常用静止轨道卫星(geostationary satellite)及相关产品

1. 数据简介

维基百科 link

地球静止轨道（或称地球赤道同步轨道，英语：geostationary orbit，简写：GEO）是指地球赤道面上方35,786km的圆形轨道，该轨道上航天器的运行方向和地球自转方向一致。在地球静止轨道上的航天器绕地球运行一周的时间和地球自转周期（一恒星日）相同，因此，在地面观测者看来，这样的航天器是在天空固定不动的。通信卫星和气象卫星一般运行在静止轨道，因此地面站天线只要对准卫星的定点位置就可以通讯，而不用转动天线。利用这个特点，把携带有可见光和近红外光传感器的海洋卫星发射到静止轨道上，这样就可以监测海洋环境的细微变化，比如GOCI卫星。

地球静止轨道是地球同步轨道的一个特例，二者之间有一些区别，地球同步轨道上的卫星每天在同样的时间通过地球上的同一个点，而地球静止轨道上的卫星一直固定在定点位置不动。

1.1 GOES-R 静止卫星

• GOES-R 系列静止气象卫星是目前NOAA 运行的最新一代静止气象卫星。该系列卫星包括GOES-R （GOES-16），GOES-S （GOES-17），GOES-T，GOES-U 等4 颗卫星，其中前两颗星分别在2016.11.19 和2018.3.1 发射运行，分别定轨在75.2°W 和137.2°W，后两颗预计在2021 年和2024 年发射（http：//www. goes-r. gov/［2020-06-08］）。

1.1.1 传感器

• GOES-R 上搭载了数个传感器, 如下图所示。

• (1) 环境类载荷包括：
  – 先进基线成像仪（Advanced Baseline Imager，ABI) 、
  – 同步闪电测绘仪（Geostationary Lightning Mapper，GLM）、
  – 远紫外与 X 射线辐照度探测器（Extreme Ultraviolet Sensor/X-Ray Sensor Irradiance Sensors，EXIS）、
  – 太阳紫外成像仪（Solar Ultraviolet Imager，SUVI）、
  – 空间环境监测器/磁力仪组（SEM/Magnetometer，SEM/MAG）、
  – 空间环境原位测量装置（Space Environment In-Situ Suite，SEISS）。

• (2) 数据收集类载荷包括：
  – 静止轨道搜索&救援系统（Geostationary Search andRescue System，GEOS&R）、
  – 数据收集系统（DCIS）等。

（A）高级基线成像仪（ABI）

• 其中，与陆面产品相关的传感器主要是高级基线成像仪（ABI）。ABI 有16 个波段，空间分辨率分别为0.5 km （0.64 μm），1.0 km （可见光—近红外）和2.0 km （>2 μm）（表1）。ABI有3种观测模式，分别对全圆盘（Full disk）、美国大陆（3000 km×5000 km）和中尺度区域（1000 km×1000 km）提供不同时间频率和空间分辨率的观测。时间分辨率分别为全圆盘15 min 以及美国大陆的5 min。中尺度区域观测由用户制定，提供特定感兴趣区域的高时频观测，可在30 或60 s 观测一次。

【补充：ABI 采用 3 种扫描模式对 3 类区域进行观测。3 种扫描模式分别为模式三、模式四和模式六；3 类扫描区域包括全圆盘区域（Full Disk）、美国本土区域（CONtinental United States，CONUS）以及中尺度区域（MESOs）。圆盘区域为以星下点和 ABI 中心连线为中心，ABI 对地方向以直径约 17.4的张角进行观测覆盖的地球表面区域，美国本土区域为3 000 km× 5 000 km 的长方形区域，不同的轨道位置对应的区域不同，具体如图 1 所示。中尺度区域为卫星观测区域内 1 000 km×1 000 km 大小的任意区域。
】

【补充： ABI 不同扫描模式对应的观测区域设置及其时间见表 7，数据大小见表 8。模式三中，Full Disk：扫描时间为 15 min；CONUS：扫描时间为 5min；中尺度区域（MESOs）：每个中尺度区域扫描时间（共2 个中尺度区域，每个区域 1 000 km×1 000 km）达到30 s，2 个中尺度区域同时扫描需 60 s。模式六除全圆盘区域扫描时间缩小为 10 min 以外，其他与模式三完全相同。模式四只完成 5 min 全圆盘扫描，美国本土观测模式可以从该模式全圆盘数据抽取得到。其中模式 6 为卫星系统默认观测模式。】

（B）地球静止轨道闪电绘图仪（GLM）

• 地球静止轨道闪电绘图仪（GLM），是一台工作在近红外谱段（777.4 nm）的 CCD 相机，空间分辨率为 8~12 km，可全天候地对发生在美国本土及近海上空的闪电进行探测和成像，补充现有陆基的测量能力。地球静止轨道闪电绘图仪能够将恶劣天气和飓风的预警时间提高到 20 min，甚至更快。GOES-16 是第一个携带闪电探测器的地球静止气象卫星。

1.1.2 标准产品

• GEOS-R产品中主要是大气产品, 有少数地面产品。 link

• （1）大气产品 link

• （2）地面产品（如地表温度，下行短波辐射，火点/热点，积雪覆盖）。如：在地面产品生产过程中，需要NOAA 国家环境预报中心（NCEP）的数值天气预报数据作为辅助（https：//www.ncep.noaa.gov/2020-06-08］）。4 个地表产品中除了下行短波辐射是经纬度坐标，其他三个都使用格网坐标和GRS80 地球椭球体，产品格式为netCDF-4，目前产品还处于试运行阶段（provisional），可通过NOAA 综合数据中心（CLASS，https：//www. class. noaa. gov［2020-06-08］）或环境信息中心（NCEI，https：//www. ncei. noaa.gov/products ［2020-06-08］）获取。

1.1.3 产品分级

• GOES-R 系列卫星产品主要包括 0 级数据、L1b数据和 L2+数据。其中 0 级数据为未经处理的原始数据；L1b 数据为经过辐射定标和地理定位后的数据，数据存储格式为 netCDF-4；L2+数据包括图像产品和定量产品，图像产品包含各个通道的云图和水汽图像，定量产品为对 L1b 数据进行环境参数反演得到的产品，以及对 L1 或 L2 数据进行时空降采样处理或者其他处理得到的数据[15-16]。各级卫星产品见表 10。

1.1.4 产品下载

产品下载的途径如下图所示 （中文教程可参考 link 和 link ）

1.1.5 Geostationary-NASA Earth Exchange (GeoNEX) (GEOS16)

• GeoNEX is a collaborative project led by scientists from NASA, NOAA, and many other institutes around the world to generate Earth monitoring products using data streams from the latest Geostationary (GEO) sensors including the GOES-16/17 Advanced Baseline Imager (ABI), the Himawari-8/9 Advanced Himawari Imager (AHI), and more.

• GEONEX DATA PORTAL link

(1) DATA DIRECTORIES

• GOES16/GEONEX-L1G/
  [6] Geostationary NASA Earth Exchange (GeoNEX) is a collaborative effort led by NASA, NOAA, and many other research institutes to explore the potential of geostationary data streams in generating rigorous and systematic science products. We introduced the GeoNEX Level 1G products, including the top-of-atmosphere (TOA) reflectance and brightness temperature, in a previous paper [8]. The GeoNEX Level 1G products include radiometrically calibrated and geometrically rectified TOA BRF and brightness temperatures reprojected onto a common grid in the geographic projection [8,9,10]. The common grid covers Earth’s surface from 60° N to 60° S and is divided into 6° × 6° tiles that were numbered from 0 to 59 horizontally and from 0 to 19 vertically. The spatial resolutions of the grid are 0.005°, 0.01°, and 0.02° in correspondence to the native 0.5 km, 1 km, and 2 km (nadir) spatial resolution of the original geostationary data in the fixed-point projection. The temporal resolution of the products is also the same as the original data, being 10–15 min for full-disk scans. The L1G data include accurate sun-target-satellite geometry information, in particular the solar zenith and azimuth angles, for each pixel in the domain at every time step. The data are archived in the EOS-HDF format and publicly accessible at the GeoNEX data portal (https://data.nas.nasa.gov/geonex; data were last accessed on 4 February 2022). More details regarding the GeoNEX L1G products can be found in [8].

  Top-of-Atmosphere (TOA) reflectance GOES-16 ABI. This product includes the Bidirectional Reflectance Factor (BRF) for Bands 1-6 and Brightness Temperature for Bands 7-16 at the top of the atmosphere measured by the satellite sensor.
  （1）BRF is defined as the ratio of the radiance reflected from the target surface (e.g., top of atmosphere) in a particular direction to the expected radiance reflected from a perfectly white Lambertian surface in the same direction under the same conditions of illumination and reflection. The phrase “bidirectional” emphasizes that the variable is a function of both solar incident and satellite-viewing angles.
  （2） Brightness temperature is the temperature of a blackbody that would emit the same amount of radiation as the targeted body in a specified spectral band. It is different from the true surface temperature because (1) the surface is not a perfect blackbody; and (2) it is influenced by atmospheric conditions. For more information on the steps used to create this product see https://www.nasa.gov/geone

• GOES16/GEONEX-L2/

  [6] The next (Level 2) step in the GeoNEX processing chain is atmospheric correction, which removes the effects of atmospheric absorption and scattering from the L1G TOA radiances to retrieve the surface reflectance (SR) and, as a by-product, the corresponding atmospheric aerosol optical depth (AOD). The algorithm we are adapting for this task is the Multiangle Implementation of Atmospheric Correction (MAIAC) system [11,12,13,14,15,16]. In this paper, however, we introduce a new algorithm, which is inspired by MAIAC and developed from its framework, to particularly exploit the diurnal variability in geostationary observations for atmospheric correction. (In this paper we do not distinguish the differences between “diurnal” and “daytime” but may use them interchangeably.) The algorithm, named GeoNEX-AC, is not intended to replace MAIAC as the operational algorithm for the GeoNEX L2G products but rather be a tool for diagnostic analyses

  GOES16 surface reflectance (Surface BRF) estimated through the process of atmospheric correction for Bands 1-6.
  （1）SR is calculated by removing the atmospheric influence (absorption and scattering by molecules and aerosols) from the TOA reflectance. The data are also filtered for unfavorable conditions (e.g., clouds and shadows) under which the retrieval of SR is not feasible or reliable. Intermediate products include cloud masks, atmospheric aerosol optical thickness (AOT), surface
  BRDF parameters, and so on.
  （2）The current surface reflectance dataset is produced with an early version of MAIAC. The accuracy of cloud detection and aerosol retrievals is not yet sufficient for science publications. We are still working on issues with the BRDF model fit and atmospheric correction.
  （3）The next improved version of MAIAC is in a stage of completion, and the updated results of processing will be available the in near future.

• GOES16/NOAA-L1B/
  GeoNEX GOES16 INPUTS, provided only as an convenience for exploration.

(2) FILE NAMING CONVENTIONS FOR L1G DATA

For all GeoNEX products, the following naming convention is used:

1. <Satellite/Sensor Code>
   GLBG_hv.hdf

2. <Satellite/Sensor Code>: HM08_AHI for Himawari 8 AHI, and GO16_ABI or GO17_ABI for GOES ABI, GK2A_AMI for GEO-KOMPSAT-2A AMI.

3. < Product ID >: A two-digit number assigned to each GeoNEX product. Currently, only the Product ID of “05” is
   available, indicating “Top-of-Atmosphere Reflectance”

4. < YYYYMMDD >: Year-Month-Day, e.g., 20180501

5. < HHMM>: Hour-Minute (UTC time), e.g., 2030

6. < hid >, < vid >: Horizontal or Vertical Tile ID

7. < version >: Product version code; currently “02”

Note, the input and convenience products may differ from the above naming convention.

(3) GEONEX-L1G数据介绍

• 数据生产

• 空间格网

• 下载

P 后面跟的是目标存储路径

(4) GEONEX-L12 surface reflectance

• 参考[6]
• The next (Level 2) step in the GeoNEX processing chain is atmospheric correction, which removes the effects of atmospheric absorption and scattering from the L1G TOA radiances to retrieve the surface reflectance (SR) and, as a by-product, the corresponding atmospheric aerosol optical depth (AOD). The algorithm we are adapting for this task is the Multiangle Implementation of Atmospheric Correction (MAIAC) system [11,12,13,14,15,16]. In this paper, however, we introduce a new algorithm, which is inspired by MAIAC and developed from its framework, to particularly exploit the diurnal variability in geostationary observations for atmospheric correction. (In this paper we do not distinguish the differences between “diurnal” and “daytime” but may use them interchangeably.) The algorithm, named GeoNEX-AC, is not intended to replace MAIAC as the operational algorithm for the GeoNEX L2G products but rather be a tool for diagnostic analyses. Before we introduce the motivations behind the development of GeoNEX-AC, let us review some of the basic ideas of atmospheric correction.
• Because the TOA radiance is a composite of solar radiations reflected by the surface and backscattered by the atmosphere, retrieving SR or/and the AOD from the TOA measurements is fundamentally under-constrained [17]. Additional information or a simplification must be introduced so that a solution may be obtained. Such information can be derived from various aspects of optical remote sensing data in spatial, spectral, angular, and temporal dimensions. Different algorithms have been developed to exploit different information sources.
  For instance, atmospheric aerosol loading generally varies smoothly over spatial scales under 50 km [18]. Therefore, we may opt to retrieve the AOD over dark pixels, where the measured TOA reflectance is mainly regulated by the atmosphere and thus the retrieved AOD is more accurate [19]. This is the basic idea underlying the MODIS Dart Target algorithm [20,21,22]. Note that pixels that are bright in some spectral bands could be dark in other bands. For instance, the visually bright deserts indeed appear dark in the 0.41 µm spectral region, which has motivated the development of the Deep Blue algorithm to retrieve the AOD over deserts and similar regions [23,24,25].
• A very common approach adopted by existing atmospheric correction algorithms to decouple the atmospheric and the surface regulation on TOA reflectance is based on the notion of spectral dependency between different bands [19]. On one hand, because the AOD generally decreases with wavelengths following the power law [26], its regulation to the TOA reflectance in the Short-Wave Infrared (SWIR) range (e.g., 2.2 µm) is relatively small. This allows us to more accurately retrieve the surface reflectance for the SWIR band. On the other hand, it was found that the ratios between the visible (i.e., Blue and Red) and the SWIR bands (e.g., 2.2 µm) over densely vegetated pixels are statistically stable [19,20,22]. When the band ratios are determined a priori, we can easily estimate the reflectance of the visible bands from the SWIR band [19] and subsequently retrieve the AOD in the visible spectral range [20,22].
• In general cases (without the restriction on dense vegetation pixels), the spectral band ratios can vary by land surface types as well as the illumination-view geometry [27,28]. Different approaches have been developed to determine the spatial and the temporal variations in the spectral band ratios. Since Collection 5, the MODIS Dark Target algorithm has used empirical models to account for land-cover and angular variations in the band ratios [20]. The latest Visible/Infrared Imager Radiometer Suite (VIIRS) AOD algorithm develops a spatial database of spectral band ratios to account for their spatial variability [29]. The MAIAC algorithm dynamically maintains band ratios at various illumination-view geometries with the latest surface reflectance retrievals at each pixel [11,16].
• Another important consideration of atmospheric correction algorithms is how to treat the anisotropy 各向异性of the surface reflectance. A full description of the angular dependence of surface reflectance (SR) requires the Bi-directional Reflectance Distribution Function (BRDF), which itself is challenging to model. The simplest approach is to neglect the surface anisotropy by assuming the surface to be Lambertian. The Lambertian assumption has the advantage of easy implementation and has been adopted in many atmospheric correction and aerosol retrieval algorithms, such as 6S [30,31], Dark Target [20], ATREM [32], ISOFIT [33], and many others [34,35]. This approach, however, neglects the angular variations in the surface reflectance, which can be an important information source for atmospheric correction for GEO observations.
• Recent atmospheric correction algorithms have emerged that explicitly account for the SR anisotropy. One example is the Aerosol and surface albedo Retrieval Using a directional Slitting method-application to GEOstationary data (AERUS-GEO) algorithm originally developed in [36,37] to process observations from the Spinning Enhanced Visible and Infra-Red Imager (SEVIRI) on board the Meteosat Second Generation (MSG) geostationary satellite. The algorithm adopts the semi-analytical Ross-Thick–Li-Sparse (RTLS) model to represent surface BRDF and evaluates its regulation on the TOA reflectance in a linearized semi-physical model at low computational cost [36]. The AERUS-GEO algorithm has been applied to process GOES ABI and Himawari AHI data [38].
• A more rigorous and comprehensive algorithm that explicitly accounts for the surface BRDF effects is MAIAC. Like AERUS-GEO, MAIAC also adopts the RTLS BRDF model. However, unlike AERUS-GEO, MAIAC rigorously derives the expression of the TOA reflectance (as a function of surface BRDF) with the help of Green’s function method [39,40]. Because the RTLS model essentially represents BRDF as a combination of kernel functions (of illumination and view geometries) weighted by a few linear parameters, MAIAC is able to efficiently and accurately exploit the angular dimension of the satellite observations.
• MAIAC also exploits temporal information from time series of satellite observations. The algorithm uses the differences in space-time variability between land surface and atmospheric optical conditions (e.g., clouds, aerosols, etc.) to separate land and atmospheric signals. It regularly monitors time series of surface reflectance in the Red, the Near Infrared (NIR), and the SWIR bands to detect potential changes in surface conditions due to weather (e.g., rain/snow) or natural disturbances (e.g., fires). The algorithm “remembers” and dynamically updates a set of physical signatures for each grid cell, including full spectral BRDF, spectral and thermal contrasts, and metrics of spatial variability from higher resolution channels with the latest measurements. These functional requirements increase the complexity of the implementation of the MAIAC algorithm, but help improve its capability in cloud/snow detection, which is a dominant error source in land remote sensing [11,12]. The data composition approach described above is based on the assumption that surface properties are stationary at short time scales. The choice of composition time windows is clearly affected by the satellite’s revisit frequency. LEO sensors such as MODIS and VIIRS complete the global coverage at daily scales. By the well-known Nyquist–Shannon sampling theorem [41], this means that these observations cannot be used to resolve surface processes that occur on bi-diurnal or shorter time scales. This time-scale limit can be much longer due to cloud contamination and issues in practice. For instance, the MAIAC algorithm composites MODIS observations with a sliding window up to 16 days [16], which is by no means ideal and only reflects the limitation of the LEO data. We will show in this paper that the 10-min data frequency of the GEO sensors provides a most promising alternative to address the issue.
• The above brief review shows that MAIAC has many advantages in processing multi-angular satellite data. For this reason, we are adapting it as the operational algorithm to generate GeoNEX Level 2 products. At the same time, because MAIAC is also applied to process data streams from many other satellites, such as MODIS and VIIRS, we are interested in the question: “what advantages do geostationary datasets have, in comparison with their LEO counterparts, that facilitate atmospheric correction?”
• This paper is set to answer this question. We will demonstrate that the diurnal variability in the geostationary TOA reflectance (or radiance), as represented by the high-temporal-resolution (i.e., ~10 min) time series, contains unique information that helps us separate the surface and the atmospheric signals without invoking other common techniques such as the spectral band ratios. We develop the GeoNEX-AC algorithm particularly to exploit the temporal and the angular information from the diurnal variations in the geostationary data. We will show that our algorithm has good performance on various, even challenging, land cover types. In addition, we will show that the retrieval results from our algorithm have the potential to help validate and refine conventional atmospheric correction techniques.

参考文献

[1] 方红亮. 基于地球静止气象卫星的地表参数遥感研究进展[J]. 遥感学报, 2021, 25(1): 109-125.
[2] Beginner’s Guide to GOES-R Series Data link
[3] 张春华, 李松, 刘洪伟, 等. 美国 GOES 系列卫星的发展[J]. 装备环境工程, 2022, 19(4): 124-131.
[4] L1G and L2 data introduction link
[5] Wang W, Li S, Hashimoto H, et al. An introduction to the geostationary-NASA earth exchange (GeoNEX) products: 1. Top-of-atmosphere reflectance and brightness temperature[J]. Remote Sensing, 2020, 12(8): 1267.
[6] Wang W, Wang Y, Lyapustin A, et al. A novel atmospheric correction algorithm to exploit the diurnal variability in hypertemporal geostationary observations[J]. Remote Sensing, 2022, 14(4): 964.