These methods, instrumentation and/or protocols apply to all data in this dataset:| Methods and protocols used in the collection of this data package |
|---|
| Description: | The GOES-R Series Advanced Baseline Imager (ABI)
The ABI is the primary Earth-observing sensor aboard GOES-R15,16. The four satellite GOES-R Series began in November, 2016 with the launch of GOES-16. GOES-16 has remained in the GOES-East position ever since. GOES-17 served as GOES-West starting in 2018, however a cooling issue on its loop heat pipe caused partial loss of imagery, and it was replaced by GOES-18 in 2022. GOES-East and GOES-West orbit at approximately 35,000 kilometers above the equator at 75.2 and 137.2 degrees West. Together they view the entire Western Hemisphere, from eastern Africa to Australia and from Alaska to Chile.
The ABI is a passive radiometer that scans the atmosphere, oceans, and Earth surface at sixteen discrete wavelengths ranging from visible to thermal infrared. In its current operational mode (Mode 6), ABI produces a full disk hemispherical image every ten minutes, a CONUS (Continental United States) or PACUS (Pacific U.S.) image every five minutes, and two mesoscale images per minute. Mesoscale regions are small movable domains that can provide detailed temporal coverage of regions with heightened meteorological interest. Twelve of the sixteen ABI bands have two kilometer (km) spatial resolution at the sub-satellite point (nadir). The shortwave bands 1, 3 and 5 have one-km resolution, while band 2 has 0.5-km resolution.
|
| | Description: | ABI Fixed Grid
Due to the geostationary orbit of GOES satellites, their position and viewing geometry relative to the Earth's surface is, ideally, unchanging. The ABI fixed grid represents each spatial domain (full disk, CONUS/PACUS, and mesoscale) as a grid of ABI scan angles which describe the North/South and East/West orientation of the ABI scan mirrors for every pixel. For each spatial resolution, any two adjacent pixels have equal angular separation. In other words, scan angles remain constant across the fixed grid. However, pixel surface area increases moving away from nadir because a constant scan angle corresponds with greater distance as the earth curves away from the sub-satellite point.
To accurately map eddy covariance tower locations onto the ABI fixed grid and obtain ABI observations, we needed to align the ABI and tower location information. For GOES-R Level 1b and most Level 2 products, geographic information for each data file is stored as horizontal (x) and vertical (y) scan angles. Converting geodetic latitude and longitude coordinates to ABI scan angle coordinates is necessary to map eddy covariance tower locations on the ABI fixed grid.
ABI fixed grid products are not terrain-corrected: there is no adjustment for the off-nadir view angle of the satellite relative to surface targets. The “parallax effect” causes the satellite to perceive high-elevation targets to be displaced from their true location by a distance that increases with target’s elevation and satellite view zenith angle (VZA). GOES satellites only have a nadir view of equatorial surface targets at the sub-satellite points (75.2 °W and 137.2 °W); all other regions require terrain-correction for proper geolocation of elevated targets. Since the present research is concerned with the eddy covariance towers at point locations, it is only necessary that the correct ABI pixel is matched with the targeted tower. The true tower location is shifted by the magnitude and direction of the parallax displacement to the location where it is perceived to be located by the ABI fixed grid, before the tower is matched with an ABI pixel.
|
| | Description: | ABI Level 2 (L2) Products
ABI scans the full disk in under ten minutes, data are processed, and individual .nc files for each data product are made available in near-real time. Most ABI products are created every time a full-disk and CONUS scan is completed, but others currently have less frequent refresh rates, such as once per hour.
|
| | Description: | L2 Cloud and Moisture Imagery (CMI)
CMI provides reflectance values or brightness temperatures at sixteen ABI channels. The primary data source for this product is the Level 1b (L1b) Radiance product, measuring solar radiation (in W m−2 sr−1) at all sixteen ABI bands. For the six reflective bands (Bands 1-6), radiance values are converted to a dimensionless reflectance factor ranging from 0 to 1.
CMI reflectances are considered top-of-atmosphere (TOA) rather than surface reflectances because they measure the total reflectance received by the satellite at the top of the atmosphere, without accounting for atmospheric scattering. For the ten emissive bands (7-16), L1b radiances are converted to brightness temperature (K) using Planck’s function. While these longer wavelength measurements are not directly used to measure vegetation, they provide critical atmospheric and environmental context such as characterizing clouds, aerosols, fire, and snow that are of importance for terrestrial carbon cycle science.
|
| | Description: | Bidirectional Reflectance Factors
The L2 bidirectional reflectance factor (BRF) product has been an operational ABI product since August 18, 2021, which provides surface reflectances as a byproduct of the L2 Land Surface Albedo (LSA) product. The LSA algorithm derives Bidirectional Reflectance Distribution Function (BRDF) parameters, which are used both to estimate broadband albedo and to simulate surface reflectance on cloudy days when it cannot be measured directly. Solving for BRDF parameters is accomplished by minimizing a cost function which relates TOA reflectances and Atmospheric Optical Depth (AOD), both of which can be computed from ABI measurements over the course of the day as the solar zenith angle changes.
The BRF algorithm has two paths available for deriving surface reflectances. The default and more accurate method, the R3 algorithm, assumes the surface is Lambertian and directly calculates surface reflectance (rs) from TOA reflectances (r) and atmospheric parameters. Transmittance (γ), and path reflectance (r0) and spherical albedo (ρ) are retrieved from a look-up table which pre-calculates these parameters given viewing geometry and AOD using MODTRAN.
A back-up method is necessary for cloudy conditions where the atmospheric parameters are not available. The R2 algorithm is used to calculate surface BRF from the BRDF parameters retrieved from the prior day’s TOA reflectance measurements to model BRF throughout the day given satellite and solar viewing geometries. Every BRF pixel is tagged with a data quality flag noting whether the R2 or R3 algorithm was used. Another data quality flag indicates the pixel’s level of cloudiness, ranging from clear-sky, to low, medium or high probability cloudiness.
Data availability is limited spatially and temporally because the BRF algorithm is dependent on viewing geometry. The algorithm is not run when either the sun or satellite stray significantly from the zenith, the highest point in the sky relative to the surface target. The VZA of a geostationary satellite to a target on the surface does not change, hence the geographical range where data gets processed is always limited to VZA < 70 degrees. This range is smaller than other full-disk ABI products. For example, the GOES-16 full disk BRF product is valid across most of the continental United States, but excludes the northwestern US, Alaska, and central-northwestern Canada. Solar zenith angle (SZA) varies throughout the day and the algorithm only runs when SZA < 67 degrees. In the Northern Hemisphere winter, when the sun is low in the sky and daylight is short-lived, BRF data are limited to a few mid-day measurements and at high latitudes, the months of December and January have no valid BRF measurements. Inversely, long summer days result in more BRF measurements due to the advantageous sun angles. Near the equator, the number of BRF measurements per day is much less variable.
The BRF product algorithm has two different methods for calculating surface reflectances, depending on whether clear-sky observations are available. Under clear-sky conditions, the surface reflectance is computed directly from the TOA reflectances and atmospheric measurements. When the surface is cloud-obstructed, the surface reflectance is modeled using BRDF parameters derived from prior day’s clear sky observations.
| | Data Source | |
| | | Description: | Land Surface Albedo
The Land Surface Albedo (LSA) product is produced in harmony with the BRF land surface reflectance product. Instantaneous broadband albedo is ideally derived from the clear-sky TOA reflectances and the prior day’s BRDF parameters, which in turn are estimated from aerosol optical depth, a daily stack of shortwave reflectances, and albedo climatology. The LSA product is limited by the same viewing geometry restrictions as the BRF product.
|
| | Description: | Downward Shortwave Radiation (DSR)
The Downward Shortwave Radiation (DSR) product measures the total instantaneous shortwave irradiance incident at the Earth’s surface integrated over visible and infrared wavelengths (0.2 to 4.0 μm). DSR consists of both direct and diffuse solar radiation, attenuated and scattered by the atmosphere, in W m-2. The DSR product is currently produced just once per hour at full disk and CONUS domains. A unique aspect of this L2 product is that DSR data is projected onto a Global Latitude and Longitude Grid, rather than the ABI Fixed Grid used for all other products discussed here.
|
| | Description: | Land Surface Temperature
Land Surface (Skin) Temperature (LST) records the instantaneous temperature of the Earth’s surface in degrees Kelvin. The LST product can only be produced under clear-sky conditions, hence cloud-obstructed observations are masked out. Like DSR, LST is also produced just once per hour. For this reason, LST and DSR were upsampled to match the half-hourly cadence of most Ameriflux time-series, and interpolated values are noted in the data files. The half-hour timestamp values were filled using cubic interpolation between consecutive existing LST data points.
|
| | Description: | Clear Sky Mask
The Clear Sky Mask, also called the Cloud Mask, provides a binary image with each pixel classified as either “clear” or “cloudy". First, the algorithm employs spectral, spatial and temporal tests on each pixel to categorize the pixel as “clear”, “probably clear”, “probably cloudy” and “cloudy.” Classifications are compared to the model outputs from the Community Radiative Transfer Model (CRTM). This four-class intermediate product is a critical input to many other ABI L2 product algorithms and the four classes are condensed into a binary mask before the final product is distributed to users.
| | Data Source | |
| | | Description: | Aerosol Detection Product
The Aerosol Detection Product (ADP) consists of three separate variable layers, each of which is a binary mask representing ‘yes detection’ or ‘no detection’. The three types of aerosol detections are dust, smoke, and aerosols generally (dust or smoke detected). There are two distinct ADP algorithm pathways for observations over land and ocean, but both begin by masking out high and optically thick clouds. Notably, an ADP product data quality flag denotes “invalid detection due to snow_ice_clouds”, information retrieved from the GOES L2 Snow/Ice product, which can be used as a proxy for masking out snow surface cover in other products.
|
| | Description: | Aerosol Optical Depth
The Aerosol Optical Depth (AOD) product retrieves aerosol optical thickness over both land and ocean. Specifically, AOD measures the extinction of solar radiation due to atmospheric aerosols at a wavelength of 550 nm. In addition, the product provides the aerosol particle size, as represented by two Ångström exponents. The algorithm relies on instantaneous TOA reflectances, and a look-up table of atmospheric parameters precalculated using a radiative transfer model. Different ABI reflectance channels are used for the land and the ocean AOD retrievals. The AOD algorithm relies on the aerosol type characterization generated by the ADP product.
|
| | Description: | Calculating NIRvP using GOES-R
To calculate NDVI, NIRv and NIRvP on a per-pixel basis, the three inputs required are ABI Band 2 (red) surface reflectance, ABI Band 3 (NIR) surface reflectance, and DSR. These values are retrieved from the L2 BRF and DSR products, respectively, and observations are filtered to remove poor quality observations using the corresponding data quality flags. The NDVI is the normalized difference between the red and NIR, which is multiplied by NIR to derive NIRv, and is then multiplied by photosynthetically active radiation (PAR) to derive NIRvP; both NIRv and NIRvP are strongly related to GPP. A photosynthetically active radiation (PAR) product is scheduled for forthcoming GOES-R data product releases and work is ongoing to provide PAR and DSR across GeoNEX52. In the interim, we estimated PAR (in W m−2) as 0.45 times DSR; we note that this will induce a small amount of uncertainty into the final NIRvP estimate as this conversion factor varies depending on atmospheric composition and solar position.
The flux community often uses photosynthetically active photon flux density with typical units of μmol m−2 s−1. PAR can be converted to photosynthetically active photon flux density PPFD by using a conversion factor of approximately 4.56 μmol J−1.
| | Data Source | |
| | | Description: | Eddy covariance
The AmeriFlux network relies on the efforts of individual tower operating teams across the western hemisphere31 which, coupled with NEON, Inc. eddy covariance towers, resulted in 314 eddy covariance towers with publicly-available data at time of writing. These data are collected by the tower-operating teams or NEON, Inc, and provide half-hourly (or in rare instances hourly) sums of carbon dioxide, water, sensible heat, and/or other trace gas fluxes and half-hourly (or hourly) averages or sums of micrometeorological variables, all quality control-checked by common algorithms and organized as .csv files. These files are updated shortly after new data are uploaded to AmeriFlux or NEON, which in practice may result in delays that can extend from months to years from the time at which data were collected.
| | Data Source | |
| | | Description: | This method step describes provenance-based metadata as specified in the LTER EML Best Practices. The source of the metadata used is the GOES-R Product Definition and Users’ Guide (PUG) Volume 5 (L2+ Products).
| | Data Source | |
| |
|
|
