Global land surface radiation budget full-factor collaborative inversion method and system of full-time space scale

Abstract

The present application belongs to the technical field of remote sensing, and relates to a global land surface radiation budget full-factor collaborative inversion method and system at all space-time scales. The method comprises: constructing a machine learning model, taking remote sensing data, meteorological reanalysis data, terrain data and sky view factor as inputs, taking global land surface radiation budget full factors as outputs, and setting a cost function to obtain a global land surface radiation budget full-factor collaborative inversion model at all space-time scales; inputting the remote sensing data, meteorological reanalysis data, terrain data and sky view factor into the global land surface radiation budget full-factor collaborative inversion model at all space-time scales to obtain target global land surface radiation budget full factors at all space-time scales. The present application can realize collaborative inversion of global land surface radiation budget full factors at all space-time scales, so that radiation budget balance is realized between different components.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing technology, specifically, it relates to a method and system for the coordinated inversion of all elements of global surface radiation balance at all time and space scales. Background Technology

[0002] The energy incident, reflected, absorbed, and emitted by the Earth system constitutes the Earth's radiation balance. The surface radiation balance represents the balance between incident and emitted radiation at the Earth's surface, and incident solar radiation includes both direct and diffuse forms. The total solar radiation budget includes direct solar radiation, diffuse solar radiation, total solar radiation incident on the surface, direct photosynthetically active radiation, diffuse photosynthetically active radiation, total photosynthetically active radiation, solar radiation emitted from the surface, incident longwave radiation, emitted longwave radiation, net shortwave radiation, net longwave radiation, and net surface radiation. The sum of direct solar radiation and diffuse solar radiation is called total solar radiation incident on the surface. The sum of direct photosynthetically active radiation and diffuse photosynthetically active radiation is called total photosynthetically active radiation. Total photosynthetically active radiation is 0.45 times the total solar radiation incident on the surface. The difference between total solar radiation incident on the surface and solar radiation emitted from the surface is called net shortwave radiation. The difference between emitted longwave radiation and incident longwave radiation is called net longwave radiation. The sum of net shortwave radiation and net longwave radiation is called net surface radiation. The Earth's surface radiation budget is crucial to the global energy balance, dominating surface temperature distribution, the allocation of latent and sensible heat, global climate models, and global hydrological processes. Accurate inversion of all elements of the global surface radiation budget is of great significance for global radiation balance, energy balance, water cycle, and carbon budget. However, current remote sensing inversions of the components of the global surface radiation budget are mostly limited to independent modeling and estimation of individual component variables. This situation makes it difficult to achieve radiation budget balance between different components in different remote sensing products.

[0003] Existing remote sensing inversion methods for each component of the global radiation budget all construct inversion models for each component independently. For example, GLASS (Global Land Surface Satellite), Hi-GLASS (High-Resolution Global Land-Atmosphere System Simulator), and FLUXCOM (Flux Comparison) products all construct remote sensing inversion models for net radiation separately. Similarly, BESS (Biophysical and Environmental Science Satellite) products construct remote sensing inversion models for incident shortwave radiation separately. Some studies construct remote sensing inversion models for incident shortwave radiation, incident longwave radiation, or incident shortwave net radiation separately. The products produced by existing technologies have significant uncertainties, resulting in inconsistencies among the components of the current global surface radiation budget and making it difficult to achieve radiation budget balance. This not only introduces significant uncertainties into the global radiation budget and its relationship with climate change, turbulent heat flux, and other changes and attribution studies, but also interferes with research on extreme weather events and regional heat island effects, increasing the difficulty of formulating measures to address regional climate change. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for the coordinated inversion of all elements of global surface radiation budget across all time and space scales.

[0005] Firstly, this invention provides a method for the coordinated inversion of all elements of global surface radiation budget across all time and space scales, including:

[0006] A machine learning model was constructed, taking remote sensing data, meteorological reanalysis data, topographic data and sky horizon factors as inputs, and global surface radiation budget elements as outputs. A cost function was set to obtain a collaborative inversion model of all elements of surface radiation budget at all time and space scales.

[0007] Remote sensing data, meteorological reanalysis data, topographic data, and sky horizon factors are input into the spatiotemporal scale-based surface radiation budget collaborative inversion model to obtain the target spatiotemporal scale-based surface radiation budget.

[0008] Secondly, this invention provides a global surface radiation balance and budget full-element collaborative inversion system at all spatiotemporal scales, including a model building unit and a data processing unit.

[0009] The model building unit is used to build a machine learning model. It takes remote sensing data, meteorological reanalysis data, topographic data and sky view factor as input, and global surface radiation budget elements as output. It sets a cost function to obtain a collaborative inversion model of surface radiation budget elements at all time and space scales.

[0010] The output unit is used to input remote sensing data, meteorological reanalysis data, topographic data and sky view factor into the full spatiotemporal scale surface radiation budget collaborative inversion model to obtain the full spatiotemporal scale surface radiation budget of the target.

[0011] Based on the above technical solution, the present invention can be further improved as follows.

[0012] Furthermore, the global surface radiation balance includes all elements of direct solar radiation, diffuse solar radiation, total solar radiation incident on the surface, direct photosynthetically active radiation, diffuse photosynthetically active radiation, total photosynthetically active radiation, solar radiation emitted from the surface, longwave radiation incident on the surface, longwave radiation emitted from the surface, net shortwave radiation from the surface, net longwave radiation from the surface, and net surface radiation.

[0013] Furthermore, based on station observation data, remote sensing data, meteorological reanalysis data, topographic data, and sky view factor, a driving dataset for the full-temporal and spatial scale surface radiation budget co-inversion model is constructed. The constructed full-temporal and spatial scale surface radiation budget co-inversion model is trained using the driving dataset to determine the optimal model parameters and obtain the optimal full-temporal and spatial scale surface radiation budget co-inversion model.

[0014] Furthermore, data required for the coordinated inversion of all elements of global surface radiation balance will be collected, including station observation data, remote sensing data, meteorological reanalysis data, topographic data, and sky view factor; topographic data includes digital elevation values, slope, and aspect.

[0015] Preprocessing and filtering of station observation data yielded effective observation data for radiation stations. This effective observation data included direct solar radiation observation data, diffuse solar radiation observation data, direct photosynthetically active radiation observation data, diffuse photosynthetically active radiation observation data, surface emitted solar radiation observation data, surface incident longwave radiation observation data, and surface emitted longwave radiation observation data. Remote sensing data included top-of-atmosphere solar radiation data, soil moisture, normalized difference vegetation index (NDVI), leaf area index, and vegetation cover. Meteorological reanalysis data included air temperature, surface temperature, and relative humidity.

[0016] Based on the effective observation data of radiation stations, the observation data of total incident solar radiation, total photosynthetically active radiation, net shortwave radiation, net longwave radiation, and net radiation at the Earth's surface are calculated. Combined with the effective observation data of radiation stations, the surface radiation observation data are obtained.

[0017] The collected remote sensing data and meteorological reanalysis data are preprocessed to obtain preprocessed remote sensing data and preprocessed meteorological reanalysis data.

[0018] Calculate digital elevation values, slope, aspect, and sky view factor based on digital elevation model data at several resolutions.

[0019] Furthermore, based on the effective observation data from radiation stations, the observation data of total solar radiation incident at the Earth's surface, total photosynthetically active radiation, net shortwave radiation at the Earth's surface, net longwave radiation at the Earth's surface, and net radiation at the Earth's surface are calculated. The observation data of total solar radiation at the Earth's surface is the sum of the observation data of direct solar radiation and the observation data of diffuse solar radiation. The observation data of total solar radiation at the Earth's surface is the sum of the observation data of direct solar radiation and the observation data of diffuse solar radiation. The observation data of total solar radiation incident at the Earth's surface is the sum of the observation data of direct solar radiation and the observation data of diffuse solar radiation. The observation data of net shortwave radiation at the Earth's surface is the sum of the observation data of net solar radiation incident at the Earth's surface and the observation data of net longwave radiation at the Earth's surface.

[0020] Furthermore, the machine learning model can be any one of the following: random forest model, gradient boosting tree model, deep neural network model, and convolutional neural network model.

[0021] Furthermore, the driving dataset for the spatiotemporal scale land surface radiation budget co-inversion model is divided into a model training dataset, a test dataset, and a validation dataset. The constructed spatiotemporal scale land surface radiation budget co-inversion model is trained using the model training dataset. The trained spatiotemporal scale land surface radiation budget co-inversion model is evaluated using the test dataset. The hyperparameters of the constructed spatiotemporal scale land surface radiation budget co-inversion model are adjusted using the validation dataset to determine the optimal spatiotemporal scale land surface radiation budget co-inversion model.

[0022] Furthermore, set Represents slope, Represents slope aspect. Represents the sky view factor. Represents digital elevation values. Represents solar radiation at the top of the atmosphere. Represents air temperature. Represents the Earth's surface temperature. Represents relative humidity. Represents soil moisture. Represents the normalized vegetation index, Represents leaf area index, Represents vegetation coverage. Represents data on direct solar radiation observed on the Earth's surface. Represents scattered solar radiation observation data, Represents the total solar radiation incident on the Earth's surface. Represents direct photosynthetically active radiation. Represents scattered photosynthetically active radiation. Represents total photosynthetically active radiation. Represents observational data of solar radiation emitted from the Earth's surface. Represents observational data of incident longwave radiation at the Earth's surface. Represents observational data of longwave radiation emitted from the Earth's surface. Represents net shortwave radiation observation data at the Earth's surface. Represents the observed data of net longwave radiation at the Earth's surface. Represents net surface radiation observation data. A comprehensive inversion model representing all elements of surface radiation balance across all time and space scales. Let the set be represented as follows: The full-scale spatiotemporal scale surface radiation budget collaborative inversion model is expressed as:

[0023]

[0024] .

[0025] Furthermore, set Represents data on direct solar radiation observed on the Earth's surface. Represents scattered solar radiation observation data, Represents the total solar radiation incident on the Earth's surface. Represents observational data of direct photosynthetically active radiation. Data representing the observed scattered photosynthetically active radiation. Represents total photosynthetically active radiation observation data, Represents observational data of solar radiation emitted from the Earth's surface. Represents observational data of incident longwave radiation at the Earth's surface. Represents observational data of longwave radiation emitted from the Earth's surface. Represents net shortwave radiation observation data at the Earth's surface. Represents the observational data of net longwave radiation at the Earth's surface. Represents net surface radiation observation data. This represents an estimated value of direct solar radiation to the Earth's surface. This represents an estimated value of scattered solar radiation. This represents an estimated value of the total solar radiation incident on the Earth's surface. This represents the estimated value of direct photosynthetically active radiation. This represents an estimated value of the scattered photosynthetically active radiation. This represents an estimated value of total photosynthetically active radiation. This represents an estimated value of solar radiation emitted from the Earth's surface. Represents an estimated value of incident longwave radiation at the Earth's surface. This represents an estimated value of longwave radiation emitted from the Earth's surface. Represents the estimated net shortwave radiation at the Earth's surface. This represents an estimated value of net longwave radiation at the Earth's surface. Represents the estimated net surface radiation. for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, If we denote a set, then the cost function is:

[0026]

[0027] The beneficial effects of this invention are as follows: By constructing several types of machine learning models and using the constructed spatiotemporal scale surface radiation budget co-inversion model to drive the dataset for model training, this invention can select the optimal spatiotemporal scale surface radiation budget co-inversion model, which is beneficial to improving the accuracy of spatiotemporal scale surface radiation budget co-inversion. In addition, this invention achieves spatiotemporal scale surface radiation budget co-inversion by collecting observation data, remote sensing data, meteorological reanalysis data, and digital elevation model data, enabling radiation budget balance among different components. Attached Figure Description

[0028] Figure 1 A schematic diagram of the principle of the global surface radiation budget full-element collaborative inversion method at all spatiotemporal scales provided in Embodiment 1 of the present invention;

[0029] Figure 2 This is a schematic diagram of the global surface radiation budget full-element collaborative inversion system provided in Embodiment 2 of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0031] Example 1

[0032] As an example, see the attached document. Figure 1 As shown, to solve the above-mentioned technical problems, this embodiment provides a method for the coordinated inversion of all elements of global surface radiation budget at all time and space scales, including:

[0033] A machine learning model was constructed, taking remote sensing data, meteorological reanalysis data, topographic data and sky horizon factors as inputs, and global surface radiation budget elements as outputs. A cost function was set to obtain a collaborative inversion model of all elements of surface radiation budget at all time and space scales.

[0034] Remote sensing data, meteorological reanalysis data, topographic data, and sky horizon factors are input into the spatiotemporal scale-based surface radiation budget collaborative inversion model to obtain the target spatiotemporal scale-based surface radiation budget.

[0035] In practical applications, the observation data come from global eddy covariance flux observation networks, terrestrial ecosystem flux observation research networks, eco-hydrological remote sensing experimental hydro-meteorological observation networks, multi-scale surface flux and meteorological element observation datasets, integrated surface process observation networks, and scientific observation networks in cold and arid regions.

[0036] Optional, the global surface radiation budget includes all elements of direct solar radiation, diffuse solar radiation, total solar radiation incident on the surface, direct photosynthetically active radiation, diffuse photosynthetically active radiation, total photosynthetically active radiation, solar radiation emitted from the surface, longwave radiation incident on the surface, longwave radiation emitted from the surface, net shortwave radiation from the surface, net longwave radiation from the surface, and net surface radiation.

[0037] Optionally, a driving dataset for the full-temporal and spatial scale surface radiation budget co-inversion model is constructed based on station observation data, remote sensing data, meteorological reanalysis data, topographic data, and sky view factor. The constructed full-temporal and spatial scale surface radiation budget co-inversion model is then trained using the driving dataset to determine the optimal model parameters and obtain the optimal full-temporal and spatial scale surface radiation budget co-inversion model.

[0038] Optionally, collect the data required for the global surface radiation balance full-element coordinated inversion, including station observation data, remote sensing data, meteorological reanalysis data, topographic data and sky view factor; topographic data includes digital elevation values, slope and aspect;

[0039] Preprocessing and filtering of station observation data yielded effective observation data for radiation stations. This effective observation data included direct solar radiation observation data, diffuse solar radiation observation data, direct photosynthetically active radiation observation data, diffuse photosynthetically active radiation observation data, surface emitted solar radiation observation data, surface incident longwave radiation observation data, and surface emitted longwave radiation observation data. Remote sensing data included top-of-atmosphere solar radiation data, soil moisture, normalized difference vegetation index (NDVI), leaf area index, and vegetation cover. Meteorological reanalysis data included air temperature, surface temperature, and relative humidity.

[0040] Based on the effective observation data of radiation stations, the observation data of total incident solar radiation, total photosynthetically active radiation, net shortwave radiation, net longwave radiation, and net radiation at the Earth's surface are calculated. Combined with the effective observation data of radiation stations, the surface radiation observation data are obtained.

[0041] The collected remote sensing data and meteorological reanalysis data are preprocessed to obtain preprocessed remote sensing data and preprocessed meteorological reanalysis data.

[0042] Calculate digital elevation values, slope, aspect, and sky view factor based on digital elevation model data at several resolutions.

[0043] Remote sensing data includes Sentinel series remote sensing data, Landsat series remote sensing data, MODIS series remote sensing products, and GLASS series remote sensing products; the spatiotemporal scales of meteorological reanalysis data include hourly, daily, and monthly scales.

[0044] Typical time scales include hourly, daily, weekly, ten-day, monthly, and yearly scales; typical spatial scales include 10m, 30m, 500m, and 1km; and typical angles include 0.05°, 0.1°, and 0.25°.

[0045] Optionally, the following data can be calculated based on the effective observation data of radiation stations: total solar radiation incident at the Earth's surface, total photosynthetically active radiation, net shortwave radiation at the Earth's surface, net longwave radiation at the Earth's surface, and net radiation at the Earth's surface. The total solar radiation observed at the Earth's surface is the sum of direct solar radiation and diffuse solar radiation observed. The total photosynthetically active radiation is the sum of direct photosynthetically active radiation and diffuse photosynthetically active radiation. The net shortwave radiation at the Earth's surface is the difference between the total solar radiation incident at the Earth's surface and the solar radiation emitted at the Earth's surface. The net longwave radiation at the Earth's surface is the difference between the longwave radiation incident at the Earth's surface and the longwave radiation emitted at the Earth's surface. The net radiation at the Earth's surface is the sum of the net shortwave radiation and the net longwave radiation at the Earth's surface.

[0046] Optionally, the machine learning model can be any one of the following: random forest model, gradient boosting tree model, deep neural network model, and convolutional neural network model.

[0047] By employing machine learning models, training any one of the following machine learning models—random forest, gradient boosting tree, deep neural network, and convolutional neural network—is beneficial for obtaining the optimal spatiotemporal scale-based surface radiation budget co-inversion model, thereby improving the accuracy of the spatiotemporal scale-based surface radiation budget co-inversion.

[0048] Optionally, the driving dataset of the spatiotemporal scale surface radiation budget co-inversion model is divided into a model training dataset, a test dataset, and a validation dataset. The constructed spatiotemporal scale surface radiation budget co-inversion model is trained using the model training dataset. The trained spatiotemporal scale surface radiation budget co-inversion model is evaluated using the test dataset. The hyperparameters of the constructed spatiotemporal scale surface radiation budget co-inversion model are adjusted using the validation dataset to determine the optimal spatiotemporal scale surface radiation budget co-inversion model.

[0049] Optional, set Represents slope, Represents slope aspect. Represents the sky view factor. Represents digital elevation values. Represents solar radiation at the top of the atmosphere. Represents air temperature. Represents the Earth's surface temperature. Represents relative humidity. Represents soil moisture. Represents the normalized vegetation index, Represents leaf area index, Represents vegetation coverage. Represents data on direct solar radiation observed on the Earth's surface. Represents scattered solar radiation observation data, Represents the total solar radiation incident on the Earth's surface. Represents direct photosynthetically active radiation. Represents scattered photosynthetically active radiation. Represents total photosynthetically active radiation. Represents observational data of solar radiation emitted from the Earth's surface. Represents observational data of incident longwave radiation at the Earth's surface. Represents observational data of longwave radiation emitted from the Earth's surface. Represents net shortwave radiation observation data at the Earth's surface. Represents the observational data of net longwave radiation at the Earth's surface. Represents net surface radiation observation data. A comprehensive inversion model representing all elements of surface radiation balance across all time and space scales. Let the set be represented as follows: The full-scale spatiotemporal scale surface radiation budget collaborative inversion model is expressed as:

[0050]

[0051] .

[0052] Optional, set Represents data on direct solar radiation observed on the Earth's surface. Represents scattered solar radiation observation data, Represents the total solar radiation incident on the Earth's surface. Represents observational data of direct photosynthetically active radiation. Data representing the observed scattered photosynthetically active radiation. Represents total photosynthetically active radiation observation data, Represents observational data of solar radiation emitted from the Earth's surface. Represents observational data of incident longwave radiation at the Earth's surface. Represents observational data of longwave radiation emitted from the Earth's surface. Represents net shortwave radiation observation data at the Earth's surface. Represents the observational data of net longwave radiation at the Earth's surface. Represents net surface radiation observation data. This represents an estimated value of direct solar radiation to the Earth's surface. This represents an estimated value of scattered solar radiation. This represents an estimated value of the total solar radiation incident on the Earth's surface. This represents an estimated value of direct photosynthetically active radiation. This represents an estimated value of the scattered photosynthetically active radiation. This represents an estimated value of total photosynthetically active radiation. This represents an estimated value of solar radiation emitted from the Earth's surface. Represents an estimated value of incident longwave radiation at the Earth's surface. This represents an estimated value of longwave radiation emitted from the Earth's surface. Represents the estimated net shortwave radiation at the Earth's surface. This represents an estimated value of net longwave radiation at the Earth's surface. Represents the estimated net surface radiation. for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, If we denote a set, then the cost function is:

[0053]

[0054] This invention constructs several types of machine learning models and uses a constructed spatiotemporal scale surface radiation budget co-inversion model to drive model training on a dataset. This allows for the selection of the optimal spatiotemporal scale surface radiation budget co-inversion model, which is beneficial for improving the accuracy of spatiotemporal scale surface radiation budget co-inversion. Furthermore, this invention achieves spatiotemporal scale surface radiation budget co-inversion by collecting observational data, remote sensing data, meteorological reanalysis data, and digital elevation model data, enabling radiation budget balance among different components.

[0055] Example 2

[0056] Based on the same principle as the method shown in Embodiment 1 of the present invention, as illustrated in the appendix. Figure 2 As shown, the embodiments of the present invention also provide a global surface radiation budget full-element collaborative inversion system at all spatiotemporal scales, including a model building unit and a data processing unit;

[0057] The model building unit is used to build a machine learning model. It takes remote sensing data, meteorological reanalysis data, topographic data and sky view factor as input, and global surface radiation budget elements as output. It sets a cost function to obtain a collaborative inversion model of surface radiation budget elements at all time and space scales.

[0058] The output unit is used to input remote sensing data, meteorological reanalysis data, topographic data and sky view factor into the full spatiotemporal scale surface radiation budget collaborative inversion model to obtain the full spatiotemporal scale surface radiation budget of the target.

[0059] Optional, the global surface radiation budget includes all elements of direct solar radiation, diffuse solar radiation, total solar radiation incident on the surface, direct photosynthetically active radiation, diffuse photosynthetically active radiation, total photosynthetically active radiation, solar radiation emitted from the surface, longwave radiation incident on the surface, longwave radiation emitted from the surface, net shortwave radiation from the surface, net longwave radiation from the surface, and net surface radiation.

[0060] Optionally, a driving dataset for the full-temporal and spatial scale surface radiation budget co-inversion model is constructed based on station observation data, remote sensing data, meteorological reanalysis data, topographic data, and sky view factor. The constructed full-temporal and spatial scale surface radiation budget co-inversion model is then trained using the driving dataset to determine the optimal model parameters and obtain the optimal full-temporal and spatial scale surface radiation budget co-inversion model.

[0061] Optionally, collect the data required for the global surface radiation balance full-element coordinated inversion, including station observation data, remote sensing data, meteorological reanalysis data, topographic data and sky view factor; topographic data includes digital elevation values, slope and aspect;

[0062] Preprocessing and filtering of station observation data yielded effective observation data for radiation stations. This effective observation data included direct solar radiation observation data, diffuse solar radiation observation data, direct photosynthetically active radiation observation data, diffuse photosynthetically active radiation observation data, surface emitted solar radiation observation data, surface incident longwave radiation observation data, and surface emitted longwave radiation observation data. Remote sensing data included top-of-atmosphere solar radiation data, soil moisture, normalized difference vegetation index (NDVI), leaf area index, and vegetation cover. Meteorological reanalysis data included air temperature, surface temperature, and relative humidity.

[0063] Based on the effective observation data of radiation stations, the observation data of total incident solar radiation, total photosynthetically active radiation, net shortwave radiation, net longwave radiation, and net radiation at the Earth's surface are calculated. Combined with the effective observation data of radiation stations, the surface radiation observation data are obtained.

[0064] The collected remote sensing data and meteorological reanalysis data are preprocessed to obtain preprocessed remote sensing data and preprocessed meteorological reanalysis data.

[0065] Calculate digital elevation values, slope, aspect, and sky view factor based on digital elevation model data at several resolutions.

[0066] Optionally, the following data can be calculated based on the effective observation data of radiation stations: total solar radiation incident at the Earth's surface, total photosynthetically active radiation, net shortwave radiation at the Earth's surface, net longwave radiation at the Earth's surface, and net radiation at the Earth's surface. The total solar radiation observed at the Earth's surface is the sum of direct solar radiation and diffuse solar radiation observed. The total photosynthetically active radiation is the sum of direct photosynthetically active radiation and diffuse photosynthetically active radiation. The net shortwave radiation at the Earth's surface is the difference between the total solar radiation incident at the Earth's surface and the solar radiation emitted at the Earth's surface. The net longwave radiation at the Earth's surface is the difference between the longwave radiation incident at the Earth's surface and the longwave radiation emitted at the Earth's surface. The net radiation at the Earth's surface is the sum of the net shortwave radiation and the net longwave radiation at the Earth's surface.

[0067] Optionally, the machine learning model can be any one of the following: random forest model, gradient boosting tree model, deep neural network model, and convolutional neural network model.

[0068] Optionally, the driving dataset of the spatiotemporal scale surface radiation budget co-inversion model is divided into a model training dataset, a test dataset, and a validation dataset. The constructed spatiotemporal scale surface radiation budget co-inversion model is trained using the model training dataset. The trained spatiotemporal scale surface radiation budget co-inversion model is evaluated using the test dataset. The hyperparameters of the constructed spatiotemporal scale surface radiation budget co-inversion model are adjusted using the validation dataset to determine the optimal spatiotemporal scale surface radiation budget co-inversion model.

[0069] Optional, set Represents slope, Represents slope aspect. Represents the sky view factor. Represents digital elevation values. Represents solar radiation at the top of the atmosphere. Represents air temperature. Represents the Earth's surface temperature. Represents relative humidity. Represents soil moisture. Represents the normalized vegetation index, Represents leaf area index, Represents vegetation coverage. Represents data on direct solar radiation observed on the Earth's surface. Represents scattered solar radiation observation data, Represents the total solar radiation incident on the Earth's surface. Represents direct photosynthetically active radiation. Represents scattered photosynthetically active radiation. Represents total photosynthetically active radiation. Represents observational data of solar radiation emitted from the Earth's surface. Represents observational data of incident longwave radiation at the Earth's surface. Represents observational data of longwave radiation emitted from the Earth's surface. Represents net shortwave radiation observation data at the Earth's surface. Represents the observational data of net longwave radiation at the Earth's surface. Represents net surface radiation observation data. A comprehensive inversion model representing all elements of surface radiation balance across all time and space scales. Let the set be represented as follows: The full-scale spatiotemporal scale surface radiation budget collaborative inversion model is expressed as:

[0070]

[0071] .

[0072] Optional, set Represents data on direct solar radiation observed on the Earth's surface. Represents scattered solar radiation observation data, Represents the total solar radiation incident on the Earth's surface. Represents observational data of direct photosynthetically active radiation. Data representing the observed scattered photosynthetically active radiation. Represents total photosynthetically active radiation observation data, Represents observational data of solar radiation emitted from the Earth's surface. Represents observational data of incident longwave radiation at the Earth's surface. Represents observational data of longwave radiation emitted from the Earth's surface. Represents net shortwave radiation observation data at the Earth's surface. Represents the observational data of net longwave radiation at the Earth's surface. Represents net surface radiation observation data. This represents an estimated value of direct solar radiation to the Earth's surface. This represents an estimated value of scattered solar radiation. This represents an estimated value of the total solar radiation incident on the Earth's surface. This represents the estimated value of direct photosynthetically active radiation. This represents an estimated value of the scattered photosynthetically active radiation. This represents an estimated value of total photosynthetically active radiation. This represents an estimated value of solar radiation emitted from the Earth's surface. Represents an estimated value of incident longwave radiation at the Earth's surface. This represents an estimated value of longwave radiation emitted from the Earth's surface. Represents the estimated net shortwave radiation at the Earth's surface. This represents an estimated value of net longwave radiation at the Earth's surface. Represents the estimated net surface radiation. for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, If we denote a set, then the cost function is:

[0073]

[0074] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for the coordinated inversion of all elements of global surface radiation budget across all time and space scales, characterized in that: include: A machine learning model was constructed, taking remote sensing data, meteorological reanalysis data, topographic data and sky horizon factors as inputs, and global surface radiation budget elements as outputs. A cost function was set to obtain a collaborative inversion model of all elements of surface radiation budget at all time and space scales. Remote sensing data, meteorological reanalysis data, topographic data, and sky view factor are input into the spatiotemporal scale surface radiation budget full-element collaborative inversion model to obtain the target spatiotemporal scale surface radiation budget full elements. The global surface radiation balance consists of all elements including direct solar radiation, diffuse solar radiation, total solar radiation incident on the surface, direct photosynthetically active radiation, diffuse photosynthetically active radiation, total photosynthetically active radiation, solar radiation emitted from the surface, longwave radiation incident on the surface, longwave radiation emitted from the surface, net shortwave radiation, net longwave radiation, and net surface radiation. The terrain data includes digital elevation values, slope, and aspect. The remote sensing data includes solar radiation data from the top of the atmosphere, soil moisture, normalized vegetation index, leaf area index, and vegetation cover. The meteorological reanalysis data includes air temperature, surface temperature, and relative humidity.

2. The method for coordinated inversion of all elements of global surface radiation budget at all spatiotemporal scales according to claim 1, characterized in that, Based on station observation data, remote sensing data, meteorological reanalysis data, topographic data, and sky horizon factors, a collaborative inversion model of all elements of surface radiation balance at all time and space scale is constructed to drive the dataset. The model was trained using a dataset driven by a spatiotemporal scale-based co-inversion model of all elements of surface radiation budget, and the optimal model parameters were determined to obtain the optimal spatiotemporal scale-based co-inversion model of all elements of surface radiation budget.

3. The method for coordinated inversion of all elements of global surface radiation budget at all spatiotemporal scales according to claim 1, characterized in that, Collect the data required for the coordinated inversion of all elements of global surface radiation balance, including station observation data, remote sensing data, meteorological reanalysis data, topographic data, and sky horizon factor; The station observation data includes direct solar radiation observation data, diffuse solar radiation observation data, direct photosynthetically active radiation observation data, diffuse photosynthetically active radiation observation data, surface emitted solar radiation observation data, surface incident longwave radiation observation data, and surface emitted longwave radiation observation data; remote sensing data includes top-of-atmosphere solar radiation data, soil moisture, normalized difference vegetation index, leaf area index, and vegetation cover; meteorological reanalysis data includes air temperature, surface temperature, and relative humidity; topographic data includes digital elevation values, slope, and aspect. Preprocessing and filtering of station observation data yields effective observation data for radiation stations; The collected remote sensing data and meteorological reanalysis data were preprocessed to obtain preprocessed remote sensing data and preprocessed meteorological reanalysis data. Calculate digital elevation values, slope, aspect, and sky view factor based on digital elevation model data at several resolutions.

4. The method for coordinated inversion of all elements of global surface radiation budget at all spatiotemporal scales according to claim 3, characterized in that, Based on effective observation data from radiation stations, the following data were calculated: total incident solar radiation, total photosynthetically active radiation, net shortwave radiation, net longwave radiation, and net surface radiation. Total solar radiation is the sum of direct and diffuse solar radiation. Total photosynthetically active radiation is the sum of direct and diffuse photosynthetically active radiation. Net shortwave radiation is the difference between total incident solar radiation and outgoing solar radiation. Net longwave radiation is the difference between incident and outgoing longwave radiation. Net surface radiation is calculated as the net shortwave radiation. The sum of radiation and net longwave radiation at the Earth's surface.

5. The method for coordinated inversion of all elements of global surface radiation budget at all spatiotemporal scales according to claim 1, characterized in that, The machine learning model can be any one of the following: random forest model, gradient boosting tree model, deep neural network model, and convolutional neural network model.

6. The method for coordinated inversion of all elements of global surface radiation budget at all spatiotemporal scales according to claim 2, characterized in that, The dataset driven by the full-time and spatial-scale surface radiation balance and expenditure co-inversion model is divided into a model training dataset, a test dataset, and a validation dataset. The constructed spatiotemporal scale surface radiation budget full-element collaborative inversion model was trained using the model training dataset; The trained spatiotemporal scale surface radiation budget co-inversion model was evaluated using the test dataset. The hyperparameters of the constructed spatiotemporal scale surface radiation budget co-inversion model were adjusted using the validation dataset to determine the optimal spatiotemporal scale surface radiation budget co-inversion model.

7. The method for coordinated inversion of all elements of global surface radiation budget at all spatiotemporal scales according to claim 1, characterized in that, Let DEM slope Represents slope, DEM aspect Represents slope aspect, SVF represents sky view factor, DEM represents digital elevation value, R g T represents solar radiation at the top of the atmosphere. a T represents air temperature. s Represents surface temperature, RH represents relative humidity, SM represents soil moisture, NDVI represents normalized difference vegetation index, LAI represents leaf area index, FVC represents vegetation cover, R s，direct R represents observational data of direct solar radiation on the Earth's surface. s，diffuse R represents the observed data of scattered solar radiation. s，down PAR represents the observed data of total solar radiation incident on the Earth's surface. direct Represents direct photosynthetically active radiation, PAR diffuse R represents scattered photosynthetically active radiation, PAR represents total photosynthetically active radiation, and R represents scattered photosynthetically active radiation. s，up R represents observational data of solar radiation emitted from the Earth's surface. l，down R represents the observational data of incident longwave radiation at the Earth's surface. l，up R represents the observational data of longwave radiation emitted from the Earth's surface. ns R represents the observed data of net shortwave radiation at the Earth's surface. nl R represents the observed data of net longwave radiation at the Earth's surface. n Represents net surface radiation observation data, Full radiation Let U() represent the set of the all-element collaborative inversion model of the total surface radiation budget across all time and space scales. Then, the all-element collaborative inversion model of the total surface radiation budget across all time and space scales is expressed as: Full radiation =U(R s,direct ,R s,diffuse ,R s，down ,COUPLE direct ,COUPLE diffuse ,PAR,R s，up ,R ns ,R l，down ,R l，up ,R nl ,R n ) f(THEY slope ,THEM aspect ,SVF,DEM,R g ,T a ,T s ,RH,SM,NDVI,LAI,FVC)。 8. The method for coordinated inversion of all elements of global surface radiation budget at all spatiotemporal scales according to claim 1, characterized in that, Let R s，direct R represents observational data of direct solar radiation on the Earth's surface. s，diffuse R represents the observed data of scattered solar radiation. s，down PAR represents the observed data of total solar radiation incident on the Earth's surface. direct Representing direct photosynthetically active radiation observation data, PAR diffuse R represents the scattered photosynthetically active radiation observation data, PAR represents the total photosynthetically active radiation observation data, and R represents the total photosynthetically active radiation observation data. s，up R represents observational data of solar radiation emitted from the Earth's surface. l，down R represents the observational data of incident longwave radiation at the Earth's surface. l，up R represents the observational data of longwave radiation emitted from the Earth's surface. ns R represents the observed data of net shortwave radiation at the Earth's surface. nl R represents the observed data of net longwave radiation at the Earth's surface. n R represents the net surface radiation observation data. s，direct ′ represents the estimated value of direct solar radiation on the Earth's surface, R s，diffuse ′ represents the estimated value of scattered solar radiation, R s，down ′ represents the estimated total solar radiation incident on the Earth's surface, PAR direct ′ represents the estimated value of direct photosynthetically active radiation, PAR diffuse ' represents the estimated value of scattered photosynthetically active radiation, PAR' represents the estimated value of total photosynthetically active radiation, and R s，up R represents the estimated value of solar radiation emitted from the Earth's surface. l，down R represents the estimated value of incident longwave radiation at the Earth's surface. l，up R represents the estimated value of longwave radiation emitted from the Earth's surface. ns ′ represents the estimated net shortwave radiation at the Earth's surface, R nl ′ represents the estimated net longwave radiation at the Earth's surface, R n ′ represents the estimated net surface radiation, and a1 is (R s，direct ′-R s，direct ) 2 The weighting coefficient, a2 is (R s，diffuse ′-R s，diffuse ) 2 The weighting coefficient, a3 is (R s，down ′-R s，down ) 2 The weighting coefficient, a4 is (PAR direct ′-PAR direct ) 2 The weighting coefficient, a5 is (PAR diffuse ′-PAR diffuse ) 2 The weighting coefficient, a6, is (PAR′-PAR). 2 The weighting coefficient, a7 is (R s，up ′-R s，up ) 2 The weighting coefficient, a8 is (R l，down ′-R l，down ) 2 The weighting coefficient, a9 is (R l，up ′-R l，up ) 2 The weighting coefficient, a 10 For (R) ns ′-R ns ) 2 The weighting coefficient, a 11 For (R) nl ′-R nl ) 2 The weighting coefficient, a 12 For (R) n ′-R n ) 2 Let U() represent the set of weights, then the cost function is:

9. A global surface radiation budget full-element collaborative inversion system at all spatiotemporal scales, characterized in that: Includes model building units and data processing units; The model building unit is used to build a machine learning model. It takes remote sensing data, meteorological reanalysis data, topographic data and sky view factor as input, and global surface radiation budget elements as output. It sets a cost function to obtain a collaborative inversion model of surface radiation budget elements at all time and space scales. The output unit is used to input remote sensing data, meteorological reanalysis data, topographic data and sky view factor into the full-temporal and spatial scale surface radiation budget collaborative inversion model to obtain the full-temporal and spatial scale surface radiation budget of the target. The global surface radiation balance consists of all elements including direct solar radiation, diffuse solar radiation, total solar radiation incident on the surface, direct photosynthetically active radiation, diffuse photosynthetically active radiation, total photosynthetically active radiation, solar radiation emitted from the surface, longwave radiation incident on the surface, longwave radiation emitted from the surface, net shortwave radiation, net longwave radiation, and net surface radiation. The terrain data includes digital elevation values, slope, and aspect. The remote sensing data includes solar radiation data from the top of the atmosphere, soil moisture, normalized vegetation index, leaf area index, and vegetation cover. The meteorological reanalysis data includes air temperature, surface temperature, and relative humidity.

Images