Home >> July 2008 Edition >> Feature: GOES Image Navigation and Registration
Feature: GOES Image Navigation and Registration

by Bruce Gibbs
Integral Systems, Inc.

The Geostationary Operational Environmental Satellites (GOES), operated by the National Oceanographic and Atmospheric Administration (NOAA), continuously track evolution of weather over almost a hemisphere. GOES primary functions are to support weather forecasting, severe storm tracking, and meteorological research.

The earliest GOES satellites, numbered 1 to 7, were 100 RPM spin-stabilized satellites where Earth imaging was accomplished using north-to-south detector step scanning on each spin. GOES-1, launched in 1975, flew Visible Infrared Spin Scan Radiometer (VISSR) and Space Environment Monitor (SEM) instruments. The VISSR provided cloud imagery and data for determining cloud and surface temperatures, and wind fields.

The next series of GOES satellites, designated GOES I-M1 and numbered 8 to 12, were first launched in 1994 ⎯ 10, 11 and 12 are still operational. These GOES are Earth-pointing, three-axis stabilized spacecraft supporting two 2-axis scanning instruments: the 5-channel visible/IR Imager with 1, 4 and 8 km resolution, and the 19-channel Sounder with 8 km resolution and 10 km sampling. The Imager and Sounder scan in an east-west (EW) direction with north-south (NS) steps at the end of east-west swaths. Other GOES I-M instruments include the Solar X-ray Imager (SXI) and SEM with magnetometer. Use of a three-axis stabilized spacecraft and 2-axis scanning instruments enable the sensors to "stare" at the Earth. GOES sensors image clouds, monitor Earth surface temperature and water vapor, and sound the atmosphere vertical thermal and vapor profiles. These capabilities allow tracking of dynamic atmospheric phenomena, particularly severe local storms and tropical cyclones.

GOES 13, the first in the NOP-series, was built by The Boeing Company and launched in May 2006; Integral Systems built most of the ground system. GOES NOP spacecraft2,3 retain the GOES I-M heritage instruments but the spacecraft differ in many respects.

Figure 1 shows the GOES 13 spacecraft configuration, antennas and instruments. GOES 13 differs operationally from GOES 8-12 in that imaging operations continue during eclipse periods, the spacecraft periodically yaw-flips to provide better instrument cooling, and daily momentum dumping maneuvers are used to offset solar torque generated by the single solar array.

GOES 13 pointing performance is greatly improved with respect to GOES 8-12 primarily because star trackers, rather than a scanning Earth sensor, are used as the attitude reference. Other improvements include closed-loop (versus open-loop) instrument compensation for spacecraft dynamic motion, and mounting of all instruments on an optical bench.

GOES-R4,5,6, the first in the next generation of GOES spacecraft, is currently in the proposal and procurement stage and is scheduled to be launched in 2015. GOES-R instruments will include the Advanced Baseline Imager (ABI), Geostationary Lightning Mapper (GLM), Solar Ultraviolet Imager (SUVI), EUVS and XRS Irradiance Sensors (EXIS), Space Environment In-Situ Suite (SEISS) and Magnetometer. The primary imaging instrument, the 16 channel ABI, has 0.5 km visible channel resolution and IR channel resolutions varying from 1 to 2 km. As explained later, GOES-R will operate quite differently than GOES I-P, and these differences will provide much greater pointing accuracy, greater scan flexibility and greatly reduced operator burden. Table 1 compares selected 3-σ visible channel INR requirements for the three GOES generations.

Image Navigation and Registration (INR)
For weather modeling purposes it is important that the Earth location of each GOES image picture element (pixel) be accurately known and that corresponding pixels from images separated in time view the same point on Earth. The later capability is important when accurately tracking severe weather or when generating movie loops.

Image navigation locates pixels relative to a fixed reference such as Earth latitude and longitude. Image registration maintains the spatial relationship between pixels within images and between images. GOES adjusts the instrument pointing so that image pixels appear to have been obtained from an “ideal” geosynchronous spacecraft located at a fixed point in space8, such as on the equator at 75o W longitude for GOES-East or 135o for GOES-West.

Figure 2 shows the EW scan pattern for such an ideal system. The term “ideal” spacecraft implies that instrument optical axis is perfectly aligned with the ideal orbital coordinates (defined by nadir and the equator), and instrument scan mirror control is perfect. The eccentricity, inclination and longitude drift of real spacecraft orbit will obviously deviate from the ideal zero values. Furthermore, instrument attitude will deviate from ideal due to high frequency spacecraft attitude control errors, low frequency spacecraft and instrument thermal distortion, and fixed-pattern and dynamic instrument servo errors.

The INR system should correct for or minimize these various error sources to obtain images that closely approximate the ideal image. The process by which this is accomplished is conceptually simple, but implementation is complicated. This article explains how the GOES I-M, NOP and R series spacecraft support INR, and shows why the later series provide much better performance. GOES navigation determines the location on the Earth of each image pixel. To accomplish this, the orientation of the instrument optical axis with respect to the spacecraft, the orientation of the spacecraft attitude with respect to the Earth, and the position of the spacecraft with respect to the Earth must be known. Then the vector representing the optical ray of each individual detector at a given time can be rotated first to the spacecraft coordinate frame, then to the Earth fixed frame, and finally the intersection of the detector ray with the Earth surface can be computed given the location of the spacecraft. This is inherently a 3-dimensional (3-D) problem, but it is difficult to understand the concepts when working in 3-D.

Figure 3 shows the relationship for a simplified 2-D case where it is assumed that the spacecraft is exactly on the equator and north-south attitude errors are zero. The real spacecraft is located at position C, and the ideal spacecraft is located at B. The true nadir direction at point C is defined by line C-A, but because spacecraft and instrument attitude control is imperfect, instrument nadir deviates slightly from true nadir.

As instrument scan angles are defined with respect to the instrument attitude reference, the optical ray from C to ground location D has a measured scan angle of α. Hence ground intercept point D can be computed given three quantities: angle α, the angle by which the instrument reference deviates from the nadir normal, and the coordinates of point C. Notice that the “ideal” scan angle for point D is β, which is not equal to α. Unfortunately, the “attitude” and “orbit” variables involved in the navigation process are time-varying, and none of them are known perfectly. The time variation occurs because (1) instrument thermal deformations affecting optical pointing vary during the day as the sun angle changes, (2) the spacecraft attitude must rotate with Earth rotation to maintain constant attitude with respect to the Earth, (3) the attitude control system cannot completely null the effects of disturbance torques, and (4) the spacecraft orbital position cannot always be maintained exactly at the desired geosynchronous reference position. Hence image navigation requires accurate knowledge of all these effects as a function of time. The method by which these effects are modeled will be explained later.

Image Registration is the second part of the problem. In the GOES system, the Earth image data broadcast to users — called GOES VARiables or GVAR9 in the I-P system — is normally registered to a “fixed grid” so that each pixel in a given frame always views the same point on the Earth (within the accuracy of the system).

The mechanism by which image registration is accomplished is quite different for GOES I-P versus GOES-R. The user wants the image to appear as if it was obtained by from an ideal spacecraft located on the equator at the reference longitude. Hence the user expects that a given fixed-grid earth point will appear in the image at an angle measured from nadir of the ideal spacecraft to the pixel earth location, i.e., angle β in Figure 3. Given a desired pixel location represented by angle β, the Earth intercept point D can be computed using simple geometry. Then the actual instrument scan angle required to view that Earth point, α, can be computed if the true orbit and attitude parameters are known.

In the GOES I-P systems, registration of the image pixels to the fixed-grid is accomplished in real-time during instrument scanning. That is, optical pointing of the instrument detectors is adjusted during each EW “swath” so that the detector output samples coincide with the ideal fixed-grid pixel locations on the Earth. Since the Imager uses an array of 8 visible detectors arranged in a NS stack and a total of 14 IR detectors arranged around the visible detectors, it is not possible to separately adjust pointing of each detector. Rather, the desired fixed-grid Earth location that corresponds to the array center (optical axis) is calculated at each sample time. Then the true instrument scan angles to the array center point on the Earth is computed taking into account the spacecraft orbit location, spacecraft pointing errors, and internal instrument pointing errors. The difference between true NS and EW scan angles and the ideal fixed-grid NS and EW scan angles for the array center is called the Image Motion Compensation (IMC) signal10.

In Figure 3, the IMC correction is angle α-β. Notice that the IMC correction is a function of the scan angle, and since the scan angles change with time, IMC is time-varying. The IMC signal is computed by the spacecraft during the EW scan and added (with appropriate scaling) to the instrument mirror gimbal commands so that the sampled detector outputs will view the desired fixed-grid Earth pixel locations. The instrument gimbals are commanded to scan exactly EW. With the IMC signal added, the true scan angle time profile curves both in NS and EW directions, as shown in Figure 4 for a sample EW scan swath using realistic orbit and attitude deviations. Notice that when the instrument scans off the Earth at the beginning and end of EW swaths, it is not possible to compute the IMC correction. Hence, the IMC signal is set to zero when scanning off the Earth; to prevent servo discontinuities, the IMC signal is tapered to zero when near the Earth edge. The Imager and Sounder can be individually configured to operate with the IMC signal applied as described above, or with it disabled. The instrument is said to function in the “fixed-gridding” mode when IMC is turned on, and in the “dynamic-gridding” mode when IMC turned off. In dynamic gridding mode, GOES users must compute the Earth location of detector samples using O&A data distributed as part of GVAR.

The maximum IMC orbit correction is about 1500 μrad when the spacecraft is located at the edge of the allocated ±0.5o latitude or longitude station-keeping box. The maximum spacecraft attitude pointing error is about 300 μrad for GOES I-M spacecraft using an Earth sensor as attitude reference, and less than 20 μrad for the stellar-inertial reference GOES NOP spacecraft. Instrument pointing bias errors can be several thousand μrad, but these are handled separately from the IMC correction. The maximum daily deviation in instrument pointing “attitude” error for the Imager or Sounder is about 600 μrad. Hence the maximum total GOES I-M IMC correction is about 2400 μrad, which is 22 times larger than the daytime navigation requirement! This emphasizes the importance of IMC in achieving the desired INR performance.

While real-time onboard IMC correction has been used very successfully on GOES since 1994, the process has undesirable practical and operational limitations. In order for the spacecraft to compute the IMC signal, it must accurately know the true orbit, spacecraft attitude and instrument attitude at all times. The GOES I-P spacecraft have no capability to independently determine orbit and instrument attitude, although spacecraft do determine spacecraft attitude within the accuracies listed in the previous paragraph. Thus the spacecraft is entirely dependent upon orbit and attitude (O&A) information provided from the ground system.

Because the O&A data is used primarily for IMC computation, the particular format used to provide O&A data is referred to as an IMC set. The GOES I-M system was designed so that a minimum of only one IMC set must be uploaded each day, although in practice it is usually necessary to upload about five because Earth sensor unpredictability requires frequent adjustment of attitude biases. GOES NOP was designed to operate using one daily IMC set upload and a second IMC upload with modified orbit parameters that adjust for daily momentum-dumping maneuvers.

To work with only one IMC upload per day, the model defined by the IMC set must accurately predict O&A values for 24 hours. Accurate prediction of a geosynchronous spacecraft orbit is relatively easy if thrusters are not fired, since other perturbing forces are small. Although IMC sets could have used standard 6-element orbital elements (e.g., Kepler, or Cartesian), it would have been necessary for the spacecraft and GOES users to time-integrate the ephemeris in order to calculate orbital position and velocity at times of interest. Instead, the IMC set uses a simpler 24 parameter orbit model that defines daily perturbations from the reference position.

24-hour prediction of instrument internal misalignments is not a trivial problem. It is assumed that instrument misalignments are due to fixed biases that do not change, and time-varying misalignments that are due to thermal deformation. Since the relative sun angle profile for one day is nearly identical to the profile for the next day, it is further assumed that the daily misalignment profile does not change from day-to-day. In truth, thermal deformation effects on pointing can change about 10 μrad between days in which solar eclipses do not occur, and more than 100 μrad during the eclipse season. Eclipse season processing is described later, and the 10 μrad daily non-repeatability is ignored. Hence the daily instrument attitude profile is modeled as a truncated Fourier series with a fundamental period equal to the solar day. In the GOES I-M system the series is truncated at orders less than or equal to 12 for each of five instrument attitude parameters: three Euler angles (roll, pitch, yaw) and two misalignment parameters that represent the effects of several internal misalignments.

The need to predict O&A profiles for 24 hours not only limits achievable accuracy, but also it imposes operational and scheduling restrictions that require significant support and limit system flexibility. Hence driving goals of the GOES-R system included greater automation, reduced operational manpower, reduced system outage time, faster scanning, and improved accuracy11.

Another design factor was the twenty-fold increase in imaging data generated by the ABI compared to the Imager and Sounder. These considerations led to the decision to abandon the GOES I-P onboard IMC concept and instead to perform INR on the ground. That is, the ABI will scan the Earth without attempting to align detector samples with fixed-grid pixels. When the data is received on the ground, the detector samples will be navigated to the fixed-grid space and the “resampled” fixed-grid pixel intensity will be computed by appropriate weighting of adjacent detector samples12,13.

As this INR process is performed in real-time just prior to image distribution, there is no need to predict O&A parameters for more than a few minutes, i.e., computation of O&A parameters can be performed almost continuously using a Kalman filter14. Other improvements of the GOES-R system include a spacecraft capability to determine orbit independently of the ground15, and to provide orbit and spacecraft attitude information to the ABI so that it can operate somewhat autonomously. Also the ABI is thermally more stable than the Imager or Sounder16 and has greater computational capabilities17,18. All these improvements help to greatly improve system accuracy and flexibility and reduce operational and scheduling requirements.

It should be noted that the Japanese MTSAT, European Union Meteosat and NOAA XGOHI geosynchronous weather satellites all implement INR on the ground. The concept is well-tested and should be low risk for GOES-R.

Orbit and Attitude Determination (OAD)
The above discussion was the “conceptually simple” part. We now address the more complicated implementation issues involving computation of the IMC sets. In the GOES I-P system, the Orbit and Attitude Determination (OAD) ground process computes O&A parameters by weighted least-squares fitting of range, star and landmark observation data taken over the previous 24 to 48 hours19,20,21.

The OAD function is executed once per day. Range measurements are obtained by inserting ranging bits in the uploaded GVAR data and measuring the time delay before the same data is received back on the ground. Hence, the range measurement is the two-way transit time between the ground station and the spacecraft, multiplied by the speed of light. A measurement preprocessing function removes known range bias errors, such as equipment delays and atmospheric group delay effects, before data are used in OAD. Notice that the range measurement is only a function of spacecraft orbital position and velocity, not spacecraft or instrument attitude.

Landmark measurements22 are obtained by correlating maps of prominent land/water interfaces, such as islands or peninsulas, with an edge-detecting transform of the received image pixel data. The shift in latitude and longitude required to maximize correlation between the shoreline map and transformed image data is added to the nominal landmark location to generate an observed landmark position. As with range measurements, a preprocessing function converts data from one form to another and applies various corrections, such as telemetry downlink time, detector offset, and IMC compensation. Because the instrument image is used to obtain the measurement, landmark observations are a function of both spacecraft orbit and spacecraft/instrument attitude. Separate landmark observations are obtained for both the visible and at least one IR channel.

Star observations are not obtained while the instruments are scanning the Earth. Several times per hour the instruments briefly stop scanning and dwell at fixed gimbal angles, selected by ground processing, to view given stars. As the spacecraft is always rotating (primarily in pitch) at Earth rotation rate to maintain Earth pointing, detector output will peak as the star crosses the detector field-of-view. The detector number, gimbal angles, and star detection time provide the information necessary to determine the star angles in instrument coordinates and to process the observation. Star preprocessing includes corrections for downlink time, servo error, origin offset, and detector offset and rotation effects. In the stellar inertial GOES NOP system, star observations are only a function of spacecraft/instrument attitude, but because GOES I-M spacecraft use an Earth sensor for attitude reference, star observations are also a weak function of spacecraft orbit.

In addition to converting/correcting observations so that they can be used by OAD, GOES preprocessing also computes the residual (difference) between the corrected observations and predicted observations based on the currently enabled IMC set. These residuals are usually interpreted as a measure of the GOES navigation errors, and landmark residuals in particular are similar (but not identical) to the navigation error seen by a GVAR user. Figures 5 and 6 show typical GOES-13 residuals obtained during a period of post-launch testing23 when the spacecraft experienced daily Earth eclipses at local midnight. If the three observation types could be grouped so that two were a function of only spacecraft orbit parameters, and one type was only a function of attitude parameters (or vice versa), then the OAD function could be performed separately for orbit and attitude. Since GOES landmark observations are a function of both orbit and attitude, it is necessary to simultaneously solve for orbit and attitude parameters using all range, landmark and star observations. The orbit determination portion of GOES OAD is similar to most batch orbit determination software, i.e., it numerically integrates epoch orbit elements to generate ephemeris, and uses weighted least-squares estimation to adjust epoch orbit elements so that residuals between actual observations and computed observations are minimized. OAD must also compute an attitude profile that minimizes observation residuals, where the attitude profile is represented as a truncated Fourier series with 24-hour fundamental. First or second-order polynomial coefficients are optionally included in the attitude model to handle trends. In routine operations, OAD simultaneously estimates 6 epoch orbit elements and 80 to 100 attitude parameters per instrument.

The GOES I-M spacecraft do not have the capability to operate during periods, lasting up to 72 minutes, in which the spacecraft passes through the Earth shadow. These spring and autumn eclipse seasons last about 48 days. GOES I-M spacecraft must stop imaging operations during each eclipse and for some period after. GOES NOP spacecraft do operate during eclipse, but because instrument thermal conditions change so rapidly, the 24-hour fundamental Fourier series attitude model is not accurate for this period. To better match the eclipse attitude profile, GOES NOP switches to a 4-hour fundamental period Fourier model for a 4 hour period around eclipse. A separate OAD fit is computed for this eclipse period, and with adjustments for changing eclipse durations, is used to compute the eclipse period IMC set for the next day.

Post yaw-flip operations are another GOES NOP capability not available in GOES I-M. The I-M series spacecraft were never intended to flip about the yaw axis, but because of a one-directional failure of the GOES-10 solar array drive, that spacecraft was operated in the inverted yaw-flip orientation rather than upright. GOES NOP spacecraft were designed to yaw-flip two times per year so that the Imager and Sounder coolers will always face away from the sun. This reduces instrument temperatures and hence detector thermal noise. However a yaw-flip maneuver changes the daily instrument thermal profile and thermal deformation. OAD solutions computed using the last 24 hour of observations cannot be used to predict the next 24 hours. Instead, a previous attitude profile for the same yaw orientation and time-of-year is used for the first 24 hours after a yaw-flip maneuver.

IMC Set Generation
Because of limited computational capabilities, the GOES I-M spacecraft do not directly use the O&A set computed by OAD. Instead, ground software time-integrates the OAD epoch orbit elements to compute spacecraft ephemeris over the 24-hour IMC prediction interval and a 24-parameter orbit perturbation model is least-squares fit to the ephemeris. OAD attitude coefficients are re-epoched and re-formatted and included with the 24 orbit parameters in the IMC set24. These IMC sets are uploaded to the spacecraft to allow on-board IMC calculations, and are also included in GVAR so that GOES users can navigate image pixels during periods when spacecraft IMC is turned off.

Whenever a maneuver is expected during the 24-hour prediction period covered by an IMC set, two IMC sets are generated: pre-maneuver and post-maneuver. The post-maneuver IMC set uses the same OAD solution as the pre-maneuver, but the post-maneuver ephemeris is computed by including the effects of the planned maneuver velocity change. This happens at least once per day for GOES NOP spacecraft because of the daily momentum-dumping maneuvers required to unload angular momentum generated by solar pressure on the single solar array.

IMC set generation is not the only system complexity caused by the GOES I-P need to predict O&A behavior for 24 hours. Instrument commands for image frames (scan patterns boundaries) and for star sensing must be uploaded in instrument coordinates that the instruments can directly convert to mirror gimbal angles. Separate star angle commands must be computed in advance and uploaded for every star sense during the 24-hour IMC prediction interval. Any change in IMC sets requires that all star sense angles after the IMC set change be re-computed. This re-computation is also necessary for all image frame boundaries whenever the spacecraft is operated with IMC turned off. Normally, IMC is on, so that it should only be necessary to compute frame boundaries once for each frame.

The proposed GOES-R system is much simpler in that there is little need to predict behavior for more than a few minutes into the future. Rather than performing OAD using batch least-squares estimation with a 24 to 48-hour span of observations, the effects of instrument thermal deformation can be computed in near real-time using a Kalman filter that processes star and optionally landmark observations as they become available25. The spacecraft will independently determine its orbit and attitude rates and supply that data to the instruments and to ground processing26,27. Star sense and frame boundary angles can be computed directly by the much more capable ABI with very little input from the ground28. These changes will greatly reduce operator workload and allow for a much more reliable and accurate system.

To perform the weather-monitoring functions for which GOES was designed, the spacecraft must image the Earth from a fixed-point relative to the Earth. Since real spacecraft and instruments cannot exactly maintain the ideal orbit and attitude position desired by users, the GOES system applies pointing corrections so that image pixels appear to have been obtained from an ideal fixed-grid system.

Requirements on INR accuracy have become progressively tighter as the GOES system has continued to evolve. GOES I-P spacecraft implement on-board pointing corrections during scanning so that the detector samples are directly registered to the fixed-grid. While this approach has been used successfully since 1994, the system is complex and requires predicting orbit and attitude behavior for 24 hours in the future. This has undesirable operational and performance limitations.

The next generation of spacecraft and instruments, starting with GOES-R, will perform INR on the ground by resampling detector samples. In this approach it is only necessary to predict O&A behavior for a few minutes into the future, so the system is much more flexible and accurate and requires less operator interaction.

We hope that this article gives readers a better appreciation of the complexity required to generate GOES images displayed on the nightly news.

About the author
Mr. Gibbs is a Senior Systems Analyst at Integral Systems, Inc., and has led the development of the GOES NOP Orbit and Attitude Tracking System (OATS) since 1998. He holds BSEE and MEE degrees from Rensselaer Polytechnic Institute and has 39 years experience in a wide variety of applications including spacecraft modeling, orbit and attitude determination, batch and recursive estimation, process control, optimization, tracking, signal processing, numerical analysis, data analysis, dynamic modeling and simulation, and software development.

1—GOES I-M Data Book, http://rsd.gsfc.nasa.gov/goes/text/goes.databook.html
2—GOES N Data Book, http://goes.gsfc.nasa.gov/text/goes.databookn.html
3—“NOAA GOES-N,O,P — The Next Generation”, NASA/NOAA, http://www.osd.noaa.gov/GOES/GOES-NOP_Brochure.pdf
4—GOES-R Mission Requirements Document, NOAA P417-R-MRD-0070, Ver. 3.2, January 2008, http://www.goes-r.gov/procurement/ground_documents/MRD_V_3_2.pdf
5—GOES-R Series Concept of Operations (CONOPS), February 2008, NOAA/NASA, http://osd.goes.noaa.gov/documents/CONOPS_V2.3.pdf
6—Krimchansky, A., D. Machi, S. Cauffman, M. Davis, “Next-generation Geostationary Operational Environmental Satellite (GOES-R series): a space segment overview”, Proceedings of SPIE -- Volume 5570, 2004, pp. 155-164, http://goes.gsfc.nasa.gov/text/Next_Generation_Geo_2006.pdf
7—Newcomb, H., R. Pirhalla, C. Sayal, C. Carson, B. Gibbs, and P. Wilkin, “First GOES-13 Image Navigation & Registration Tests Confirm Improved Performance”, SpaceOps 2008 Conference, May 2008, AIAA-2008-3464
8—Earth Location User’s Guide (ELUG), NOAA, DRL 504-11, Rev. 2, July 2005, http://www.osd.noaa.gov/gvar/documents/Earth_Location_Users_Guide.pdf
9—GOES Operations Ground Equipment (OGE) Interface Specifications, DRL 504-02 Part 1, Revision 2, Section 3.0, GVAR Transmission Format, http://www.osd.noaa.gov/gvar/documents/G023_504.02_DCN3_Sect_3.pdf (current 10 April 2008)
10—GOES I-M Data Book, op cit.
11—GOES-R Series Concept of Operations (CONOPS), op cit.
12—“GOES-R Series Advanced Baseline Imager Performance and Operational Requirements Document (PORD)”, NOAA, May 2004, http://www.osd.noaa.gov/rpsi/Baseline_PORD_Ver2.pdf
13—Ormiston, J., J. Blume, J. Ring, J. Yoder, “GOES Advanced Baseline Imager – Ground Processing Development System”, P1.7, 5th GOES Users’ Conference, New Orleans, January 2008, http://ams.confex.com/ams/88Annual/techprogram/paper_136059.htm
14—Ellis, K., D. Igli, K. Gounder, P. Griffith, J. Ogle, V. Virgilio, A. Kamel, “GOES Advanced Baseline Imager Image Navigation and Registration”, P1.27, 5th GOES Users’ Conference, New Orleans, January 2008, http://ams.confex.com/ams/88Annual/techprogram/paper_136060.htm
15—GOES-R Mission Requirements Document, op cit.
16—Ellis, K., D. Igli, K. Gounder, P. Griffith, J. Ogle, V. Virgilio, A. Kamel, op cit.
17—Krimchansky, A., D. Machi, S. Cauffman, M. Davis, op cit.
18—GOES-R Mission Requirements Document, op cit.
19—Ong, K. and S. Lutz, “GOES orbit and attitude determination — theory, implementation, and recent results”, Proc. SPIE, v2812, p 652-663, Oct. 1996
20—Kelly, K., J. Hudson, and N. Pinkine, “GOES 8/9 Image Navigation and Registration Operations”, GOES-8 and Beyond, 1996 International Symposium on Optical Science, Engineering, and Instrumentation, SPIE, Denver, Co., August 1996; describes measurements and preprocessing
21—Gibbs, B., J. Carr, D. Uetrecht and C. Sayal, “Analysis of GOES-13 Orbit and Attitude Determination”, SpaceOps 2008 Conference, May 2008, AIAA-2008-3222. 22—Madani, H., J. Carr, and C. Schoeser, “Image Registration using AutoLandmark”, Proceedings of the 2004 IEEE International Geoscience and Remote Sensing Symposium, 20-24 September 2004
23—Carson, C., J. Carr, and C. Sayal, “GOES-13 End-to-End INR Performance Verification and Post Launch Testing”, 5th GOES Users Conference, Annual Meeting of the American Meteorological Society, 20-24 January 2008, http://ams.confex.com/ams/pdfpapers/135921.pdf
24—GOES I-M Data Book, op cit; IMC set.
25—Ormiston, J., J. Blume, J. Ring, J. Yoder, op cit.
26—Ellis, K., D. Igli, K. Gounder, P. Griffith, J. Ogle, V. Virgilio, A. Kamel, op cit.
27—GOES-R Series Concept of Operations (CONOPS), op cit.
28—Ellis, K., K. Gounder, P. Griffith, E. Hoffman, D. Igli, J. Ogle, and V. Virgilio, “The ABI star sensing and star selection”, P1.28, 5th GOES Users’ Conference, New Orleans, January 2008, http://ams.confex.com/ams/88Annual/techprogram/paper_136061.htm

This article does not reflect the perspective of NASA/GSFC GOES-R Project— all information presented was gathered from the public domain by the author.