News & Updates
JCSDA News and Newsletters
Preparing the NCEP GDAS/GFS for Atmospheric Motion Vectors Derived from the Future GOES-R Advanced Baseline Imager (ABI)
From the September 2013 issue of the JCSDA Quarterly
A new Atmospheric Motion Vector (AMV) nested tracking algorithm has been
developed for the Advanced Baseline Imager (ABI) to be flown on NOAA's
future GOES-R satellite [Bresky et al., 2012]. This algorithm is very
different from the AMV algorithm used operationally at NOAA/NESDIS today.
The new AMV algorithm was designed to capture the dominant motion in each
target scene from a family of local motion vectors derived for each target
scene. Capturing this dominant motion is achieved through use of a
two-dimensional clustering algorithm that segregates local displacements
into clusters. The dominant motion is taken to be the average of the local
displacements of points belonging to the largest cluster. This approach
prevents excessive averaging of motion that may be occurring at multiple
levels or at different scales which may lead to a slow speed bias and
a poor quality AMV. A representative height is assigned to the dominant
motion vector through exclusive use of cloud heights
[Heidinger & Pavolonis, 2009 and Heidinger et al., 2010] from pixels
belonging to the largest cluster. This algorithm has been demonstrated
to significantly reduce the slow speed bias associated with winds at upper
levels, something that is commonly observed in AMVs derived from satellite
imagery. Given the significant improvement in quality we have observed
with AMVs derived from this new AMV algorithm, we are hopeful that future
GOES-R AMVs will positively impact the accuracy of NCEP GFS forecasts.
Figure 1. Density plot of normalized AMV speed departure from GFS background and the nested tracking parameter PCT1 for cirrus clusters in June 2012. The black line at each PCT1 bin shows the mean speed departure value. AMV number in the x/y bin range from 0 to 5000.
To estimate the impact of the addition of GOES-R AMVs to the observing system on the GFS analysis state and forecast skill, an experiment suite was constructed to include two seasons for both the experiment which includes GOES-R AMV proxy data and the control which does not use AMVs from SEVIRI. The pre-implementation version of the Hybrid Ensemble Kalman Filter GDAS/GFS was selected for this study. Initial results from the first season, May-July 2012, revealed an observation minus analysis wind vector difference RMSE which is not ideal for the ABI Channel 14 infrared AMVs. The large RMSE for the GOES-R AMVs indicates the quality control is too lenient and/or the specified observation error is too large. Tuning experiments, which varied the AMV observation error within the GDAS/GFS, showed a positive response by the wind vector difference RMSE for all 4 AMV types when the error was reduced. Repeating the first season run with the AMV error set at 75% of current GOES AMV observation error has lowered the vector difference RMSE (Figure 2). Also included in the new summer season run is the application of a log normal vector difference threshold to replace the current wind component departure check. Analysis of the AMVs impact is underway and initial results indicate the GOES-R AMV data are behaving as expected in the NCEP GDAS/GFS.
Figure 2. Vertical profiles of speed bias (GFS-AMV) and wind vector difference RMSE for IR AMVs with respect to the GFS analysis for June 2012. Initial QC represents results using the current GOES AMV observation error. Final QC represents results with a 25% reduction in observation error.
(Sharon Nebuda and Dave Santek, CIMSS; Jim Jung, CIMSS/JCSDA; Jaime Daniels, NOAA/STAR; Wayne Bresky, IMSG)