Surface water heat flux integrated estimation method, system and device

Abstract

The application provides a kind of ground surface water heat flux integrated estimation method, system and equipment, related to hydrology meteorological remote sensing technical field, the method comprises: according to target area target ground surface temperature data, target net radiation data, target air temperature data, target vegetation coverage data and target soil moisture data, the target ground surface environment key parameter corresponding to target area is constructed;Target ground surface environment key parameter is input into ground surface water heat flux integrated estimation model, and the target ground surface sensible heat, latent heat, soil heat flux estimation value corresponding to target area in daily time scale is obtained by ground surface water heat flux integrated estimation model simultaneously output, wherein, ground surface water heat flux integrated estimation model is based on sample ground surface environment key parameter and sample ground surface environment key parameter corresponding sample ground surface water heat flux observation data, and the physical information neural network machine learning model is trained to obtain.The application improves the accuracy of ground surface water heat flux estimation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hydro-meteorological remote sensing, and in particular to a surface water heat flux integrated estimation method, system and device. BACKGROUND

[0002] The surface water heat flux refers to the exchange of water and heat between the earth's surface and the atmosphere, which describes the dynamic exchange process of energy and matter between the surface and the atmosphere, including sensible heat flux (sensible heat flux), latent heat flux and soil heat flux.

[0003] The existing water heat flux remote sensing estimation method is to estimate the latent heat flux, sensible heat flux and soil heat flux respectively, which has many limitations, for example, the empirical regression method depends on the calibrated coefficient and has poor regional applicability; the Penman formula is complex in parameterization; the sensible heat flux is affected by the estimation of roughness length; the soil heat flux ignores heat conduction; the energy balance residual method has error transmission, etc. It is difficult to accurately estimate the surface water heat flux estimation result under complex surface conditions by using the existing water heat flux remote sensing estimation method.

[0004] Therefore, there is an urgent need for a surface water heat flux integrated estimation method, system and device to solve the above problems. SUMMARY

[0005] In view of the problems existing in the prior art, the present application provides a surface water heat flux integrated estimation method, system and device.

[0006] The present application provides a surface water heat flux integrated estimation method, comprising:

[0007] According to the target surface temperature data, the target net radiation data, the target air temperature data, the target vegetation coverage data and the target soil moisture data of the target region, the target surface environment key parameters corresponding to the target region are constructed;

[0008] The target surface environment key parameters are input into the surface water heat flux integrated estimation model to obtain the target surface water heat flux estimation value corresponding to the target region at the daily time scale output by the surface water heat flux integrated estimation model, wherein the target surface water heat flux estimation value includes a sensible heat flux estimation value, a latent heat flux estimation value and a soil heat flux estimation value; the surface water heat flux integrated estimation model is obtained by training a physical information neural network machine learning model based on sample surface environment key parameters and sample surface water heat flux observation data corresponding to the sample surface environment key parameters.

[0009] According to the surface water heat flux integrated estimation method provided by the present application, the surface water heat flux integrated estimation model is obtained by the following steps:

[0010] Obtain sample surface temperature data and sample net radiation data corresponding to the intraday time scale of the sample area;

[0011] Obtain sample temperature data corresponding to the intraday time scale of the sample area;

[0012] Obtaining sample vegetation coverage data and sample soil moisture data corresponding to the intraday time scale of the sample area;

[0013] Performing time resolution conversion processing on the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data, and the sample soil moisture data to obtain sample surface temperature data, sample net radiation data, sample air temperature data, sample vegetation cover data, and sample soil moisture data with the same time resolution; and constructing key parameters of the sample surface environment based on the sample surface temperature data, sample net radiation data, sample air temperature data, sample vegetation cover data, and sample soil moisture data with the same time resolution;

[0014] Based on the ground flux station observation data, obtaining sample surface water heat flux observation data of the sample area at the intraday time scale, wherein the sample surface water heat flux observation data includes sensible heat flux observation value, latent heat flux observation value and soil heat flux observation value;

[0015] Constructing surface water heat flux labels corresponding to the sample surface environment key parameters based on the sample surface water heat flux observation data, and constructing a training sample set based on the sample surface environment key parameters marked with the surface water heat flux labels;

[0016] Based on the training sample set, the physical information neural network machine learning model is trained, and after determining that the training result satisfies the preset data loss function, the surface water heat flux integrated estimation model is obtained, wherein the preset data loss function is based on the error between the surface water heat flux prediction value and the sample surface water heat flux observation data, and the surface energy balance equation.

[0017] According to the integrated surface water heat flux estimation method provided by the present invention, the preset data loss function is specifically:

[0018] ;

[0019] in, represents the preset data loss function, N represents the total number of samples, Indicates the samples, represents the sample surface temperature data, represents the sample temperature data, represents the sample soil moisture data, represents the sample vegetation coverage data, represents the sample net radiation data, represents the observed value of the sensible heat flux, represents the predicted value of sensible heat flux, represents the predicted value of latent heat flux, represents the predicted soil heat flux, represents the surface energy balance weight coefficient, represents the latent heat flux observation value, represents the soil heat flux observation value.

[0020] According to a method for estimating surface water heat flux integration provided by the present invention, before performing time resolution conversion processing on the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data, and the sample soil moisture data to obtain the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data, and the sample soil moisture data with the same time resolution, the method further includes:

[0021] The sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data and the sample soil moisture data are subjected to data preprocessing to obtain the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data and the sample soil moisture data after data preprocessing, wherein the data preprocessing includes at least missing data filling processing and invalid value removal processing.

[0022] According to a surface water heat flux integrated estimation method provided by the present invention, the sample surface temperature data and the sample net radiation data are obtained based on geostationary satellites; the sample air temperature data are obtained based on atmospheric reanalysis data; and the sample vegetation cover data and the sample soil moisture data are obtained based on polar-orbiting satellites.

[0023] According to a surface water heat flux integrated estimation method provided by the present invention, after inputting the target surface environment key parameters into the surface water heat flux integrated estimation model and obtaining the target surface water heat flux estimation value corresponding to the target area on the intraday time scale output by the surface water heat flux integrated estimation model, the method further includes:

[0024] Get preset time scale information;

[0025] Based on the preset time scale information, the target land surface water heat flux estimation value is accumulated and averaged to obtain the land surface water heat flux estimation value corresponding to different time scales.

[0026] The application also provides a land surface water heat flux integrated estimation system, comprising:

[0027] A data processing module is configured to construct target land surface environment key parameters corresponding to a target region based on target land surface temperature data, target net radiation data, target air temperature data, target vegetation coverage data and target soil moisture data of the target region.

[0028] A flux estimation module is configured to input the target land surface environment key parameters into a land surface water heat flux integrated estimation model to obtain a target land surface water heat flux estimation value corresponding to an intraday time scale of the target region output by the land surface water heat flux integrated estimation model, wherein the target land surface water heat flux estimation value comprises a sensible heat flux estimation value, a latent heat flux estimation value and a soil heat flux estimation value; and the land surface water heat flux integrated estimation model is obtained by training a physical information neural network machine learning model based on sample land surface environment key parameters and sample land surface water heat flux observation data corresponding to the sample land surface environment key parameters.

[0029] The application also provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the land surface water heat flux integrated estimation method according to any one of the above embodiments when executing the program.

[0030] The land surface water heat flux integrated estimation method, system and device provided by the application can improve the accuracy of water heat flux estimation under complex land surface conditions by integrating land surface temperature, net radiation, air temperature, vegetation coverage and soil moisture and other multi-source data of a target region, constructing land surface environment key parameters, inputting the key land surface environment parameters into a water heat flux integrated estimation model trained based on sample land surface environment parameters and synchronous observation water heat flux data, and outputting intraday time scale sensible heat, latent heat and soil heat flux estimation values. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0032] Figure 1 The flowchart of the land surface water heat flux integrated estimation method provided by the present application is shown.

[0033] Figure 2 The overall flow schematic diagram of the ground surface water heat flux integrated estimation method based on stationary satellite provided by the present application is shown in the figure;

[0034] Figure 3 The estimation results of sensible heat, latent heat and soil heat flux obtained by the ground surface water heat flux integrated estimation model provided by the present application are shown in the figure;

[0035] Figure 4 The structure schematic diagram of the ground surface water heat flux integrated estimation system provided by the present application is shown in the figure;

[0036] Figure 5 The structure schematic diagram of the electronic device provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0037] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0038] Sensible heat flux, latent heat flux and soil heat flux are important components of the ground surface energy balance. Among them, sensible heat flux is driven by the temperature gradient between the ground surface and the atmosphere, and energy is transferred in the form of heat conduction and convection, which directly affects the spatial and temporal distribution of atmospheric temperature; latent heat flux is derived from the evaporation or condensation process of water, which absorbs or releases heat during phase change, regulates atmospheric humidity and drives cloud and precipitation formation; soil heat flux is driven by the vertical gradient of soil temperature, which reflects the heat storage or release process between soil and ground surface, and is an important buffer link of regional energy budget.

[0039] Stationary meteorological satellites can implement continuous and high-frequency observation on target areas by being deployed in geosynchronous orbit, which provides unique technical support for quantifying ground surface water heat flux. The high-frequency observation capability of stationary meteorological satellites can capture the diurnal dynamic changes of water heat flux, breaking through the spatial and temporal limitations of traditional ground observation, and providing key data support for extreme weather warning (such as high temperature heat wave, rainstorm and flood), climate change assessment (such as regional energy cycle evolution), agricultural water resource management (such as crop water consumption monitoring) and ecological sustainable development (such as urban heat island effect regulation). By analyzing the water heat flux characteristics in satellite remote sensing data, the scientificity and accuracy of multi-scale climate prediction and water resource management decision can be significantly improved.

[0040] The existing water heat flux remote sensing estimation method estimates latent heat flux, sensible heat flux and soil heat flux respectively. Among them, the latent heat flux in the surface water heat flux, that is, the evapotranspiration, is the core parameter of the surface energy balance and water balance.

[0041] The existing estimation method of latent heat flux can be divided into direct parameterization of latent heat flux and energy balance residual method based on sensible heat flux estimation. Among them, the direct parameterization method includes empirical regression method and method based on Penman formula. Specifically, the empirical regression method mainly estimates by directly establishing the empirical relationship between latent heat flux and influencing variables. Evapotranspiration is the result of the comprehensive action of surface energy, temperature, vegetation, soil and atmosphere, etc. The variables that can be used at present include atmospheric driving wind temperature humidity pressure, vegetation index, surface temperature and solar radiation, etc. Under the background of the development of machine learning technology, such method has been further improved.

[0042] The Penman formula is a method of directly parameterizing latent heat flux, which introduces the concept of canopy resistance into the Penman formula to represent the influence of vegetation physiological action and soil water supply condition on latent heat flux, so as to estimate the actual evaporation of non-saturated underlying surface. The energy balance residual method is to obtain the surface net radiation, soil heat flux and sensible heat flux by using the remote sensing inversed parameters, and then the surface latent heat flux is obtained by the residual of energy balance formula. Sensible heat flux is mainly estimated by using the bulk transfer formula according to the Monin-Obukhov similarity theory. Soil heat flux is mainly estimated by assuming the proportional relationship with net radiation and the correlation with vegetation coverage.

[0043] In the existing ground surface water heat flux estimation method, for the latent heat flux estimation of the empirical regression method, although the method is simple, the input variables are few, and it is convenient to operate, but it depends on the regression coefficient, and the coefficient determination usually needs to be calibrated by ground measurement data, which limits its application in different ground surface conditions at the regional scale;The latent heat flux estimation based on the Penman formula only needs one atmospheric parameter, but the essence is the calculation of various impedances, the input parameters are many, the model is complex, and the parameterization of the impedance is very complex and uncertain under the complex and inhomogeneous underlying surface, which affects the estimation accuracy and application of the latent heat flux;The estimation of sensible heat flux depends on the heat transfer roughness length and the difference between the radiation temperature and the aerodynamic temperature, and needs to parameterize the additional impedance, when the ground surface is partially covered with vegetation, the water and heat exchange between the soil and the vegetation may not be at the same height, which brings difficulty to the estimation of the heat transfer roughness length, and hinders the estimation accuracy and application of the sensible heat flux;The estimation of soil heat flux only assumes the correlation with net radiation and vegetation coverage, ignores the thermal conductivity of the soil itself, and affects the accuracy and application of the estimation;For the latent heat flux estimation of the energy balance residual method, the estimation errors of the sensible heat flux and the soil heat flux will be transferred to the estimation of the latent heat flux. In addition, the existing ground surface water heat flux estimation method is mainly based on the instantaneous observation of the polar orbit satellite, or the estimation on the daily scale, without fully utilizing the characteristics of the continuous observation of the stationary satellite on the ground surface, and the diurnal variation information of the ground surface water heat flux cannot be obtained.

[0044] In view of the problems that the input parameters in the existing sensible heat, latent heat and soil heat flux parameterization scheme are many and difficult to obtain, and the impedances, heat transfer roughness length and soil thermal conductivity are difficult to accurately estimate under complex ground surface conditions, the application provides a ground surface water heat flux integrated estimation method based on a stationary satellite, fully utilizes the characteristics of high-frequency observation of the stationary satellite, solves the equation parameters based on the ground surface energy balance equation through the machine learning method, no longer estimates the parameters such as impedance, heat transfer roughness length and soil thermal conductivity which are difficult to quantify, and realizes the simultaneous estimation of the diurnal variation information of the sensible heat, latent heat and soil heat flux.

[0045] Figure 1 The flowchart of the ground surface water heat flux integrated estimation method provided by the application is shown in Figure 1 As shown in the figure, the application provides a ground surface water heat flux integrated estimation method, which comprises:

[0046] Step 101, according to the target ground surface temperature data, the target net radiation data, the target air temperature data, the target vegetation coverage data and the target soil moisture data of the target area, the target ground surface environment key parameters corresponding to the target area are constructed.

[0047] In the present application, the target land surface temperature data is the actual temperature of the target area land surface, which can be obtained by geostationary satellite and reflects the thermal state of the land surface at a specific time and space. As one of the important indicators of land surface radiant energy, land surface temperature directly affects the energy exchange process between the land surface and the atmosphere. For example, higher land surface temperature accelerates the evaporation of land surface water, affecting soil moisture and vegetation growth. At the same time, land surface temperature anomalies may also be related to urban heat island effect, drought and other environmental problems, and are an important information indispensable to the construction of key parameters of land surface environment.

[0048] The target net radiation data is the sum of shortwave net radiation and longwave net radiation, representing the actual net energy obtained by the land surface. Shortwave net radiation refers to the difference between total solar radiation reaching the land surface and solar radiation reflected by the land surface, and longwave net radiation refers to the difference between atmospheric longwave radiation and land surface longwave radiation. The target net radiation data can also be obtained by geostationary satellite. Net radiation data is the main energy source driving various physical, chemical and biological processes of the land surface, determining the trend of land surface temperature change and the migration and transformation of water, heat and other substances. For example, in areas with higher net radiation, land surface water evaporation and vegetation transpiration may be more intense, thereby affecting the regional water cycle and climate characteristics.

[0049] The target air temperature data refers to the temperature of the air in the atmosphere of the target area, which can be obtained by atmospheric reanalysis data, reflecting the thermal state of the atmosphere. Changes in air temperature will affect the heat budget of the land surface. For example, in cold nights, the decrease in air temperature may cause the land surface temperature to drop rapidly, affecting the soil freezing depth and the cold resistance of vegetation. At the same time, the land surface temperature will also be fed back to the atmosphere through radiation and convection, affecting the distribution and change of air temperature.

[0050] The target vegetation coverage data refers to the percentage of the vertical projection area of vegetation (including herbaceous plants, shrubs and trees, etc.) in the total area of the statistical region in the target area, which can be obtained by polar orbit satellite. Vegetation coverage reflects the vegetation condition and ecological environment quality of the land surface. Vegetation plays a key role in water cycle, energy balance and climate regulation of the land surface through physiological processes such as photosynthesis and transpiration. For example, areas with high vegetation coverage usually have better water and soil conservation capacity, which can reduce the speed of land surface runoff and reduce soil erosion. At the same time, vegetation can also improve the local microclimate by blocking sunlight and reducing land surface temperature.

[0051] The target soil moisture data represent the content of water in the soil of the target area, and can also be obtained by a polar orbit satellite. Soil moisture is an important part of the terrestrial ecosystem, directly affecting the growth, development and distribution of vegetation. Suitable soil moisture conditions can provide sufficient water supply for plants, promote photosynthesis and material metabolism of plants. Excessive or insufficient soil moisture will adversely affect plant growth, such as causing plant root hypoxia, wilting, and even death. In addition, soil moisture also participates in the water cycle process of the ground surface, affecting the links of surface runoff, infiltration and evaporation, etc.

[0052] After collecting the above data, the collected raw data is subjected to quality inspection and correction to remove abnormal values and noise interference, so as to ensure the accuracy and reliability of the data and facilitate subsequent analysis and comparison.

[0053] In step 102, the target surface environment key parameters are input into the integrated surface water and heat flux estimation model to obtain the target surface water and heat flux estimation value corresponding to the target area at the intraday time scale output by the integrated surface water and heat flux estimation model, wherein the target surface water and heat flux estimation value includes sensible heat flux estimation value, latent heat flux estimation value and soil heat flux estimation value; the integrated surface water and heat flux estimation model is obtained by training a physical information neural network machine learning model based on sample surface environment key parameters and sample surface water and heat flux observation data corresponding to the sample surface environment key parameters.

[0054] In the present application, in the construction stage of the integrated surface water and heat flux estimation model, the machine learning model is trained by using sample surface environment key parameters and corresponding sample surface water and heat flux observation data; in the application stage, the target surface environment key parameters of the target area are input into the integrated surface water and heat flux estimation model, and the integrated surface water and heat flux estimation model outputs the target surface water and heat flux estimation value of the target area at the intraday time scale, and the estimation value includes sensible heat flux, latent heat flux and soil heat flux.

[0055] Specifically, in the present application, the target surface environment key parameters are a series of parameters reflecting the key characteristics and states of the surface environment of the target area, including surface temperature, net radiation, air temperature, vegetation coverage and soil moisture, etc. These parameters comprehensively describe the surface environment conditions of the target area and are important input information for estimating the surface water and heat flux, helping the integrated surface water and heat flux estimation model to understand the surface environment characteristics of the target area, so as to more accurately estimate the surface water and heat flux.

[0056] The sample ground surface environment key parameter is ground surface environment key parameter data collected from multiple representative regions, which should have certain diversity in geographical environment, climate condition and vegetation type, etc., to ensure that the integrated estimation model of ground surface water and heat flux can learn the variation law of water and heat flux under different ground surface environments.

[0057] The sample ground surface water and heat flux observation data is actual ground surface water and heat flux data of the sample ground surface obtained through field observation, including sensible heat flux, latent heat flux and soil heat flux, which are used as real labels of the integrated estimation model of ground surface water and heat flux to train and verify the accuracy of the model.

[0058] In the training process, the sample ground surface environment key parameter diurnal variation continuous information is taken as input, and the corresponding sample ground surface water and heat flux observation data diurnal variation is taken as output, and the physical information neural network machine learning model is trained. The physical information neural network machine learning model learns the complex nonlinear relationship between the input parameters and the output results by continuously adjusting its own parameters until the physical information neural network machine learning model can achieve good prediction accuracy on the training set and the validation set. The trained physical information neural network machine learning model (i.e. the integrated estimation model of ground surface water and heat flux) can quickly and accurately estimate the ground surface water and heat flux according to the input ground surface environment key parameters.

[0059] In the present application, the integrated estimation model of ground surface water and heat flux estimates the ground surface water and heat flux on the diurnal time scale, which refers to the time range within a day, divided by hours, half hours or even smaller time intervals, which can more carefully understand the daily variation process of energy and water exchange between the ground surface and the atmosphere, such as the water and heat flux characteristics difference of daytime and nighttime, different time periods, etc., which is helpful for in-depth analysis of the daily variation law of ground surface water and heat flux and its influencing factors, such as the daily variation of solar radiation, the daily rhythm of vegetation physiological activity, etc. In the present application, the integrated estimation model of ground surface water and heat flux outputs the estimated value of target ground surface water and heat flux of target region on the diurnal time scale, including the estimated value of sensible heat flux, the estimated value of latent heat flux and the estimated value of soil heat flux, which are presented in the form of time series to show the variation of ground surface water and heat flux at different times.

[0060] Further, the output estimated results are analyzed, such as calculating the average value, maximum value and minimum value of water and heat flux at different time periods, drawing the daily variation curve of water and heat flux, etc.; at the same time, the estimated results can be compared and verified with the measured data (if any) or the estimated results of other models to evaluate the accuracy and reliability of the model. If the estimated results have large deviation, the reasons need to be further analyzed, such as the quality of input data, the parameter setting of the model, etc., and the model needs to be optimized and improved.

[0061] The application provides a ground surface water heat flux integrated estimation method, which integrates ground surface temperature, net radiation, air temperature, vegetation coverage and soil moisture and other multi-source data of a target area to construct ground surface environment key parameters, and then inputs the key ground surface environment parameters into a water heat flux integrated estimation model trained by sample ground surface environment parameters and synchronous observed water heat flux data, so as to output sensible heat, latent heat and soil heat flux estimation values in an intraday time scale, and significantly improve the accuracy of water heat flux estimation under complex ground surface conditions.

[0062] On the basis of the above embodiment, the ground surface water heat flux integrated estimation model is trained by the following steps:

[0063] Obtain sample ground surface temperature data and sample net radiation data of a sample area in the intraday time scale;

[0064] Obtain sample air temperature data of the sample area in the intraday time scale;

[0065] Obtain sample vegetation coverage data and sample soil moisture data of the sample area in the intraday time scale;

[0066] Perform time resolution conversion processing on the sample ground surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation coverage data and the sample soil moisture data to obtain sample ground surface temperature data, sample net radiation data, sample air temperature data, sample vegetation coverage data and sample soil moisture data with the same time resolution; and based on the sample ground surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation coverage data and the sample soil moisture data with the same time resolution, sample ground surface environment key parameters are constructed;

[0067] Based on ground flux station observation data, obtain sample ground surface water heat flux observation data of the sample area in the intraday time scale, wherein the sample ground surface water heat flux observation data includes sensible heat flux observation value, latent heat flux observation value and soil heat flux observation value;

[0068] According to the sample ground surface water heat flux observation data, construct ground surface water heat flux labels corresponding to the sample ground surface environment key parameters, and based on the sample ground surface environment key parameters after the ground surface water heat flux labels are marked, a training sample set is constructed;

[0069] Based on the training sample set, a physical information neural network machine learning model is trained, and after a training result is determined to satisfy a preset data loss function, the integrated estimation model of the surface water heat flux is obtained, wherein the preset data loss function is constructed based on an error between a surface water heat flux prediction value and the sample surface water heat flux observation data, and a surface energy balance equation.

[0070] In the present application, the time resolution of the obtained sample surface temperature, net radiation, air temperature, vegetation coverage and soil moisture data may be inconsistent due to different data collection time intervals of different data sources. Through time resolution conversion processing, these data are unified to the same time interval, such as 1 hour, 30 minutes, etc. Specifically, in the present application, interpolation methods (such as linear interpolation, spline interpolation) can be used to downscale resample the data with higher time resolution, or to upscale resample the data with lower time resolution. For example, if the surface temperature data is collected every half hour, and the vegetation coverage data only has one value per day, the vegetation coverage data with the same time resolution as the surface temperature data can be obtained by interpolating the average vegetation coverage within the day or combining other auxiliary information (such as a vegetation growth model).

[0071] Further, the sample surface temperature, net radiation, air temperature, vegetation coverage and soil moisture data subjected to time resolution conversion processing are integrated together to form a key parameter set that can comprehensively reflect the characteristics of the sample area. These parameters provide input features for subsequent construction of the integrated estimation model of the surface water heat flux, and by analyzing the mutual relationship between them, it is helpful for the model to better understand the influence mechanism of the surface environment on the water heat flux.

[0072] In the present application, the sample surface water heat flux observation data is obtained based on ground flux site observation data, including sensible heat flux observation value, latent heat flux observation value and soil heat flux observation value. The ground flux site calculates the sensible heat flux and latent heat flux between the surface and the atmosphere by measuring parameters such as wind speed, temperature and humidity, and the soil heat flux is measured by a heat flux plate. In the present application, the surface water heat flux observation values at different times within a day are obtained, which can describe the diurnal variation process of energy and water exchange between the surface and the atmosphere in detail, such as the relative size change of sensible heat flux and latent heat flux during the day, and the direction and intensity change of soil heat flux at night.

[0073] Further, the sample ground surface water heat flux observation data (sensible heat flux observation value, latent heat flux observation value and soil heat flux observation value) are taken as labels corresponding to the sample ground surface environment key parameters, that is, the ground surface environment key parameters at each time point are matched with the ground surface water heat flux true value. Then, the sample ground surface environment key parameters labeled with the ground surface water heat flux labels are combined to form a training sample set, each sample containing a group of ground surface environment key parameters (input) and corresponding ground surface water heat flux observation value (output) for subsequent machine learning model training.

[0074] In the present application, the physical information neural network machine learning model can process complex nonlinear relationships and is suitable for estimating the ground surface water heat flux. Specifically, the training sample set is input into the physical information neural network machine learning model, and the physical information neural network machine learning model learns the mapping relationship between the input ground surface environment key parameters and the output ground surface water heat flux observation value by continuously adjusting its own parameters (such as the weights and biases of the neural network, etc.). During the training process, the training sample set is divided into a training set and a validation set, the training set is used for updating the model parameters, and the validation set is used to evaluate the generalization ability of the model to prevent overfitting.

[0075] In the present application, the preset data loss function is based on the error between the ground surface water heat flux prediction value and the sample ground surface water heat flux observation data, and is constructed based on the ground surface energy balance equation. Not only the difference between the prediction value and the observation value (such as mean square error MSE, mean absolute error MAE) is considered, but also the prediction result is constrained through the ground surface energy balance equation to ensure that the estimated ground surface water heat flux satisfies the law of conservation of energy. On the basis of the above embodiment, the preset data loss function is specifically:

[0076] ;

[0077] Wherein, represents the preset data loss function, N represents the total number of samples, represents the i-th sample, represents the sample ground surface temperature data, represents the sample air temperature data, represents the sample soil moisture data, represents the sample vegetation coverage data, represents the sample net radiation data, represents the sensible heat flux observation value, represents the sensible heat flux prediction value, represents the latent heat flux prediction value, represents the soil heat flux prediction value, represents the soil heat flux prediction value, represents the surface energy balance weight coefficient, represents the latent heat flux observation value, represents the soil heat flux observation value.

[0078] During the training process, the value of the loss function is continuously calculated. When the value of the loss function meets the preset threshold (that is, the training result meets the preset data loss function), it is considered that the machine learning model has learned the relationship between the input parameters and the output results. At this time, the training is stopped and the integrated surface water and heat flux estimation model is obtained, which can be used to estimate the surface water and heat flux in other regions or unobserved periods.

[0079] On the basis of the above embodiment, before performing time resolution conversion processing on the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data, and the sample soil moisture data to obtain the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data, and the sample soil moisture data with the same time resolution, the method further includes:

[0080] The sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data and the sample soil moisture data are subjected to data preprocessing to obtain the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data and the sample soil moisture data after data preprocessing, wherein the data preprocessing includes at least missing data filling processing and invalid value removal processing.

[0081] In the present application, after obtaining the sample land surface temperature, sample net radiation, sample air temperature, sample vegetation coverage and sample soil moisture data, due to the influence of various factors such as equipment precision, environmental interference and transmission error in the data collection process, there are often missing values and invalid values in the original data. These abnormal data will interfere with the subsequent model training and analysis, and reduce the accuracy and reliability of the model. Therefore, the present application uses data preprocessing to clean and correct these data, ensuring the quality and integrity of the data, and providing high-quality input data for building an accurate integrated estimation model of land surface water and heat flux. For example, during data transmission, due to network problems, signal interference and other reasons, part of the data may not be successfully transmitted to the data center, resulting in data missing. For time series data, if the data changes relatively smoothly within a certain period of time before and after the missing data, the average value of the data within a certain time window before and after the missing data can be used to fill in. When calculating the diurnal variation of land surface temperature, if the data of a certain hour is missing, the average value of the land surface temperature of the two hours before and after the hour can be taken as the filling value. If the data is linearly changed near the missing point, the value of the missing point can be calculated according to the values of the two known data points before and after the missing point and the time interval between them. For example, if the air temperature of a certain area is known to be 20 degrees Celsius at 10 am and 24 degrees Celsius at 12 pm, the air temperature at 11 am can be calculated by linear interpolation to be 22 degrees Celsius.

[0082] In the present application, the data collection equipment may have measurement errors, resulting in data collected beyond a reasonable range. For example, the accuracy of temperature sensors is limited, and in extreme environments, it may measure obviously unreasonable temperature values. And, during data transmission, due to data packet loss, damage or encoding errors, the received data may appear garbled or invalid values. Or, some environmental factors may interfere with the normal operation of the data collection equipment, producing invalid data, for example, strong electromagnetic interference may affect the signal transmission of the sensor, causing the collected data to appear abnormal fluctuations or invalid values.

[0083] Based on the invalid values generated by the above reasons, the present application can set reasonable upper and lower threshold values according to the physical meaning and actual distribution of the data. For data beyond the threshold range, it is considered as invalid value and removed. For example, it is known that the normal variation range of air temperature in a certain area within a year is-20 degrees Celsius to 40 degrees Celsius, then the air temperature data lower than-20 degrees Celsius or higher than 40 degrees Celsius is regarded as invalid value and removed. Or, combined with the internal logical relationship of the data, for example, the land surface temperature in clear sky during the day usually will not be lower than the air temperature, if the land surface temperature is obviously lower than the air temperature and does not conform to the actual physical process, it can be determined as invalid value.

[0084] After the missing data filling and invalid value removal processing, the obtained sample land surface temperature data, sample net radiation data, sample air temperature data, sample vegetation coverage data and sample soil moisture data have higher quality and integrity. These high-quality data can provide reliable input features for subsequent construction of integrated estimation model of land surface water and heat flux, and help to improve the training effect and prediction accuracy of the model. At the same time, the complete data set is also convenient for more in-depth data analysis and research, such as analyzing the correlation between key parameters of the land surface environment and the land surface water and heat flux, exploring the daily and seasonal variation of the land surface water and heat flux, etc.

[0085] On the basis of the above embodiment, the sample land surface temperature data and the sample net radiation data are obtained based on a stationary satellite; the sample air temperature data is obtained based on atmospheric reanalysis data; and the sample vegetation coverage data and the sample soil moisture data are obtained based on a polar orbit satellite.

[0086] The stationary satellite has the same running period as the earth rotation period and remains stationary relative to the earth surface. This characteristic enables the stationary satellite to continuously and uninterruptedly observe a specific area and obtain high temporal resolution data. In the present application, for the two parameters of land surface temperature and net radiation which need to capture the intra-day dynamic changes, based on the multiple sensors carried on the stationary satellite, physical quantities related to land surface temperature and net radiation can be measured. For example, the thermal infrared sensor on the stationary satellite can detect the infrared radiation emitted by the earth surface and convert the infrared radiation energy into land surface temperature values. These sensors have high spatial resolution and temporal resolution and can provide land surface temperature data at different time scales (such as hours, minutes), meeting the demand of monitoring the dynamic changes of land surface temperature for land surface water and heat flux estimation.

[0087] Atmospheric reanalysis data is a set of global or regional scale atmospheric state variable data sets generated by combining observation data (including ground meteorological station observations, satellite remote sensing observations and sounding data, etc.) with numerical weather prediction models using data assimilation technology. Atmospheric reanalysis data comprehensively considers the information of multiple observation data sources and fills in the gaps of observation data through model simulation, providing continuous and complete meteorological element field data. In the present application, for the sample air temperature data required for land surface water and heat flux estimation, atmospheric reanalysis data can provide intra-day continuous air temperature sequence. These data not only have high temporal resolution (such as every hour), but also have good spatial coverage, which can make up for the spatial and temporal limitations of ground meteorological station observations. For example, in some areas where ground meteorological stations are sparsely distributed, atmospheric reanalysis data can provide relatively reliable air temperature information, ensuring the integrity of the input variables required for land surface water and heat flux estimation.

[0088] Polar orbit satellites can provide global high-resolution observation data. Polar orbit satellites are usually equipped with multiple types of sensors, including multispectral sensors and microwave sensors, etc., wherein the multispectral sensors can obtain the spectral reflection information of the ground surface at different wavebands, and the vegetation coverage can be inverted through vegetation index algorithms such as vegetation index; the microwave sensor is sensitive to soil moisture and can penetrate clouds and vegetation to obtain soil surface humidity information.

[0089] In the present application, based on the multispectral data of the polar orbit satellite, the normalized difference vegetation index (NDVI), the enhanced vegetation index (EVI) and other vegetation indexes can be used to calibrate and verify the model combined with the ground measured data, and the vegetation coverage data at different time scales (such as multi-day or daily scale) can be inverted, which can reflect the growth condition and spatial distribution of the ground vegetation.

[0090] The microwave sensor (such as active microwave radar and passive microwave radiometer) on the polar orbit satellite is the main means to obtain soil moisture data. The active microwave radar transmits microwave signals and receives ground scattering echoes to obtain soil surface roughness and dielectric constant information, and then inverses soil moisture; the passive microwave radiometer measures the microwave radiation energy emitted by the ground, and inverses soil moisture combined with the radiation transfer model. In the present application, the soil moisture data product obtained by the polar orbit satellite has high spatial resolution and certain time resolution (such as several days to one week), which can reflect the spatial distribution and dynamic change of soil moisture.

[0091] The present application comprehensively utilizes multiple data sources, so that the input of the model can be obtained by remote sensing data, to realize the large-area estimation of the surface water and heat flux, reduce the error and uncertainty of a single data source, and improve the accuracy and reliability of the integrated estimation of the surface water and heat flux.

[0092] On the basis of the above-mentioned embodiments, after the target ground environment key parameter is input into the integrated estimation model of the surface water and heat flux, and the target ground water and heat flux estimation value corresponding to the target region at the intraday time scale output by the integrated estimation model of the surface water and heat flux is obtained, the method further comprises:

[0093] Obtaining preset time scale information;

[0094] Based on the preset time scale information, the target ground water and heat flux estimation value is accumulated and averaged to obtain the ground water and heat flux estimation value corresponding to different time scales.

[0095] In the present application, different application scenarios have different requirements for the time scale of surface water heat flux estimation. For example, in agricultural irrigation management, it is necessary to understand the daily variation of surface water heat flux in order to reasonably arrange the irrigation time and water quantity, and to ensure the water and heat conditions required for crop growth. In the field of climate change research, more attention is often paid to the long-term variation trend of monthly or annual surface water heat flux, which is used to analyze the energy balance and water cycle characteristics of global or regional climate systems.

[0096] Further, by presetting time scale information, the specific time range is determined, such as the daily scale can be set as a natural day (from 0 o'clock to 24 o'clock of the day) or a meteorological day (usually from 20 o'clock of the previous day to 20 o'clock of the day, used for meteorological business statistics); the monthly scale can be divided according to the month, or can be self-defined according to actual needs (such as agricultural seasonal months); the annual scale can be a calendar year or a hydrological year (the start and end times are determined according to the local hydrological characteristics). In addition to determining the time range, the time interval, i.e. the interval time between adjacent two estimation time points, also needs to be determined, for example, in the daily scale estimation, the time interval is 1 hour or half an hour; in the daily scale estimation, the time interval is 1 day; in the monthly scale estimation, the time interval is 1 month; in the annual scale estimation, the time interval is 1 year.

[0097] After obtaining the sensible heat, latent heat and soil heat flux information within a day, the water heat flux estimation values within the corresponding time range are accumulated and averaged according to the preset time scale information, for example, in the daily scale estimation, the sensible heat flux, latent heat flux and soil heat flux estimation values at each time point (such as every hour) within a day are added and averaged respectively to obtain the daily average sensible heat flux, latent heat flux and soil heat flux. The accumulation and averaging process reflects the average of the water heat flux within the time scale, and can directly reflect the income and expenditure of surface energy and water, obtaining the estimation values of different time scales. The average processing method can be selected as arithmetic mean and weighted mean according to specific needs.

[0098] In an embodiment, the accumulation and averaging process of the daily scale is described, assuming that the sensible heat flux, latent heat flux and soil heat flux are calculated every hour, the sensible heat flux estimation values within 24 hours are added to obtain the total sensible heat flux of the day; similarly, the 24-hour estimation values of latent heat flux and soil heat flux are added respectively to obtain the total latent heat flux and total soil heat flux of the day. Then, the total sensible heat flux is divided by 24 hours to obtain the average sensible heat flux of the day; the total latent heat flux and total soil heat flux are respectively divided by 24 hours to obtain the average latent heat flux and average soil heat flux of the day.

[0099] The present application expands the time scale of surface water heat flux estimation by accumulation and averaging processing, which makes the estimation value more reflect the true characteristics of surface water heat flux of different time scales.

[0100] Figure 2 The overall flow diagram of the integrated estimation method of surface water heat flux based on geostationary satellite provided by the present invention can be referred to Figure 2 The figure below provides an overview of the integrated surface water heat flux estimation method provided by the present invention. In the present invention, the geostationary satellites can be selected from the Fengyun series, MSG series, GOES series, or Sunflower series, without any specific limitation. Furthermore, continuous observations from networked polar-orbiting satellites can also provide the input data required by the present invention.

[0101] In this invention, there are five main input variables, including surface temperature ( ), net radiation ( )、Temperature( ), vegetation coverage ( ), soil moisture ( ), among which, the surface temperature and net radiation can be directly obtained from geostationary satellite observations. The three variables of air temperature, vegetation cover and soil moisture are obtained by using the daily continuous air temperature of atmospheric reanalysis data, and the multi-day or daily scale data of vegetation cover and soil moisture based on polar-orbiting satellites.

[0102] Furthermore, the input variable data of the present invention is required to be continuous within a day, and the data time interval is half an hour or hour. The continuous data within a natural day is taken as a valid sample, and the geostationary satellite observation data can meet this time scale requirement; the temperature data from the atmospheric reanalysis data is usually also on the hourly scale, which can meet the input requirements; the vegetation coverage data from the polar-orbiting satellite is usually on the daily or 8-day time scale. Since the vegetation coverage is not sensitive to time changes, for the daily scale data, it can be directly assumed that the time within the day is unchanged. For the 8-day time scale, the present invention uses the interval data of two consecutive times to obtain the daily scale data through linear interpolation, and then assumes that it is unchanged within the day; soil moisture usually does not change much on the daily scale. Assuming that the soil moisture is unchanged within the day, it is directly obtained using the daily scale soil moisture data.

[0103] In this invention, the output estimated data is determined to be hourly or half-hourly based on the final estimation goal, thereby unifying the temporal resolution of the input daily data for surface temperature, air temperature, net radiation, vegetation cover, and soil moisture. In particular, for half-hourly data, the present invention averages the data from two adjacent moments. Because thermal infrared surface temperature data is affected by clouds, direct satellite inversion data often produces invalid values. This invention excludes these invalid values ​​from the calculation.

[0104] In the present invention, the above five input variables are selected to be closely related to the surface water heat flux.H latent heat (Q LE ) and soil heat flux (Q G ) are expressed by the following equations:

[0105] Equation (1)

[0106] Equation (2)

[0107] Equation (3)

[0108] The function relationships in Equations (1), (2) and (3) , and , the present application uses ground flux station observation data to solve parameters by machine learning method, the ground flux station observation data includes sensible heat, latent heat and soil heat flux, the three variables are target variables, the station observation data is usually semi-hourly or hourly scale, by processing the target variables, continuous time scale data (i.e. sample land surface water heat flux observation data of daily time scale) within a day is obtained.

[0109] Since machine learning is a method of solving parameters driven by data, the present application uses available station data and satellite long-time available data consistent with the same, by physical information neural network machine learning method, while accurately optimizing the sensible heat, latent heat and soil heat flux, the physical constraint of the surface energy balance equation is obeyed, a preset data loss function is constructed, which is expressed by the following equation:

[0110] Equation (4)

[0111] In the machine learning process of the neural network, the error of the above-mentioned Equation (4) loss function is ensured to be minimum, considering that the scale ranges of the observation instruments of the station measured land surface sensible heat, latent heat and soil heat flux are different, there is a problem of non-closed energy balance, the error of the energy imbalance is about 20%, the present application adds a small land surface energy balance weight coefficient to constrain the error caused by the energy balance. In the machine learning process, by continuously training sample data, the number of neural network layers, the number of training times, the number of processing samples, the land surface energy balance weight coefficient and the corresponding parameters for estimating the sensible heat, latent heat and soil heat flux, i.e. the parameters corresponding to the functions , and in Equations (1), (2) and (3) are obtained.

[0112] Finally, the diurnal continuous surface temperature data, net radiation data, air temperature data, vegetation coverage data and soil moisture data of the target area are input into the integrated estimation model of surface water and heat flux, and the function relationship learned by the integrated estimation model of surface water and heat flux is used to calculate the diurnal variation information of sensible heat, latent heat and soil heat flux of the target area.

[0113] Figure 3 The estimation result diagram of sensible heat, latent heat and soil heat flux obtained based on the integrated estimation model of surface water and heat flux provided by the present application is shown in Figure 3 The present application uses the time-continuous data of surface temperature, net radiation, air temperature, vegetation coverage and soil moisture simulated by the process model, and obtains the estimation result of sensible heat, latent heat and soil heat flux by the integrated estimation model of surface water and heat flux, and the error of the estimation result and the measured data is The following is better than the currently commonly used single variable method for estimating the surface water and heat flux, and the estimation result of the present application can well capture the diurnal variation of the surface water and heat flux.

[0114] The present application aims at the problems of single variable estimation, complex model, multiple input parameters, impedance calculation and large estimation uncertainty of the existing integrated estimation method of surface water and heat flux, uses high-frequency observation information of stationary satellite, simplifies the parameterization process of sensible heat, latent heat and soil heat flux, based on the physical constraints of surface energy balance, converts the integrated estimation of surface water and heat flux into solving mathematical equations, acquires unknown parameters by machine learning method, and estimates the diurnal variation of water and heat flux with surface temperature, air temperature, net radiation, vegetation index and soil moisture as input data. In the present application, high-frequency observation information of stationary satellite is used, and fewer input variables are used to solve the impedance calculation problem of relying on many parameter inputs in the existing surface water and heat flux estimation process, and the impedance does not need to be calculated when estimating the surface water and heat flux, and the input variables are less. Unlike the existing single variable estimation method of surface water and heat flux, the present application realizes the simultaneous estimation of sensible heat, latent heat and soil heat flux, avoids the transmission of errors, accurately estimates the sensible heat, latent heat and soil heat flux, and the three also comply with the energy balance. Moreover, unlike the existing surface water and heat flux estimation which mainly focuses on the water and heat flux estimation at the passing moment or daily scale of polar orbit satellite, the present application provides the diurnal variation information of surface water and heat flux based on stationary satellite, which can meet the research of the variation law of surface water and heat flux in a large area.

[0115] The integrated estimation system of surface water and heat flux provided by the present application is described below, and the integrated estimation system of surface water and heat flux described below can be correspondingly referred to the integrated estimation method of surface water and heat flux described above.

[0116] Figure 4A structure schematic diagram of a surface water heat flux integrated estimation system provided by the present application is shown in the figure, and the present application provides a surface water heat flux integrated estimation system, which comprises a data processing module 401 and a flux estimation module 402. Figure 4 The present application provides a surface water heat flux integrated estimation system, which comprises a data processing module 401 and a flux estimation module 402.

[0117] The present application provides a surface water heat flux integrated estimation system, which comprises a data processing module 401 and a flux estimation module 402.

[0118] The system provided by the present application is used for executing the above-mentioned method embodiments, and the specific process and detailed content are referred to the above-mentioned embodiments, which will not be described here.

[0119] Figure 5 A structure schematic diagram of an electronic device provided by the present application is shown in the figure, and the present application provides an electronic device, which comprises a processor 501, a memory 502 and a communication interface 503. Figure 5As shown, the electronic device can include a processor 501, a communications interface 502, a memory 503, and a communications bus 504, wherein the processor 501, the communications interface 502, and the memory 503 complete mutual communication through the communications bus 504. The processor 501 can invoke a logical instruction in the memory 503 to execute the integrated estimation method of the surface water heat flux, which includes: constructing a target ground environment key parameter corresponding to a target region according to target ground temperature data, target net radiation data, target air temperature data, target vegetation coverage data, and target soil moisture data of the target region; inputting the target ground environment key parameter into a surface water heat flux integrated estimation model to obtain a target ground water heat flux estimation value corresponding to the target region at an intraday time scale output by the surface water heat flux integrated estimation model, wherein the target ground water heat flux estimation value includes a sensible heat flux estimation value, a latent heat flux estimation value, and a soil heat flux estimation value; and the surface water heat flux integrated estimation model is obtained by training a physical information neural network machine learning model based on sample ground environment key parameters and sample ground water heat flux observation data corresponding to the sample ground environment key parameters.

[0120] The apparatus embodiments described above are merely illustrative, wherein the units shown as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment scheme. Those skilled in the art can understand and implement without creative labor.

[0121] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0122] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit the same; and although the present application has been described in detail with reference to the foregoing embodiments, it should be appreciated by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features thereof can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A surface water heat flux integrated estimation method, characterized in that: include: According to the target surface temperature data, target net radiation data, target air temperature data, target vegetation coverage data and target soil moisture data of the target area, the target surface environment key parameters corresponding to the target area are constructed; Inputting the target surface environment key parameters into the surface water heat flux integrated estimation model, obtaining the target surface water heat flux estimation value corresponding to the target area on the intraday time scale output by the surface water heat flux integrated estimation model, wherein the target surface water heat flux estimation value includes the sensible heat flux estimation value, the latent heat flux estimation value, and the soil heat flux estimation value; the surface water heat flux integrated estimation model is obtained by training a physical information neural network machine learning model based on the sample surface environment key parameters and the sample surface water heat flux observation data corresponding to the sample surface environment key parameters; The surface water and heat flux integrated estimation model is trained through the following steps: Obtain sample surface temperature data and sample net radiation data corresponding to the intraday time scale of the sample area; Obtain sample temperature data corresponding to the intraday time scale of the sample area; Obtaining sample vegetation coverage data and sample soil moisture data corresponding to the intraday time scale of the sample area; Performing time resolution conversion processing on the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data, and the sample soil moisture data to obtain sample surface temperature data, sample net radiation data, sample air temperature data, sample vegetation cover data, and sample soil moisture data with the same time resolution; and constructing key parameters of the sample surface environment based on the sample surface temperature data, sample net radiation data, sample air temperature data, sample vegetation cover data, and sample soil moisture data with the same time resolution; Based on the ground flux station observation data, obtaining sample surface water heat flux observation data of the sample area at the intraday time scale, wherein the sample surface water heat flux observation data includes sensible heat flux observation value, latent heat flux observation value and soil heat flux observation value; Constructing surface water heat flux labels corresponding to the sample surface environment key parameters based on the sample surface water heat flux observation data, and constructing a training sample set based on the sample surface environment key parameters marked with the surface water heat flux labels; Based on the training sample set, the physical information neural network machine learning model is trained, and after determining that the training result satisfies a preset data loss function, the surface water heat flux integrated estimation model is obtained, wherein the preset data loss function is constructed based on the error between the surface water heat flux prediction value and the sample surface water heat flux observation data, and the surface energy balance equation; The preset data loss function is specifically: ； in, represents the preset data loss function, N represents the total number of samples, Indicates the samples, represents the sample surface temperature data, represents the sample temperature data, represents the sample soil moisture data, represents the sample vegetation coverage data, represents the sample net radiation data, represents the observed value of the sensible heat flux, represents the predicted value of sensible heat flux, represents the predicted value of latent heat flux, represents the predicted soil heat flux, represents the surface energy balance weight coefficient, represents the latent heat flux observation value, represents the soil heat flux observation value.

2. The surface water heat flux integrated estimation method according to claim 1, characterized in that: Before performing time resolution conversion processing on the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data, and the sample soil moisture data to obtain the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data, and the sample soil moisture data with the same time resolution, the method further includes: The sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data and the sample soil moisture data are subjected to data preprocessing to obtain the sample surface temperature data, the sample net radiation data, the sample air temperature data, the sample vegetation cover data and the sample soil moisture data after data preprocessing, wherein the data preprocessing includes at least missing data filling processing and invalid value removal processing.

3. The integrated surface water heat flux estimation method according to claim 1, characterized in that: The sample surface temperature data and the sample net radiation data are obtained based on geostationary satellites; the sample air temperature data are obtained based on atmospheric reanalysis data; and the sample vegetation cover data and the sample soil moisture data are obtained based on polar-orbiting satellites.

4. The integrated surface water heat flux estimation method according to claim 1, characterized in that: After inputting the target surface environment key parameters into the surface water heat flux integrated estimation model to obtain the target surface water heat flux estimation value corresponding to the target area on a daily time scale output by the surface water heat flux integrated estimation model, the method further includes: Get preset time scale information; Based on the preset time scale information, the target surface water heat flux estimation value is cumulatively averaged to obtain the surface water heat flux estimation values ​​corresponding to different time scales.

5. A surface water heat flux integrated estimation system based on the surface water heat flux integrated estimation method according to any one of claims 1 to 4, characterized in that: include: A data processing module is used to construct target surface environment key parameters corresponding to the target area based on target surface temperature data, target net radiation data, target air temperature data, target vegetation coverage data and target soil moisture data of the target area; A flux estimation module is used to input the target surface environment key parameters into the surface water heat flux integrated estimation model to obtain the target surface water heat flux estimation value corresponding to the target area on the daily time scale output by the surface water heat flux integrated estimation model, wherein the target surface water heat flux estimation value includes the sensible heat flux estimation value, the latent heat flux estimation value and the soil heat flux estimation value; the surface water heat flux integrated estimation model is obtained by training a physical information neural network machine learning model based on the sample surface environment key parameters and the sample surface water heat flux observation data corresponding to the sample surface environment key parameters.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the integrated surface water heat flux estimation method according to any one of claims 1 to 4 is implemented.

Images