Data Fusion to mitigate Loop Heat Pipe data dropouts with GOES-17

February 13th, 2019 |

GOES-17 Band 10 (7.34 µm) imagery, 0900-1700 UTC on 13 February 2018 (Click to animate)

(The Experimental GOES-17 Data Fusion link is here).

GOES-17 is now operational as GOES-West at 137.2º W Longitude, as noted here and elsewhere. However, Loop Heat Pipe (LHP) problems persist (as noted here and here and here), with the effect varying seasonally. Because of the Loop Heat Pipe malfunction, the Advanced Baseline Imager warms up, emitting radiation at wavelengths similar to those being sensed from the Earth. This does not happen during the day when the ABI instrument is pointing towards the Earth and the Sun is behind the satellite. During the night, however, the sun illuminates the ABI instrument and warms it. This effect is most noticeable (and detrimental to the imagery) around the solstices. The animation above shows the effects of the Loop Heat Pipe on Band 10, the low-level water vapor imagery. Starting at about 1015 UTC, there is a perceptible warming of the image, globally, shortly after that the image integrity is lost and the data become unuseable. At 1615-1630 UTC, the instrumentation cools enough to produce, again, useable imagery.

The chart below shows how the temperature of the focal plane in the ABI changes throughout the course of the year. The differences between Northern Hemisphere Vernal and Autumnal Equinoxes occur because the Earth is closer to the Sun at the Northern Hemisphere Vernal Equinox. The chart also shows the effects of ‘Eclipse’ — when the satellite moves through the Earth’s shadow. When the ABI is in the Earth’s shadow, it is not being heated by the Sun (and that’s why the GOES-R Series carries batteries to power the satellite at that time). The reduced heating occurs between 26 February and 14 April in the Spring, and between 30 August and 16 October in the Fall. The chart also contains for the longwave infrared bands the temperature at which the heat from the satellite increases the likelihood of bad data.

Warmest Focal Plane Temperature as a function of Year. Also included: the threshold temperatures when the ABI Detection is affected by the warmer Focal Plane. The step in values near both Equinoxes occurs when a Yaw Flip is performed on the satellite (Click to enlarge)

Note in the plot above that Band 13 and Band 14, the clean window and longwave infrared window bands at 10.3 µm and 11.2 µm, respectively, is relatively unaffected by warming. This 10.3 µm animation (spanning the same times as the animation above) shows subtle features that are related to warming because of the faulty Loop Heat Pipe, but overall, data integrity is preserved.

The GOES-17-only fusion solution for mitigating data outages uses GOES-17 ABI Clean Window (10.3 µm) radiances to create missing spectral band radiances. In this method, the last full set of calibrated radiances at time t0 is marked (For GOES-17 at present this is done at 0900 UTC). Then, for subsequent times when LHP issues cause missing data, a so-called k-d tree search (discussed in this paper: Weisz, E., B. A. Baum, and W. P. Menzel, 2017: Construction of high spatial resolution narrowband infrared radiances from satellite-based imager and sounder data fusion. J. Appl. Remote Sens. 11 (3), 036022, doi: 10.1117/1.JRS.11.036022) is performed on time t0 infrared window band 13 measurements to find the five t0 pixels that best match each time t infrared window band 13 pixel; the five k-d tree time t0 matches for each pixel are then averaged for all ABI IR spectral bands to estimate ABI bands at time t for that pixel.

In other words, the fusion method (1) finds the five best matches of time t0 Band 13 measurements for each time t Band 13 measurement and then (2) uses the average of those t0 matches for each missing spectral band to create a fusion estimate at time t. Using time 0900 UTC for t0, when Loop Heat Pipe issues subsequently impact data quality, the Band 13 image at that time is used to predict what the other bands would look like — based on the relationship established for Band 13 measurements between time t and time t0 at 0900 UTC. The fusion method is most challenging for the Band 10 imagery — because it is the least correlated with Band 13 (especially in regions of clear skies). In contrast, a window channel like Band 11 is highly correlated with Band 13 and the Data Fusion product has few artifacts. This animation compares Band 13 and Band 11 — very similar! — and Band 13 and Band 10 — not that similar, except in regions of clouds.

A resultant image is shown below. The left-most image is the unusable GOES-17 Band 10 image. The middle image shows the GOES-17 water vapor image produced via data fusion; the right-most image is the corresponding GOES-16 Band 10 image over the same region (a region that is midway between the two satellite nadirs so that view angle differences are minimized). There are some subtle differences; in particular the Fusion product shows values that are cooler than GOES-16 in some regions, but the qualitative aspects of the match are good.

(The GOES-17 Imagery below pre-dates the 1800 UTC 12 February 2019 time when GOES-17 became operational and should therefore be considered preliminary and non-operational)

GOES-17, GOES-17 Data Fusion, and GOES-16 Low-Level Water Vapor (Band 10, 7.3 µm) Imagery, 1300 – 1445 UTC on 4 February 2019 (Click to play mp4 animation)

Data Fusion Imagery is available at the SSEC Geo Browser (Link).  It is a separate drop-down menu (‘GOES-17 Fusion’) and should be considered experimental.  The animation below shows the transition between non-fusion and fusion data.  Here is a full-disk animation from 13 February (1200 UTC to 1700 UTC) that shows Fusion Data transitioning back to GOES-17 data as the Loop Heat Pipe issues end in the morning.

GOES-17 Low-Level water vapor imagery (7.3 µm) at 1100, 1115 and 1130 UTC. Data Fusion was used to produce the 1130 UTC image (Click to enlarge)

Seattle, Washington as viewed by 4 GOES

February 13th, 2019 |

Visible images, centered at Seattle, from (left to right) GOES-17, GOES-15, GOES-16 and GOES-13 [click to play animation | MP4]

Visible images, centered at Seattle, from (left to right) GOES-17, GOES-15, GOES-16 and GOES-13 [click to play animation | MP4]

The GOES-13 satellite was brought out of storage for annual maintenance activities on 13 February 2019 — allowing for a unique view of the Seattle, Washington area from that satellite as well as GOES-15, GOES-16 (GOES-East) and GOES-17 (GOES-West). After receiving significant snowfall during the previous several days, snowcover was abundant across that region. The brighter-white snow-covered mountain peaks south and southeast of Seattle (especially that of Mount Rainier) were also apparent on visible imagery from all 4 satellites.

Note that visible images from the older GOES-13/GOES-15 are not as bright as those from the newer GOES-16/GOES-17 — performance of visible detectors on the previous generation of satellites degraded over time, while the new GOES-R series benefits from on-orbit calibration of the visible detectors to mitigate this effect.

Using a spare rooftop antenna, staff at the SSEC Data Center were able to ingest and process this data from GOES-13 (in addition to the other 3 GOES satellites). GOES-13 will be placed back into storage on 25 February 2019.

A toggle between larger-scale images using the 5 spectral bands of the GOES-13 Imager are shown below.

GOES-13 Visible (0.63 µm), Shortwave Infrared (3.9 µm), Water Vapor (6.5 µm), Infrared Window (10.7 µm) and Infrared CO2 Absorption (13.3 µm) images at 2015 UTC [click to enlarge]

GOES-13 Visible (0.63 µm), Shortwave Infrared (3.9 µm), Water Vapor (6.5 µm), Infrared Window (10.7 µm) and Infrared CO2 Absorption (13.3 µm) images at 2015 UTC [click to enlarge]

Turbulence over the western United States

February 13th, 2019 |

GOES-17 Low-Level Water Vapor Imagery (7.3 µm) on 13 February 2019, 1932-2012 UTC (Click to animate)

A strong storm system affecting the western third of the United States on 13 February 2019 was responsible for turbulence that caused damage and injuries and forced an ERJ-170 aircraft flying from John Wayne Airport (in Orange County CA) to Seattle WA to make an unscheduled stop in Reno NV (News report). (Flight Aware information on the flight) Three passengers were hospitalized.

The water vapor imagery animation above, annotated so that the departure airport (SNA) and Reno (RNO) are shown in the first frame. During the second run-through of the animation, the approximate location of the turbulence is noted with a red square (Flight Aware data suggests it occurred very near 2030 UTC — Flight Aware times are EST), and the animation slows. Water Vapor imagery suggests that the aircraft encountered a convective element that likely was associated with the Sierra Nevada. A stepped animation between 2027 and 2032 UTC is shown below.

GOES-17 Low-Level Water Vapor Imagery (7.3 µm) on 13 February 2019, 2027 and 2032 UTC (Click to enlarge)