CN106501186B - A Downscaling Method for Soil Water Content Products - Google Patents

Abstract

The invention discloses a method for downscaling soil water content products, including: A. acquiring passive microwave soil water content products and optical remote sensing image data at the same time in the area to be studied; The optical remote sensing image data is used to extract the soil spectrum; C. Utilize GA-PLS to establish a quantitative relationship model between the soil spectral reflection characteristics and the soil moisture content obtained from the passive microwave soil moisture content product; D. Based on the quantitative relationship model , using the Taylor series expansion form to build a soil water content downscaling model to obtain high spatial resolution soil water content data. The present invention utilizes the advantages of both passive microwave remote sensing data and optical remote sensing data in temporal and spatial resolution, and effectively integrates the two to obtain soil water content data with high spatial resolution, which can satisfy large-scale watershed-scale regional research and realize watershed-scale soil Real-time or quasi-real-time dynamic monitoring of water content, high accuracy, easy to set up, saving time and effort.

Description

A Downscaling Method for Soil Water Content Products

technical field

The invention relates to the technical field of geographic surveying and mapping, in particular to a downscaling method for soil water content products.

Background technique

As one of the important indicators of climate and environmental drought, soil moisture content affects the physical and chemical properties of soil and the growth of vegetation, which in turn affects my country's grain production. At the same time, soil water content is an important part of surface energy balance and water cycle, and is an important monitoring factor in global change research. At present, various soil moisture data sets based on global observation site data are insufficient in the density and spatial representation of observation points, and the accuracy of simulation and forecast cannot meet the application requirements.

Advanced Microwave Scanning Radiometer AMSR-2 (Advanced Microwave Scanning Radiometer for the Earth Observing System), soil moisture and ocean salinity sensor SMOS (Soil Moisture and Ocean Salinity), SMAP (Soil Moisture Active Passive) and Fengyun-3 meteorological satellite all have all-day The outstanding advantages of time, all-weather, large observation scale, and short revisit cycle can provide global soil moisture data with high coverage and timeliness. However, the spatial resolution (9-40km) of these data is low, which cannot meet the needs of spatial-temporal dynamic monitoring of soil water content at the watershed scale. The spatial resolution of optical remote sensing data can reach less than 1km. The characteristics of high spatial resolution, low temporal resolution, and vulnerability to weather are just the opposite of passive microwave remote sensing data.

In order to solve the problem of low spatial resolution of passive microwave soil water content products, domestic and foreign scholars have also proposed various downscaling methods: one type of downscaling model is based on soil physical parameters obtained by remote sensing technology, such as using optical Soil evapotranspiration retrieved from remote sensing data is used to establish a downscaling algorithm, but the limitation of this algorithm is that there is a strong difference between the soil water content retrieved from microwave remote sensing data and the effective soil evaporation retrieved from optical remote sensing data. linear relationship. Another type of model is the four-dimensional variational assimilation method, which analyzes soil water content on the fourth-dimensional scale of microwave remote sensing data. However, this method has higher requirements on the acquisition of surface data.

It can be seen that there is an urgent need to develop a new downscaling method for soil water content products, establish a feasible spatial downscaling model, gradually improve the spatial resolution of soil water content data, and promote the spatiotemporal dynamics of soil water content at the watershed scale monitor.

Contents of the invention

The technical problem to be solved by the present invention is to provide a method for downscaling soil water content products, which can improve the spatial resolution of passive microwave soil water content data products, meet the application requirements of water resources and agricultural management at the watershed scale, and overcome the current soil water content. The lack of spatial representation and low spatial resolution of water volume product datasets.

In order to solve the above-mentioned technical problems, the present invention provides a method for downscaling soil water content products, said method comprising the following steps:

A. Obtain passive microwave soil moisture content products and optical remote sensing image data at the same time in the area to be studied;

B. Extracting the soil spectrum from the optical remote sensing image data based on the multi-terminal mixed pixel decomposition method;

C. Utilize GA-PLS to establish the quantitative relationship model between the soil moisture content obtained from the soil spectral reflectance feature and the passive microwave soil moisture content product;

D. Based on the quantitative relationship model established in step C, a soil water content downscaling model is constructed using Taylor series expansion to obtain high spatial resolution soil water content data.

As an improvement of the present invention, the passive microwave soil water content product in the step A uses SMAP soil water content data; the optical remote sensing image data uses MODIS image data.

As a further improvement, the method for extracting the soil spectrum in the step B is: resampling the high spatial resolution MODIS image to the same low spatial resolution as the SMAP data, and resampling the high spatial resolution MODIS image and resampling MODIS with low spatial resolution after sampling applies the MESMA method to extract soil spectra.

As a further improvement, the MESMA method includes the steps of spectral library creation, optimal spectral library selection and multi-terminal mixed pixel analysis,

The creation of the spectral library includes creating a spectral library based on ROI, making spectral library metadata and managing the spectral library;

The selection of the optimal spectral library includes creating a square array and optimal spectral library;

The multi-terminal mixed pixel analysis adopts the vegetation-impermeable surface-soil model, and the optimal vegetation, impermeable surface, and soil spectral sets are combined to form a 2EM, 3EM, and 4EM mixed pixel analysis model, based on the optimal spectral library The optimized results of the MESMA results are shaded and normalized to obtain the endmember abundance values and the root mean square error representing the accuracy of the results, and then use the endmember abundance values and the following formula to obtain the soil endmember spectrum in the study area ,

Among them, R _s (λ) is the reflectivity of the soil spectrum in the band λ, R(λ) is the reflectivity of the pixel in the band λ, R(i, λ) is the reflection of the i-th end member in the band λ rate, f _i is the i-th endmember abundance value, N is the number of endmembers, ε _λ is the residual, and the sum of the abundance values of all endmember components is defined as 1.

As a further improvement, the step C establishes the quantitative relationship model between the soil spectral reflectance feature and the soil water content based on the reflectance, band ratio, and curvature of each band in the soil spectrum of each pixel calculated based on the soil spectrum Quantitative relationship with the soil moisture content.

Further improvement, the expression of Taylor series expansion in the step D is:

Among them, θ _n-1 and θ _n represent low spatial resolution and high spatial resolution soil water content respectively, R _n-1 (λ _i ) _s and R _n (λ _i ) _s represent low spatial resolution and high spatial resolution The reflectance of the i-th band of the high-resolution soil spectrum, Ratio _n-1 (j) _s ) and Ratio _n (j) _s represent the ratio of the low-spatial-resolution and high-spatial-resolution soil spectrum bands, Curv _n-1 (k) _s and Curv _n (k) _s represent the soil spectral curvature at high and low spatial resolution, respectively, and M, N, and L represent the total number of variables for reflectance i, band ratio j, and curvature k, respectively.

As a further improvement, the method for establishing the downscaling model in the step D adopts a step-by-step downscaling method to perform the downscaling operation, and gradually meet the requirement of high spatial resolution.

As a further improvement, the soil water content data with high spatial resolution in the step D refers to the soil water content data with a spatial resolution of 500m.

As a further improvement, in the step A, the auxiliary soil data of the area to be studied is also obtained at the same time, and the auxiliary soil data includes: land use or land cover classification data, and DEM data.

As a further improvement, the GA-PLS model established in step C can also add DEM data or slope data of each pixel as an input parameter of the GA-PLS model.

Adopt above-mentioned technical scheme, the present invention has following advantage at least:

(1) Based on optical remote sensing and passive microwave remote sensing data, the present invention obtains a method of soil moisture content with high spatial resolution, which can satisfy large-scale watershed-scale regional research, has high accuracy, is easy to establish, and saves time and effort.

(2) The present invention has high scalability. In the process of application, the auxiliary data or the increase or decrease of the relationship items of the quantitative inversion model of soil moisture can be carried out according to the actual situation, and the stepwise regression downscaling method can also be used. Improve the spatial resolution of passive microwave soil water content products in a step-by-step manner, and continuously improve the calculation accuracy of the model.

Description of drawings

The above is only an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

Fig. 1 is a principle flow chart of the soil water content product downscaling method of the present invention.

Fig. 2 is a result diagram of soil optimal spectrum selection in the present invention.

Fig. 3 is a diagram of the optimal spectrum selection result of vegetation in the present invention.

Fig. 4 is the MESMA result map of the 9km spatial resolution MODIS image in the present invention.

Fig. 5 is a distribution diagram of sample points in the present invention.

Fig. 6 is the accuracy evaluation result of the 9km scale GA-PLS model in the present invention.

Fig. 7 is the MESMA result map of the 5km spatial resolution MODIS image in the present invention.

Fig. 8 is the evaluation result of downscaling accuracy of 5km soil water content in the present invention.

Fig. 9 is a MESMA result diagram of a MODIS image with a spatial resolution of 500m in the present invention.

Fig. 10 is the accuracy evaluation result of the 5km scale GA-PLS model in the present invention.

Fig. 11 is the evaluation result of downscaling accuracy of 500m soil water content in the present invention.

Detailed ways

On the basis of the prior art, the present invention utilizes the reflectance spectrum characteristics of the soil, that is, the soil water content is an important factor affecting the soil spectral reflectance, and because the soil water content has very large temporal and spatial variability, the high spatial resolution Soil spectral information can better reflect the change characteristics of soil water content in the scope of time and space, so the present invention uses the quantitative relationship between the spectral characteristics of soil and soil water content to complete the downscaling research of soil water content, which is soil water content The high spatiotemporal resolution monitoring provides a new way of thinking.

In view of the advantages of passive microwave remote sensing in global soil moisture data acquisition, a feasible spatial downscaling model is created to gradually improve the spatial resolution of soil moisture data. Through the comprehensive application of optical remote sensing and passive microwave remote sensing, passive microwave The practicability of microwave soil water content data products will eventually promote the temporal and spatial dynamic monitoring of watershed-scale soil water content. The specific downscaling method of the soil moisture content product is as follows.

With reference to shown in accompanying drawing 1, the downscaling method of soil water content product of the present invention mainly comprises the following steps:

A. Obtain passive microwave soil moisture content products and optical remote sensing image data at the same time;

The specific data acquisition method is as follows:

Obtain the SMAP soil moisture content product of the area to be studied. The SMAP satellite was launched by NASA in January 2015. It is equipped with an L-band radar and an L-band radiometer. The following examples select the daily average soil water data of SMAP L4 with a spatial resolution of 9 km. The download address is: https://ns idc.org/ .

At present, the time resolution of optical images with high spatial resolution is often low. Due to the influence of meteorological conditions, the success rate of image acquisition is low. In contrast, optical images with medium and low spatial resolution are easier to obtain due to their high temporal resolution. The optical remote sensing image data in the following examples are selected from the MOD09 ground albedo data in the MODIS land product with a spatial resolution of 500m that is synchronized with the SMAP soil water content data. Among them, the 1-7 band (620~670, 841~876, 459~479, 545~565, 1230~1250, 1628~1652, 2105~2155nm) data of this product were used to establish the soil water content inversion model.

In this step, the soil auxiliary data of the area to be studied is also obtained, which includes: (1) land use/land cover classification data, which can obtain MODIS land cover type products, which are used for the selection and selection of endmember components in the process of mixed pixel decomposition. Verification, to remove the influence of vegetation and other ground features on the agricultural soil spectrum to the greatest extent; (2) DEM data, in areas affected by terrain, this data can be used for correction and verification of the downscaling model of the present invention.

B. Extraction of soil spectrum based on multi-terminal mixed pixel decomposition method;

Multiple Endmember Spectral Mixture Analysis (MESMA), based on the physical meaning of endmembers, selects multiple spectra for each type of ground object, and generates multiple endmember combinations (each The endmember combination is composed of a certain spectrum in different ground objects), and then find the endmember combination with the least error of the least square method for each pixel, and then calculate the endmember ratio of each pixel.

In the present invention, the MODIS image with a spatial resolution of 500m is resampled to the same spatial resolution as the SMAP data of 9km, and the soil spectrum is extracted by applying the MESMA method to the MODIS image of 500m and the MODIS of 9km after resampling respectively to obtain a 9km space resolution and soil spectra at 500m spatial resolution.

Among them, the MESMA processing and analysis process mainly includes three parts: spectral library creation, optimal spectral library selection, and multi-terminal mixed pixel analysis. The creation of its spectral library includes creating a spectral library based on ROI, making spectral library metadata, and spectral library management; the selection of its optimal spectral library is the key to the success of multi-terminal mixed pixel decomposition. The optimal spectral library selection process includes creating a square array and spectral library optimization, the preferred method is to combine COB (Count-based Endmember Selection), EAR (Endmember Average RMES), MASA (Minimum Average Spectral Angle) calculations to obtain the spectral library that can best represent various ground features; The pixel decomposition process mainly adopts the V-I-S (vegetation-impermeable surface-soil) model, and the optimal vegetation, impermeable surface, and soil spectral sets are combined to form a 2EM, 3EM, and 4EM mixed pixel analysis model, and the MESMA results are analyzed based on the optimized results. Shade normalization processing, and finally get the mixed pixel decomposition results of different endmember combinations, that is, the root mean square error (RMSE) of the abundance value of each endmember and the accuracy of the result.

Then use the endmember abundance value obtained by the above MESMA and the following formula (1) to obtain the endmember spectrum of the soil in the study area.

Among them, R _s (λ) is the reflectivity of the soil spectrum in the band λ, R(λ) is the reflectivity of the pixel in the band λ, R(i, λ) is the reflection of the i-th end member in the band λ rate, f _i is the i-th endmember abundance value. N is the number of end members, ε _λ is the residual. The sum of the abundance values of all endmember components was defined as 1.

In this embodiment, an automatic extraction method of pure pixel spectrum is also designed in combination with ENVI and Matlab. Use ENVI software to extract the abundance value, RMSE and residual of the endmember combination model of each sampling point, determine the endmember combination model selected by the sampling point and output the combination abundance value f _i of the sampling point.

Firstly, in the randomly selected sampling points, the sampling points containing impervious surfaces are eliminated according to the land use classification type, and only the S-V model is established for the image, and attribute values are assigned to the combination model of the S-V spectrum, and the soil endmembers and vegetation in each combination model are extracted The coordinate values of the end members, extract the coordinate values of each spectrum in the spectral library, and match the two to obtain the combined spectrum of the S-V spectral combination model; extract the attribute values of each sampling combination model, and match them with the attribute values in the S-V spectral combination model library. The combined spectrum R(i,λ) of the S-V endmembers at each sampling point. The soil spectrum of each sampling point is calculated according to the above mixed pixel decomposition formula (1).

C. Using GA-PLS to establish a quantitative relationship model between soil spectra and soil moisture obtained by passive microwaves;

On the basis of the acquisition of the soil spectrum with a spatial resolution of 9 km above, based on the characteristic law of soil spectrum changing with soil moisture content, the partial least squares-genetic algorithm (GA-PLS) was introduced to construct a quantitative relationship model between soil moisture content and the spectrum , as the basis for downscaling.

Among them, PLS is a regression modeling method of multiple dependent variables on multiple independent variables, which can be applied to the analysis of full spectrum data or partial spectrum data. It combines the decomposition and regression of the data matrix to obtain the eigenvalue vectors related to the predicted components, which makes it applicable to complex analysis systems and obtains more robust results. The basic idea of genetic algorithm (GA) is to carry out natural selection in the evolution process, which is realized by gene recombination and mutation. The most notable advantage of this algorithm is that it avoids the problem of initial value selection. In order to reduce the amount of calculation and improve the precision of the PLS model, the present invention uses a genetic algorithm (GA) to screen spectral parameters, and eliminates variables that do not respond obviously to changes in water content.

The present invention uses GA-PLS model to set up soil water content reflection spectrum characteristic model, as the MODIS reflectance of 9km, band ratio, curvature and SMAP L4 soil aquatic products are used as the input variable of GA-PLS model, obtain the MODIS reflection on 9km scale The GA-PLS model between rate, band ratio, curvature and soil water content was calculated, and the accuracy was evaluated.

D. Based on the quantitative relationship model established in step C, a soil water content downscaling model was constructed using Taylor series expansion to obtain soil water content data with a spatial resolution of 500m.

The basis of the present invention's downscaling is the quantitative relationship between soil water content and soil spectral reflectance, namely set forth in step C based on passive microwave soil aquatic products and MODIS soil reflection characteristics such as reflectance of each band, band ratio (such as GA-PLS model of MODIS3/MODIS1), curvature (eg: MODIS3×MODIS1/(MODIS2) ² ). In the process of changing the spatial scale, the relationship between soil water content and reflectance, band ratio, curvature, etc. changes is nonlinear, so the present invention uses the form of Taylor series expansion, such as formula (2), to construct different scales The relationship between soil moisture content.

Among them, θ _n-1 and θ _n represent low spatial resolution (eg: 9km) and high spatial resolution (eg: 500m) soil water content respectively, R _n-1 (λ _i ) _s and R _n (λ _i ) _s represent the reflectance of the i-th band of the soil spectrum with low spatial resolution and high spatial resolution, respectively, Ratio _n-1 (j) _s ) and Ratio _n (j) _s represent low spatial resolution and high spatial resolution The soil spectral band ratio of , Curv _n-1 (k) _s and Curv _n (k) _s represent the soil spectral curvature of high spatial resolution and low spatial resolution, respectively. M, N, and L represent the total number of variables for reflectivity (i), band ratio (j), and curvature (k), respectively. The relationship f represents the relationship between the low spatial resolution soil spectral reflectance, band ratio and curvature and soil water content obtained by the GA-PLS model.

In this step D, the downscaling operation can be carried out in a gradually decreasing manner according to the situation. For example, first select a scale interval t (such as 4km) to reduce the 9km soil water content data to the (9-t) scale, and use the formula (2) repeatedly , use the high spatial resolution data calculated in the previous step as the low spatial resolution data in the next formula to calculate the soil water content with higher spatial resolution again, and finally gradually approach the requirement of the final 500m spatial resolution. The optimal scale interval t can be determined according to the actual data acquisition and result accuracy.

Another purpose of this step-by-step downscaling method is to continuously control the inversion accuracy of the soil water content results during the downscaling calculation process. The soil water content results of each downscale scale inversion will establish a correlation with the measured data of soil water content at the same spatial resolution. If the correlation of the verification data is high, the downscaling calculation will continue; The scale operation is stopped in due course, and the influencing factors are analyzed. If the accuracy of downscaling is affected by factors such as terrain and vegetation, new relationship items can be added to the downscaling model, and adjustments and optimizations can be made in time. Finally, the downscaling results of the model will be validated by synchronous field soil moisture test data.

The process of the method for downscaling the soil moisture content product based on the soil optical properties of the present invention is described below with a specific example of downscaling the SMAP soil moisture content data in the Central Plains of the United States.

In this example, SMAP L4 data with a spatial resolution of 9 km is selected for the passive microwave soil moisture content product, and MODIS 09A1 data with a spatial resolution of 500 m is selected for the optical image.

After the data is acquired, the MODIS data is first preprocessed by using ENVI software for projection conversion, mosaic, cropping, and resampling.

Then select the optimal spectral library, establish soil and vegetation type spectral libraries by visual recognition, and use COB (Count-based Endmember Selection), EAR (Endmember Average RMES), MASA (Minimum Average Spectral Angle) to calculate and optimize The spectral curves that best represent soil and vegetation were established to establish an optimal spectral library, and finally 13 optimal spectra for soil were selected, as shown in Figure 2, and 6 optimal spectra for vegetation, as shown in Figure 3.

Then, the optimized vegetation and soil spectral sets were combined to form a 3EM mixed pixel analysis model, and the results were normalized by shadow processing, and finally the mixed pixel decomposition results of different endmember combinations were obtained. The MODIS images with a spatial resolution of 500m were resampled to obtain images with a spatial resolution of 9km and 5km.

MESMA analysis was performed on the MODIS image with a resolution of 9 km, as shown in Figure 4. The reflectance, band ratio and curvature of the 9km resolution soil spectrum were calculated. Randomly select 51 sampling points, as shown in Figure 5, among them, 36 points are used as the calibration set, 15 points are used as the verification set, and the soil spectrum and SMAP soil water content of 51 sampling points are extracted respectively, and they are used as parameters Input to build a GA-PLS model between soil spectra and water content. The results show that the coefficient of determination of the calibration set is 0.78, and the coefficient of determination of the verification set is 0.72, as shown in Figure 6.

The GA-PLS model is used as the basis of the downscaling model to establish a step-by-step spatial downscaling model. Step-by-step downscaling selects the spatial interval t=4km, and first reduces the soil moisture content to 5km. It is necessary to perform MESMA processing and analysis on the MODIS data resampled to 5km resolution, as shown in Figure 7, and further calculate the 5km-scale soil spectrum The reflectivity, band ratio and curvature of the above formula (2) are used to obtain the soil water content with a 5km spatial resolution. It has been verified that the correlation between the 5km resolution water content data and the measured data is not lower than that of the 9km. The accuracy R ² of the comparative analysis is 0.54, as shown in Figure 8.

Repeat the previous step to establish a GA-PLS model of soil spectrum and water content at a scale of 5 km. A total of 89 sampling points were selected, as shown in Figure 5, including 62 calibration set samples and 27 validation set samples. The results It shows that the coefficient of determination of the calibration set at the 5km scale is 0.71, and the coefficient of determination of the verification set is 0.67, as shown in Figure 9. The 500m resolution MODIS data is processed and analyzed by MESMA, as shown in Figure 10, and the reflectance, band ratio and curvature of the soil spectrum at the 500m scale are calculated. Using the GA-PLS model established at the 5km scale, the formula (2 ) to obtain soil water content data with a spatial resolution of 500m. Using the correlation analysis between the measured soil water content and the 500m soil water content, the accuracy R2 of the comparative analysis with the measured data obtained at the site is 0.49, as shown in Figure ¹¹ .

The present invention utilizes the advantages of both passive microwave remote sensing data and optical remote sensing data in temporal and spatial resolution, effectively integrates the two to conduct downscaling research on soil moisture content products, and realizes real-time or quasi-real-time dynamic monitoring of watershed-scale soil moisture content .

The above is only a preferred embodiment of the present invention, and does not limit the present invention in any form. Those skilled in the art make some simple modifications, equivalent changes or modifications by using the technical content disclosed above, all of which fall within the scope of the present invention. within the scope of protection of the invention.

Claims (8)

1. a kind of soil moisture content product NO emissions reduction method, which is characterized in that the described method comprises the following steps:

A. the passive microwave soil moisture content product in area to be studied and the optical remote sensing image data of same time are obtained；

B. based on multiterminal member mixed pixel decomposition method the optical remote sensing image data are carried out with the extraction of soil spectrum；

C. the soil spectrum reflectance signature is established using GA-PLS to obtain with from the passive microwave soil moisture content product Soil moisture content between causes, the causes be calculated based on the soil spectrum it is every The reflectivity of each wave band in the soil spectrum of a pixel, band ratio, curvature and the soil moisture content quantitative relationship；

D. the causes established based on step C drop ruler using Taylor series expansion form building soil moisture content Model is spent, the soil moisture content data of high spatial resolution are obtained；

Wherein, the expression formula of Taylor series expansion form are as follows:

Wherein, θ_n-1And θ_nRespectively represent low spatial resolution and high spatial resolution soil moisture content, R_n-1(λ_i)_sAnd R_n(λ_i)_s Respectively indicate reflectivity of the soil spectrum of low spatial resolution and high spatial resolution in the i-th wave band, Ratio_n-1(j)_sWith Ratio_n(j)_sRespectively indicate the soil spectrum band ratio of low spatial resolution and high spatial resolution, Curv_n-1(k)_sWith Curv_n(k)_sThe soil spectrum curvature of high spatial resolution and low spatial resolution is respectively indicated, M, N and L respectively represent reflectivity I, the total number of variable of band ratio j and curvature k.

2. soil moisture content product NO emissions reduction method according to claim 1, which is characterized in that passive in the step A Microwave soil moisture content product uses SMAP soil moisture content data；The optical remote sensing image data use MODIS image number According to.

3. soil moisture content product NO emissions reduction method according to claim 2, which is characterized in that extracted in the step B The method of the soil spectrum are as follows: by the MODIS image resampling of high spatial resolution to low spatial same as SMAP data Resolution ratio, respectively to the MODIS application MESMA of low spatial resolution after the MODIS image of the high spatial resolution and resampling The extraction of method progress soil spectrum.

4. soil moisture content product NO emissions reduction method according to claim 3, which is characterized in that the MESMA method packet Library of spectra creation, the selection of optimal library of spectra and multiterminal member mixed pixel analytical procedure are included,

The library of spectra creation includes based on ROI creation library of spectra, the production of library of spectra metadata and library of spectra management；

The selection of the optimal library of spectra includes that creation square array and library of spectra are preferred；

The multiterminal member mixed pixel analysis uses vegetation-impervious surface-soil model, by preferred vegetation, impervious surface, soil The combination of earth spectra collection constitutes 2EM, 3EM, 4EM mixed pixel analysis model, the preferred result pair based on the optimal library of spectra MESMA result carries out shade normalized, obtains each end member Abundances and indicates the root-mean-square error of result precision, recycles Each end member Abundances and following equation obtain research area's soil endmember spectra,

Wherein, R_s(λ) is reflectivity of the soil spectrum in wave band λ, and R (λ) is reflectivity of the pixel on wave band λ, and R (i, λ) is the Reflectivity of the i end member on wave band λ, f_iFor i-th of end member Abundances, N is end member number, ε_λIt is residual error, all end-member compositions The sum of Abundances be defined as 1.

5. soil moisture content product NO emissions reduction method according to claim 1, which is characterized in that drop ruler in the step D The method for building up for spending model carries out NO emissions reduction operation using gradually decreasing fashion, progressively reaches the requirement of high spatial resolution.

6. soil moisture content product NO emissions reduction method according to claim 5, which is characterized in that the step D high and medium Between the soil moisture content data of resolution ratio refer to the soil moisture content data of 500m spatial resolution.

7. soil moisture content product NO emissions reduction method according to claim 1, which is characterized in that also same in the step A When obtain the soil auxiliary data in the area to be studied, the soil auxiliary data includes: land use or land cover classification Data and dem data.

8. soil moisture content product NO emissions reduction method according to claim 7, which is characterized in that the step C was established GA-PLS model can also increase the input parameter of the dem data or Gradient of each pixel as GA-PLS model.