PERFORMANCE OF THE TROPICAL Z/R RELATIONSHIP DURING THE OCTOBER 8-11, 1997 SOUTH TEXAS FLOOD EVENT
Steve Pfaff and Chris Jacobson
NWSO Corpus Christi
During the four day period of October 8-11, 1997, parts of the Coastal Bend region of South Texas experienced the most widespread and severe flooding event since Hurricane Beulah made landfall near Port Mansfield on September 20, 1967. The highest rainfall totals were across the Corpus Christi metropolitan area where many areas had rainfall ranging from 15 to 20 inches. Unfortunately this excessive rainfall caused two deaths, well over $5 million in damage to Corpus Christi and surrounding areas alone and flood damage to hundreds of homes and hundreds of cars. Elsewhere across the NWSO Corpus Christi county warning area, rainfall totals of 10 to 15 inches were common, resulting in numerous road closures and significant property damage.
The WSR-88D radar (KCRP), commissioned at NWSO Corpus Christi in September 1996, was a critical component in the office's performance during this very dangerous and costly flood event. Radar rainfall estimates proved crucial in providing the public with timely and accurate flood and flash flood warnings and statements. Based on a highly tropical environment (common to South Texas during the summer and early fall), forecasters at NWSO Corpus Christi decided to utilize the tropical Z/R relationship (Z=250R1.2) during this rain event. This adaptation had proved to enhance the rainfall estimation under similar circumstances (Wood 1997) in place of the default Z/R (Z=300R1.4) relationship normally in operation. The decision to adapt the Z/R parameter proved to be very advantageous in the assessment and verification of this tropical rainfall event.
The purpose of this paper is to demonstrate that, under certain atmospheric conditions, the local adaptation of Z/R relationships can and should be utilized to better exploit WSR-88D capabilities. The advantage to radar operators will be enhanced WSR-88D derived precipitation products which are critical in hydrologic warning decision processes.
3. Synoptic and Mesoscale Discussion
The prevailing synoptic pattern indicated a large mid-level ridge anchored over the southeast United States. A rather deep vortex was located off the northwest U.S. coast with an associated trough over western states. Outflow from Tropical Cyclone Pauline located over the eastern North Pacific was spreading across Mexico, enhancing a Tropical Upper Tropospheric Trough or TUTT circulation over the western Gulf of Mexico Fig. 1. Water vapor imagery revealed deep moisture advecting into South Texas from across central Mexico as a result of these features. Surface charts revealed the development and establishment of an inverted coastal trough Fig. 2 on October 9 which persisted across the area until a strong cold front swept across the region on October 13. The trough acted as the primary focus for the deep tropical moisture which channeled widespread convection into the Texas Coastal Plains near Corpus Christi.
Initial gridded data from the Eta model runs from October 8 at 1200 UTC through October 12 at 0000 UTC depicted broad difluence aloft, with increasing upward vertical velocities and a significant amount of low level speed convergence Fig. 3 through the period of this rainfall event. Also, a distinct theta-e ridge and theta-e advection was noted along the lower and middle Texas Coast throughout the period. Moisture convergence from the strong east and southeast low level flow off the Gulf of Mexico coupled with the surface trough appeared to be the primary lifting mechanisms.
4. Sounding Data
Analyses of Corpus Christi radiosonde observations showed deep tropical moisture reflective of very high precipitable water values and moderate instability present through the entire event. Typical precipitable water values ranged from 2.20 inches to 2.40 inches with a peak of 2.62 inches at 1200 UTC October 11.
Convective parameters revealed a moderately unstable airmass. Lifted indices ranged from -3 to -6C; the 850mb based K index ranged from 35 to 39; Convective Available Potential Energy (CAPE) ranged from 1275 J/Kg to 2635 J/Kg. There were no reports of hail at anytime during this event, which would confirm the high wet-bulb zero and freezing level heights. Very weak cap strengths persisted throughout the event along with very low lapse rates averaging around 5.0 degrees Celsius/kilometer. This is typical of a highly tropical air-mass.
5. Data and Methodology
Although the allocation of a single Z/R relationship cannot thoroughly represent specific situations due to significant fluctuations in drop size distribution (Hunter 1996) it was hoped that local radar adaptations could enhance the radar rainfall estimation products. Due to curvature and range effects on a radar beam, only rain gauges within 90 nm will be considered for Gauge Rate (GR)/Radar Rate (RR) (Klazura, et al. 1995) comparisons in this study. Other potential radar estimation problems that may adversely affect rainfall rates include calibration, anomalous propagation, and attenuation (Hunter 1996). Hunter also reveals that rain gauge measuring inaccuracies may occur due to strong wind as well as gauge locations.
Forty-seven rain gauges were used to analyze the performance of the tropical Z/R relationship for this event. The gauge net was comprised of local government, state extension agency, cooperative observers, and NWS employee rain gauges. Since the KCRP WSR-88D employed the tropical Z/R relationship during this event, Archive II base data was used to obtain sets of standard and tropical Z/R storm total precipitation products. The KCRP WSR-88D Archive II data was processed using a set of radar data processing programs (Jendrowski 1997). Unfortunately, a failure in Archive II recording occurred as a result of changing power sources at the Radar Data Acquisition (RDA) system. This failure and associated loss of data resulted in the generation of two separate time sets of Z/R comparisons instead of one set of continuous data for the entire event.
The rain gauge to tropical Z/R derived storm total precipitation ratios yielded several very positive results about the performance of the tropical Z/R during the event. Overall, derived storm total precipitation products yielded a gauge/radar estimation ratio of 1.46 for all locations within 90 nm of the RDA. Surprisingly, the radar performed best within 15 nm of the RDA with a ratio of 1.18. This contradicts the hypothesis that underestimations occur less than 30 nm from the RDA because of beam elevations employed by the precipitation processing subsystem (Hunter 1996). Improved rainfall radar estimation near the radar may have been due to the radar's ability to capture the coalescence process associated with tropical convection Fig. 4. Overall, radar rainfall estimates were poor across western parts of the study area as large overestimates and underestimates occurred. This poor performance may be due to a result of tropical air-mass modification with an ambiently drier air-mass entrenched across northeast Mexico and the Texas Rio Grande Plains region.
Very high underestimations are found when comparing the tropical Z/R storm total precipitation products to the standard Z/R products. This is seen nicely in figures 5a and 5b, near Lake Corpus Christi (northwest of the RDA) the tropical Z/R generated values of 6 to 10 inches while the standard Z/R yielded 4 to 6 inches. Rain gauges in the vicinity of this precipitation maximum coincided more with the tropical Z/R values. This result can also be viewed in figures 6a and 6b.
Although radar rainfall algorithm limitations and sampling errors may adversely affect precipitation derived products, it is hoped that the local manipulation of adaptable parameters such as the Z/R parameter may enhance the performance of the WSR-88D. This enhanced capability will provide a radar operator with a better correlation between "ground truth" and radar estimation products, leading to improvements in short fuse warnings and warning decision processes.
Hunter, S.M., 1996: WSR-88D radar rainfall estimation: capabilities, limitations and potential improvements. National Weather Digest, June 1996, Vol. 20, 26-38.
Jendrowski, P., 1997: Programs to create WSR-88D products. NWSFO Forecast Office, Honolulu HI.
Klazura, G.E. and Kelly, 1995: A comparison of high resolution rainfall accumulation estimates from the WSR-88D precipitation algorithm with rain gage data. Preprints, 27th Conf. on Radar Met., Amer. Meteor. Soc., Boston, 31-34.
Wood, L.T., 1997: Using the tropical Z-R relationship to improve precipitation estimates during a heavy rain event in southeast Texas. Preprints, 28th Conf. on Radar Met., Amer. Meteor. Soc., Boston, 208-209.