Method for estimating high-resolution soil organic matter and soil moisture by using dielectric proxy

Abstract

The present invention relates to a method for estimating high-resolution soil organic matter and soil moisture by using a dielectric proxy, the method comprising the steps of: acquiring initial soil data through satellite data; acquiring soil moisture data from the initial soil data at a preset resolution; observing a dielectric constant in a preset band on the basis of the acquired soil moisture data, and generating a dielectric proxy related to an organic matter content on the basis of the observed dielectric constant value; generating an estimated soil moisture initial value by using the dielectric proxy; and correcting the estimated soil moisture initial value through a pre-trained machine learning model. According to the present invention, a dielectric proxy that indirectly reflects the relationship with organic matter is generated through SAR data so that dynamic and high-resolution soil organic matter and moisture information can be provided.

Description

High-resolution soil organic matter and soil moisture estimation method using permittivity surrogates

[0001] The present invention relates to a method for estimating soil organic matter and soil moisture at high resolution using a dielectric surrogate, and more specifically, to a method for estimating soil organic matter and soil moisture at high resolution using a dielectric surrogate, which generates a dielectric surrogate by acquiring soil moisture information through a Synthetic Aperture Radar (SAR) satellite and estimates the distribution of organic matter in the soil at high resolution in space and time using the generated surrogate.

[0002] Providing precise soil organic matter maps is essential for climate change management, agricultural productivity, and environmental control. However, most conventional static organic matter maps have limitations, such as low spatiotemporal resolution and an inability to respond to dynamic changes. These issues hinder the effective operation of agricultural and environmental monitoring systems.

[0003] Therefore, to address these issues, it is necessary to develop a method utilizing SAR to track dynamic changes in soil organic matter and generate high-resolution organic matter maps while minimizing signal disturbances caused by clouds and vegetation in satellite observations.

[0004] Currently, most SAR satellites provide important soil moisture information for agriculture and water resource management, but technology to observe soil organic matter content has not yet been developed. This is because there is no clear model between soil organic matter and backscattered albedo.

[0005] SAR is sensitive to soil moisture, but there are limitations in accurately estimating organic matter due to a lack of models for physical relationships regarding organic matter.

[0006] Therefore, an approach is needed to integrate SAR data into microwave permittivity models of different wavelengths that account for organic matter.

[0007] Furthermore, since soil moisture information obtained from microwaves is affected by the uncertainty of organic matter, the currently provided SAR soil moisture also contains errors.

[0008] According to the present invention, the invention provides a method for estimating soil organic matter and soil moisture at high resolution using a dielectric surrogate, wherein soil moisture information is obtained through a Synthetic Aperture Radar (SAR) satellite to generate a dielectric surrogate, and the distribution of organic matter in the soil is estimated spatiotemporally at high resolution using the generated surrogate.

[0009] According to one embodiment of the present invention for achieving such technical challenges, a method for estimating high-resolution soil organic matter and soil moisture using a permittivity surrogate comprises: a step of acquiring initial soil data through satellite data; a step of acquiring soil moisture data from the initial soil data at a preset resolution; a step of observing the permittivity in a preset band based on the acquired soil moisture data and generating a permittivity surrogate for organic matter content based on the observed permittivity value; a step of generating an initial soil moisture estimation value using the permittivity surrogate; and a step of correcting the estimated initial soil moisture value through a preset machine learning model.

[0010] As such, according to the present invention, a dielectric surrogate that indirectly reflects the relationship with organic matter through SAR data can be generated to provide dynamic and high-resolution information on soil organic matter and moisture.

[0011] Accordingly, high-resolution organic matter information can be provided even in areas with weather changes and vegetation, which can contribute to climate change and the improvement of agricultural productivity.

[0012] FIG. 1 is a diagram illustrating the flow of a high-resolution soil organic matter and soil moisture estimation method using a dielectric surrogate according to one embodiment of the present invention.

[0013] Figure 2 is a diagram illustrating an example of estimating organic matter using multiple methods.

[0014] Figure 3 is a diagram illustrating an example of testing the accuracy of soil moisture estimation using the dielectric surrogate of the present invention using SMAPVEX12.

[0015] Figure 4 is a diagram illustrating an example of the organic matter estimation results according to the resolution of the organic matter map and the presence or absence of errors in soil moisture.

[0016] Figure 5 is a diagram showing the results of post-processing the estimated value of soil moisture according to one embodiment of the present invention.

[0017] Preferred embodiments according to the present invention will be described in detail below with reference to the attached drawings. In this process, the thickness of lines or the size of components shown in the drawings may be exaggerated for clarity and convenience of explanation.

[0018] Furthermore, the terms described below are defined in consideration of their functions within the present invention, and these may vary depending on the intent or practice of the user or operator. Therefore, the definitions of these terms should be based on the content throughout this specification.

[0019] Hereinafter, a method for estimating high-resolution soil organic matter and soil moisture using a dielectric surrogate according to an embodiment of the present invention will be described in detail with reference to FIGS. 1 to 4.

[0020] FIG. 1 is a diagram illustrating the flow of a high-resolution soil organic matter and soil moisture estimation method using a dielectric surrogate according to an embodiment of the present invention, FIG. 2 is a diagram illustrating an example of estimating organic matter using a plurality of methods, FIG. 3 is a diagram illustrating an example of testing the accuracy of soil moisture estimation using the dielectric surrogate of the present invention using SMAPVEX12, FIG. 4 is a diagram illustrating an example of organic matter estimation results according to the resolution of the organic matter map and the presence or absence of an error in soil moisture, and FIG. 5 is a diagram illustrating the result of post-processing the estimated value of soil moisture according to an embodiment of the present invention.

[0021] At this time, the high-resolution soil organic matter and soil moisture estimation method using the dielectric surrogate shown in Fig. 1 can be performed by a processor.

[0022] First, as illustrated in FIG. 1, a method for estimating high-resolution soil organic matter and soil moisture using a dielectric surrogate according to one embodiment of the present invention obtains initial soil data through satellite data (S110).

[0023] In this case, the satellite may be a Synthetic Aperture Radar (SAR) capable of acquiring soil data in real time.

[0024] Next, soil moisture data is acquired from the initial soil data at a preset resolution (e.g., 1 km) (S120).

[0025] In this case, if soil moisture information is not provided in the satellite data, a machine learning model can be created by training existing soil moisture products together with other input data, including backscattered albedo observation data.

[0026] Through this, it is possible to calculate soil moisture from satellite data that does not provide soil moisture information.

[0027] In this case, the soil moisture data may contain errors.

[0028] Next, the dielectric constant in a preset band (e.g., 50 MHz) is observed based on the acquired soil moisture data, and a dielectric surrogate for organic matter content is generated based on the observed dielectric constant (S130).

[0029] At this time, the permittivity surrogate can be generated through the following mathematical formula 1.

[0030]

[0031] At this time, SM sensor is data collected from a sensor and includes information on soil moisture and organic matter.

[0032] In addition, a, b, and c are coefficients estimated through regression analysis based on actual data.

[0033] Next, an initial value for soil moisture estimation is generated using a permittivity surrogate (S140).

[0034] More specifically, first, the maximum permittivity of bound water and free water is generated to generate the initial values ​​for soil moisture estimation as shown in Equation 2 below.

[0035]

[0036] At this time, is the maximum permittivity of the number of bonds, and ε is the maximum permittivity of free water, SOM is the weight ratio of organic matter, and a bound , b bound , c bound is an empirical constant, and v clay is the volume ratio of clay.

[0037] Next, using the maximum permittivity of the generated bound water and free water, the permittivity of the bound water and free water permittivity Calculates.

[0038]

[0039] At this time, is the minimum permittivity of the number of bonds, and is the minimum permittivity of free numbers.

[0040] Next, initial values ​​for soil moisture estimation are generated through iterative estimation as shown in mathematical formula 4 below.

[0041]

[0042] At this time, is the effective dielectric constant of dry organic soil, and ε is the permittivity of air, P is the saturation point, WP is the wilting point, and H is the damping coefficient with a value less than or equal to 1.

[0043] At this time, in the case of severely attenuated SAR microwaves, the value becomes 0.8 or less.

[0044] also, ε₀ is the permittivity for the cases where soil moisture is less than the wilting point, greater than the wilting point and below the saturation point, and greater than the saturation point, respectively, and is the simulated permittivity, and SM iterative It can be calculated through iterative estimation using an organic matter map as input data.

[0045] However, during the subsequent process of estimating organic matter using a permittivity model, if the accuracy of the initial soil moisture value is relatively higher than that of the initial organic matter map value, the performance of the optimal soil organic matter estimation may rapidly deteriorate.

[0046] Therefore, an error is applied to the soil moisture data as shown in mathematical formula 5 below.

[0047]

[0048] At this time, can be a value between 0.4 and 1.8, excluding 1, and can be a value between -0.1 and 0.1.

[0049] Figure 2(a) is a diagram showing the result of estimating organic matter using gravimetric soil moisture as the initial value based on actual measurements, Figure 2(b) is a diagram showing the result of estimating organic matter using soil moisture estimated from permittivity as the initial value, and Figure 2(c) is a diagram showing the result of estimating organic matter using a value with an error applied to soil moisture estimated from permittivity as the initial value.

[0050] In addition, Figure 3(a) is a diagram showing the relationship between soil moisture estimated by a dielectric sensor and actual soil moisture, Figure 3(b) is a diagram showing the relationship between the sensor soil moisture and actual soil moisture with an artificial error applied, and Figure 3(c) is a diagram showing the final stage where the soil moisture with an error applied approaches the actual measurement after optimal estimation.

[0051] Next, set the initial value of soil organic matter.

[0052] At this time, since the optimization performance may vary depending on the initial value of soil organic matter, high-resolution soil organic matter maps (e.g., 250m resolution maps) and low-resolution soil organic matter maps (e.g., 1000m resolution maps) can be selected and used according to the characteristics and purpose of the region.

[0053] For example, if the map information in the target area is very accurate and highly reliable, a high-resolution organic map can be used.

[0054] High-resolution maps provide more detailed and precise information, which can equalize the degree of error between soil moisture and soil organic matter during the optimization process.

[0055] On the other hand, if there is high uncertainty in the organic map of the target area, a low-resolution organic map can be used.

[0056] When optimizing high-resolution maps with excessive detail in areas of high uncertainty, prior organic information can actually cause more errors; therefore, optimizing with low-resolution, general values ​​can yield more accurate information.

[0057] After selecting the organic matter map, the initial value of soil organic matter is extracted through the selected soil organic matter map.

[0058] Next, as shown in Equation 6 below, the optimal soil moisture value is obtained by optimally estimating the initial soil moisture value with an error applied to the initial organic matter value as a starting point.

[0059]

[0060] At this time, as shown in mathematical formula 7 below, the organic matter corresponding to the optimal estimated soil moisture becomes the optimal organic matter estimate.

[0061]

[0062] In addition, the optimally estimated organic matter obtains an organic carbon value through the organic matter-organic carbon conversion coefficient as shown in Equation 8 below.

[0063]

[0064] Figure 4(a) is a diagram showing the result of optimizing organic matter by setting the organic matter map to high resolution and inputting actual soil moisture, and comparing it with the actual organic matter measurement; Figure 4(b) is a diagram showing the result of optimizing organic matter after applying an error to the soil moisture of the sensor in the high-resolution organic matter map; Figure 4(c) is a diagram showing the result of optimizing organic matter by inputting actual soil moisture measurements in the low-resolution organic matter map; and Figure 4(d) is a diagram showing the result of optimizing organic matter from the soil moisture sensor moisture with an error applied to the low-resolution organic matter.

[0065] As shown in Figure 4, using a low-resolution organic matter map rather than a high-resolution organic matter map allows for a prediction that is closer to the actual value when estimating organic matter, and enables the production of results closer to reality when estimating organic matter using data that includes an appropriate error rather than the actual soil moisture value.

[0066] That is, as shown in FIGS. 2, 3, and 4, by introducing an error to the soil moisture, noticeably precise results can be obtained in the optimal estimation of organic matter.

[0067] Figure 4 shows that the accuracy of optimal organic matter estimation can be improved by expanding the search range of soil organic matter (OM) by introducing an artificial error to the initial value of soil moisture.

[0068] This method has the disadvantage of having to sacrifice some of the accuracy of the final SM.

[0069] The initial iteration SM initial value shown in Fig. 4(a) already has a relatively lower error than the OM map initial value, so it does not provide sufficient motivation to estimate SM-OM simultaneously.

[0070] On the other hand, in Fig. 4(b), artificial errors are injected into the SM initial values ​​to expand the parameter search space, thereby inducing the OM values ​​to be adjusted more actively during the optimal estimation process.

[0071] However, this strategy increases the variance of the SM estimation results, leading to an increase in the error (RMSE) compared to the existing iterative method.

[0072] Next, as can be seen in Fig. 4(c), the variance remains somewhat large even in the optimized SM, showing a trade-off between reflecting the uncertainty of the OM and maintaining SM accuracy.

[0073] When aiming to estimate SM and OM simultaneously, one must consider that the initial value of SM can be intentionally disturbed to improve the convergence of OM.

[0074] To improve the accuracy of soil moisture estimation, a soil moisture error correction model using machine learning methods can be created as shown in Equation 9 below.

[0075] Optimal soil moisture estimate (SM opt After generating ), the optimal soil moisture estimate is corrected through the previously trained machine learning model (S150).

[0076] At this time, the machine learning model can receive high-resolution organic matter data as input and learn the error of SAR-based soil moisture estimates.

[0077] More specifically, the machine learning model can learn the difference between satellite soil moisture and ground soil moisture observations according to optimal estimated organic matter, clay map, and soil temperature sampled in areas where ground soil moisture observations exist, as shown in Equation 9 below.

[0078]

[0079] At this time, OM opt is the optimal estimated organic matter, clay map is the clay map, ST is the soil temperature, and SM insitu is the observed ground soil moisture value, SM opt is the optimal estimated soil moisture observation.

[0080] The machine learning model trained in this way predicts the optimal estimate of the error that may occur in the soil moisture estimate, as shown in Equation 10 below.

[0081]

[0082] In this case, the input data required for the machine learning model to make a prediction is the optimally estimated organic matter, clay map, and soil temperature for any region or time period that were not used for training, regardless of the presence or absence of surface soil moisture.

[0083] After predicting the optimal error estimate, post-processing is performed to correct the error in soil moisture by applying the predicted optimal error estimate to the soil moisture estimate as shown in Equation 11 below.

[0084]

[0085] Corrected soil moisture can correct existing satellite soil moisture in all spacetimes.

[0086] In other words, the machine learning model analyzes the input high-resolution organic matter data to learn the errors that may occur in the soil moisture estimate, and can calculate the final corrected soil moisture value in conjunction with the soil moisture estimate.

[0087] As shown in Figure 5, when soil moisture is post-processed in SMAP, the accuracy of soil moisture can be improved and can also be applied to soil moisture estimation based on satellite data.

[0088] In conclusion, by adding high-resolution organic matter data to which a soil moisture error machine learning model has been applied to the soil moisture estimates, it is possible to obtain more accurate and reliable corrected soil moisture values, thereby enabling a more accurate reflection of soil conditions.

[0089] As such, according to one embodiment of the present invention, a dielectric surrogate that indirectly reflects the relationship with organic matter through SAR data can be generated to provide dynamic and high-resolution information on soil organic matter and moisture.

[0090] Accordingly, high-resolution organic matter information can be provided even in areas with weather changes and vegetation, which can contribute to climate change and the improvement of agricultural productivity.

[0091] The present invention has been described with reference to the embodiments illustrated in the drawings, but this is merely illustrative, and those skilled in the art will understand that numerous modifications and equivalent alternative embodiments are possible therefrom. Accordingly, the true technical scope of protection of the present invention should be determined by the technical spirit of the following claims.

Claims

1. A step of acquiring initial soil data through satellite data; A step of acquiring soil moisture data from the above initial soil data at a preset resolution; A step of observing the dielectric constant in a preset band based on the soil moisture data obtained above, and generating a dielectric constant surrogate regarding organic matter content based on the observed dielectric constant value; A step of generating an initial soil moisture estimation value using the above permittivity surrogate; and A method for estimating high-resolution soil organic matter and soil moisture using a permittivity surrogate, comprising the step of correcting the estimated initial soil moisture value through a previously trained machine learning model.

2. In Paragraph 1, The above satellite is a method for estimating high-resolution soil organic matter and soil moisture using a dielectric surrogate, which is a synthetic aperture radar (SAR) capable of acquiring soil data in real time.

3. In Paragraph 1, A method for estimating high-resolution soil organic matter and soil moisture using a permittivity surrogate, wherein, when soil moisture information is not provided in the above satellite data, a machine learning model is constructed by training existing soil moisture products together with other input data including backscattered albedo observation data to calculate soil moisture.

4. In Paragraph 1, The step of generating the above permittivity surrogate is, The following mathematical formula Generate a permittivity surrogate through, and SM sensor A high-resolution soil organic matter and soil moisture estimation method using a permittivity surrogate, wherein is data collected from a sensor and a, b, and c are coefficients estimated through regression analysis based on actual data.

5. In Paragraph 1, The above machine learning model is the following mathematical formula As shown above, it takes high-resolution organic matter data as input, learns the error of SAR-based soil moisture estimates, and OM opt is the optimal estimated organic matter, clay map is the clay map, ST is the soil temperature, and SM insitu is the observed ground soil moisture value, SM opt A high-resolution soil organic matter and soil moisture estimation method using a permittivity surrogate, which is the optimal estimated soil moisture observation.

6. In Paragraph 5, The step of correcting the above soil moisture estimate is the following mathematical formula A high-resolution soil organic matter and soil moisture estimation method using a permittivity surrogate to predict the optimal estimate of an error that may occur in soil moisture estimates, as shown above.

7. In Paragraph 6, The step of correcting the above soil moisture estimate is the following mathematical formula A high-resolution soil organic matter and soil moisture estimation method using a dielectric surrogate that performs post-processing to correct errors in soil moisture by applying the predicted error optimal estimate to the soil moisture estimate.

Images