Passive microwave soil moisture product downscaling method fused with surface heterogeneity

Abstract

The invention relates to the technical field of microwave remote sensing, and provides a passive microwave soil moisture product downscaling method fusing earth surface heterogeneity, which comprises the following steps: determining respective earth surface heterogeneity and earth surface environment data under a first spatial resolution and a second spatial resolution based on original earth surface environment data; inputting the surface environment data and the surface heterogeneity under the first spatial resolution into a downscaling relation model to obtain downscaling soil moisture data under the first spatial resolution output by the downscaling relation model; the scale of the second spatial resolution is thicker than that of the first spatial resolution; the downscaling relation model is obtained by training based on the earth surface environment data and earth surface heterogeneity under the second spatial resolution and the downscaling soil moisture data under the second spatial resolution, so that the defects of poor downscaling precision and reliability at present are overcome, the spatial resolution of the passive microwave soil moisture product is effectively improved, and the method is suitable for popularization and application. And the application requirement of the regional scale can be met.

Description

Technical Field

[0001] The present invention relates to the field of microwave remote sensing technology, and in particular to a passive microwave soil moisture product downscaling method integrating surface heterogeneity. Background Art

[0002] As a key variable in terrestrial ecosystems and the hydrological cycle, soil moisture plays a vital role in climate change, crop yield estimation, and water resource management. Obtaining high-precision, high-temporal and spatial resolution soil moisture information is crucial for understanding ecosystems and guiding agricultural production. Passive microwave remote sensing, due to its all-day, all-weather coverage and its immunity to cloud and fog, has become an important tool for large-scale soil moisture monitoring. However, current passive microwave soil moisture products suffer from low spatial resolution, making them inadequate for regional-scale applications. To address this issue, downscaling methods have emerged as an effective means of improving the spatial resolution of passive microwave soil moisture products.

[0003] Currently, downscaling methods for passive microwave soil moisture products primarily include empirical, semi-empirical, and physical downscaling methods. Physical downscaling methods have a strong theoretical foundation, but they involve large amounts of data, are complex to run and debug, and are difficult to apply in practice. Semi-empirical methods have a certain physical background, but simplify the model or some parameters, resulting in a lack of a complete description of complex physical processes and an inability to accurately reflect the intrinsic relationship between soil moisture and downscaling factors. In contrast, empirical methods, particularly machine learning, have greater application potential. However, the downscaling results of current machine learning methods in practical applications deviate significantly from the actual situation, resulting in poor downscaling accuracy and reliability. Summary of the Invention

[0004] The present invention provides a downscaling method for passive microwave soil moisture products that integrates surface heterogeneity, so as to solve the problem that the downscaling methods in the prior art have poor accuracy and reliability and cannot effectively improve the spatial resolution of passive microwave soil moisture products.

[0005] The present invention provides a downscaling method for passive microwave soil moisture products integrating surface heterogeneity, comprising: Obtain original surface environmental data; Based on the original surface environmental data, determining surface heterogeneity and surface environmental data at a first spatial resolution and a second spatial resolution, respectively; Inputting the surface environmental data and surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; The scale of the second spatial resolution is coarser than that of the first spatial resolution; the downscaling relationship model is trained based on surface environmental data and surface heterogeneity at the second spatial resolution, and downscaled soil moisture data at the second spatial resolution.

[0006] According to a downscaling method for passive microwave soil moisture products integrating surface heterogeneity provided by the present invention, the downscaling relationship model is determined based on the following steps: Determining an initial downscaling model, where the initial downscaling model is constructed based on a random forest method; The downscaling initial model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model.

[0007] According to a method for downscaling a passive microwave soil moisture product integrating surface heterogeneity provided by the present invention, the downscaling initial model is trained based on the surface environment data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model, including: constructing a surface heterogeneity vector based on the surface environmental data and surface heterogeneity at the second spatial resolution; Training the downscaling initial model based on the surface heterogeneity vector and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model; The training goal of the downscaling initial model is to search for optimal parameters through a grid search method, and determine the downscaling relationship model based on the optimal parameters.

[0008] According to a downscaling method for passive microwave soil moisture products integrating surface heterogeneity provided by the present invention, there are multiple surface heterogeneities at each spatial resolution; The step of inputting the surface environmental data and the surface heterogeneity at the first spatial resolution into the downscaling relationship model to obtain the downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model includes: Inputting the surface environmental data and surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain first soil moisture data output by the downscaling relationship model; sorting and screening the surface heterogeneity at the first spatial resolution by importance, and obtaining the most important surface heterogeneity as the target heterogeneity; Based on the target heterogeneity at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

[0009] According to a method for downscaling a passive microwave soil moisture product integrating surface heterogeneity provided by the present invention, the method comprises performing residual correction on the first soil moisture data based on target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution, including: aggregating the first soil moisture data to the second spatial resolution to obtain second soil moisture data; determining a residual at the second spatial resolution based on the second soil moisture data and the downscaled soil moisture data at the second spatial resolution; Resampling the residual at the second spatial resolution to the first spatial resolution to obtain the residual at the first spatial resolution; Based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

[0010] According to a method for downscaling a passive microwave soil moisture product integrating surface heterogeneity provided by the present invention, the method comprises performing residual correction on the first soil moisture data based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution, including: determining a correction residual at the first spatial resolution based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution; Based on the correction residual at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

[0011] According to the present invention, a passive microwave soil moisture product downscaling method integrating surface heterogeneity is provided, wherein the target heterogeneity is terrain heterogeneity, and the surface heterogeneity also includes surface type heterogeneity, vegetation cover heterogeneity and soil texture heterogeneity; The original surface environmental data includes numerical environmental variable data and categorical environmental variable data. The surface heterogeneity is determined by measuring the standard deviation of the numerical environmental variable data, and the surface heterogeneity is obtained by characterizing the categorical environmental variable data through the Gini-Simpson index.

[0012] The present invention also provides a passive microwave soil moisture product downscaling device integrating surface heterogeneity, comprising: A data acquisition unit, used for acquiring original surface environment data; a heterogeneity determination unit, configured to determine surface heterogeneity and surface environment data at a first spatial resolution and a second spatial resolution, respectively, based on the original surface environment data; a data downscaling unit, configured to input the surface environmental data and the surface heterogeneity at the first spatial resolution into a downscaling relationship model, and obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; The scale of the second spatial resolution is coarser than that of the first spatial resolution; the downscaling relationship model is trained based on surface environmental data and surface heterogeneity at the second spatial resolution, and downscaled soil moisture data at the second spatial resolution.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements any of the above-described methods for downscaling passive microwave soil moisture products that integrate surface heterogeneity.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for downscaling passive microwave soil moisture products that incorporate surface heterogeneity.

[0015] The present invention provides a method for downscaling a passive microwave soil moisture product by integrating surface heterogeneity. The method determines surface heterogeneity and surface environmental data at a first spatial resolution and a second spatial resolution, respectively, based on original surface environmental data. The surface environmental data and surface heterogeneity at the first spatial resolution are input into a downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model. The downscaling relationship model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution and the downscaled soil moisture data at the second spatial resolution. This method overcomes the defects of poor downscaling accuracy and reliability in traditional schemes. In the process of constructing the downscaling relationship model, the influence of surface heterogeneity on the spatial distribution of soil moisture is fully considered, thereby making the constructed model more accurate. Downscaling processing based on this method can effectively improve the spatial resolution of passive microwave soil moisture products, meet regional-scale application requirements, and provide more accurate and reliable soil moisture information for soil moisture-related research. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 Schematic diagram of the process of downscaling the passive microwave soil moisture product integrating surface heterogeneity provided by the present invention; Figure 2 is a schematic diagram of the residual correction process based on terrain heterogeneity provided by the present invention; Figure 3 This is the overall framework diagram of the passive microwave soil moisture product downscaling method integrating surface heterogeneity provided by the present invention; Figure 4 This is a comparison chart of the impact of adding heterogeneity on model accuracy provided by the present invention; Figure 5 2. It is a schematic structural diagram of a passive microwave soil moisture product downscaling device integrating surface heterogeneity provided by the present invention; Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0019] As a key variable in terrestrial ecosystems and the hydrological cycle, soil moisture plays a vital role in many fields, including climate change, crop yield estimation, and water resource management. Accurately acquiring soil moisture information with high precision and high temporal and spatial resolution is of paramount importance for understanding ecosystems, guiding agricultural production, and planning water resources. Among the many soil moisture monitoring methods, passive microwave remote sensing has become one of the core methods for large-scale soil moisture monitoring due to its significant advantages, such as its ability to operate around the clock and in all weather conditions and its immunity to cloud and fog interference. However, current passive microwave soil moisture products generally suffer from low spatial resolution, which greatly limits their effectiveness and accuracy in regional-scale applications and makes it difficult to meet the demand for fine-scale soil moisture information in practical work.

[0020] To address these issues, downscaling methods have emerged as an effective way to improve the spatial resolution of passive microwave soil moisture products. Currently, downscaling methods for passive microwave soil moisture products are primarily categorized into three types: empirical, semi-empirical, and physical. Physical model-based downscaling methods have a strong theoretical foundation and employ methods such as assimilating high-resolution data and using land surface process models. While they can provide relatively accurate results, they involve complex models and numerous parameters, making model operation and debugging complex. They also require extensive and accurate input data, making practical application difficult. While semi-empirical methods have a certain physical background, they simplify the model or some parameters and lack a complete description of complex physical processes, resulting in an inability to accurately reflect the relationship between soil moisture and downscaling factors in some cases. In contrast, empirical methods, particularly machine learning methods, have shown greater potential for downscaling soil moisture products and are more readily applicable in practical applications.

[0021] However, it is worth noting that the current machine learning method also has obvious shortcomings, namely, there is a large deviation between the downscaling results in practical applications and the actual situation, the accuracy and reliability of the downscaling are poor, and it is impossible to effectively improve the spatial resolution of passive microwave soil moisture products, and it is still difficult to meet the application needs at the regional scale.

[0022] To this end, the present invention provides a downscaling method for passive microwave soil moisture products that integrates surface heterogeneity, aiming to overcome the limitations of current downscaling methods, consider the surface heterogeneity that affects the spatial distribution of soil moisture, and integrate it into the downscaling process to improve the accuracy of downscaling and reduce the uncertainty in the downscaling process, thereby improving the spatial resolution of passive microwave soil moisture products and meeting regional-scale application requirements.

[0023] Figure 1 This is a flow chart of the method for downscaling passive microwave soil moisture products that incorporates surface heterogeneity, as provided by the present invention. Figure 1 As shown, the method includes: Step 110, obtaining original surface environment data; Step 120 , determining surface heterogeneity and surface environment data at a first spatial resolution and a second spatial resolution, respectively, based on the original surface environment data; Step 130: Inputting the surface environmental data and the surface heterogeneity at the first spatial resolution into the downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; The scale of the second spatial resolution is coarser than that of the first spatial resolution; the downscaling relationship model is trained based on surface environmental data and surface heterogeneity at the second spatial resolution, as well as downscaled soil moisture data at the second spatial resolution.

[0024] Specifically, considering that the current passive microwave soil moisture product downscaling method has poor downscaling accuracy and reliability and cannot effectively improve the spatial resolution of passive microwave soil moisture, in the embodiment of the present invention, the surface heterogeneity that affects the spatial distribution of soil moisture is taken into consideration, and downscaling is performed based on this, which effectively solves the problem of low spatial resolution of passive microwave soil moisture products, improves the downscaling accuracy and reliability, can meet regional scale application needs, and can provide more accurate and reliable soil moisture information for soil moisture-related research. It has important application potential in the field of passive microwave soil moisture product downscaling.

[0025] Specifically, when downscaling, the low spatial resolution of passive microwave soil moisture products leads to significant spatial heterogeneity within the satellite footprint. This means that surface attributes, such as vegetation cover and land type, have a complex spatial distribution. This heterogeneity has a significant impact on the downscaling of soil moisture products. Surface heterogeneity can cause spatial variations in soil moisture distribution and variation. Ignoring this heterogeneity can lead to significant deviations between the downscaling results and the actual situation. Therefore, to ensure effective downscaling, the embodiments of the present invention require determining the environmental variables that influence the spatial distribution of soil moisture, namely, surface heterogeneity and surface environmental data, before downscaling. Surface heterogeneity here can include heterogeneity in land type, vegetation cover, terrain, and soil texture. Surface heterogeneity reflects the spatially uneven distribution of surface features (such as terrain, vegetation, and soil texture).

[0026] Surface heterogeneity can be obtained using high-resolution environmental variables (surface environmental data). Specifically, since high-resolution surface environmental data can effectively characterize the distribution characteristics of surface conditions at coarse-resolution satellite pixels, in embodiments of the present invention, high-resolution surface environmental data can be used to calculate low-resolution surface heterogeneity and corresponding surface environmental data.

[0027] Specifically, here we can first determine the high-resolution surface environment data, that is, the original surface environment data, which may include land cover type data (LC), leaf area index data (LAI), digital elevation model (DEM), soil texture data (ST), etc. at higher spatial resolutions; such as land cover type data at 500m spatial resolution, leaf area index data at 500m spatial resolution, digital surface elevation data at 500m spatial resolution, and soil texture data at 1km spatial resolution.

[0028] Furthermore, surface heterogeneity and surface environmental data at the first and second spatial resolutions can be calculated based on the original surface environmental data. Specifically, surface heterogeneity can be calculated based on the data type of the original surface environmental data, i.e., the type of environmental variable, to obtain surface heterogeneity at the first and second spatial resolutions. Surface environmental data can be directly calculated based on the original surface environmental data through spatial averaging to obtain surface environmental data at the first and second spatial resolutions.

[0029] It should be noted here that the first spatial resolution and the second spatial resolution are both lower than the spatial resolution of the original surface environment data; and the first spatial resolution is higher than the second spatial resolution, which can also be understood as the scale of the second spatial resolution being coarser than the scale of the first spatial resolution, such as the first spatial resolution can be a 5km spatial resolution, and the second spatial resolution can be a 25km spatial resolution; of course, it can also be a spatial resolution of other scales as long as the above restrictions are met, and the embodiments of the present invention do not make specific limitations on this.

[0030] After obtaining the surface heterogeneity and surface environmental data at two spatial resolutions, in an embodiment of the present invention, downscaling can be performed based on this to obtain downscaled soil moisture data. In this way, the downscaling processing of the corresponding passive microwave soil moisture product is completed, and a soil moisture product with higher spatial resolution is obtained, thereby improving the spatial resolution of the passive microwave soil moisture product and better meeting the application requirements at the regional scale.

[0031] Specifically, a relationship model for downscaling, i.e., a downscaling relationship model, can be obtained by first training data at a lower resolution, i.e., surface heterogeneity and surface environmental data at a second spatial resolution. Specifically, an initial model can be determined during the training process. The initial model can be a machine learning model, such as a random forest model, a gradient boosting decision tree, or a feedforward neural network. Next, the surface heterogeneity and surface environmental data at the second spatial resolution can be used to train the initial model, thereby obtaining a trained model, i.e., a downscaling relationship model. The training method used for model training can be conventional supervised training or other training methods, such as semi-supervised training or self-supervised training, which are not specifically limited in the present embodiment.

[0032] Preferably, in an embodiment of the present invention, a supervised training method is used to obtain the downscaling relationship model. Specifically, labels corresponding to the input sample data can be first determined, namely, downscaled soil moisture data corresponding to the surface heterogeneity and surface environmental data at the second spatial resolution, also referred to as downscaled soil moisture data at the second spatial resolution. The downscaled soil moisture data at the second spatial resolution can be obtained by manual labeling or downloaded from the Internet. For example, after determining the sample data, spatial association (via longitude and latitude) and / or temporal association (via date) can be performed based on the sample data to search and download corresponding labels from currently available soil moisture data websites, shared satellite soil moisture datasets, etc. Other methods can also be used, such as those provided by other institutions, which are not specifically limited in the embodiment of the present invention. Subsequently, the sample data and corresponding labels (the surface heterogeneity and surface environmental data at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution) can be used to train the initial model, thereby obtaining a trained downscaling relationship model.

[0033] Furthermore, after obtaining the downscaling relationship model, in embodiments of the present invention, this downscaling relationship model can be applied to downscale the data at the first spatial resolution, thereby obtaining downscaled soil moisture data at the first spatial resolution, i.e., downscaled soil moisture data at the first spatial resolution. Specifically, surface environmental data and surface heterogeneity at the first spatial resolution can be input into the downscaling relationship model, so that the downscaling relationship model performs downscaling processing accordingly and outputs downscaled soil moisture data, i.e., downscaled soil moisture data at the first spatial resolution. This achieves downscaling of the passive microwave soil moisture product corresponding to the first spatial resolution, effectively improving its spatial resolution, thereby better meeting regional-scale application requirements.

[0034] Compared with the current downscaling method that fails to consider the influence of surface heterogeneity, the embodiment of the present invention fully considers the large heterogeneity of soil moisture in spatial distribution. When establishing the downscaling relationship model, the surface heterogeneity affecting the spatial distribution of soil moisture is considered, which can make the constructed model more accurate and reliable, and closer to reality. Therefore, in the further downscaling process, more accurate, reliable and spatially resolved soil moisture data can be obtained, which effectively improves the accuracy and reliability of downscaling, solves the problems of poor downscaling accuracy and reliability existing in the current downscaling method, and realizes the improvement of the spatial resolution of passive microwave soil moisture products.

[0035] The present invention provides a method for downscaling a passive microwave soil moisture product by integrating surface heterogeneity. The method determines surface heterogeneity and surface environmental data at a first spatial resolution and a second spatial resolution, respectively, based on original surface environmental data. The surface environmental data and surface heterogeneity at the first spatial resolution are input into a downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model. The downscaling relationship model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution and the downscaled soil moisture data at the second spatial resolution. This method overcomes the defects of poor downscaling accuracy and reliability in traditional schemes. In the process of constructing the downscaling relationship model, the influence of surface heterogeneity on the spatial distribution of soil moisture is fully considered, thereby making the constructed model more accurate. Downscaling processing based on this method can effectively improve the spatial resolution of passive microwave soil moisture products, meet regional-scale application requirements, and provide more accurate and reliable soil moisture information for soil moisture-related research.

[0036] Based on the above embodiment, the downscaling relationship model is determined based on the following steps: Determine the initial downscaling model, which is constructed based on the random forest method; Based on the surface environment data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution, the downscaling initial model is trained to obtain the downscaling relationship model.

[0037] Specifically, the process of determining the downscaling relationship model may include the following steps: First, it is necessary to determine the initial model in the training process, which is referred to herein as the downscaled initial model. The downscaled initial model can be constructed based on the machine learning model. Specifically in the embodiment of the present invention, considering that the random forest method can effectively overcome the multicollinearity problem in the data compared with the traditional model and shows strong adaptability to data of different scales, the random forest method is selected here to construct the downscaled initial model, which can also be referred to as a random forest model.

[0038] Then, the sample data and corresponding labels, i.e., the surface environment data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution, can be applied to train the random forest model to obtain a downscaling relationship model. Specifically, the surface heterogeneity and surface environment data at the second spatial resolution can be input into the random forest model so that the random forest model performs downscaling processing accordingly and outputs predicted downscaled soil moisture data, i.e., predicted downscaled soil moisture data at the second spatial resolution; thereafter, a loss metric can be performed based on the predicted downscaled soil moisture data and the downscaled soil moisture data at the second spatial resolution to measure the loss of the random forest model in this downscaling task, and the parameters of the model can be adjusted based on the loss so that the prediction results output by the model after the parameter adjustment are as close as possible to, or even consistent with, the labels, and finally a trained model, i.e., a downscaling relationship model, can be obtained.

[0039] Based on the above embodiment, the downscaling initial model is trained based on the surface environment data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution to obtain a downscaling relationship model, including: Based on the surface environmental data and surface heterogeneity at the second spatial resolution, a surface heterogeneity vector is constructed; Based on the surface heterogeneity vector and the downscaled soil moisture data at the second spatial resolution, the downscaling initial model is trained to obtain the downscaling relationship model; The training goal of the initial downscaling model is to search for optimal parameters through a grid search method, and determine the downscaling relationship model based on the optimal parameters.

[0040] Specifically, the process of training the downscaling initial model using the surface environment data and surface heterogeneity at the second spatial resolution and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model may specifically include: First, the sample data can be vectorized to represent it in vector form, thereby obtaining the corresponding surface heterogeneity vector.

[0041] Specifically, vector construction may be performed based on surface heterogeneity and surface environment data at the second spatial resolution to obtain a plurality of surface heterogeneity vectors, and the plurality of surface heterogeneity vectors may constitute a surface heterogeneity vector space.

[0042] Then, the constructed multiple surface heterogeneity vectors can be input into the random forest model, so that the random forest model can perform downscaling processing according to the input vectors and output the predicted downscaled soil moisture data, that is, the predicted downscaled soil moisture data at the second spatial resolution.

[0043] Afterwards, a loss metric can be performed based on the predicted downscaled soil moisture data and the downscaled soil moisture data at the second spatial resolution to measure the loss of the random forest model in this downscaling task, and the parameters of the model can be adjusted based on this loss so that the prediction results output by the model after parameter adjustment are as close as possible to, or even consistent with, the labels. Finally, a trained model, namely the downscaling relationship model, can be obtained.

[0044] It should be noted that the training objective of the model in the above training process is to search for optimal parameters through a grid search method and determine the downscaling relationship model based on these optimal parameters. Specifically, after constructing the initial downscaling model / random forest model through the random forest method, during training, to simultaneously consider both dynamic and static data in the input data, eight days of data can be selected from the sample data as a group to train a model. Furthermore, during training, the optimal parameters are automatically searched through the grid search method.

[0045] Based on the above embodiment, the process of training the downscaled initial model can be expressed by the following formula: Where, represents the predicted downscaled soil moisture data at the second spatial resolution, Represents surface heterogeneity and surface environment data at the second spatial resolution The constructed surface heterogeneity vector, represents the surface heterogeneity vector space composed of multiple surface heterogeneity vectors, Represents the dimension, represents the constructed initial downscaling model, Represents random error.

[0046] Based on the above embodiments, there are multiple types of surface heterogeneity at each spatial resolution; Step 130 includes: Inputting the surface environmental data and surface heterogeneity at the first spatial resolution into the downscaling relationship model to obtain first soil moisture data output by the downscaling relationship model; The surface heterogeneity at the first spatial resolution is sorted and screened by importance to obtain the most important surface heterogeneity as the target heterogeneity; Based on the target heterogeneity at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

[0047] Specifically, considering that the current downscaling method not only fails to consider the impact of surface heterogeneity on the spatial distribution of soil moisture, but also fails to process the residuals before and after downscaling, or the processing is imperfect, resulting in the inability to effectively reduce the uncertainty in the downscaling process, thereby affecting the accuracy and reliability of passive microwave soil moisture products, in the embodiments of the present invention, it is proposed that the heterogeneity can be used in the downscaling process to correct the residuals of the downscaling results to further reduce the uncertainty in the downscaling process, thereby further improving the downscaling accuracy and enhancing the spatial resolution of the passive microwave soil moisture product.

[0048] Specifically, in an embodiment of the present invention, surface environmental data and surface heterogeneity at a first spatial resolution may be first input into a downscaling relationship model, so that the downscaling relationship model performs downscaling processing accordingly and outputs downscaled soil moisture data. Since the downscaled soil moisture data obtained at this point requires further residual correction, it can be referred to as preliminary soil moisture data, i.e., first soil moisture data. Specifically, during the downscaling process, the input data can also be grouped into groups of 8 days for input, corresponding to the training process, to obtain the first soil moisture data output by the model.

[0049] After this, residual correction can be performed on the first soil moisture data to obtain corrected soil moisture data, i.e., soil moisture data at the first spatial resolution. Specifically, the surface heterogeneities at the first spatial resolution are first ranked, i.e., ranked according to the importance of the features considered by the random forest method or the downscaling relationship model, and the surface heterogeneity with the highest importance is selected as the target heterogeneity. Then, residual correction processing can be performed using this target heterogeneity as a residual correction indicator. In other words, residual correction can be performed on the first soil moisture data based on this target heterogeneity, thereby obtaining corrected soil moisture data, i.e., soil moisture data at the first spatial resolution. This can reduce uncertainty in the downscaling process and improve downscaling accuracy.

[0050] Compared with current downscaling methods that fail to consider the impact of surface heterogeneity and fail to process residuals before and after downscaling, the method provided by the embodiments of the present invention not only establishes a downscaling relationship model that considers surface heterogeneity, but also uses surface heterogeneity to correct residuals of the preliminary downscaling results, further reducing uncertainty and improving the spatial resolution of passive microwave soil moisture products.

[0051] In the embodiment of the present invention, the surface heterogeneity that affects the spatial distribution of soil moisture is taken into account when establishing the downscaling relationship model, which can make the downscaling relationship model established thereby more accurate and reliable. Furthermore, the basis for residual correction is selected according to the importance ranking of the random forest method, and the most important terrain heterogeneity is used as the residual correction indicator, which effectively reduces the uncertainty in the downscaling process and improves the accuracy of downscaling, thereby better meeting the application needs of regional scale.

[0052] Based on the above embodiment, residual correction is performed on the first soil moisture data based on the target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution, including: aggregating the first soil moisture data to a second spatial resolution to obtain second soil moisture data; determining a residual at the second spatial resolution based on the second soil moisture data and the downscaled soil moisture data at the second spatial resolution; Resampling the residual at the second spatial resolution to the first spatial resolution to obtain the residual at the first spatial resolution; Based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution, the first soil moisture data is subjected to residual correction to obtain downscaled soil moisture data at the first spatial resolution.

[0053] Specifically, the process of performing residual correction on the first soil moisture data according to the target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution may include: First, the first soil moisture data may be aggregated to a second spatial resolution to obtain aggregated first soil moisture data at the second spatial resolution, referred to herein as second soil moisture data.

[0054] Then, the residual at the second spatial resolution can be calculated based on the second soil moisture data and the original downscaled soil moisture data at the second spatial resolution; that is, the second soil moisture data and the downscaled soil moisture data at the second spatial resolution are subtracted, and the difference is used as the residual at the second spatial resolution.

[0055] Next, the residual obtained in the previous step may be resampled to obtain a residual at the first spatial resolution. Specifically, the residual at the second spatial resolution may be resampled to the first spatial resolution to obtain a residual at the first spatial resolution.

[0056] Subsequently, residual correction can be performed on the first soil moisture data based on the residual at the first spatial resolution to obtain soil moisture data at the first spatial resolution. Specifically, residual correction can be performed on the first soil moisture data based on the residual at the first spatial resolution and in combination with target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution. This can reduce uncertainty in the downscaling process and improve downscaling accuracy.

[0057] Based on the above embodiment, residual correction is performed on the first soil moisture data based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution, including: Determining a correction residual at the first spatial resolution based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution; Based on the correction residual at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

[0058] Specifically, the process of performing residual correction on the first soil moisture data based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution to obtain the downscaled soil moisture data at the first spatial resolution may include: After obtaining the residual at the first spatial resolution, for a single pixel, window division can be performed, divided into 5×5 windows, and the central pixel (3, 3) is used as the target pixel for residual correction. The target heterogeneity at the corresponding position is used as the weight, and the corrected residual is determined by weighted averaging, that is, the corrected residual at the first spatial resolution.

[0059] After this, residual correction is performed on the first soil moisture data based on the correction residual to obtain downscaled soil moisture data at the first spatial resolution. In other words, residual correction is performed on the first soil moisture data on a pixel-by-pixel basis within a window. The result of this correction is the final downscaling result, i.e., the downscaled soil moisture data at the first spatial resolution.

[0060] The following describes the residual correction process in detail, assuming that the first spatial resolution is 5 km (expressed as 0.05°) and the second spatial resolution is 25 km (expressed as 0.25°): Figure 2 is a schematic diagram of the residual correction process based on terrain heterogeneity provided by the present invention, such as Figure 2As shown in the figure, after obtaining the first soil moisture data through the downscaling relationship model, the first soil moisture data can be aggregated to a spatial resolution of 0.25° to obtain the second soil moisture data. Then, based on the second soil moisture data and the downscaled soil moisture data at a spatial resolution of 0.25°, the residual at a spatial resolution of 0.25° is calculated. This process can be expressed by the following formula: Where, represents the residual at 0.25° spatial resolution, represents the first soil moisture data, represents the second soil moisture data, Represents downscaled soil moisture data at 0.25° spatial resolution.

[0061] Then, the residuals at the 0.25° spatial resolution can be resampled to the 0.05° spatial resolution to obtain the residuals at the 0.05° spatial resolution.

[0062] Next, for a single pixel, a 5×5 window can be selected, with the central pixel (3, 3) as the target pixel for residual correction. The target heterogeneity at the corresponding position (standard deviation elevation (SDE)) is used as the weight. The residual at a spatial resolution of 0.05° after correction is determined by weighted averaging, i.e., the corrected residual at a spatial resolution of 0.05°. This process can be expressed as follows: Where, is the correction residual at 0.05° spatial resolution, is the first residual window at a spatial resolution of 0.05°. pixels, is the terrain heterogeneity of the corresponding location, i.e., the weight, is the amount of valid data in the window, and is the number of residuals and SDEs that have values ​​at the same time.

[0063] After that, the residual correction of the first soil moisture data can be performed pixel by pixel in a window-by-pixel manner to obtain the downscaled soil moisture data at a spatial resolution of 0.05°. This process can be expressed by the following formula: Where, It is the downscaled soil moisture data of a certain pixel at a spatial resolution of 0.05°.

[0064] Based on the above embodiment, the target heterogeneity is terrain heterogeneity, and the surface heterogeneity also includes surface type heterogeneity, vegetation cover heterogeneity and soil texture heterogeneity; The original surface environmental data include numerical environmental variable data and categorical environmental variable data. The surface heterogeneity of the numerical environmental variable data is determined by measuring the standard deviation, and the surface heterogeneity of the categorical environmental variable data is characterized by the Gini-Simpson index.

[0065] Specifically, in the importance ranking process of the features considered using the random forest method or downscaling relationship model, the most important surface heterogeneity selected is topographic heterogeneity. This is because topographic heterogeneity can directly affect the accuracy of soil moisture inversion in soil moisture monitoring.

[0066] In addition, the surface heterogeneity in the embodiment of the present invention may also include surface type heterogeneity, vegetation cover heterogeneity, and soil texture heterogeneity.

[0067] Here, surface type heterogeneity, vegetation cover heterogeneity, terrain heterogeneity, and soil texture heterogeneity are key factors affecting soil moisture distribution and dynamic changes, which describe the complexity and heterogeneity of the surface environment from the perspectives of surface cover type, vegetation distribution, terrain characteristics, and soil physical properties, respectively.

[0068] Surface heterogeneity specifically refers to the diversity and uneven spatial distribution of land cover types. Surface types include bare land, farmland, forestland, grassland, urban land, and water bodies. These different types of land cover significantly influence soil moisture retention, infiltration, and evaporation.

[0069] Vegetation cover heterogeneity refers to the uneven spatial distribution of vegetation type, density, height, and biomass. Vegetation cover heterogeneity can lead to significant spatial differences in soil moisture, for example, higher soil moisture content in areas with dense vegetation and lower soil moisture content in areas with sparse or bare vegetation.

[0070] Topographic heterogeneity refers to the spatial variation in surface elevation, slope, aspect, and relief. It is an important factor affecting the spatial distribution of soil moisture, especially in mountainous or hilly areas.

[0071] Soil texture heterogeneity refers to the uneven distribution of sand, silt, and clay in the soil. This heterogeneity can lead to significant spatial variations in soil moisture; for example, areas with clay soils have higher soil moisture contents, while areas with sandy soils have lower soil moisture contents.

[0072] Among them, the surface heterogeneity of each region can be obtained by the following methods: For the original surface environmental data, such as surface cover type data LC, vegetation leaf area index data LAI, digital surface elevation data DEM and soil texture data ST, they can be divided into two categories, namely numerical type and categorical type. For continuous numerical environmental variable data LAI and DEM, the standard deviation can be used to characterize the size of heterogeneity, that is: Where, Indicates the first spatial resolution and the second spatial resolution. OK Surface heterogeneity within a column pixel calculated from LAI and DEM, Indicates the sub-grid at the spatial resolution corresponding to the original surface environment data LAI data and DEM data, is the standard deviation.

[0073] For categorical environmental variable data ST and LC, the Gini-Simpson Index (GSI) can be used to characterize the size of heterogeneity, namely: Where, Indicates the first spatial resolution and the second spatial resolution. OK Surface heterogeneity within a column pixel calculated by ST and LC, Indicates the ratio of the area occupied by a certain type at the first spatial resolution and the second spatial resolution.

[0074] It should be noted that during the calculation, the valid data in a coarse-resolution grid is guaranteed to be greater than 50%. In order to further reduce the impact of abnormal data, for the calculated surface heterogeneity, the data outside the 98th percentile are set to the value corresponding to the 98th percentile and then normalized.

[0075] Figure 3 This is the overall framework diagram of the passive microwave soil moisture product downscaling method integrating surface heterogeneity provided by the present invention, such as Figure 3 As shown, the method includes: First, obtain the original surface environment data.

[0076] Then, based on the original surface environmental data, surface heterogeneity and surface environmental data at the first and second spatial resolutions are determined. Here, the original surface environmental data includes numerical environmental variable data and categorical environmental variable data. For the numerical environmental variable data, surface heterogeneity is determined by measuring the standard deviation, while for the categorical environmental variable data, surface heterogeneity is characterized by the Gini-Simpson index.

[0077] Next, the surface environmental data and surface heterogeneity at the first spatial resolution are input into the downscaling relationship model to obtain the first soil moisture data output by the downscaling relationship model. The surface heterogeneity at the first spatial resolution is then ranked and screened by importance, with the most important surface heterogeneity being selected as the target heterogeneity. Here, the target heterogeneity is terrain heterogeneity, while surface heterogeneity also includes surface type heterogeneity, vegetation cover heterogeneity, and soil texture heterogeneity.

[0078] Subsequently, based on the target heterogeneity at the first spatial resolution, the first soil moisture data is residual corrected to obtain the downscaled soil moisture data at the first spatial resolution.

[0079] The method comprises the following steps: performing residual correction on the first soil moisture data based on the target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution, including: aggregating the first soil moisture data to the second spatial resolution to obtain second soil moisture data; determining the residual at the second spatial resolution based on the second soil moisture data and the downscaled soil moisture data at the second spatial resolution; resampling the residual at the second spatial resolution to the first spatial resolution to obtain the residual at the first spatial resolution; and performing residual correction on the first soil moisture data based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution to obtain the downscaled soil moisture data at the first spatial resolution.

[0080] Here, based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution, including: determining the correction residual at the first spatial resolution based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution; and performing residual correction on the first soil moisture data based on the correction residual at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution.

[0081] The scale of the second spatial resolution is coarser than the scale of the first spatial resolution.

[0082] The downscaling relationship model is determined based on the following steps: Determine the initial downscaling model, which is constructed based on the random forest method; Based on the surface environmental data and surface heterogeneity at the second spatial resolution, a surface heterogeneity vector is constructed; Based on the surface heterogeneity vector and the downscaled soil moisture data at the second spatial resolution, the downscaling initial model is trained to obtain the downscaling relationship model; The training goal of the initial downscaling model is to search for optimal parameters through a grid search method, and determine the downscaling relationship model based on the optimal parameters.

[0083] The method provided by an embodiment of the present invention considers environmental variables that affect the spatial distribution of soil moisture during the downscaling process, including surface type heterogeneity, vegetation cover heterogeneity, terrain heterogeneity, and soil texture heterogeneity, to improve the accuracy of downscaling. Furthermore, after downscaling, the most important terrain heterogeneity is selected based on the importance ranking of surface heterogeneity. Residual correction is performed using a residual correction method based on terrain heterogeneity, further reducing uncertainty in the downscaling process.

[0084] However, it is worth noting that after obtaining the downscaled soil moisture data at the first spatial resolution, its accuracy can also be verified. That is, the accuracy of the downscaled soil moisture data at the first spatial resolution is verified using the measured soil moisture data obtained from the International Soil Moisture Network (ISMN). Here, the evaluation indicators selected for accuracy verification include root mean square error (RMSE), average bias, unbiased root mean square error (ubRMSE), and correlation coefficient. R .

[0085] The following is an explanation of the downscaling process and accuracy verification process provided by the present invention based on a specific example: First, the original surface environmental data can be used to determine surface heterogeneity and surface environmental data at both 0.05° and 0.25° spatial resolutions. Next, a downscaling relationship model is established using the surface heterogeneity and surface environmental data at 0.25° spatial resolution and the downscaled soil moisture data at 0.25° spatial resolution. The passive microwave soil moisture product can then be downscaled to a 0.05° spatial resolution, yielding the first soil moisture data at 0.05° spatial resolution. Finally, residual correction of the first soil moisture data at 0.05° spatial resolution can be performed using topographic heterogeneity to yield the downscaled soil moisture data at 0.05° spatial resolution. Finally, the downscaling results (downscaled soil moisture data at 0.05° spatial resolution) are verified for accuracy using field-measured soil moisture data from the International Soil Moisture Observation Network (ISMN).

[0086] Figure 4 This is a comparison chart of the impact of adding heterogeneity on model accuracy provided by the present invention, such as Figure 4As shown in the figure, under the same data volume, the RMSE value of the model decreases after adding surface heterogeneity, indicating that the error of the model in predicting downscaled soil moisture data has been effectively controlled. In addition, the mean absolute error (MAE) value also shows a corresponding decrease. MAE is less sensitive to outliers than RMSE, which indicates that the overall performance of the model is more stable and accurate after adding surface heterogeneity. In addition, the coefficient of determination is R ² has also improved, indicating that the model can more accurately capture the dynamic trend of soil moisture over time.

[0087] Table 1: Validation results of the downscaling method on surface measured soil moisture data.

[0088] The table above shows the accuracy of the original passive microwave soil moisture product, verified using surface measured soil moisture data, and the downscaling method for the passive microwave soil moisture product (referred to as the downscaling method) proposed in this invention, both before and after residual correction. The table shows that the downscaling method proposed in this invention improves the spatial resolution of the passive microwave soil moisture product while maintaining or even slightly improving the accuracy of the original passive microwave soil moisture product (e.g., reducing both ubRMSE and Bias).

[0089] The method provided by the embodiments of the present invention fully considers the impact of surface heterogeneity on soil moisture downscaling, effectively improving downscaling accuracy. Furthermore, while preserving the original spatial distribution, the downscaling results effectively increase spatial resolution and enhance texture information. Furthermore, verification based on field-measured soil moisture data from a soil moisture observation network demonstrates that the present invention effectively improves the spatial resolution of passive microwave soil moisture products (from 0.25° to 0.05°), resulting in improved accuracy compared to the original passive microwave soil moisture products.

[0090] The passive microwave soil moisture product downscaling device integrating surface heterogeneity provided by the present invention is described below. The passive microwave soil moisture product downscaling device integrating surface heterogeneity described below and the passive microwave soil moisture product downscaling method integrating surface heterogeneity described above can be referenced to each other.

[0091] Figure 5 This is a schematic diagram of the structure of the passive microwave soil moisture product downscaling device that integrates surface heterogeneity provided by the present invention. Figure 5 As shown, the device includes: The data acquisition unit 510 is used to acquire original surface environment data; A heterogeneity determination unit 520 is configured to determine surface heterogeneity and surface environment data at a first spatial resolution and a second spatial resolution, respectively, based on the original surface environment data; A data downscaling unit 530 is configured to input the surface environmental data and the surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; The scale of the second spatial resolution is coarser than that of the first spatial resolution; the downscaling relationship model is trained based on surface environmental data and surface heterogeneity at the second spatial resolution, and downscaled soil moisture data at the second spatial resolution.

[0092] The passive microwave soil moisture product downscaling device provided by the present invention integrates surface heterogeneity. Based on original surface environmental data, the device determines surface heterogeneity and surface environmental data at a first spatial resolution and a second spatial resolution, respectively. The surface environmental data and surface heterogeneity at the first spatial resolution are input into a downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model. The downscaling relationship model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution and the downscaled soil moisture data at the second spatial resolution. This overcomes the defects of poor downscaling accuracy and reliability in traditional solutions. In the process of constructing the downscaling relationship model, the influence of surface heterogeneity on the spatial distribution of soil moisture is fully considered, thereby making the constructed model more accurate. Downscaling processing based on this can effectively improve the spatial resolution of passive microwave soil moisture products, meet regional-scale application requirements, and provide more accurate and reliable soil moisture information for soil moisture-related research.

[0093] Based on the above embodiment, the device further includes a model training unit, which is used to: Determining an initial downscaling model, where the initial downscaling model is constructed based on a random forest method; The downscaling initial model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model.

[0094] Based on the above embodiment, the model training unit is used to: constructing a surface heterogeneity vector based on the surface environmental data and surface heterogeneity at the second spatial resolution; Training the downscaling initial model based on the surface heterogeneity vector and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model; The training goal of the downscaling initial model is to search for optimal parameters through a grid search method, and determine the downscaling relationship model based on the optimal parameters.

[0095] Based on the above embodiments, there are multiple types of surface heterogeneity at each spatial resolution; The data downscaling unit 530 is used to: Inputting the surface environmental data and surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain first soil moisture data output by the downscaling relationship model; sorting and screening the surface heterogeneity at the first spatial resolution by importance, and obtaining the most important surface heterogeneity as the target heterogeneity; Based on the target heterogeneity at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

[0096] Based on the above embodiment, the data downscaling unit 530 is configured to: aggregating the first soil moisture data to the second spatial resolution to obtain second soil moisture data; determining a residual at the second spatial resolution based on the second soil moisture data and the downscaled soil moisture data at the second spatial resolution; Resampling the residual at the second spatial resolution to the first spatial resolution to obtain the residual at the first spatial resolution; Based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

[0097] Based on the above embodiment, the data downscaling unit 530 is configured to: determining a correction residual at the first spatial resolution based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution; Based on the correction residual at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

[0098] Based on the above embodiment, the target heterogeneity is terrain heterogeneity, and the surface heterogeneity also includes surface type heterogeneity, vegetation cover heterogeneity and soil texture heterogeneity; The original surface environmental data includes numerical environmental variable data and categorical environmental variable data. The surface heterogeneity is determined by measuring the standard deviation of the numerical environmental variable data, and the surface heterogeneity is obtained by characterizing the categorical environmental variable data through the Gini-Simpson index.

[0099] Figure 6 An example of a physical structure diagram of an electronic device is shown below. Figure 6 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communications bus 640. The processor 610, the communications interface 620, and the memory 630 communicate with each other via the communications bus 640. The processor 610 may invoke logic instructions in the memory 630 to execute a passive microwave soil moisture product downscaling method that incorporates surface heterogeneity. The method includes: obtaining original surface environmental data; determining surface heterogeneity and surface environmental data at a first spatial resolution and a second spatial resolution based on the original surface environmental data; inputting the surface environmental data and surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; wherein the scale of the second spatial resolution is coarser than the scale of the first spatial resolution; and the downscaling relationship model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution, as well as the downscaled soil moisture data at the second spatial resolution.

[0100] Furthermore, the logic instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0101] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the passive microwave soil moisture product downscaling method that integrates surface heterogeneity provided by the above methods. The method includes: obtaining original surface environmental data; based on the original surface environmental data, determining the surface heterogeneity and surface environmental data at a first spatial resolution and a second spatial resolution respectively; inputting the surface environmental data and surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain the downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; wherein the scale of the second spatial resolution is coarser than the scale of the first spatial resolution; the downscaling relationship model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution.

[0102] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the passive microwave soil moisture product downscaling method provided by the above-mentioned methods that integrates surface heterogeneity, the method comprising: obtaining original surface environmental data; determining surface heterogeneity and surface environmental data at a first spatial resolution and a second spatial resolution based on the original surface environmental data; inputting the surface environmental data and surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; wherein the scale of the second spatial resolution is coarser than the scale of the first spatial resolution; the downscaling relationship model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution.

[0103] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0104] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A passive microwave soil moisture product downscaling method incorporating surface heterogeneity, characterized by: include: Obtain original surface environmental data; Based on the original surface environmental data, determining surface heterogeneity and surface environmental data at a first spatial resolution and a second spatial resolution, respectively; Inputting the surface environmental data and surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; wherein the scale of the second spatial resolution is coarser than the scale of the first spatial resolution; The downscaling relationship model is trained based on the surface environment data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution.

2. The passive microwave soil moisture product downscaling method integrating surface heterogeneity according to claim 1 is characterized in that: The downscaling relationship model is determined based on the following steps: Determining an initial downscaling model, where the initial downscaling model is constructed based on a random forest method; The downscaling initial model is trained based on the surface environmental data and surface heterogeneity at the second spatial resolution and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model.

3. The passive microwave soil moisture product downscaling method integrating surface heterogeneity according to claim 2 is characterized in that: The downscaling initial model is trained based on the surface environment data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model, including: constructing a surface heterogeneity vector based on the surface environmental data and surface heterogeneity at the second spatial resolution; Training the downscaling initial model based on the surface heterogeneity vector and the downscaled soil moisture data at the second spatial resolution to obtain the downscaling relationship model; The training goal of the downscaling initial model is to search for optimal parameters through a grid search method, and determine the downscaling relationship model based on the optimal parameters.

4. The passive microwave soil moisture product downscaling method according to any one of claims 1 to 3, characterized in that: There are many types of surface heterogeneity at each spatial resolution; The step of inputting the surface environmental data and the surface heterogeneity at the first spatial resolution into the downscaling relationship model to obtain the downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model includes: Inputting the surface environmental data and surface heterogeneity at the first spatial resolution into a downscaling relationship model to obtain first soil moisture data output by the downscaling relationship model; sorting and screening the surface heterogeneity at the first spatial resolution by importance, and obtaining the most important surface heterogeneity as the target heterogeneity; Based on the target heterogeneity at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

5. The passive microwave soil moisture product downscaling method integrating surface heterogeneity according to claim 4 is characterized in that: The performing residual correction on the first soil moisture data based on the target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution includes: aggregating the first soil moisture data to the second spatial resolution to obtain second soil moisture data; determining a residual at the second spatial resolution based on the second soil moisture data and the downscaled soil moisture data at the second spatial resolution; Resampling the residual at the second spatial resolution to the first spatial resolution to obtain the residual at the first spatial resolution; Based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

6. The passive microwave soil moisture product downscaling method integrating surface heterogeneity according to claim 5 is characterized in that: The performing residual correction on the first soil moisture data based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution to obtain downscaled soil moisture data at the first spatial resolution includes: determining a correction residual at the first spatial resolution based on the residual at the first spatial resolution and the target heterogeneity at the first spatial resolution; Based on the correction residual at the first spatial resolution, residual correction is performed on the first soil moisture data to obtain downscaled soil moisture data at the first spatial resolution.

7. The passive microwave soil moisture product downscaling method integrating surface heterogeneity according to claim 4 is characterized in that: The target heterogeneity is terrain heterogeneity, and the surface heterogeneity also includes surface type heterogeneity, vegetation cover heterogeneity and soil texture heterogeneity; The original surface environmental data includes numerical environmental variable data and categorical environmental variable data. The surface heterogeneity is determined by measuring the standard deviation of the numerical environmental variable data, and the surface heterogeneity is obtained by characterizing the categorical environmental variable data through the Gini-Simpson index.

8. A passive microwave soil moisture product downscaling device integrating surface heterogeneity, characterized in that: include: A data acquisition unit, used for acquiring original surface environment data; a heterogeneity determination unit, configured to determine surface heterogeneity and surface environment data at a first spatial resolution and a second spatial resolution, respectively, based on the original surface environment data; a data downscaling unit, configured to input the surface environmental data and the surface heterogeneity at the first spatial resolution into a downscaling relationship model, and obtain downscaled soil moisture data at the first spatial resolution output by the downscaling relationship model; wherein the scale of the second spatial resolution is coarser than the scale of the first spatial resolution; The downscaling relationship model is trained based on the surface environment data and surface heterogeneity at the second spatial resolution, and the downscaled soil moisture data at the second spatial resolution.

9. 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 passive microwave soil moisture product downscaling method integrating surface heterogeneity is implemented as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for downscaling passive microwave soil moisture products integrating surface heterogeneity according to any one of claims 1 to 7 is implemented.

Images