阅读说明：本技术 一种基于地表反射率库的沙尘暴监测方法 (Sand storm monitoring method based on earth surface reflectivity library ) 是由 苏庆华 孙林 贾臣 于会泳 王春香 于 2019-06-26 设计创作，主要内容包括：本发明公开了一种基于地表发射率库的沙尘暴监测方法,传统沙尘暴监测方法利用沙尘与其它地物光谱特性的差异,设定固定的阈值,实现沙尘信息识别。用最小值法合成月地表反射率数据库,通过6S模拟不同大气参数条件下晴空像元的最大表观反射率阈值,卫星遥感数据表光反射率大于模拟表观反射率,判别为沙尘像元,实现动态阈值法沙尘暴监测。通过验证结果分析比较,沙尘提取范围与真彩色影像沙尘区域一致,吻合较好,96.8%的地面观测站点沙尘天气与监测结果相同,OMI AI数据与沙尘监测结果基本一致。验证结果说明该方法有效降低了混合像元及大气环境等因素的影响,能够以较高精度实现沙尘暴监测。(The invention discloses a sand storm monitoring method based on a ground surface emissivity library. And synthesizing a lunar surface reflectance database by using a minimum method, simulating the maximum apparent reflectance threshold of the clear sky pixel under different atmospheric parameters by 6S, judging the satellite remote sensing data surface light reflectance to be a sand pixel if the surface light reflectance is greater than the simulated apparent reflectance, and realizing the sand storm monitoring by using a dynamic threshold method. Through the analysis and comparison of verification results, the extraction range of the sand dust is consistent with that of a true color image sand dust area, the coincidence is better, the sand dust weather of 96.8% of ground observation stations is the same as the monitoring result, and the OMI AI data is basically consistent with the sand dust monitoring result. The verification result shows that the method effectively reduces the influence of factors such as mixed pixels, atmospheric environment and the like, and can realize the sand storm monitoring with higher precision.)

1. A sand storm monitoring method based on a surface reflectivity library is characterized by comprising the following steps: the sandstorm monitoring method based on the earth surface reflectivity library comprises the following steps:

step 1, constructing a surface reflectance database;

step 2, obtaining apparent reflectivity through the surface reflectivity;

step 3, determining the functional relation between the earth surface reflectivity and the simulated apparent reflectivity;

and 4, solving the simulated apparent reflectivity corresponding to the pixel, and comparing the simulated apparent reflectivity with the actual pixel value of the remote sensing image to identify the sand-dust pixel.

2. The method for monitoring the sandstorm based on the earth's surface reflectivity library as claimed in claim 1, wherein: step 1, constructing a surface reflectance database; the specific mode is as follows:

the MOD09A1 earth surface reflectivity data synthesized in 8 days are downloaded from a data and information system (EOSDIS) of an earth observation system, four MOD09A1 data product images can exist in each month, and the pixel with the lowest earth surface reflectivity in the four images is taken as the pixel value of the month by using a minimum value synthesis technology, wherein the formula is as follows:

I(i,j)＝Min(I₁(i,j),I₂(i,j),I₃(i,j),I₄(i,j))

wherein I is a synthetic image I₁、I₂、I₃、I₄Four MOD09a1 data product images in a month, i and j being the rows and columns of an image.

3. The method for monitoring the sandstorm based on the earth's surface reflectivity library as claimed in claim 1, wherein: step 2, obtaining apparent reflectivity through the surface reflectivity; the specific mode is as follows:

as can be seen from the atmospheric radiation transmission equation, assuming that the conditions such as aerosol mode, atmosphere and observation geometry are known, the apparent reflectivity can be obtained from the surface reflectivity, and the atmospheric radiation transmission equation is:

where ρ is_aIs the intrinsic reflectivity of the atmosphere (range radiation). From the formula, it can be seen that the apparent light reflectance is a function of the atmospheric optical thickness and the surface reflectance. Thus, at a known apparent reflectance ρ^*And the surface reflectivity rho, the rho of the atmospheric information can be obtained_aS and T (τ)_a,μ_s)T(τ_a,μ_v). Similarly, if the surface reflectivities ρ, ρ are at_aS and T (τ)_a,μ_s)T(τ_a,μ_v) Given this knowledge, the corresponding apparent reflectance ρ can be calculated^*。

4. The method for monitoring a sand storm based on a surface reflectivity library of claim 1, wherein the step 3. determining a functional relationship between the surface reflectivity and the simulated apparent reflectivity; the specific mode is as follows: respectively simulating four wave bands of 1, 3, 6 and 7 of MODIS data to obtain the apparent reflectivity of the earth surface reflectivity of each pixel under different sun and satellite zenith angles under the clear air atmosphere condition, taking the maximum value of the simulated apparent reflectivity under various conditions, determining the functional relation between the earth surface reflectivity and the simulated apparent reflectivity,

sand and dust monitoring dynamic threshold equation

。

5. The method for monitoring the sandstorm based on the earth's surface reflectivity library as claimed in claim 1, wherein: step 4, calculating the simulated apparent reflectivity corresponding to the pixel, comparing with the actual pixel value of the remote sensing image, and identifying the sand-dust pixel; the specific mode is as follows:

the sand dust discrimination satisfies the following conditions:

R＝R_B∪R_R∪R_SWRI1∪R_SWIR2

in the formula

6. The method for monitoring the sandstorm based on the earth's surface reflectivity library as claimed in claim 5, wherein: and 4, when the sand storm is identified by the sand monitoring dynamic threshold method, the cloud in the atmosphere is judged as the sand storm by mistake, so that possible cloud pixels need to be separated.

7. The method for monitoring the sandstorm based on the earth's surface reflectivity library as claimed in claim 6, wherein: the normalized sand dust discrimination index NDDI (normalized Difference DustIndex) is constructed by adopting a normalization processing method:

in the formulaAnd

Technical Field

The invention belongs to the field of environmental research, and particularly relates to a sand storm monitoring method based on a surface reflectivity library.

Background

Sandstorm (sandstorm) is a general name of both sandstorm and duststorm, and refers to a severe windy and sandy weather phenomenon that a large amount of sandstorm substances on the ground are blown up and drawn into the air by strong wind, so that the air is particularly turbid and the horizontal visibility is less than one kilometer. The sandstorm is a regional strong disaster weather phenomenon, has strong destructive power to an ecosystem, can accelerate land desertification, causes serious pollution to the atmospheric environment, obviously reduces the air quality, has negative effects on human health, agricultural production, transportation, communication and the like, can supplement nutrient substances such as iron, nitrogen and the like for oceans through the sedimentation of the sandstorm, changes the marine ecosystem and biochemical cycle, and indirectly influences regional and global climate change. The research means of the sand storm mainly comprises remote sensing monitoring and ground fixed-point observation, the sand storm mainly occurs in deserts and arid and semiarid regions near the deserts, and the distribution density of ground observation stations is sparse, so that the research by using ground observation station network data has great limitation, and the multi-source property, the dynamic property, the instantaneity and the accuracy of the remote sensing data not only make up the deficiency of the time-space resolution of the ground observation data, but also can be mutually verified and supplemented with the ground observation data in precision, and can be widely applied to the research of the monitoring, the tracking, the forecasting, the disaster assessment and the like of the sand storm.

The current remote sensing monitoring technology mainly uses visible light, near infrared and thermal infrared data of satellites to monitor the sand storm. The Ackerman monitors the sand storm by utilizing the characteristic that the bright temperature difference of the sand at the positions of 11 mu m and 12 mu m of the thermal infrared band is negative; huxiuqing and the like are used for constructing an infrared Difference sand Dust index IDDI (acquired Difference Dust index) to monitor a sand storm according to the fact that infrared signals emitted to a space from the ground surface can be attenuated by sand Dust particles, and the real-time target brightness temperature is observed by a satellite to subtract the ground surface background brightness temperature; gunie et al uses 0.46 μm and 2.1 μm of MODIS data to construct a normalized sand discrimination index NDSI >0, and simultaneously uses 3.75 μm and 8.55 μm to construct another sand discrimination index DSI >33K to identify sand storms; the Lujingning and the like give a dust intensity index (DDI) based on a 1.6 mu m channel e index and an infrared splitting window channel ratio index, the DDI is more than 1 and is used as a dust identification basis, and the larger the value is, the stronger the dust intensity is; di et al uses the apparent reflectivities of visible light 0.65 μm and short wave infrared band 1.63 μm, and the optical thicknesses of bright and atmospheric aerosols of infrared band 3.9 μm and 10.8 μm to construct enhanced Dust index EDI (enhanced Dust index), which is identified as a Dust pixel by EDI > 0.

The remote sensing monitoring method selects one or more wave bands by analyzing the spectral characteristics of the sand particles in different wave bands such as visible light, near infrared, thermal infrared and the like, adopts different algorithms, gives appropriate threshold values and extracts sand dust information from different ground objects, thereby realizing the monitoring and the forecast of the sand storm, and the remote sensing monitoring method giving the fixed threshold value is called as a fixed threshold value method. The fixed threshold value method is convenient and quick to extract the sand dust information, has certain precision and stability, and is widely applied to the application of the remote sensing monitoring business of the sand dust storm. However, the remote sensing monitoring precision of the sand storm depends on the spectral characteristics of sand particle minerals, the characteristics of an underlying surface through which the sand passes, the sand density, the type and density of cloud, the sensor characteristics, the height of a sand layer, the sand monitoring technology and the like, and the traditional fixed threshold method is influenced by complex factors such as mixed pixels, complex earth surface composition, atmospheric environment background and the like, so that the monitoring is not ideal in some areas and conditions, and the problems of difficult monitoring of a thin sand area, small monitoring range of sand blowing weather, easy misjudgment of high-reflectivity earth surface, confusion of sand and cloud and the like exist.

Disclosure of Invention

The method for monitoring the sand storm by the dynamic threshold supported by the earth surface reflectance database is used for reducing the influence of factors such as mixed pixels, atmospheric environment and the like and improving the sand storm monitoring precision and stability. The method carries out deep research in the fields of aerosol, atmospheric correction, cloud detection and the like, and better solves the problems of mixed pixels and complicated atmospheric environment influences which bother remote sensing monitoring.

In order to achieve the purpose, the invention is realized by the following technical scheme: the sandstorm monitoring method based on the earth surface reflectivity library comprises the following steps:

step 1, constructing a surface reflectance database;

step 2, obtaining apparent reflectivity through the surface reflectivity;

step 3, determining the functional relation between the earth surface reflectivity and the simulated apparent reflectivity;

and 4, solving the simulated apparent reflectivity corresponding to the pixel, and comparing the simulated apparent reflectivity with the actual pixel value of the remote sensing image to identify the sand-dust pixel.

Preferably, step 1, constructing a surface reflectance database; the specific mode is as follows:

the MOD09A1 earth surface reflectivity data synthesized in 8 days are downloaded from a data and information system (EOSDIS) of an earth observation system, four MOD09A1 data product images can exist in each month, and the pixel with the lowest earth surface reflectivity in the four images is taken as the pixel value of the month by using a minimum value synthesis technology, wherein the formula is as follows:

I(i,j)＝Min(I₁(i,j),I₂(i,j),I₃(i,j),I₄(i,j))

wherein I is a synthetic image I₁、I₂、I₃、I₄Four MOD09a1 data product images in a month, i and j being the rows and columns of an image.

Preferably, the step 2 obtains the apparent reflectivity through the surface reflectivity; the specific mode is as follows:

as can be seen from the atmospheric radiation transmission equation, assuming that the conditions such as aerosol mode, atmosphere and observation geometry are known, the apparent reflectivity can be obtained from the surface reflectivity, and the atmospheric radiation transmission equation is:

where ρ is_aIs the intrinsic reflectivity of the atmosphere (range radiation). From equation (3), it can be seen that the apparent light reflectance is a function of the atmospheric optical thickness and the surface reflectance. Thus, at a known apparent reflectance ρ^*And the surface reflectivity rho, the rho of the atmospheric information can be obtained_aS and T (τ)_a,μ_s)T(τ_a,μ_v). Similarly, if the surface reflectivities ρ, ρ are at_aS and T (τ)_a,μ_s)T(τ_a,μ_v) Given this knowledge, the corresponding apparent reflectance ρ can be calculated^*。

Preferably, step 3. determining the functional relationship between the earth surface reflectivity and the simulated apparent reflectivity; the specific mode is as follows: respectively simulating four wave bands of 1, 3, 6 and 7 of MODIS data to obtain the apparent reflectivity of the earth surface reflectivity of each pixel under different sun and satellite zenith angles under the clear air atmosphere condition, taking the maximum value of the simulated apparent reflectivity under various conditions, determining the functional relation between the earth surface reflectivity and the simulated apparent reflectivity,

sand and dust monitoring dynamic threshold equation

。

Preferably, the step 4. calculating the simulated apparent reflectivity corresponding to the pixel, and comparing with the actual pixel value of the remote sensing image to identify the sand-dust pixel; the specific mode is as follows:

the sand dust discrimination satisfies the following conditions:

R＝R_B∪R_R∪R_SWRI1∪R_SWIR2

in the formula

For different wavesSimulated apparent reflectance of a segment, ρ_iIn order to know the reflectivity of the earth surface, alpha is the zenith angle of the sun, beta is the zenith angle of the satellite,

to apparent reflectance, R_iAnd R is a sand monitoring result. If the apparent reflectivity of the image data pixel

Greater than the simulated apparent reflectanceNamely R_iIf the number is positive, the pixel is determined to be a sand pixel.

Preferably, in step 4, when the sand storm is identified by the sand monitoring dynamic threshold method, the cloud in the atmosphere is misjudged as sand, so that possible cloud pixels need to be separated.

Preferably, a normalization processing method is adopted to construct a normalized sand Dust discrimination index NDDI (normalized Difference Dust index):

in the formula

And

the apparent reflectivities of the 7 th and 3 rd bands of MODIS data, respectively, and NDDI is the normalized dust index if NDDI>0, the pixel is the sand dust, otherwise, the pixel is the cloud pixel.

The invention has the beneficial effects that:

the method is characterized in that a MOD09A1 earth surface reflectivity 8-day synthesis product is adopted, a global earth surface reflectivity database is constructed by a minimum method synthesis technology, and dynamic threshold identification of each pixel is realized by simulating the relationship between the earth surface reflectivity and the apparent reflectivity under different conditions through 6S. The sand storm remote sensing monitoring dynamic threshold method is applied to four times of sand storm events, the sand extraction range is visually interpreted to be consistent with a real-color image sand area, the sand weather of 96.8 percent of MICAPS ground observation stations is the same as the monitoring result, the OMI AI data is basically consistent with the sand monitoring result, and the verification result shows that the sand storm monitored by the dynamic threshold method has higher monitoring precision, so that the influence of factors such as mixed pixels, the atmospheric environment and the like is effectively reduced.

Drawings

FIG. 1 is a graph of the spectral response of different feature types in MODIS (1-7 bands);

FIG. 2 is a graph of the results of surface reflectivity and apparent reflectivity simulations for different zenith angles of the sun and satellite;

fig. 3, fig. 3 is a comparison of the sandstorm monitoring results with MODIS raw image, MICAPS data and OMI AI product data;

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the accompanying drawings and examples, which are not intended to limit the present invention.

It can be seen from fig. 1 that the spectral curves of the sand (thick sand and thin sand) and the desert (gobi) are consistent, and the reflectance values increase with increasing wavelength, are lowest at 0.469 μm in the blue light band, reach a maximum at 1.64 μm, and decrease slightly at 2.13 μm. The reflectance values of clouds and snow (ice) are high in the visible and near infrared, usually above 0.50, but reach a minimum at 2.13 μm with a tendency to fall off linearly in the short-wave infrared. The spectral curve of the vegetation has no obvious change, and the reflectivity value of each wave band is below 0.16.

Through the spectral curve change characteristic analysis of various types of objects, the reflectivity value of the dust is in visible light and near infrared wave bands, generally between clouds and the ground surface, higher than vegetation with the reflectivity below 0.16 and lower than the reflectivity generally above 0.5, and the cloud and snow (ice) are in a certain range; in a short wave infrared band, the spectral curve change of the dust is just opposite to that of the cloud and the snow (ice), the reflectivity value of the dust in the area is higher and is usually 0.30, and the reflectivity of the cloud and the snow (ice) is just opposite to that of the vegetation; the reflectivity values of the desert (gobi) are obviously lower than those of the thick dust, but the desert (gobi) is not obviously limited to the thin dust area.

According to the difference of spectral curves of the sand dust and different ground objects in different wave bands, wave bands with large difference of reflectivity values are selected, a fixed threshold value is set, the sand dust is separated from the ground objects such as clouds, ice, snow, vegetation and the like, and the monitoring of sand dust storms is realized.

The premise of determining the threshold value by the traditional fixed threshold value method is to extract information from a pure pixel (pure pixel), but the actual remote sensing image is rarely composed of a single uniform ground object and is mostly formed by mixing several different ground object types. The spectral information of a single pixel in the remote sensing image is not only the spectral feature of a certain ground feature but also the mixed information of the spectrums of several ground features, the reflectivity of the observed mixed pixel (mixed pixel) is a function of the spectrums of the ground features in the pixel and the areas of the spectrums, and the formula is as follows:

in which the reflectivity of P pixels, the total number of N ground object types, e_iIs the reflectivity of each feature in the picture element, c_iIs e_iThe percentage of pixels, n, is not a deterministic factor (random noise).

As can be seen from the formula (4), the pixel value is determined by the total number of the feature types, the reflectivity of the features, the percentage of the features, and the independency factor. The pixels in the remote sensing image may be composed of sand, dust and desert, may be a combination of sand, cloud and gobi, may also be composed of ice, snow and vegetation, and the like, and the reflectivity value of each pixel represents the sum of the ground object radiation energy of the pixels with different properties. The mixed pixels are a common problem of remote sensing images, the reflectivity values of pixels formed by different ground objects in different areas are different, and a large amount of phenomena of missing judgment and erroneous judgment can be generated when the uniform fixed threshold value is used for monitoring the dust. The mixed pixel problem makes the remote sensing monitoring of the sand storm by the traditional fixed threshold value method more difficult.

MODIS (model Resolution Imaging Spectrophotometer) data have 36 wave bands, the spectral range is 0.4-14.44 mu m, the spectral range covers visible light, near infrared and middle and far infrared wave bands, the spatial Resolution is 250m, 500m and 1km respectively, the bandwidth is 2330km, and the earth observation coverage can be realized once in one to two days. The MODIS data has the characteristics of high spatial resolution, high temporal resolution, high spectral resolution and the like, and can provide rich information for monitoring the sand storm. According to the spectral curve characteristic analysis of different surface feature types in fig. 1, a wave band with a sand dust reflectivity value having a larger difference with other surface feature types is selected, five wave bands of 1, 3, 4, 6 and 7 are selected in visible light, near infrared and short wave infrared regions for sand dust storm monitoring research, and table1 is detailed parameters of the selected MODIS data of the wave bands of 1-7.

Table1 MODIS data 1-7 wave band parameter

Constructing a surface reflectance database

Construction of a surface reflectance database the MOD09a1 product was used. MOD09A1 is a MODIS earth surface reflectivity 8-day synthetic product obtained by Terra satellites, has a spatial resolution of 500m, covers 7 wave bands including visible light, near infrared and short wave infrared, and is obtained by correcting MODIS1B product 1-7 wave band data through atmosphere, aerosol and cirrus. An MOD09A1 earth surface reflectivity data product synthesized for 8 days is downloaded from a data and information system (EOSDIS) of a NASA (national aeronautics and astronautics administration) earth observation system, the earth surface reflectivity is supposed not to change within a certain period, four MOD09A1 data product images can exist in each month, a minimum value synthesis technology is used, and a pixel with the lowest earth surface reflectivity in the four images is used as a pixel value of the month, and the formula is as follows:

I(i,j)＝Min(I₁(i,j),I₂(i,j),I₃(i,j),I₄(i,j))

wherein I is a synthetic image I₁、I₂、I₃、I₄Four MOD09a1 data product images in a month, i and j being the rows and columns of an image.

Dynamic threshold method

From the atmospheric radiation transport equation, the apparent reflectivity can be obtained from the surface reflectivity, assuming that the conditions of the aerosol mode, atmosphere and observation geometry are known. Inputting parameters such as an atmospheric mode, an aerosol optical thickness, an observation geometry, an aerosol mode and the like into a 6S (the Second Simulation of the Satellite Signal in the aerosol spectrum) model, respectively simulating four wave bands of 1, 3, 6 and 7 of MODIS data to obtain the apparent reflectivity of the earth surface reflectivity of each pixel under different sun and Satellite zenith angles under clear air conditions, taking the maximum value of the simulated apparent reflectivity under various conditions, and determining the functional relationship between the earth surface reflectivity and the simulated apparent reflectivity (see equation 6-10).

The sand storm monitoring dynamic threshold method is characterized in that according to the atmospheric radiation transmission theory, according to the function relation between the earth surface reflectivity and the simulated apparent reflectivity, the earth surface reflectivity of a pixel obtained from a constructed earth surface emissivity database is substituted into a function equation, the simulated apparent reflectivity corresponding to the pixel is solved, then the simulated apparent reflectivity is compared with the actual pixel value of the remote sensing image, if the actually measured apparent reflectivity of the satellite is larger than the simulated apparent reflectivity, the pixel is considered as a sand pixel, and the dynamic threshold identification of each pixel is realized.

The sand monitoring dynamic threshold value method equation 6-10:

the sand dust discrimination satisfies the following conditions:

R＝R_B∪R_R∪R_SWRI1∪R_SWIR2 (10)

in the formula

Simulated apparent reflectance, p, for different wavebands_iIn order to know the reflectivity of the earth surface, alpha is the zenith angle of the sun, beta is the zenith angle of the satellite,

to apparent reflectance, R_iAnd R is a sand monitoring result. If the apparent reflectivity of the image data pixel

Greater than the simulated apparent reflectanceNamely R_iIf the number is positive, the pixel is determined to be a sand pixel.

Cloud identification

When a sand storm occurs, the sand storm is often accompanied by cloud generation. When the sand storm is identified by the sand monitoring dynamic threshold method, the cloud in the atmosphere is judged as sand by mistake, so that possible cloud pixels need to be separated. As can be seen from fig. 1, the cloud reflectivity is higher in the visible light and near infrared bands, and in the 3 rd band of the blue light band, the cloud reflectivity is represented as a high-value region, and in the 7 th band of the short-wave infrared region, the cloud reflectivity is the lowest, and the cloud is identified by using the characteristic that the difference between the cloud reflectivities of the cloud in the 7 th band and the 3 rd band is negative. Therefore, a normalized sand dust discrimination index nddi (normalized Difference dustindex) is constructed by a normalization processing method:

in the formula

And

the apparent reflectivities of the 7 th and 3 rd bands of MODIS data, respectively, and NDDI is the normalized dust index if NDDI>0, the pixel is the sand dust, otherwise, the pixel is the cloud pixel.

Fig. 3 is a comparison of the sand storm monitoring result with the MODIS original image, the MICAPS data and the OMI AI product data, wherein fig. 3-a is the MODIS true color composite image (RGB:1, 4, 3) of four sand storms, fig. 3-b is the sand storm monitoring result, fig. 3-c is the actual sand storm data of the MICAPS foundation site, fig. 3-d is the OMI AI product data, and the imaging and acquisition time of the related data is shown in table 2:

table2 sandstorm original image and verification data acquisition time statistical Table

Through visual interpretation and comparison of the true color synthetic image in the graph 3-a and the sand monitoring result in the graph 3-b, both thick sand dust and thin sand dust can be identified more accurately, particularly, the thin sand dust extraction effect is better on the low-planted covered high-reflectivity earth surface, the shape and the texture of the thin sand dust are clear and visible, and the interference problem of factors such as clouds, high-brightness earth surface and the like is reduced. The extraction range of the four sand storms is consistent with the sand dust area of the true color image, and the matching is better. The 96.8% MICAPS ground observation station sand weather data (figure 3-c) is the same as the monitoring result, the OMI AI data (figure 3-d) is basically consistent with the sand monitoring result, and the verification result shows that the dynamic threshold method has higher monitoring precision when monitoring the sand storm.

The technical content of the present invention is further explained by the examples only, so as to facilitate the understanding of the reader. The general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not limited to the embodiments shown herein, and any technical extension or re-creation performed according to the present invention is protected by the present invention.