News & Updates
The 14th JCSDA Workshop on Satellite Data Assimilation
JCSDA News and Newsletters
Impact of a Degraded Polar-orbiting Satellite Observing System on NOAA Global NWP
From the Fall 2015 issue of the JCSDA Quarterly
The current global environmental satellite observing system consists of a complex arrangement of geostationary and low Earth-orbiting platforms, providing a multitude of space-borne sensors capable of remotely measuring quantities of the Earth's atmosphere and surface across the visible, infrared, and microwave electromagnetic spectra. These observations consist of observations from passive and active microwave (PMW/AMW), along with narrow-band and hyperspectral infrared (IR) sensors, configured in three primary orbits for optimal global coverage: the early morning (early-AM), mid-morning (mid-AM), and afternoon (PM), labeled according to the platforms' equatorial crossing times.
Coverage in each primary orbit could be considered quasi-redundant, due to the presence of multiple platforms currently providing operational observations. The coverage is provided in the early-AM by the Defense Meteorological Satellite Program (DMSP) platforms of the U.S. Department of Defense, in the mid-AM by the Exploitation of Meteorological Satellites (EUMETSAT) MetOp series, and in the PM by the U.S. National Oceanic and Atmospheric Administration (NOAA) Polar-orbiting Operational Environmental Satellite (POES) program, which is in the process of transitioning to the next-generation Joint Polar Satellite System (JPSS) program beginning with the Suomi-National Polar-orbiting Partnership (SNPP) satellite, in partnership with the U.S. National Aeronautic and Space Administration (NASA).
In the current polar-orbiting satellite constellation, the DMSP F17, F18, and F19 comprise the early-AM platforms; MetOp-A and B provide the mid-AM coverage, and the NOAA-15, 18, 19, along with the SNPP and NASAAqua satellites, provide PM coverage. However, since their launch, the NOAA-15 and 18 satellites have drifted closer to an early-AM orbit.
The continued health of many of the PMW/AMW and IR sensors on these platforms, well beyond their design life, has led to robust and quasi-redundant coverage as it relates to observations assimilated in Numerical Weather Prediction (NWP) data assimilation (DA) systems. For the operational NOAA Global Data Assimilation System/Global Forecast Model (GDAS/GFS), this includes all observations +/-3 hours centered on the analysis cycle time.
As satellite programs transition to the next generation of sensors and launch schedules change, however, there is a risk of losing the quasi-redundant coverage or even all observations in one of the primary orbits. This shift and other factors could lead to only one platform in each primary orbit for extended periods of time, and any delays in follow-on missions could result in no observations being provided by one of the primary orbits.
Specifically, for the latter scenario, there currently exists a risk that the JPSS-1 satellite may not launch before the failure of all other current POES, SNPP, and Aqua satellites. This so-called JPSS data gap could have an impact on NWP forecast skill. In support of NOAA data gap mitigation activities, the JCSDA performed a number of Observing System Experiments (OSEs) over the summer 2014 season to 1) assess the impact on NWP forecast skill from a series of evolved global satellite constellations, and 2) establish a baseline of forecast skill to gauge other mitigation activities designed to increase forecast skill. For these OSEs, a control run (cntrl) was performed using all conventional and satellite observations available to the GDAS/GFS.
A series of data-denial experiments was then executed, including: a configuration where all quasi-redundant polar-orbiting platforms were removed, leaving only one satellite in each primary orbit (3polar); a configuration like the 3polar experiment but removing further only the PM coverage provided by SNPP, leaving only observations in the early- and mid-AM (2polar); and the 3polar experiment but altering the Global Positioning System Radio Occultation (GPSRO) observations to assess the impact of the COSMIC-2 (3pgps).
Specifically, the future COSMIC-2 mission will provide dense GPRSO coverage in the tropical latitudes (to +/-24° latitude) with the phase 1 launch in 2016, but the COSMIC-2 phase 2 which will provide GPSRO observations from a high-inclination orbit is uncertain. Therefore, the 3pgps experiment assesses the impact if only phase 1 is launched.
Table 1 summarizes the satellite constellation configuration for each experiment.
Table 1. Satellite data assimilated in the cntrl (current operational during the experiment period), 3polar, 2polar, and 3pgps experiments (green), and denied (red).
To execute the OSEs, we use a version of the GDAS/GFS almost identical to the current system implemented into NOAA operations in January 2015, including the full model and assimilation (Hybrid 3DVAR/EnKF) resolutions at T1534/574. The experiments were run for May 15-August 7, 2014, and the impact assessment period is from May 25-August 7, 2014, assessing only the GFS 00Z forecast cycle.
Figure 1 shows both the Northern Hemisphere (NH) and Southern Hemisphere (SH) 500 mb Height Anomaly Correlation (AC) as a function of forecast time, using the cntrl analysis for verification. The bottom panels of the plots show the difference with respect to the cntrl forecast, and the points outside of the error boxes are beyond the 95 percent significance level. The figure shows that in both hemispheres, removal of the quasi-redundant satellite observations (3polar) has a negative impact on the Height forecast, though it is only slightly significant at short lead times in the NH. Removal of the GPSRO on top of the redundancy (3pgps) further degrades the Height forecast. The degradation is significant at all lead times in the NH, but only slightly significant at short lead times in the SH. However, removal of the PM polar data on top of the redundancy (2polar) shows a significant impact at all lead times in both hemispheres. It is interesting to note that although the NH has more conventional observations assimilated, the forecast impacts for the 3polar and 3pgps experiments are more negative than for the SH, while impacts for the 2polar between hemispheres are similar.
Figure 1.500 mb Geopotential Height Anomaly Correlation as a function of forecast hour for Northern Hemisphere (left) and Southern Hemisphere (right). The differences with respect to the Control analysis are shown in the lower panels. Points outside the boxes are beyond the 95% significance level. Verification is performed against the Control analysis.
Figure 2 shows the Tropical Wind Vector RMSE as a function of forecast hour for 200 mb and 850 mb, verified against ECMWF analysis. The bottom panels show the RMSE difference with respect to the cntrl forecast. The degradation in wind forecasts by removing quasi-redundant polar information (3polar) is significant at both levels and at all lead times, except at 200 mb beyond day 4. The impact from further removal of GPSRO (3pgps) does not further degrade the forecast except at 200 mb beyond day 4. This is expected since only the GPSRO observations poleward of +/- 24° are reduced in the experiment. For exclusion of PM polar data on top of the 3polar configuration (2polar), there is a significant and much larger negative impact on wind forecasts at both levels.
Figure 2. Tropical wind vector RMSE as a function of forecast hour for 200 mb (left) and 850 mb (right). Bottom panels show the difference in RMSE with respect to the Control analysis. Points outside the boxes are beyond the 95% significance level. Verification is performed against ECMWF analysis.
Verification for a number of parameters was performed against both the control run and ECMWF analyses, and yielded similar results to those illustrated above. The results show that even removal of information that has been labeled as quasi-redundant in the global polar-orbiting satellite observing system has a negative impact on forecast skill. The differing orbital characteristics, especially those due to drift, has resulted in unique observations being provided by each platform in the primary orbits. Further alteration of the 3polar configuration, including removal of extratropical GPSRO or PM polar data, has further negative impact on forecast skill, with that from the 2polar configuration being more global and significant.
A journal publication with more comprehensive assessments, including impact on hurricane track error and overall forecast accuracy, is currently in progress.
Kevin Garrett, Sid Boukabara, and Krishna Kumar (JCSDA)