high-precision, robust, and localized inversion model applicable to the target plateau region, thereby significantly improving the inversion accuracy. Physical constraints suppress the drift of inversion results, and joint emissivity estimation enhances the model's generalization ability to complex surfaces. This provides an effective technical system for solving the problem of localized application of remote sensing algorithms in areas with sparse data and special environments.
[0024] Figure 1 A flowchart of a method 100 for inverting temperature and humidity profiles in high-altitude areas based on an airborne temperature and humidity profiler, according to an embodiment of this disclosure, is shown. Figure 1 As shown, method 100 includes: S110 acquires brightness temperature data for high-altitude regions based on an airborne temperature and humidity profiler.
[0025] In some embodiments, an airborne temperature and humidity profiler with multi-band microwave receiving capability, high-stability temperature control, and resistance to high-altitude low-temperature and low-pressure characteristics is used, with a UAV as the airborne platform, to acquire brightness temperature data in high-altitude areas.
[0026] S120 inputs the brightness temperature data of the plateau region into the pre-trained plateau region atmospheric temperature and humidity profile inversion model to obtain the corresponding atmospheric temperature and humidity profile.
[0027] like Figure 2 As shown, the pre-trained atmospheric temperature and humidity profile inversion model for plateau regions is obtained through the following steps: S121. Brightness temperature data of the plateau region is obtained using an airborne temperature and humidity profiler, and the true atmospheric temperature and humidity profile data of the corresponding region is obtained simultaneously using a drop-in radiosonde. The brightness temperature data of the plateau region and the true atmospheric temperature and humidity profile data are spatiotemporally matched to construct a synchronous matching dataset.
[0028] Specifically, using a drone as the airborne platform, a coordinated mode of grid-like flight and fixed-point radiosonde deployment is adopted. That is, when the airborne platform reaches the center point of each grid, one drop-type radiosonde is deployed simultaneously (the deployment interval matches the grid flight cycle, about 30 minutes / time). This ensures that the horizontal distance between the deployment position of the radiosonde and the brightness temperature collection position of the airborne temperature and humidity profiler is ≤10km. After being deployed from the airborne platform, the radiosonde collects temperature, humidity and pressure data at 50m vertical intervals as it descends with the airflow, and transmits the data back to the ground receiving station in real time via wireless signal.
[0029] In some embodiments, during the acquisition process, the airborne central control unit and the ground receiving station synchronously perform real-time quality screening, convert the format of the acquired brightness temperature raw data and complete outliers, and uniformly interpolate the non-uniform vertical interval temperature and humidity data acquired by the radiosonde to the same vertical resolution as the target height layer for brightness temperature data inversion, and remove true data that do not conform to the laws of atmospheric physics.
[0030] In some embodiments, based on the radiosonde deployment time, brightness temperature data collected by the airborne temperature and humidity profiler within 5 minutes before and after deployment are selected to form a temporal candidate brightness temperature set. The horizontal distance between each brightness temperature collection point in the temporal candidate brightness temperature set and the radiosonde deployment point is calculated, and brightness temperature data with a distance ≤10km are retained to form a spatiotemporal candidate brightness temperature set. If multiple brightness temperature data exist in the spatiotemporal candidate brightness temperature set, the brightness temperature data closest to the radiosonde deployment time is selected as the matching object to construct a synchronization matching dataset. ,in The brightness temperature vector, This is the true value vector corresponding to the temperature / humidity profile.
[0031] S122, Obtain the pre-trained general atmospheric temperature and humidity profile inversion model.
[0032] S123, using the synchronous matching dataset, the general atmospheric temperature and humidity profile inversion model is trained to obtain a pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
[0033] In some embodiments, the general atmospheric temperature and humidity profile inversion model is based on a neural network and has been trained to parameter convergence on a global atmospheric temperature and humidity profile dataset; Using the network parameters of a general atmospheric temperature and humidity profile inversion model as initial parameters, the model is iteratively fine-tuned using a synchronous matching dataset until the independent evaluation index meets the preset performance threshold, thus obtaining a pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
[0034] Specifically, the general atmospheric temperature and humidity profile inversion model is built on a three-layer BP neural network. This model contains a fully connected architecture of input layer, hidden layer, and output layer. Input layer, containing There are 1 node, and each node corresponds to 1 microwave channel brightness temperature data from an airborne temperature and humidity profiler. The input vector is... ; Hidden layer, containing Each node maps the original brightness temperature signal of the input layer to intermediate features characterizing the vertical structure of the atmosphere, and the activation function is the Sigmoid function. —This function enhances the model's ability to analyze weak signals in a high-altitude, low-moisture environment through nonlinear transformation, avoiding the limitations of linear models in fitting complex atmospheric signals; Output layer, containing There are 10 nodes, each corresponding to the temperature and humidity data of one altitude layer. The output vector is... ; The model has been trained using a global atmospheric temperature and humidity profile dataset and has achieved parameter convergence. The converged network parameters are denoted as follows: ,in For input layer to hidden layer Weight matrix, From hidden layer to output layer Weight matrix, and These are the bias vectors for the hidden layer and the output layer, respectively.
[0035] In some embodiments, the network parameters of a general atmospheric temperature and humidity profile inversion model are used as initial parameters, and a synchronous matching dataset is utilized. Multiple rounds of iterative fine-tuning were carried out.
[0036] In some embodiments, a fine-tuning strategy for high-altitude environments is adopted, employing a phased fine-tuning strategy, including: Freeze the network parameters from the input layer to the hidden layer, update only the output layer parameters, and train using the first preset learning rate; Unfreeze the network parameters of the input layer and hidden layer, and fine-tune the entire network with a second preset learning rate; wherein the second preset learning rate is less than the first preset learning rate.
[0037] Specifically, to avoid catastrophic forgetting and overfitting under small sample conditions, a two-stage training method is adopted: Phase 1: Freeze all connection parameters between the input layer and the hidden layer, i.e., the weight matrix. and bias vector Maintain the initial parameters of the pre-trained general model It remains unchanged and does not participate in subsequent gradient calculations and parameter updates; only the connection parameters between the hidden layer and the output layer are optimized and updated, including the weight matrix. and bias vector A larger learning rate is used to achieve fast convergence of the output layer parameters; the specific value range is [range to be filled in]. Based on synchronous matching dataset The training continues for 10-50 epochs. In each iteration, the predicted profile is calculated through forward propagation, and the loss value is calculated in conjunction with the loss function, only for the weight matrix. and bias vector Backpropagation gradient updates are performed until the output layer parameters initially converge, aiming to quickly adapt to the model's output space. Phase II: Building upon Phase I, the feature extraction capabilities of the hidden layers are finely optimized. The parameters between the input and hidden layers are unfrozen, allowing all network parameters to participate in gradient calculation and updates. This ensures that the parameters of each layer can adapt to the plateau features. A second preset learning rate is adopted, which is lower than the first preset learning rate, to avoid significant oscillations in the already converged output layer parameters, while simultaneously ensuring fine-tuning of the hidden layer parameters. The specific value range is as follows: The entire network is jointly trained end-to-end. In each iteration, the predicted profile is calculated through forward propagation, and the total loss is calculated by combining the loss function. The final loss is calculated through backpropagation. The gradient, and according to the second preset learning rate. All parameters were updated to fine-tune the network's ability to extract features from general to localized features.
[0038] In some embodiments, the matching dataset will be synchronized. The dataset is divided into training and validation sets in a 7:3 ratio to avoid overfitting of the training set. Mini-batch processing of data with sizes of 32–128 is used to balance training efficiency and parameter update stability. The maximum number of iterations is set to 200 to avoid overtraining.
[0039] In some embodiments, in each iteration, training set samples are input into the model for forward propagation to obtain predicted values for temperature and humidity profiles.
[0040] Specifically, the brightness temperature vector Input the network and calculate the predicted output. .
[0041] Calculate each node in the hidden layer (based on node). Net input (for example) and output : in, It is a weight matrix elements, It is a bias vector elements, It is a hidden layer activation function; Calculate the output layer nodes (in terms of nodes) Net input (for example) and output : in, It is a weight matrix elements, It is a bias vector elements, It is the output layer activation function (linear function: Finally, the predicted vector is obtained. .
[0042] In some embodiments, fine-tuning training employs a joint loss function, which is: in, For joint losses, The loss is an adaptive hierarchical weighted loss based on SNR / sensitivity. For physical consistency regularization, For joint estimation of surface emissivity and prior constraints, These are the weight parameters.
[0043] The adaptive hierarchical weighted loss for SNR / sensitivity is: Among them, weight For the first Error weights for each height layer For the first Each height level corresponds to the signal-to-noise ratio of the microwave channel. For the first Each height level corresponds to the signal-to-noise ratio of the microwave channel. For the first Sensitivity factors at each height level, For the first Sensitivity factors at each height level, For hyperparameters, This represents the total number of height layers in the temperature and humidity profile. For the first Predicted temperature and humidity values for each altitude level For the first True values of temperature and humidity at each altitude level.
[0044] In some embodiments, a weight lower bound is applied to the humidity-dependent layer. (like This prevents the model from completely abandoning its learning of humidity under extremely dry conditions.
[0045] In some embodiments, to ensure the physical plausibility of the inversion results, the forward model residuals are introduced as a constraint, which mandates that the inversion profile be derived from the inversion model. Simulated brightness temperature It should be consistent with the actual observed brightness temperature Consistent; The physical consistency regularization term is: in, This represents the total number of microwave channels. For the forward model of radiative transfer in the channel Brightness temperature simulation, For surface emissivity, This represents the actual observed brightness temperature in channel c. The temperature and humidity profile prediction vector is the surface emissivity. It is adjusted by the weighting coefficient λ∈[0.05,0.5].
[0046] In some embodiments, to decouple atmospheric and surface signals, the surface emissivity is... As an additional output or parameter to be optimized in the network, prior constraints based on land surface type (LC) are added to prevent ill-conditioned solutions. The joint estimation of surface emissivity and the prior constraints are as follows: in, It is the prior emissivity obtained from the surface type library. μ It is the constraint strength. For the surface emissivity, the emissivity ϵ can be jointly optimized with the profile parameters, or alternately optimized every r epochs (e.g., r=1~5).
[0047] In some embodiments, the prediction error is calculated using the joint loss function described above, and finally, the error is calculated against all network parameters through backpropagation. gradient And update the network parameters based on the learning rate of the corresponding stage. After fine-tuning training stops, save the finally converged network parameters. A pre-trained inversion model of atmospheric temperature and humidity profiles in plateau regions was obtained.
[0048] According to embodiments of this disclosure, a phased freeze-thaw fine-tuning strategy is adopted to achieve efficient model transfer on small-sample localized datasets, solving the problem of poor plateau adaptability of general models. Through SNR / sensitivity adaptive weighted loss, the constraint of channel saturation on humidity inversion is effectively alleviated. By utilizing the overall optimization of the joint loss function, the output temperature and humidity profiles are ensured to conform to the atmospheric thermodynamics and radiative transfer characteristics of plateau, suppressing result drift, solving the interference of uncertain emissivity of complex underlying surfaces, improving inversion accuracy and stability, enhancing the plateau application value of airborne detection methods, and filling data gaps.
[0049] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this disclosure is not limited to the described order of actions, because according to this disclosure, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this disclosure.
[0050] The above is an introduction to the method embodiments. The following describes the present disclosure further through device embodiments.
[0051] Figure 3 A block diagram of a plateau region temperature and humidity profile inversion system 300 based on an airborne temperature and humidity profiler, according to an embodiment of this disclosure, is shown. Figure 3 As shown, the device 300 includes: The acquisition module 301 is used to acquire brightness temperature data of plateau areas based on an airborne temperature and humidity profiler; The inversion module 302 is used to input the brightness temperature data of the plateau region into the pre-trained plateau region atmospheric temperature and humidity profile inversion model to obtain the corresponding atmospheric temperature and humidity profile. The pre-trained atmospheric temperature and humidity profile inversion model for plateau regions is obtained through the following steps: Brightness temperature data of the plateau region was obtained using an airborne temperature and humidity profiler, and atmospheric temperature and humidity profile data of the corresponding region were obtained simultaneously using a drop-in radiosonde. Spatiotemporal matching was performed on the brightness temperature data of the plateau region and the atmospheric temperature and humidity profile data to construct a synchronous matching dataset. Obtain a pre-trained general atmospheric temperature and humidity profile inversion model; The general atmospheric temperature and humidity profile inversion model was trained using a synchronous matching dataset to obtain a pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
[0052] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0053] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.
[0054] Figure 4 A schematic block diagram of an electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.
[0055] Electronic device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in ROM 402 or a computer program loaded into RAM 403 from storage unit 408. RAM 403 can also store various programs and data required for the operation of electronic device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. I / O interface 405 is also connected to bus 404.
[0056] Multiple components in electronic device 400 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of displays, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0057] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as method 100. For example, in some embodiments, method 100 may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by the computing unit 401, one or more steps of method 100 described above may be performed. Alternatively, in other embodiments, the computing unit 401 may be configured to perform method 100 by any other suitable means (e.g., by means of firmware).
[0058] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0059] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0060] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0061] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including voice input, speech input, or tactile input).
[0062] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.
[0063] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.
[0064] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.
[0065] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.
Claims
1. A method for inverting temperature and humidity profiles in plateau regions based on an airborne temperature and humidity profiler, characterized in that, include: Brightness temperature data for plateau regions were obtained using an airborne temperature and humidity profiler. The brightness temperature data of the plateau region is input into a pre-trained plateau region atmospheric temperature and humidity profile inversion model to obtain the corresponding atmospheric temperature and humidity profile. The pre-trained atmospheric temperature and humidity profile inversion model for plateau regions is obtained through the following steps: Brightness temperature data of the plateau region was obtained using an airborne temperature and humidity profiler, and atmospheric temperature and humidity profile data of the corresponding region were obtained simultaneously using a drop-in radiosonde. The plateau region brightness temperature data and the atmospheric temperature and humidity profile data were spatiotemporally matched to construct a synchronous matching dataset. Obtain a pre-trained general atmospheric temperature and humidity profile inversion model; The general atmospheric temperature and humidity profile inversion model is trained using the synchronous matching dataset to obtain a pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
2. The method according to claim 1, characterized in that, The step of training the general atmospheric temperature and humidity profile inversion model using the synchronous matching dataset includes: The general atmospheric temperature and humidity profile inversion model is based on a neural network and has been trained to parameter convergence on a global atmospheric temperature and humidity profile dataset. Using the network parameters of the general atmospheric temperature and humidity profile inversion model as initial parameters, the model is iteratively fine-tuned using the synchronous matching dataset until the independent evaluation index meets the preset performance threshold, thus obtaining the pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
3. The method according to claim 2, characterized in that, The fine-tuning adopts a phased fine-tuning strategy, including: Freeze the network parameters from the input layer to the hidden layer, update only the output layer parameters, and train using the first preset learning rate; Unfreeze the network parameters of the input layer and hidden layer, and fine-tune the entire network with a second preset learning rate; wherein the second preset learning rate is less than the first preset learning rate.
4. The method according to claim 3, characterized in that, The fine-tuning training employs a joint loss function, which is: in, For joint losses, The loss is an adaptive hierarchical weighted loss based on SNR / sensitivity. For physical consistency regularization, For joint estimation of surface emissivity and prior constraints, These are the weight parameters.
5. The method according to claim 4, characterized in that, The SNR / sensitivity adaptive hierarchical weighted loss is: Among them, weight For the first Error weights for each height layer For the first Each height level corresponds to the signal-to-noise ratio of the microwave channel. For the first Each height level corresponds to the signal-to-noise ratio of the microwave channel. For the first Sensitivity factors at each height level, For the first Sensitivity factors at each height level, For hyperparameters, This represents the total number of height layers in the temperature and humidity profile. For the first Predicted temperature and humidity values for each altitude level For the first True values of temperature and humidity at each altitude level.
6. The method according to claim 4, characterized in that, The physical consistency regularization term is: in, This represents the total number of microwave channels. For the forward model of radiative transfer in the channel Brightness temperature simulation, For surface emissivity, Indicates in the channel Actual observed brightness temperature This is the prediction vector for the temperature and humidity profile.
7. The method according to claim 4, characterized in that, The joint estimation of surface emissivity and the prior constraints are as follows: in, It is the prior emissivity obtained from the surface type library. μ It is the constraint strength. It represents the surface emissivity.
8. A plateau region temperature and humidity profile inversion system based on an airborne temperature and humidity profiler, characterized in that, include: The acquisition module is used to acquire brightness temperature data for plateau regions based on an airborne temperature and humidity profiler. The inversion module is used to input the brightness temperature data of the plateau region into a pre-trained plateau region atmospheric temperature and humidity profile inversion model to obtain the corresponding atmospheric temperature and humidity profile. The pre-trained atmospheric temperature and humidity profile inversion model for plateau regions is obtained through the following steps: Brightness temperature data of the plateau region was obtained using an airborne temperature and humidity profiler, and atmospheric temperature and humidity profile data of the corresponding region were obtained simultaneously using a drop-in radiosonde. The plateau region brightness temperature data and the atmospheric temperature and humidity profile data were spatiotemporally matched to construct a synchronous matching dataset. Obtain a pre-trained general atmospheric temperature and humidity profile inversion model; The general atmospheric temperature and humidity profile inversion model is trained using the synchronous matching dataset to obtain a pre-trained atmospheric temperature and humidity profile inversion model for plateau regions.
9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1-7.
10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method described in any one of claims 1-7.