Global surface maximum relative humidity inversion method based on physical constraint and machine learning

Abstract

The invention provides a global surface maximum relative humidity inversion method based on physical constraint and machine learning, and relates to the technical field of hydrology and computer science. The method comprises the following steps: constructing a candidate data set, and obtaining actually measured latent heat flux and synchronous meteorological elements of a flux observation station; inputting the candidate data set and the synchronous meteorological elements into a physical constraint model, calculating and simulating latent heat flux, and selecting a value with the minimum error to determine the optimal maximum relative humidity of the earth surface; training a machine learning inversion model by taking the optimal humidity and the synchronous meteorological elements as samples; and fusing the trained model with remote sensing surface temperature, surface coverage type and reanalysis meteorological data, and calculating and outputting a global rasterization result. According to the method, the inversion of the maximum relative humidity of the earth surface under the global scale is realized, and a new quantitative mode is provided.

Description

Technical Field

[0001] The present invention relates to the fields of hydrology and computer science and technology, and in particular to a method for inverting global surface maximum relative humidity based on physical constraints and machine learning. Background Art

[0002] The maximum relative humidity of the surface directly determines the extreme distribution of surface water vapor pressure and its daily variation characteristics by limiting the threshold effect of surface saturated water vapor pressure. It is restricted by thermal conditions such as surface temperature and air pressure, and has obvious temporal and spatial heterogeneity due to factors such as topography, vegetation cover and urbanization. Therefore, it has a multi-level and multi-scale regulatory effect on the material and energy exchange process between land and atmosphere. Accurately quantifying the maximum relative humidity of the surface can not only provide key parameters for revealing the internal mechanism of global-scale precipitation distribution and atmospheric circulation pattern, but also lay a theoretical foundation for evaluating the frequency and intensity changes of climate events such as extreme droughts and heavy rains under the background of climate warming and analyzing their causes.

[0003] In existing studies, the estimation of maximum relative humidity on the surface mostly relies on empirical values: in areas with dense vegetation or high soil moisture content, it is often assumed to be 0.8-1.0; in areas with low vegetation coverage or drought, it is usually taken as 0.6 or lower. Although this method is easy to operate, it is highly subjective, has a limited applicable scale, and ignores the joint effects of multiple key factors such as vegetation physiological conditions, soil moisture dynamics, and ambient temperature, making it difficult to obtain accurate and reliable maximum relative humidity on the surface. In fact, the maximum relative humidity on the surface is a basic parameter in the water cycle and energy cycle, and has an important impact on turbulent mixing in the atmospheric boundary layer, hydrological processes, precipitation formation, and surface evapotranspiration. However, there is currently no method that can take into account the coupling mechanism of the large-scale atmosphere-vegetation-soil system and achieve high-precision quantification under all-weather, all-time and space conditions. This gap not only limits the in-depth understanding of the mechanism of related processes, but also restricts the improvement of the accuracy of environmental monitoring and ecohydrological models to a certain extent. Therefore, proposing a high-precision quantitative method for maximum relative humidity on the surface that takes environmental factors into account, has wide applicability and all-weather operation capabilities has become a scientific problem that needs to be solved urgently. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a global surface maximum relative humidity inversion method based on physical constraints and machine learning, taking the dynamic interaction relationship between the maximum surface relative humidity and evapotranspiration as the physical constraint, and using a machine learning algorithm as the inversion model construction method, to achieve the inversion of the maximum surface relative humidity on a global scale, which can provide a new method for accurately quantifying the maximum surface relative humidity.

[0005] To achieve the above object, the present invention provides the following solutions: A method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning, comprising: Construct a candidate dataset for the surface maximum relative humidity; the candidate dataset for the surface maximum relative humidity includes multiple candidate values of the surface maximum relative humidity; Obtain the measured latent heat flux and synchronous meteorological elements at the flux observation site, and obtain the remote sensing surface temperature, surface cover type and reanalysis meteorological data; Input the candidate dataset for the surface maximum relative humidity and the synchronous meteorological elements into the physical constraint model, calculate the simulated latent heat flux corresponding to each candidate value, and compare the simulated latent heat flux with the measured latent heat flux to select the candidate value with the smallest error in the corresponding simulated latent heat flux, and determine the candidate value with the smallest error as the optimal surface maximum relative humidity of the flux observation site; Train a machine learning inversion model with the optimal surface maximum relative humidity and synchronous meteorological elements as samples according to the surface cover type; Use the trained machine learning inversion model to fuse with the remote sensing surface temperature, the surface cover type and the reanalysis meteorological data to calculate the surface maximum relative humidity, and output the global rasterized result.

[0006] Preferably, the candidate values of the surface maximum relative humidity in the candidate dataset for the surface maximum relative humidity are generated by discretization within a preset range.

[0007] Preferably, the expression of the physical constraint model includes: Wherein, h s,max is the surface maximum relative humidity; e s is the surface vapor pressure; Ω is the land-air coupling factor; e a is the actual water vapor pressure; e s * is the surface saturation vapor pressure; Δ is the slope of the saturation water vapor pressure curve; γ is the psychrometric constant; r s is the surface impedance; r ae is the aerodynamic impedance; ρ is the air density; C p is the specific heat of air;T s is the surface temperature; T a is the air temperature; R n is the surface net radiation; G is the soil heat flux; ET is the latent heat flux.

[0008] Preferably, input the candidate dataset of the maximum surface relative humidity and the synchronous meteorological elements into a physical constraint model, calculate the simulated latent heat flux corresponding to each candidate value, and compare the simulated latent heat flux with the measured latent heat flux to select the candidate value with the minimum error of the simulated latent heat flux, and determine the candidate value with the minimum error as the optimal surface maximum relative humidity of the flux observation site, including: Combine the candidate maximum relative humidity value with the synchronous meteorological elements to form a multi-dimensional input feature array; Input the multi-dimensional input feature array into the physical constraint model to calculate the simulated latent heat flux corresponding to each candidate value by using the physical constraint model; Compare the simulated latent heat flux with the measured latent heat flux one by one to calculate the error of each candidate value; Select the candidate value with the minimum error and determine it as the optimal surface maximum relative humidity of the flux observation site.

[0009] Preferably, use the optimal surface maximum relative humidity and synchronous meteorological elements as samples to train a machine learning inversion model, including: Classify the samples composed of the optimal surface maximum relative humidity and synchronous meteorological elements according to the surface cover type to form sample sets of multiple categories; Pair the optimal surface maximum relative humidity and the corresponding synchronous meteorological elements with the sample sets of each category to construct feature vectors; Perform data augmentation on the feature vectors, and divide the feature vectors after data augmentation into a training set and a validation set; Use the training set to fit a preset initial model, and use the validation set for model evaluation to optimize the model parameters to obtain the final machine learning inversion model.

[0010] Preferably, the machine learning inversion model includes any one of a random forest, a support vector machine, and a neural network.

[0011] Preferably, the method of model evaluation is the cross-validation method.

[0012] Preferably, the surface cover types include: forest, grassland, city, farmland, and water body.

[0013] Preferably, the verification metrics of the machine learning inversion model include root mean square error and coefficient of determination.

[0014] Preferably, the numerical range of the candidate dataset of the maximum surface relative humidity is [0, 1].

[0015] The present invention discloses the following technical effects: Based on the physical constraint relationship between the maximum surface relative humidity and the latent heat flux, the present invention calculates and screens the optimal maximum surface relative humidity by using the constructed candidate dataset of the maximum surface relative humidity and the data of observation stations; and constructs an inversion model by using a machine learning algorithm at the station scale, further combines remote sensing surface temperature data, remote sensing data of surface cover types and reanalysis meteorological data products, and realizes the inversion of the global maximum surface relative humidity by using the inversion model, which provides a new method for accurately quantifying and precisely monitoring and simulating the global surface water vapor exchange process. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is the flowchart of the method provided by the embodiment of the present invention; Figure 2 It is the schematic diagram of the technical route provided by the embodiment of the present invention. Detailed Embodiments

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0019] The purpose of the present invention is to provide a method for inverting the global maximum surface relative humidity based on physical constraints and machine learning. Taking the dynamic interaction relationship between the maximum surface relative humidity and evapotranspiration as the physical constraint and the machine learning algorithm as the method for constructing the inversion model, the inversion of the maximum surface relative humidity at the global scale is realized, which provides a new method for accurately quantifying the maximum surface relative humidity.

[0020] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0021] Figure 1 A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides a method for inverting the global maximum relative humidity on the surface based on physical constraints and machine learning, comprising: Step 100: constructing a candidate data set of maximum relative humidity on the surface; the candidate data set of maximum relative humidity on the surface includes a plurality of candidate values of maximum relative humidity on the surface; Step 200: Obtain the measured latent heat flux and synchronous meteorological elements of the flux observation site, and obtain remote sensing surface temperature, surface cover type and reanalysis meteorological data; Step 300: Input the candidate data set of maximum relative humidity on the surface and the synchronous meteorological elements into the physical constraint model, calculate the simulated latent heat flux corresponding to each candidate value, and compare the simulated latent heat flux with the measured latent heat flux to select the candidate value with the smallest error of the corresponding simulated latent heat flux, and determine the candidate value with the smallest error as the optimal maximum relative humidity on the surface of the flux observation station; Step 400: training a machine learning inversion model based on the land cover type, taking the optimal land surface maximum relative humidity and synchronous meteorological elements as samples; Step 500: Utilize the trained machine learning inversion model to fuse with remote sensing surface temperature, land cover type and reanalysis meteorological data, estimate the maximum relative humidity of the surface in the unobserved area, and output the global gridded result.

[0022] Specifically, Figure 2 As shown, the technical route of this embodiment includes the following steps: Step 1: Construct a candidate dataset of maximum relative humidity on the surface; Step 2: Obtain global flux observation site data, surface temperature remote sensing data, remote sensing land cover type data and reanalysis meteorological data products; Step 3: Based on steps 1 and 2, the optimal maximum relative humidity of the surface is calculated and screened using the candidate data set of maximum relative humidity of the surface and observation site data according to the physical constraint model between the maximum relative humidity of the surface and the latent heat flux; Step 4: Based on the optimal maximum relative humidity result obtained in step 3, the meteorological variables are screened by combining correlation analysis, and the maximum relative humidity inversion model of each land cover type is further constructed using machine learning algorithm; Step 5: Based on the inversion model of the maximum relative humidity of the surface for each land cover type obtained in step 4, the global maximum relative humidity of the surface can be obtained by using the global surface temperature remote sensing data, remote sensing land cover type data and reanalysis meteorological data products collected in step 2.

[0023] Optionally, in step 1, the candidate dataset of the maximum surface relative humidity is: the candidate values of the maximum surface relative humidity with a self-constructed dataset range of [0, 1] and a step size of 0.01.

[0024] Optionally, in step 2, the global flux observation site data, surface temperature remote sensing data, and reanalysis meteorological data products include: latent heat flux data at the site scale, as well as air temperature, surface temperature, soil heat flux, net radiation, vapor pressure deficit, relative humidity, absolute humidity, specific humidity, wind speed, and atmospheric pressure data at the site and global scales, and global remote sensing surface cover type data, including types such as forests, farmlands, and grasslands.

[0025] Furthermore, in step 3, the physical constraint model between the maximum surface relative humidity and the latent heat flux is given by formulas (1) - (4): ; ; ; ; Wherein, h s,max is the maximum surface relative humidity; e s is the surface vapor pressure; Ω is the land - atmosphere coupling factor; e a is the actual water vapor pressure; e s * is the surface saturation vapor pressure; Δ is the slope of the saturation water vapor pressure curve; γ is the psychrometric constant; r s is the surface impedance; r ae is the aerodynamic impedance; ρ is the air density; C p is the specific heat of air; T s is the surface temperature; T a is the air temperature; R n is the surface net radiation; G is the soil heat flux; ET is the latent heat flux.

[0026] Furthermore, in step 3, the optimal maximum surface relative humidity is calculated and screened using the candidate dataset of the maximum surface relative humidity and the observation site data Even further, based on the calculated and screened optimal maximum surface relative humidity, the specific implementation steps are: Step 3011, obtain input data; obtain candidate values from the basic surface maximum relative humidity dataset, denoted as: where represents the j th h s,max candidate value discretely generated within the closed interval [0, 1] with a step size of 0.01; Obtain latent heat flux data and meteorological data (air temperature, surface temperature, soil heat flux, net radiation, vapor pressure deficit, relative humidity, absolute humidity, specific humidity, wind speed, and atmospheric pressure data) from global flux observation stations; denoted as: where , respectively represent the latent heat flux and meteorological data corresponding to the i th station; is the total number of stations.

[0027] Step 3012, model calculation; based on the physical constraint model (denoted as: ), for each , calculate the simulated latent heat flux under the same meteorological conditions , denoted as: Step 3013, error comparison; based on the simulated results of the latent heat flux corresponding to each candidate value , compare with the measured latent heat flux data of the global flux observation stations, and calculate the absolute error between the two, denoted as: Step 3014, determine the optimal value; based on the absolute error between the simulated results and the measured data, for the i th station, select the item that minimizes from all candidate values as the optimal surface maximum relative humidity, and finally form a result set at the station scale, denoted as: Optionally, in step 4, according to the optimal surface maximum relative humidity values of each observation station, the corresponding meteorological data and the station surface cover type label , model using machine learning algorithms by surface cover type.

[0028] Furthermore, the machine learning algorithm is a gradient boosting decision tree (XGBoost, eXtreme GradientBoosting) algorithm.

[0029] Furthermore, an inversion model is constructed based on the algorithm, and the specific implementation steps are as follows: Step 4011, site grouping; for each site i , read its surface cover category identification (such as forests, grasslands, farmlands, etc.) All sites are grouped according to coverage type to form several subsets, which are recorded as: in T is the total number of coverage types.

[0030] Step 4012, grouping sample set construction; for each type t , construct a training set of this type, recorded as: in, .

[0031] Step 4013, feature selection; for each subset Correlation tests (Pearson coefficient) were performed on all the variables, some variables with low correlation were eliminated, and the main features were retained. middle and The correlation coefficient between them can be obtained by the following formula: in, n For subset The amount of data included; For subset Middle i The meteorological variable value corresponding to each data point; For this variable in the subset The mean value in ; For subset The i-th data point in value; For subset middle Mean.

[0032] Furthermore, each subset Correlation test is performed on both and (including correlation coefficients between air temperature, surface temperature, soil heat flux, net radiation, vapor pressure deficit, relative humidity, absolute humidity, specific humidity, wind speed and atmospheric pressure data).

[0033] Step 4014, model training: Using the selected highly correlated meteorological variables, divide the data into a training set and a validation set according to a ratio of 7:3, and use the gradient boosting decision tree (XGBoost) algorithm to train the surface maximum relative humidity inversion model for each type t denoted as: wherein, is the surface type t corresponding XGBoost model; are hyperparameters, including but not limited to the depth of the tree, learning rate, and subsampling rate.

[0034] Furthermore, take | r xy |>0.5 as the criterion to judge the high or low correlation between the meteorological variable and .

[0035] Step 4015, model optimization: For each type of inversion model, optimize the core hyperparameters of XGBoost (including learning rate, maximum depth of the tree, subsampling ratio, column sampling ratio, and number of weak learners) based on the Bayesian optimization algorithm. At the same time, introduce a regularization term (L1 / L2) to control the model complexity and reduce overfitting. Minimize the loss function value (mean square error) to improve the accuracy and generalization ability of the model.

[0036] Specifically, determine the optimal hyperparameters of the inversion model for each land type based on the minimization of the mean square error, and the form of the mean square error is: Step 4016, model output: For each type t , generate the final surface maximum relative humidity inversion model, denoted as: wherein, is the surface maximum relative humidity inversion model for the corresponding surface type t ; is the meteorological variable with high correlation with ; are the optimal hyperparameters.

[0037] Step 4017, model validation: Use an independent validation dataset to validate the inversion model.

[0038] Furthermore, the validation metrics are the root mean square error and the coefficient of determination.

[0039] Even further, the training process will stop when any of the following conditions is met: When the value of the loss function is continuously QTraining stops when there is no significant improvement within a certain number of rounds.

[0040] When the change range of the value of the loss function is less than a preset threshold within consecutive R number of rounds, training stops.

[0041] If the loss of the validation set continuously Z rises for consecutive rounds, training stops.

[0042] When the number of training reaches the preset maximum number of training rounds, training stops.

[0043] Based on the physical constraint relationship between the surface maximum relative humidity and the latent heat flux, the present invention calculates and screens the optimal surface maximum relative humidity by using the constructed candidate dataset of the surface maximum relative humidity and the data of observation stations; and constructs an inversion model by using a machine learning algorithm at the station scale, further combines remote sensing surface temperature data, remote sensing data of surface cover types and reanalysis meteorological data products, and realizes the inversion of the global surface maximum relative humidity by using the inversion model, which can provide a new method for accurately quantifying, precisely monitoring and simulating the global surface water vapor exchange process.

[0044] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same and similar parts among the various embodiments can be referred to each other.

[0045] Specific examples are used in this article to elaborate on the principles and implementation manners of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning, characterized in that Including: Construct a candidate dataset of the surface maximum relative humidity; the candidate dataset of the surface maximum relative humidity includes multiple candidate values of the surface maximum relative humidity; Obtain the measured latent heat flux and synchronous meteorological elements of the flux observation site, and obtain the remote sensing surface temperature, surface cover type, and reanalysis meteorological data; Input the candidate dataset of the surface maximum relative humidity and the synchronous meteorological elements into the physical constraint model together, calculate the simulated latent heat flux corresponding to each candidate value, and compare the simulated latent heat flux with the measured latent heat flux to select the candidate value with the smallest error in the corresponding simulated latent heat flux, and determine the candidate value with the smallest error as the optimal surface maximum relative humidity of the flux observation site; Train a machine learning inversion model with the optimal surface maximum relative humidity and synchronous meteorological elements as samples according to the surface cover type; Use the trained machine learning inversion model to fuse with the remote sensing surface temperature, the surface cover type, and the reanalysis meteorological data, calculate the surface maximum relative humidity, and output the global rasterized result.

2. The global surface maximum relative humidity inversion method based on physical constraints and machine learning according to claim 1, wherein The candidate values of the surface maximum relative humidity in the candidate dataset of the surface maximum relative humidity are generated by discretization within a preset range.

3. The method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning according to claim 1, characterized in that The expression of the physical constraint model includes: wherein, h s,max is the maximum relative humidity at the ground surface; e s is the ground surface vapor pressure; Ω is the land-air coupling factor; e a is the actual water vapor pressure; e s * is the saturated vapor pressure at the ground surface; Δ is the slope of the saturated water vapor pressure curve; γ is the psychrometric constant; r s is the surface impedance; r ae is the aerodynamic impedance; ρ is the air density; C p is the specific heat of air; T s is the ground surface temperature; T a is the air temperature; R n is the net radiation at the ground surface; G is the soil heat flux; ET is the latent heat flux.

4. The method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning according to claim 1, wherein Input the candidate dataset of the surface maximum relative humidity and the synchronous meteorological elements into the physical constraint model together, calculate the simulated latent heat flux corresponding to each candidate value, and compare the simulated latent heat flux with the measured latent heat flux to select the candidate value with the smallest error in the corresponding simulated latent heat flux, and determine the candidate value with the smallest error as the optimal surface maximum relative humidity of the flux observation site, including: Combine the candidate maximum relative humidity value and the synchronous meteorological elements to form a multi-dimensional input feature array; Input the multi-dimensional input feature array into the physical constraint model to calculate the simulated latent heat flux corresponding to each candidate value by using the physical constraint model; Compare the simulated latent heat flux with the measured latent heat flux one by one to calculate the error of the simulated latent heat flux corresponding to each candidate value; Select the candidate value with the smallest error and determine it as the optimal surface maximum relative humidity of the flux observation site.

5. The method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning according to claim 1, wherein Train a machine learning inversion model with the optimal surface maximum relative humidity and synchronous meteorological elements as samples, including: Classify the samples composed of the optimal surface maximum relative humidity and synchronous meteorological elements according to the surface cover type to form sample sets of multiple categories; Pair the optimal surface maximum relative humidity and the corresponding synchronous meteorological elements with the sample sets of each category to construct feature vectors; Perform data augmentation on the feature vectors, and divide the data-augmented feature vectors into a training set and a validation set; Use the training set to fit a preset initial model, and use the validation set for model evaluation to optimize the model parameters to obtain the final machine learning inversion model.

6. The method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning according to claim 1, wherein The machine learning inversion model includes any one of random forest, support vector machine, and neural network.

7. The method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning according to claim 1, wherein The method of model evaluation is the cross-validation method.

8. The method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning according to claim 1, characterized in that, The surface cover types include: forest, grassland, urban area, farmland, and water body.

9. The method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning according to claim 1, wherein The verification metrics of the machine learning inversion model include root mean square error and coefficient of determination.

10. The method for retrieving the global surface maximum relative humidity based on physical constraints and machine learning according to claim 1, wherein The numerical range of the candidate dataset of the maximum surface relative humidity is [0, 1].

Images