Extended Abstract (464K)7.2
Radar precipitation estimates in mountainous regions: corrections for partial beam blockage and general radar coverage limitations
Dennis A. Miller, NOAA/NWS, Silver Spring, MD; and D. H. Kitzmiller, S. Wu, and R. Setzenfand
Radar based quantitative precipitation estimation (QPE) in mountainous areas is often hampered by terrain induced beam blockages, beam overshooting of precipitation from shallow cloud layers, and range limitations on coverage. We have recently concluded two studies seeking operational solutions to these problems. The first study involved assessing the limitations on methods for correcting for terrain blockages, using both reflectivity-power augmentation in the existing, operational, WSR-88D precipitation-estimation algorithm and an alternative, dual-polarization algorithm that is less sensitive to beam blockage. The second study sought to objectively define a radar data quality map that could be used to determine where radar estimates should be blended with or replaced by estimates from a spatially-continuous source such as numerical prediction model forecasts or satellite estimates. Together, the two studies aim to improve QPE quality from individual radar units, and to create better composites from multiple radars or from radar-multisensor systems that incorporate satellite estimates or numerical prediction model forecasts.
We undertook the first study to investigate the possible application of specific differential phase (Kdp) in zones of partial beam blockage and to compare its effectiveness as a precipitation estimator to that of horizontally-polarized reflectivity (Zh) corrected for beam blockage. The Zh correction is effected by augmenting the returned power by an amount proportional to the degree of blockage. The motivation for this study is that the WSR-88D radar will soon be upgraded from a single polarization to a dual polarization system. In anticipation, a new, Dual-Polar QPE algorithm has been incorporated into the latest-release WSR-88D code (“Build 12”) and will be assessed for national, operational implementation. That algorithm, developed by staff of the NOAA/National Severe Storms Laboratory, utilizes the Kdp field as the basis for precipitation estimates at places and times when a preliminary algorithm had determined the predominant hydrometeor type to be hail or rain mixed with hail. Also, the Kdp moment, which is determined as the first-order spatial derivative of the dual-polar differential phase, is expected to remain effective as a precipitation estimator in the presence of beam blockage.
Our study data were collected by the National Center for Atmospheric Research's S-band, dual-polarization radar unit known as “S Pol”, while that unit was deployed in the vicinity of Boulder, Colorado during June-August 2006. The coverage umbrella of this radar site features large sectors with partial beam blockage to the south and north. For reference and verification, precipitation estimates were taken from the WSR 88D unit near Denver International Airport, which has an unobstructed view of these areas. For purposes of the comparison, S Pol estimates were prepared from the prototype WSR 88D R(Kdp) algorithm and from the standard convective R(Zh) algorithm, the latter with and without correction for beam blockage. The blockage-correction was determined from a digital elevation model and assumptions of standard beam refraction, and was applied to compensate for beam blockage up to 90%.
Our results to date indicate that R(Zh) and R(Kdp) yield estimates of comparable quality in areas with minimal beam blockage. However, as blockage increases towards and above 50%, R(Kdp) is seen to become negatively biased – a result not necessarily consistent with the theoretical behavior of that estimator under such circumstances. Meanwhile, R(Zh) with compensation for beam blockage continues to provide relatively unbiased precipitation estimates, even as blockage increases towards 90%. Overall, R(Zh) corrected for beam blockage yields higher-quality estimates than R(Kdp) in sectors with high-percentage beam blockage.
These findings have some practical implications, since the current operational practice for the WSR 88D precipitation processing system ignores reflectivity data collected wherever beam blockage exceeds 50%. For such locations, reflectivity from the next higher unblocked antenna elevation is used in the precipitation processing. Our results indicate that further investigation is warranted to determine if the blockage limitation should be extended to 90%.
The second study was designed to define a spatially continuous radar QPE quality field. It was focused on the mountainous, northwestern United States during the cool season November-March, where beam blockages and beam overshooting of precipitation lead to large coverage gaps. Current interactive radar processing software within the Advanced Weather Interactive Processing System utilizes binary radar coverage maps, defined subjectively, for each point in the radar umbrella. We wish to define such coverage maps objectively, and with a numerically-continuous quality index which could be used to weight radar QPE proportionally to its quality.
Our basis for assessing radar QPE quality at any one place is the long-term correlation between the radar estimates and another set of estimates not affected by radar visibility, such as ones derived from rain gauge observations. While radar-raingauge correlations could be used to define radar data quality maps in flat areas, terrain beam blockages often cause large azimuthal variations in radar quality in mountains, precluding the possible use of spatially-interpolated radar quality information. However, we found that at individual points, the long-term correlation between 24-h radar QPE and rain gauge reports can be approximated by the correlation between radar estimates and precipitation forecasts from the NCEP North American Mesoscale (NAM) model. Since NAM forecasts are available at all points within the United States, we were able to derive a spatially-continuous map of radar-NAM QPE correlation that closely models known radar coverage problems. This correlation map could be used to determine which, among multiple radars with overlapping coverage, has the best overall estimates at a particular point, or to statistically weight data from radar and other sources. While these results were obtained from observations in the cool season, they could be extended to the warm season, when convective precipitation is prevalent, by replacing the NAM forecasts with satellite estimates.
Session 7, Remote Sensing of Hydrometeorological Observations
Wednesday, 20 January 2010, 8:30 AM-10:00 AM, B304
Previous paper Next paper