FY1999 Progress – FY2000 Plans


1999 Annual Report –
FY 1999 Progress/FY 2000 Plans

Peter J. Lamb, Director
Randy A. Peppler, Associate Director

The University of Oklahoma (OU) and NOAA established the Cooperative Institute for Mesoscale Meteorological Studies (CIMMS) in 1978. Through mid-1995, CIMMS promoted cooperation and collaboration on problems of mutual interest among research scientists in the NOAA Environmental Research Laboratories (ERL) National Severe Storms Laboratory (NSSL), and faculty, postdoctoral scientists, and students in the School of Meteorology and other academic departments at OU.

The Memorandum of Agreement (MOA) between OU and NOAA that established CIMMS was updated in 1995 to include the National Weather Service (NWS). This expanded the formal OU/NOAA collaboration to the Operational Support Facility (OSF) for the WSR-88D (NEXRAD) Program, the NCEP (National Centers for Environmental Prediction) Storm Prediction Center (SPC), and the NWS Forecast Office, all located on the OU campus in Norman, Oklahoma.

Through CIMMS, OU faculty and NOAA ERL/NWS scientists collaborate on research supported by NOAA programs and laboratories as well as other agencies such as the National Science Foundation (NSF), the U.S. Department of Energy (DOE), the Federal Aviation Administration (FAA), and the National Aeronautics and Space Administration (NASA).

The present 5-year cooperative agreement between OU and NOAA for CIMMS funding took effect on July 1, 1996. Under this agreement, CIMMS concentrates its efforts and resources on the following five principal research themes, the fifth of which is new under the current plan: (1) basic convective and mesoscale research, (2) forecast improvements, (3) climate effects of/controls on mesoscale processes, (4) socioeconomic impacts of mesoscale weather systems and regional-scale climate variations, and (5) Doppler weather radar research and development.

This document describes research progress made by CIMMS scientists assigned at OU and at our cooperative NOAA units during fiscal year 1999 (July 1, 1998 through June 30, 1999) and presents research plans for fiscal year 2000 (July 1, 1999 through June 30, 2000), and as such represents the third annual report written under the present agreement. The NOAA units where much of the research documented here was performed are indicated as throughout.


Progress – FY 99

Parameterization of Cloud Microphysics and Radiation

A new full-moment cloud/drizzle scheme for stratiform clouds is currently under development at CIMMS. The important difference between the new scheme and the commonly used Kessler-type parameterizations is that the new scheme does not depend on the threshold radius dividing the cloud and rain water. As a result it is more accurate and easier to generalize for all cloud types. Its use of empirically observed physical prognostic variables is an additional advantage. The parameters of the scheme have been determined using regression analysis of data from case studies of stratiform clouds observed during several field programs. The parameterization has been tested in a one-dimensional model, its refinement and testing in a more realistic 3D model is planned for the next year.

We also completed the development of a parameterization of the cloud drop effective radius designed and tested for drizzling marine stratocumulus. The parameterization is based on the parameters that are predicted in the mesoscale model and represents a generalization of an empirical parameterization valid only for the case on non-precipitating clouds. The new effective radius parameterization will result in more accurate calculations of radiative parameters in regional and global circulation models.

Implementation of the CIMMS Cloud Parameterization into NWP Regional Models

The previously developed CIMMS cloud physics parameterization has been implemented into the U.S. Navy Coupled Ocean-Atmosphere Mesoscale Prediction System (COAMPS). This project is supported by a grant to OU from the Office of Naval Research. The CIMMS parameterization accounts for the difference in the ambient CCN concentration between clean and polluted environments and provides more accurate calculations of autoconversion and accretion rates for stratiform clouds. A mesoscale simulation of the cloud/fog system over the central California coastal region has been performed. The preliminary analysis showed that COAMPS with the new drizzle scheme captures many important features visible in fine scale LES simulations. The main goal of the simulation is to improve the cloud/drizzle/fog forecast by investigating the feedbacks between mesoscale conditions and cloud/drizzle parameters.

Development of a Cloud Microphysics Retrieval Algorithm

The performance of various cloud liquid water content (LWC) retrieval algorithms was evaluated at CIMMS using data from a three-dimensional large-eddy simulation model with explicit size-resolving microphysics. A new retrieval algorithm has been developed based on measurements of radar reflectivity and independently measured liquid water path. The new algorithm proved to significantly reduce the errors in the retrieved profile of LWC. Optimal algorithm parameters have been determined and tested for drizzling marine stratocumulus cloud conditions. A case study of stratiform cloud layer observed over ARM Southern Great Plains Cloud and Radiation Testbed site showed that the algorithm performs equally well over the continent.

The Effects of Horizontal Radiative Transport on Cloud Absorption

The CIMMS LES model of stratocumulus clouds with explicit liquid phase microphysics and a 3D Monte Carlo model were linked to study the spectral dependence of the horizontal transport and its effect on absorption estimates. It was shown that:

horizontal transport depends markedly on the wavelength. As wavelength changes by roughly 1 m m, the variance of the horizontal transport may change by 15-20%,
the spectral dependence of the horizontal transport is primarily determined by spectral dependency of water drops and water vapor absorption parameters, and
the neglect of spectral dependency of the horizontal transport may lead to considerable (up to 100%) errors in absorption estimates.

Electrical Structure of Storms and Storm Systems

During the summer and continuing into early fall of 1998, storm intercept operations in conjunction with the MEaPRS field program were conducted by CIMMS, NSSL, and OU School of Meteorology personnel. On seven days, 12 balloon-borne electric-field meters, three balloon-borne particle charge detectors, five electric-field-change antennas, and one x-ray detector were launched. The principal tasks involved the electric-field meters and particle charge devices, though sometimes these were launched with a field-change antenna on the same balloon train. Data reduction and analysis, as well as modeling, by graduate students, co-principal investigators and collaborators, and comparisons with National Lightning Detection Network (NLDN) data, Lightning Mapping Array (LMA) data (New Mexico Tech), and radar data, are all progressing.

The results of the preliminary analyses of the data obtained during summer and fall of 1998 are very promising. Multiple soundings of electric field were obtained through various regions of a mesoscale convective system, and individual soundings of electric field were obtained in storms that produced frequent positive ground flashes. For example, we think that the observed variation in the distribution of charged regions from one portion of the storm to another will, when examined in light of radar and NLDN data, lead to insights about the mechanisms for production of positive cloud-to-ground lightning in various circumstances. The modeling work will provide the means to pull the observations together in order to focus on the essential processes that lead to positive cloud-to-ground lightning.

A mesoscale convective system on 25 May 1998 produced a series of eight bow-shaped radar echoes that moved rapidly during the night from 0000 UTC to 1200 UTC. Two electric field profiles were obtained in the convective regions of two successive convective bow echoes. The magnitude of the maximum measured electric field, E, was approximately 100 kV per meter. The majority of the ground flashes in the convective region were positive. A third sounding was made in the transition zone about 100-km north-northeast of the soundings in the convective region. There have been very few such transition-zone soundings reported in the literature. This one showed a peak electric field of about 100 kV per meter and five separate charged regions, inferred from the vertical profile of E. The lowest charged region was positive. Another profile was obtained toward the back of the stratiform precipitation region about 55 km north-northeast and downwind of the transition zone launch site. In that sounding, the peak E magnitude was about 25 kV per meter, and there were three alternating charge regions, with negative charge lowest. The final sounding was at the back edge of the stratiform region. The electric field was less than 5 kV per meter along the flight path. The maximum reflectivity in the stratiform rain region was 35-40 dBZ, about typical for mid-latitude mesoscale convective systems. We are continuing to analyze the electric field and inferred charge regions from the soundings in comparison with horizontal winds and reflectivities derived from Doppler radar data.

Preliminary results of the analysis of electric-field changes observed at altitudes of 12 km to 15 km MSL suggest that the observed field changes could be related to ground flashes in at least two cases positive ground flashes, but uncertainties in the timing and time resolution make it difficult to determine unambiguously what discharge process is directly responsible for the observed field changes. Two papers have been accepted for publication.

An analysis of variations in cloud-to-ground lightning flash rates versus variations in radar-derived storm parameters was completed for three days during MEaPRS in which storms produced many positive cloud-to-ground flashes. The goal was to try to quantify the association between positive ground flash occurrence and large hail that had been reported previously. In isolated storms during MEaPRS, trends in large hail production inferred from radar appeared similar to trends in positive ground flash rates, but this was not true of mesoscale convective systems that also produced many positive ground flashes. However, too few storms have been analyzed to determine whether this dependence on storm type is generally valid.

Initial analysis of the soundings and lightning data acquired from some severe storms that produced frequent positive ground flashes suggests that they had a different vertical sequence of polarity in the charge distribution than has been observed previously in severe storms that have no positive ground flashes. However, data from some severe storms that produced positive ground flashes could be interpreted to support a completely different hypothesis: that upper positive charge extends horizontally beyond lower negative charge. It may be that positive ground flashes can be produced by multiple, fundamentally different charge distributions. These inferences will be tested by further data analysis and modeling.

Electrification parameterizations developed by Ziegler, Straka, and MacGorman have been studied in two models: the ARPS model provided by the CAPS and a model by Straka. Mansell has developed additional parameterizations of electrification mechanisms and lightning. Both models give similar results for electrification in small thunderstorms and through the first few flashes of supercell storms. However, they differ substantially at later times in the simulations of supercell storms and other large thunderstorms, because the more sophisticated ice microphysics of the Straka model produces longer-lasting electrification and more realistic maturing and dissipation of large storms. We have shown that the latter model can produce supercell structure and electrification similar to that produced by a kinematic model of the Binger, OK supercell storm studied by Ziegler and MacGorman (1994). Furthermore, tests of the lightning parameterization developed by MacGorman et al. (1999), with a few new features by Mansell, have shown that it can produce both cloud and cloud-to-ground flashes that appear similar in many respects to actual mapped flashes.

Editing and analysis of airborne pseudo dual-Doppler radar data for three storm days during MEaPRS has been finished at NSSL.  In addition, analysis of airborne dual PRF Doppler data from MEaPRS have been used to evaluate the Doppler radar upgrades on the NOAA P-3 aircraft.

VORTEX-Related Studies

Analysis work was completed at NSSL on two VORTEX tornadic storms (Newcastle, TX,  May 28, 1994; Dimmitt, TX, June 2, 1995).  In the latter case, a comprehensive multi-platform study was conducted of the tornadogenesis phase and three phases of the tornado: intensifying, transition, and weakening.  It has been documented that the tornado formed as a descending cyclonic/anticyclonic vortex pair, straddling the rear-flank downdraft, and interacted with a near-ground vortex along the rear-flank gust front produced through tilting and stretching of extremely helicity-rich air.  This established a several-kilometer diameter vortex from the ground through the lower levels of the storm.  Convergence and stretching led to tornado formation and intensification.  The tornado cyclone gradually became engulfed in downdrafts that transported air downward containing smaller angular momentum.  The demise of the tornado was associated with the entire tornado cyclone becoming a downdraft.

Analysis work has also been performed on the non-tornadic supercell that occurred during VORTEX on June 8, 1995.

Using data obtained during VORTEX and several subsequent focused field efforts, we are also examining nearly 30 tornadic and non-tornadic mesocyclones for differences in mobile-mesonet observed surface conditions.  We are specifically looking for differences in the coldness of the rear-flank downdraft outflow, as measured through buoyancy.

SouthWest Area Monsoon Project (SWAMP)

The NSSL Western Intermountain Storm & Hydrometeorology (WISH) group has continued to study the synoptic patterns that bring severe summer thunderstorms to south and central Arizona. Findings that were published during this period showed that a surprisingly large number of days during the summer “monsoon” at Phoenix are actually quite dry; specifically, of the 70% of the summer days that are moist, only about 50% have thunderstorms over low desert areas. Also, routine sounding data taken at Tucson do not discriminate between storm and no storm days over the central deserts, but lightning data indicate that storms are much more widespread over the entire state (in spite of the lack of signal in the Tucson data) on low desert storm days.

A microburst storm that occurred near Phoenix has been studied in depth and a paper has been provided to the OSF for their training website. It discusses the structures and likely microphysics of this event. The case illustrates clearly that the collapse of the elevated reflectivity core (presumed due to small hail) lags substantially behind in time the occurrence of the microburst at the surface. This has important implications for methodologies to warn for these events. This study is being finalized and submitted for formal publication.

Quantitative Precipitation Estimation Using Multiple Sensors

The NSSL WISH group has advised a CIMMS graduate student on a project that has led to a prototype precipitation estimation scheme incorporating both WSR-88D PPS data and GOES satellite data to improve rain estimations over regions of complex terrain. The technique also has great potential for application over flatlands. This effort has demonstrated that the combined data approach can mitigate the bright band contamination that makes the operational precipitation estimates nearly useless during winter storms.

Three -Dimensional Simulation of a Composite, Derecho-Producing Convective System

During the late spring and summer months, a type of widespread convective windstorm (a derecho) is an occasional occurrence east of the Rocky Mountains. This study at NSSL focused on the progressive derecho, which usually consists of a single, large mesoscale convective system in a relatively benign synoptic-scale environment. Given the current inability of operational numerical models and observing systems to consistently resolve weakly forced convective events and our lack of a dynamical understanding, these occurrences remain a significant forecasting and warning problem.

Sounding data from twelve derechos that occurred in weakly forced large-scale environments were composited and objectively analyzed, using the sum of a lowpass and a bandpass analysis, in an attempt to include important large-scale features in the pre-convective environment. This analysis, which agrees favorably with prior observational studies of progressive derecho environments, was used to initialize the Pennsylvania State University/National Center for Atmospheric Research Mesoscale Model Version V (MM5). Major differences in the initial conditions from past simulations include a horizontally inhomogeneous environment, unidirectional shear in upper-levels with weaker shear in low-levels, and a relatively dry mid-troposphere.

The model develops an explicitly resolved, derecho-producing convective system, that resembles observations on many scales. Past numerical simulations suggest that the strength and longevity of squall lines results from a balance between the vorticity associated with the environmental low-level shear and the baroclinic generation of vorticity at the leading edge of the cold pool. In contrast, it is found that the production of a deep cold pool and the maintenance of upper-level shear (and associated critical layer) through convective feedbacks are important to the strength and longevity of this simulated derecho. The cold pool, acting as an effective barrier to the flow, provides deep convergence throughout the lower troposphere and the critical layer maintains the updrafts at a favorable location above the gust front, as long as there is significant CAPE within the inflow layer. A strong and mobile meso-high associated with the deep cold pool is the main mechanism that maintains the strong winds over large distances. These results suggest that internal parameters are more important to the evolution of the simulated convective system than its interaction with the initial low-level wind profile, which may partially explain the difficulty in forecasting derecho events.

Conditional Symmetric Instability

A manuscript on conditional symmetric instability based on research at NSSL has been accepted by Monthly Weather Review for publication.  Further research has involved investigation of the synoptic and mesoscale aspects of a heavy snowband in northern Oklahoma and examination of the structure and evolution of a damaging squall line over the eastern U.S.

Frontal Structure over the Western United States

Several manuscripts have been written and submitted by scientists at NSSL examining the structure and evolution of fronts and cyclones over the western United States.

The Formation and Climatological Distribution of Tornadoes within Quasi-Linear Convective Systems

This NSSL project pertains to the formation and climatological distribution of tornadoes within quasi-linear convective systems (QLCSs). The horizontal extent of viable tornado-breeding sites is an order of magnitude larger in QLCSs than in individual supercells. QLCS tornadoes can be strong and produce extensive damage despite “conventional wisdom” that suggests otherwise. Also, QLCS tornadogenesis appears to occur, on average, more rapidly than does supercell tornadogenesis from the perspective of Doppler radar. The geographical, seasonal, and diurnal distributions of QLCS tornadoes are unknown.

A complementary, two-part study has been initiated. In phase 1 of the project, the annual number of U.S. tornadoes associated with QLCSs will be estimated using existing radar and verification data. QLCS tornado attributes such as average duration and damage-based intensity then will be determined, and possible geographical, seasonal, and diurnal dependencies will be explored. Some of this work was accomplished recently by a student participating in the SOARS (Significant Opportunities in Atmospheric Research and Science) program. In phase 2 of the project, hypothesized mechanisms of QLCS tornadogenesis will be investigated numerically and theoretically, as will limitations on subsequent tornado intensity and duration. In collaboration with scientists at NCAR, some of the preliminary model experimentation has begun.

Structure and Evolution of an Oklahoma Winter-Precipitation Event as Inferred from Dual-Polarization and Dual-Doppler Radar Data

A significant winter-precipitation event occurred on March 7-9, 1994 in Oklahoma. Snow accumulations greater than 30 cm (12 in) were measured within a ~50-km wide corridor in northern Oklahoma. On the synoptic- and meso-scale, a correspondence between large snow accumulations and 600-hPa frontogenesis was revealed; the precipitation was formed above the cold-frontal surface, owing to midtropospheric ascent associated with the cross-frontal circulation in a region of elevated conditional instability. The location of such a narrow corridor of large accumulations was not, however, disclosed by any patterns in the radar reflectivity data. Indeed, during this event, a snow-accumulation “band” was not associated with a persistent “band” of enhanced reflectivity and vice versa.

Dual-polarization and dual-Doppler radar data allowed for a novel analysis at NSSL of storm-scale winter precipitation features, within the context of the larger-scale diagnosis. It was possible, in order of decreasing poleward distance from the surface cold front, to identify (1) elevated convective elements, which presumably functioned as ice crystal “generator” cells and were embedded within a broad region of generally stratiform precipitation; (2) a reflectivity band and associated rain-snow transition zone, the evolution and structure of which apparently were coupled to the effects of melting precipitation and strong vertical wind shear; and (3) a highly-tilted, prolific lightning-producing, non-elevated cell that was sustained in the post-frontal air in part by virtue of its rotational dynamics.

Utilizing Mesoscale Numerical Weather Prediction (NWP) Models for Diagnostic Purposes

Scientists at NSSL have used both the Penn State/NCAR mesoscale model (MM5) and EMC’s mesoscale Eta model as diagnostic tools, primarily to gain a retrospective understanding of weather systems that have caused forecasting problems.  For example, these models are currently being utilized for diagnostic analyses of the January 21, 1999 tornado outbreak in Arkansas.  In this case, supercell storms were anticipated, but they formed several hours earlier than expected.  Diagnostic analysis of model output is being used to identify the physical mechanisms responsible for early convective initiation.

Development of a Shallow Convective Parameterization for Meteorological and Air Quality Models

In collaboration with scientists at Penn State and the U.S. EPA, scientists at NSSL led the development of a mass-flux-based parameterization of shallow convective clouds. This work has been completed.  This parameterization represents the meteorological effects of the complete life cycle of non-precipitating convective clouds.  It has been fully integrated into the Penn State/NCAR mesoscale model (MM5) and has been designed to be completely compatible with existing parameterizations of deep, precipitating convection as well as non-convective precipitation processes.

Studies of Balanced and Unbalanced Mesoscale Dynamics

The geostrophic coordinate transformation was applied at NSSL to the viscous semigeostrophic (SG) Eady wave solution. In the transformed SG space, the inversion of the geostrophic potential vorticity (GPV) becomes a linear problem, so the development of the Eady wave can be clearly interpreted by the interaction between the upper and lower GPV anomalies. The generation of interior GPV anomaly plumes along the upper and lower fronts in the viscous SG solution can be easily interpreted through its analogy to the intrusion of frontal discontinuities into the interior fluid in the inviscid solution after the surface front collapses.

The classic adjoint theory derived for differentiable systems of equations is not applicable to systems with parameterized discontinuities. The classic theory was recently generalized, and the generalized adjoint formulations were further developed to deal with various complex situations in numerical models. The problems can be avoided by introducing coarse-grain tangent linearization and adjoint without modifying the traditional discretization, although the coarse-grain gradient check can be performed only for finite perturbations. Generalized adjoint with modified discretization and generalized coarse-grain adjoint were derived for a vector system of equations that contains parameterized on/off switches. With vector examples, it was shown that the conventional adjoint minimization might have a convergence problem in multi-dimensional space. The problem can be solved by the generalized adjoint with modified discretization or by the generalized coarse-grain adjoint without modifying the traditional discretization in the forward model.

Assimilation of Doppler Radar, Surface Mesonet, and Satellite Data

Existing Doppler wind and thermodynamic retrieval packages in the ARPS ADAS model have been tested with the NRL COAMPS model at NSSL. Currently, the codes are being upgraded with terrain coordinates using COAMPS backgrounds.

Static Stability and its Evolution Prior to Convective Outbreaks

This NSSL project involves examination of two aspects of low- and mid-level tropospheric static stability over the U.S. — its climatological distribution, and the forcing mechanisms responsible for its evolution prior to and during convective outbreaks.

The climatology of static stability is complete and the results suggest that the Rocky Mountains play an important role in the creation of low stability during most of the year.  The synoptic-scale flow interacts with and modifies this area of low stability to create an area in the Plains where severe convection is possible when combined with other favorable ingredients. To investigate the physical processes responsible for modifying static stability during severe weather outbreaks, a severe event in November 1992 was simulated using the Penn State/UCAR mesoscale model.

Synoptic Setting of Destructive Tinker AFB Tornadoes of 1948

A study was completed at NSSL of the synoptic setting that led to the two destructive tornadoes at Tinker AFB in 1948, the second of which became known as the first successful forecast of a tornado. This work has recently appeared in the formal literature.

Plans – FY 00

Parameterization of Cloud Microphysics and Radiation

Work will continue at CIMMS on the refinement of the new full moment cloud/drizzle scheme for stratiform clouds. The new scheme will be implemented into the CIMMS LES model and verified in case studies against explicit microphysical simulations.

The parameterization of the cloud drop effective radius will be reformulated in terms of parameters predicted by the Kessler-type schemes, such as cloud and drizzle water. This will allow implementing the parameterization in mesoscale models that are using Kessler-type cloud parameterizations.

Implementation of the CIMMS Cloud Parameterization into NWP Regional Models

The comprehensive analysis of the mesoscale simulation of the cloud/fog system over the central California coastal region will be performed at CIMMS. The focus of the analysis will be on understanding the physical mechanisms and feedbacks between mesoscale and cloud/drizzle parameters with the objective of improving the cloud/drizzle/fog forecast. We will also look at the processes that represent sources of CCN and ways to formulate these processes in NWP models.

Development of a Cloud Microphysics Retrieval Algorithms

Work will continue at CIMMS to further the evaluation and improvement of cloud radar and microwave radiometer microphysical retrievals based on the continental stratiform cloud case observed during the spring 1998 intensive operations period conducted at the ARM SGP CART site. The analyzed data set will include in-situ and remote observations, as well as CIMMS LES model simulations.

The Effects of Horizontal Radiative Transport on Cloud Absorption

A new project is planned in at CIMMS collaboration with scientists from NASA and Los Alamos National Laboratory. Its objective is to study the 3D radiation effects on the dynamics and cloud parameters of the marine boundary layer. A proposal to NSF will be prepared and submitted; previous work on the effects of cloud inhomogeneity (done in collaboration with scientists from the Pacific Northwest National Laboratory) will be prepared for publication.

Electrical Structure of Storms and Storm Systems

Because there were fewer storms than expected during our observation period in 1998 and because of unexpected support for some aspects of the MEaPRS field campaign from other sources, less than what was budgeted was spent on equipment and on supplies and services. Though we acquired some very good data sets, the number and variety of cases was limited. Therefore, for the second year of the original grant, in addition to continued data analysis and preparation of publications, we hope to be able to seize cost-effective opportunities to obtain additional data on electric fields and field changes in storms for which other relevant meteorological data will be readily available. With partial support from this grant, NSSL participated in balloon launches at Langmuir Laboratory during July and August 1999 and will attempt to participate in launches in Utah in winter thunderstorms during the Intermountain Precipitation Experiment (IPEX). Also, if the STEPS (Severe Thunderstorm Electrification and Precipitation Studies) field program is funded, it will provide the supplementary data needed by this project. Thus, we plan to participate in STEPS, if it is funded, and will use some resources from our existing grant to supplement new funds to participate.

As discussed in the original proposal and in the activities and findings reported above, we launched a few instruments to observe electric-field changes at altitude in conjunction with the electric field and particle-charge soundings. In one case, we observed field changes coincident with the occurrence of positive ground flashes, though we are unable to determine exactly what lightning processes were responsible for the actual field changes we saw. We believe our original objective, to understand the distribution of charge and electric field in storms that produce positive cloud-to-ground lightning, would be advanced if we could make more such field-change observations in conjunction with electric-field soundings.

This year, graduate students at OU and assigned to NSSL will continue or finish their research projects. Microphysical data will be analyzed from standard probes on the P-3 in some of the same storms observed in MEaPRS for which there are electric field soundings, lightning ground strike locations, and radar data. Planned model simulations will be completed for a dissertation, with emphasis on simulating the electrification and lightning production of different types of severe storms. Research was just completed comparing lightning ground flash rates with radar-derived storm parameters for three storm days during MEaPRS, with a finished thesis and defense near. And, an MS thesis on research analyzing electric field profiles acquired at Langmuir Laboratory during the summer of 1999 will be finished this year.

VORTEX-Related Studies

Additional multi-platform analyses, using techniques developed for the Dimmitt analysis, will be performed at NSSL on the non-tornadic supercell of June 8, 1995.  This storm contained a tornado cyclone, but not a tornado.  A comparison of angular momentum budgets and near-ground conditions associated with this tornado cyclone versus the Dimmitt tornado cyclone will be performed.  Several other tornado cyclones will be examined using the multi-platform analysis techniques, including those that occurred near Friona, TX on June 2, 1995, and Wheeler, TX on June 8, 1995.

General Supercell Studies

NSSL will commence a major new study to investigate the interactions of supercell microphysics and dynamics that are associated with changes in low-level vorticity near the updraft.  This study will encompass the causes of the supercell precipitation spectrum, the nature of hook echoes and forcing of rear-flank downdrafts, and the ability of rear flank downdrafts to produce tornado cyclones and tornadoes.  This study will utilize new and existing multi-parameter Doppler radar data sets, the Straka/Rasmussen microphysics package in Straka’s cloud model, and multi-platform observational data sets (photographic, mobile mesonet, airborne/mobile Doppler, etc.).  New observational data will be obtained through participation in the STEPS experiment in the High Plains in spring 2000.


During the coming year the focus of this NSSL work will shift toward documenting the wide variety of synoptic conditions that can bring increased moisture and storms to all of southern Arizona. A presentation of many aspects of forecasting severe thunderstorms is planned at the SPC.

Quantitative Precipitation Estimation Using Multiple Sensors

During the coming year this NSSL work will continue with goals of developing gridded, multi-sensor data sets that can be used easily for precipitation estimation over the Salt and Verde drainages of Arizona, avoiding the use of the operational PPS system. The work will also be expanded to develop techniques for determining brightband height, differentiating regions of rain versus snow and estimating snow accumulations.

Frontal Structure over the Western United States

The evolution of the midlatitude cyclone of December 12-14, 1988 will be described from its appearance over the eastern Pacific Ocean, to landfall on the coast of western North America, to the progression of a lower-tropospheric cold front across the Pacific Northwest, and to its redevelopment on the lee side of the Rocky Mountains.  This research at NSSL, though primarily observation-based using the standard observational network, will also employ the nonhydrostatic MM5 to assist in the analysis.  In this way, an understanding should be gained of the evolution of a cyclone and its associated fronts over the western United States using the same tools available to operational meteorologists in real time.  Each stage in the evolution of this event addresses scientific issues relevant to both operational and research meteorologists.  For example, in the landfalling stage, the structure of the mid- and upper-tropospheric baroclinic zone undergoes an evolution previously documented only for upper-level baroclinic zones in northwesterly flow.  As the system moves through the Pacific Northwest, a nonfrontal pressure trough and wind shift occur at the surface, apparently associated with the upper-level baroclinic zone.  These observations are relevant to the issues of surface frontal analysis addressed by Sanders and Doswell (1995) and Sanders (1999).  As the system moves through the Rockies, frontogenesis results on a variety of scales (mountain, meso, and synoptic), leading to the development of a warm front in Montana.  The movement of the cold frontal zone will also described.

The Formation and Climatological Distribution of Tornadoes within Quasi-Linear Convective Systems

Work will continue at NSSL on the study of the formation and climatological distribution of tornadoes within quasi-linear convective systems.

Utilizing Mesoscale NWP Models for Diagnostic Purposes

Scientists at NSSL will continue to utilize NWP models as tools to facilitate basic and applied meteorological research.  They will continue to cultivate a working relationship with SPC forecasters and to seek guidance from them in defining research topics that are particularly relevant to operational forecasting problems.

Studies of Balanced and Unbalanced Mesoscale Dynamics

The generation of gravity waves by the unbalanced dynamics associated with mesoscale fronts will be examined at NSSL. In addition, the potential applications of the balanced and unbalanced dynamics in mesoscale data assimilation will be explored.

Assimilation of Doppler Radar, Surface Mesonet, and Satellite Data

Wind and thermodynamic retrieval algorithms will be developed and tested at NSSL using COAMPS backgrounds having terrain coordinates. The spline smoother will be compared to a penalty function smoother for possible improvements of the retrieval schemes.

Static Stability and Its Evolution Prior to Convective Outbreaks

Model output from the Penn State/UCAR model simulation will be analyzed at NSSL and two papers will be submitted for publication over the next year.  Preliminary results suggest that horizontal advection of low stability is the dominant process for decreasing stability; other less significant processes include vertical advection of stability, vertical stretching, vertical motion, differential diabatic heating, and differential thermal advection.

The Intermountain Precipitation Experiment (IPEX)

The Intermountain Precipitation Experiment (IPEX) will be conducted in February 2000 in northern Utah. It is a field and research program designed to improve the understanding, analysis, and prediction of precipitation and precipitation processes in complex terrain. The project is being lead by CIMMS and NSSL and will involve scientists from the University of Utah, Desert Research Institute (DRI), and staff from the NWS forecast office in Salt Lake City. Students from the Universities of Utah, Nevada and Oklahoma will also participate. The field phase is scheduled for January 31-February 25, 2000. The major scientific objectives of IPEX are (1) to advance fundamental knowledge of orographic precipitation, with an emphasis on the narrow, steeply sloped Wasatch Mountains of northern Utah, (2) to improve knowledge of lake-effect precipitation of the Great Salt Lake, (3) to validate and improve high-resolution data-assimilation systems, mesoscale model performance, and quantitative-precipitation forecasts over complex terrain, and (4) to validate and improve quantitative-precipitation estimates produced by WSR-88D’s located at high elevation.

These objectives are directly related to two major foci of the U. S. Weather Research Program (USWRP): (1) quantitative precipitation forecasting and (2) the optimal mix of observations in numerical weather prediction. Results from IPEX will also have positive scientific and socioeconomic benefits for the intermountain west, including Salt Lake City, host of the 2002 Winter Olympics.

IPEX observing platforms include two NCAR Integrated Sounding Systems (ISS), two NSSL mobile laboratories with cross-chain LORAN Atmospheric Sounding Systems (CLASS), two Doppler weather radars on wheels (DOW), a NOAA P-3 research aircraft, 28 portable surface mesonet stations from the U.S. Army Dugway Proving Grounds, two surface-based microwave radiometers, a suite of ground-based microphysical probes, and supplemental soundings from NWS upper-air observing sites. These platforms will enhance an existing surface observing system known as the Utah Mesonet. Data collected during six to eight intensive observational periods will allow project scientists to examine a number of questions and testable hypotheses concerning the interaction of dynamical and microphysical processes during orographic precipitation events, including factors controlling the distribution and intensity of precipitation across a narrow, steeply sloped mountain range like the Wasatch Mountains. The complex interactions between thermally- and terrain-driven circulations that produce lake-effect snowbands of the Great Salt Lake will also be examined. Other project activities will validate and improve mesoscale model quantitative precipitation forecasts and quantitative precipitation estimates using WSR-88D radars.


Progress – FY 99

Improving Quantitative Precipitation Forecasting by Numerical Weather Prediction Models

Techniques for improving quantitative precipitation forecasting (QPF) using the NCEP Environmental Modeling Center’s (EMC) Eta model have been investigated at NSSL. An experimental configuration of the model was run in forecast mode at NSSL in parallel with the operational model at EMC, and in collaboration with EMC scientists. It was configured with the Kain-Fritsch convective parameterization and higher-order numerical diffusion than the operational model contains, both of which are designed to allow the model to produce and retain mesoscale structures. After a series of refinements, the experimental configuration of the model achieved comparable scores on traditional measures of skill for QPF, while providing higher resolution mesoscale guidance than the operational model. New verification techniques are being developed so that the accuracy of the model in producing finer-scale features can be better evaluated. Further model refinements are likely to be concentrated in the parameterizations of turbulent mixing, microphysics, and moist convection, all of which have been shown to have a significant impact on QPF. Forecasters from the SPC will be involved in the identification of relevant model output fields as well as the development and implementation of new verification techniques. Daily numerical predictions, comparisons with operational EMC models, and verification statistics have been made available on the World Wide Web.

OSF-Operations Training Branch Distance Delivery Evaluation

From the inception of the WSR-88D network in 1991 up until 1997, the Operations Training Branch (OTB) of the NWS OSF used traditional instructor-led, classroom based, residential training to transfer research knowledge into skills to achieve documented success in improved weather predictions and warnings. However, because of structural and budgetary changes within the NWS, the OSF had to move rapidly to new distance-delivery systems employing an array of real-time audio graphics, web-based training, and computer-based training on CD-ROM. As part of standard NWS procedure, a collaboration among CIMMS, the OSF/OTB, and the Center for Distance Education (CDE) of the College of Continuing Education (CCE) of the OU are evaluating whether the shift to distance-delivery systems has adversely impacted overall training results, and especially whether it has degraded forecast and warning skills. The CDE has utilized its expertise in the academic field of Instructional System Design to research the effectiveness of the new training delivered to operational field forecasters of the NWS.

Although the preliminary validation study completed by CDE in March 1999 found that the distance delivery systems had no adverse impact on training effectiveness, weaknesses in some areas of NWS distance delivery systems were noted. Many of these recommendations were addressed in the development of courses to be delivered from October 1999 through March 2000.

OSF Operations Branch Support of WSR-88D Operators

The OSF Operations Branch has been supporting WSR-88D field sites in improving the reliability of access to WSR-88D data and assisting operators in interpretation and application of WSR-88D data, products, and algorithm output. These efforts improve the ability of forecasters to accurately apply radar data to their forecast operations and the decision process of issuing severe weather and tornado warnings.

SPC Forecast Verification

During the last year scientists at SPC have continued work toward developing the forecast verification program at the SPC. Old methods of verification have been applied to SPC products for the period 1970-1997 so that the SPC has a climatological record of forecast verification through which it may judge forecaster performance. Most of the verification work so far has focused on the Convective Watches (severe thunderstorm and tornado) that comprise one of the main public information services that SPC provides.

Improving Short Term Forecasting and Warning of Severe Weather using Satellite and Radar Data

Development and testing continued at NSSL of algorithms to determine storm severity based on cloud top temperature, height, and/or growth. Analyses of several supercell thunderstorms support previous research showing correlations between cloud top temperature characteristics and storm severity. Software was modified to track cold cloud areas in mid-latitude thunderstorms and evaluated it extensively.

Hazardous Winter Weather Climatologies

This project is designed to provide the meteorological community, particularly NWS forecasters at the SPC, climatological information on hazardous winter weather phenomena over the United States.  This project started four years ago at NSSL because of the need to analyze and publish climatological data related to these hazardous winter events.

During the past year, serious errors in the original processing of the research dataset were found, requiring it to be completely recreated, using programs that had to be rewritten by two CIMMS programmers.  This reprocessing continues.

Warning Decision Support System (WDSS)

For six years, scientists at NSSL have been developing and testing the Warning Decision Support System (WDSS). The WDSS includes enhanced Doppler-radar algorithms, data integration and imaging techniques, innovative algorithm product display capabilities, and a severe-weather warning generation system. It has been tested at 18 NWS forecast offices since 1994. Many of the concepts developed in the WDSS have received very favorable comments from the operational forecasting community. During this fiscal year, CIMMS and NSSL personnel worked closely with the Techniques Development Laboratory to implement WDSS functionality in the NWS Advanced Weather Interactive Processing System (AWIPS). Many of the best features of WDSS were implemented into AWIPS and are slated for release in 2000.


The NSSL WISH group has continued to work directly with the staff of a number of NWS forecast offices, the SPC and the OSF/OTB to improve the use of WSR-88D data to identify and warn for severe thunderstorms in the western United States. Efforts have been made to evaluate the performance of the SCIT and HDA algorithms in the west and numerous workshops have been held to convey the results to field forecasters. Results of this effort have affected the evolution of several WSR-88D algorithms and helped lead to a real time evaluation of the NSSL WDSS system at the Tucson forecast office during summer 1999. Problems with clutter filtering have also been identified and we are working directly with the OSF and the Phoenix forecast office to both improve the local clutter filter tables and to identify causes for apparent failure of the clutter filter at ranges less than 50 km. This work also led to interactions with NWS operations headquarters regarding the concurrent failure of the ground echo filter and the PPS system, leading to non-real, huge precipitation accumulations from the Phoenix radar. These kinds of failures affect long-term research since the false precipitation is archived and not flagged and thus resides in the databases as if real rain events occurred. The impacts of military chaff releases and their false echoes have also been documented and discussed at a number of Department of Defense meetings.

Areal Mean Basin Estimated Rainfall (AMBER) Algorithm

Meteorologists at the Pittsburgh forecast office originally developed the Areal Mean Basin Estimated Rainfall (AMBER) algorithm. AMBER accumulates precipitation on a basin level for comparison to flash flood guidance values to determine the likelihood of flash flooding. During 1998, the algorithm was implemented at NSSL to run in real-time as part of WDSS, including the development of a display that shows these accumulations and guidance values in geographic, tabular, and plot formats.  This implementation marked the first Flash Flood decision support tool to be implemented for real-time operational feedback within the NWS.  During 1999, testing has continued at the Sterling and Tulsa forecast offices.  The feedback obtained from the testing and evaluation of the algorithm and display will be utilized in implementing this algorithm or similar applications in AWIPS.

A procedure manual was written to assist NWS offices on delineating basins, and the computer program ArcView GIS was used to create necessary data files for the AMBER algorithm. To date, the Salt Lake City, Tucson, and Jackson forecast offices have successfully completed the tasks of basin delineation and AMBER file creation. Also, an improved basin delineation procedure was developed using the recently available River Reach Files Version 3-Alpha (RF3) from the U.S. EPA. This is used to provide quality control for the accuracy of basin definitions derived from digital elevation model data. This improved procedure is being used to delineate basins for the Phoenix and Flagstaff forecast offices.

Flash Flood Algorithm Development

Scientists at NSSL made necessary modifications to the OUFEAHRM, a finite-element analysis hydrologic runoff model developed at OU, to enable its use in real-time forecasting. Data from several case studies were obtained for the Blue River test basin in Oklahoma, including Level II archive radar data, raingage measurements, and streamgage measurements.

Multi-Sensor Estimation of Precipitation

A Neural Network QPE (quantitative precipitation estimation) scheme based on radar, satellite, lightning, and environmental data has been proposed at NSSL to tackle the problem of the multi-sensor estimation of precipitation. Multi-sensor observations for all rainfall events during the STormscale Operational and Research Meteorology-Weather data Assimilation and Verification Experiment (STORM-WAVE) have been obtained.  The STORM-WAVE project was conducted in a region of 30°N to 45°N and 109°W to 86°W and spanned a three-month time period (April-June 1995).  The data sources include raingage, radar, satellite, and lightning observations.  Numerical model analysis and forecast data were also obtained. Software has been written for ingesting raingage, radar, and model data.  Some statistical analyses have been performed on the raingage data.  Also, software has been developed to stratify radar and model (environmental) data based on raingage observations. Some intercomparisons have been performed between raingage, radar and satellite rainfall estimations for the Dallas hailstorm case of May 5-6, 1995) using the VisAD software. Software was also developed for re-mapping WSR-88D radar data onto a 3D regular Cartesian grid.

An Evaluation of Algorithms that Infer Precipitation Type near the Ground

This NSSL project examines several algorithms that infer precipitation type near the ground using thermodynamic variables obtained from rawinsondes and numerical models.  Although the U.S. and many other countries use one or more of these algorithms to forecast precipitation type, a rigorous statistical evaluation of these programs has not been done prior to this project. Computer code has been obtained and written for five algorithms that infer precipitation type near the ground.  Some data have been processed and prepared for use in testing the algorithms.

During the next year this research will be completed.  Data will be obtained from datasets at NSSL and the evaluation procedure will be determined.  Research results will be presented at a Canadian conference and a manuscript will be prepared for submission to an American Meteorological Society journal.

System for Convection Analysis and Nowcasting (SCAN) Implementation and Testing

For the third year, the System for Convection Analysis and Nowcasting (SCAN) was tested in a real-time operational setting in the Sterling and Tulsa forecast offices.  SCAN is being developed and designed as a future AWIPS component to provide short-term guidance to forecasters on the probability of severe weather. NSSL, the National Center for Atmospheric Research (NCAR), and the NWS Techniques Development Laboratory (TDL) are undertaking the development of the system. The 1999 test will run from June 1 – October 30 and will provide feedback on the utility of new severe weather applications for operational warning situations.

Common Operations and Development Environment (CODE) Development

During 1998-1999, development of the Common Operations and Development Environment (CODE) began at NSSL.  CODE is being designed to provide application developers with an environment for developing, testing, evaluating, scoring and documenting their applications.  CODE will provide data access services for WSR-88D data, satellite data, and other observational data.  These access services will be available for offline development and testing as well as real-time testing and evaluation.  In addition, CODE will provide a library of common calculations that application developers regularly use and a data analysis and display tool.

Application of Sounder and RUC-derived Environmental Conditions to Forecasting the Demise of Organized Convective Systems: A Feasibility Study

NSSL investigated the feasibility of applying environmental conditions related to the sustenance of organized convection to convective demise forecasts. Specifically, we investigated the feasibility for using the environmental conditions observed by the GOES-8 Sounder and forecast by the RUC model. Analysis of long-lived convective events showed that practical use of RUC forecast output and GOES sounder data are limited by forecast errors and cloud cover in the inflow regions of mature and dissipating convection.  Owing to the relatively few combinations of useful GOES-8 sounder data and RUC forecasts, the study of relationships between environmental conditions, derived from these sources and convective decay, was not recommended.

Explicit Modeling of Convection in the Terminal Area

The objective of the project is to evaluate the predictive capability of the Advanced Regional Prediction System (ARPS), the storm-scale numerical weather prediction system created by CAPS over the past ten years primarily under funding through the National Science Foundation. Specifically, ARPS precipitation forecasts were evaluated against radar observations for five selected severe weather cases over the past year and yielded, on average, a POD of 0.80 (hits divided by the number of hits and misses) through the 2-hour forecast. A hit was determined if forecast reflectivity of 30 dBZ or higher at a point was verified anywhere in the same county (loosely, within 25 km). An algorithm that numerically evaluates the forecasts was later created following the “fuzzy” logic method of Hallowell. In this approach, a “hit” is determined if precipitation forecast at a point is observed to occur within a radius of about 7.5-km. The ARPS forecast for the May 3 Oklahoma City area tornado case was evaluated and netted a POD of 0.7 for the 1-hour forecast and 0.5 for the 2-hour forecast. The 2-hour forecast error improved somewhat when phase shifting was taken into account. Small-scale forecasts pose a difficulty in verification because any small time lag error will cause the forecast to score badly even when the overall set of features (intensity, areal coverage, and duration) are forecasted well. A technique was invoked from Brewster that moves the forecasted storm complex to overlay the observed fields and thus discount phase errors.

Using a Soil Hydrology Model to Initialize Soil Moisture Profiles for Numerical Weather Prediction Models

Scientists at NSSL modified a Soil Hydrology Model (SHM) and daily predictions of soil volumetric water content were made at 38 Oklahoma Mesonet sites during July 1997. These predictions were then compared to soil moisture observations made at the Mesonet sites at 5, 25, 60, and 75 cm.

Comparisons of time series between the observed and model-predicted volumetric water content at 5 cm revealed similar phase and amplitude changes, a coefficient of determination (R2) of 0.64, and small mean bias and root-mean-square errors (MBEs and RMSEs) of 0.03 and 0.09, respectively. At 25, 60, and 75 cm, the model performance was slightly worse, with R2 values between 0.27-0.40, MBEs between -0.01-0.02, and RMSEs between 0.11-0.13. The model response to changes in soil water at these levels was sluggish, possibly due to a lack of ability to model preferential downward water flow through cracks in the soil.

Sensitivity tests revealed the importance of specifying the proper infiltration depth, while modifying a previous assertion that the model output is independent of initial soil water content. Tests also showed that the modeling of the conductive processes was relatively insensitive to the specification of the soil hydraulic parameters; and that the processes of precipitation infiltration and evapotranspiration overpower the changes in soil water content associated with vertical water transfer.

Finally, the viability of using a much simpler model to initialize soil moisture in the topsoil was demonstrated. However, since much of the water that is extracted from the ground comes from beneath the topsoil, the use of SHM is recommended to resolve the complete vertical profile of soil water. The ability to model the phase and amplitude changes of soil moisture is unique and provides an opportunity to initialize both weather and storm-runoff models with realistic soil moisture values based upon currently available observations.

Improvements in Land Use Specification in MM5-PLACE

Sensitivity tests were performed at NSSL on the Parameterization for Land-Atmosphere-Cloud Exchange (PLACE) module of the Pennsylvania State University – National Center for Atmospheric Research Mesoscale Model version 5 (MM5) in order to determine the importance of the individual land use parameters. The coverage and thickness of green vegetation (as manifested by the fractional green vegetation and leaf area index values) had the greatest effect on the relative magnitudes of sensible and latent heat fluxes, which in turn determine the depth and character of the daytime boundary layer. Variations in albedo and surface roughness length were found to have much smaller effects on the surface fluxes.

Land use parameters derived from the Advanced Very High Resolution Radiometer (AVHRR) sensor were then inserted into PLACE and the improvements in model-predicted surface energy fluxes in Oklahoma during July 1997 were documented. Two-week average values for fractional vegetation coverage, albedo, and leaf area index at 1-km resolution were all available for use. Since PLACE allows for incorporation of 12 separate surface energy budgets (mosaic tiles) within each model grid box, it is possible to take advantage of this high-resolution land use data set. Previous land-surface models have simply used climatological values of these crucial land use parameters. The ability to improve model predictions of surface energy fluxes in a diagnostic sense provides promise for future attempts at ingesting pseudo-real-time land use data into numerical models. These model improvements would be most helpful in predictions of extreme temperature events, where current numerical weather prediction models often perform poorly.

Plans – FY 00

Improving Quantitative Precipitation Forecasting by Numerical Weather Prediction Models

Scientists at NSSL will continue daily numerical forecasts and testing with the experimental version of the Eta model, along with the development of verification strategies that focus on mesoscale predictability.  Model refinements are likely to be concentrated in the parameterizations of turbulent mixing, microphysics, and moist convection, which have been shown to have a significant impact on QPF.  Verification strategies will involve primarily QPF, but will also include evaluations of model skill in producing the mesoscale precursors to disruptive and severe weather.  Forecasters from the SPC will be involved in the identification of relevant model-output fields as well as the development and implementation of new verification techniques.

OSF-Operations Training Branch Distance Delivery Evaluation

The collaboration of CIMMS, the NWS OSF, and the CDE of the College of Continuing Education at OU plan to continue evaluating the training effectiveness of distance learning. The focus of this follow-on research is threefold:

Are students in the newly developed distance learning courses meeting training objectives at the same high rate as students trained in past residential courses?
Are the implementations of recommendations made in the March 1999 validation study adequate?
Are materials presented in “train-the-trainer” workshops actually reaching field forecasters in an effective manner? If not, what recommendations might be made for enhancement and further investigation?

This collaborative training evaluation effort requires extensive interactions (including electronic communication) between the OU and NWS participants, and will especially capitalize on the facilities and flexibility of the diverse OU meteorological and continuing education environments. The goal of the project is to enhance the training of NWS meteorologists, hydrologists, and meteorological technicians so that they might more effectively utilize the WSR-88D to produce better forecasts and warnings. The results will be summarized in research notes and reports.

OSF Operations Branch Support of WSR-88D Operators

Project plans include continued support of WSR-88D operators and technical personnel world wide in interpretation and application of WSR-88D data, products, and algorithm output. In addition, the OSF Operations Branch will:

Respond to WSR-88D technical and scientific inquires from the field;
Evaluate operator applications of WSR-88D clutter mitigation capabilities through surveys, comparisons, and summaries of site performance in weather detection effectiveness and publish technical articles to assist sites in improving their clutter mitigation;
Support OSF efforts to optimize WSR-88D calibration network wide; and
Work to identify possible causes of precipitation estimation errors in the WSR-88D Precipitation Processing Subsystem at numerous sites.

These efforts are aimed at further improving WSR-88D data quality and thus weather forecasts and warnings of severe and tornadic weather.

SPC Forecast Verification

Work will continue on SPC forecast verification. This work will include the verification of SPC Convective Outlooks in addition to Convective Watches. Further, work will begin to implement the verification schemes in near real-time so that forecasters can judge their performance as quickly as possible. Once the current SPC verification scheme is operational, work will begin on a next generation set of forecast verification schemes in collaboration with scientists at NSSL. These will begin to address some of the problems with the current verification scheme as well as develop schemes for the verification of (future) SPC probabilistic forecasts.

Local Software Development within the SPC

Research continues in looking at the local probabilities of severe weather. In the next decade, the SPC will likely begin to issue probabilistic guidance for forecast products rather than using arbitrary thresholds for severe weather occurrence and intensity. Preliminary work is being done to look at the feasibility of developing a national climatology of severe weather occurrence as it relates to SPC Convective Watches and Convective Outlooks. These probabilities, once developed, would give NWS forecasters an idea of the likelihood of severe weather in their area based on the SPC decision to include their area in certain categories of Convective Watches or Outlooks.

Owing to uncertainties and limitations contained within numerical model data, SPC forecasters continue to rely heavily on observational data when making short term decisions related to Convective Watches. There remains a need to continue to develop software that allows the SPC forecasters to interrogate all of the data to its fullest extent. This development continues.

Climatology of SPC Products

Interaction with NWS forecasters has shown that they are very interested in seeing a climatology of SPC Convective Watches so that they may have a general idea of when and where to expect the SPC to issue watches across the U.S. on a monthly basis. Currently, the SPC Convective Watches verify with severe weather approximately 90% of the time. By having a watch climatology, the NWS forecasters would then use this research to adjust staffing levels and for other advanced planning, which would help provide better service to the local communities that they are tasked with providing guidance.

Improving Short Term Forecasting and Warning of Severe Weather using Satellite and Radar Data

The next step for this NSSL project is to automate the storm severity algorithms for implementation into the NWS ‘s System for Convection, Analysis and Nowcasting (SCAN).

Hazardous Winter Weather Climatologies

Processing of the research dataset should be complete by October 1999.  Once the research dataset is available, many climatological studies that require use of these data will be undertaken and completed during the next fiscal year: 1) a freezing rain climatology in the Great Lakes region of the U.S. (already conditionally accepted for publication in Monthly Weather Review); 2) a winter weather climatology for North America (in collaboration with Canadian and NCAR meteorologists); 3) a study to examine synoptic conditions associated with areas in the U.S. that experience a high frequency of freezing rain; and 4) a climatology of lightning occurrences at subfreezing temperatures.

Warning Improvement through Better Knowledge of Mesocyclone Climatology

The goal of this new project at OU is to establish a mesocyclone climatology for Oklahoma and the surrounding region. This will represent the first step toward establishing a national climatology on mesocyclones, as observed by the WSR-88D radar. We plan  to adapt the existing mesocyclone detection algorithm (used in WDSS) to identify, follow, and gather statistics on all of the mesocyclones that formed during one or two seasons. The adapted algorithms will use the base radar data streams being collected under the CRAFT project. Some preliminary meetings have been held to discuss the organization of the work and a graduate research assistant will begin her work this fall.


The WDSS is now being improved to WDSS-Integrated Information (WDSS-II) at NSSL. The WDSS-II will support multiple data types to allow for easier inclusion of new algorithms, and it will provide a new user interface for making more informed warning decisions. Version 1.0 of WDSS-II is expected to be ready for field testing during spring 2000.


During the coming year the focus of this NSSL work will on providing input to the development and improvement of a damaging downburst algorithm and formal documentation of the filter and false echo problems, with particular emphasis on the southwestern U.S.

AMBER Algorithm

Testing of the AMBER algorithm at the Sterling and Tulsa forecast offices will continue at NSSL. Testing of the AMBER algorithm will start at the Tucson, Salt Lake City, and Jackson forecast offices.  AMBER will be implemented as part of the new WDSS-II system.

Flash Flood Monitoring and Prediction Program

The Flash Flood Monitoring and Prediction (FFMP) program is scheduled to be included in the next build of AWIPS. It will include flash flood warning decision guidance similar to that provided by the AMBER program. The overall goal of this proposed work is to develop and implement tools and algorithms that, in real-time, identify individual basins at risk of flooding due to intense, though short duration, summertime thunderstorm events. Use of FFMP will require basins to be defined to a minimum size of two square miles for every radar site in the U.S. NSSL has been contracted to delineate basins for 25 radar sites in the Eastern Region during the next 18 months. The tasks to be completed as part of this project include editing the EPA’s RF3 files to be used for quality control, delineating and editing basin boundaries, assembling various information associated with each basin, creating the data files necessary for the FFMP, and saving all data sets for a site on a CD-ROM so that individual forecast offices can make changes or additions as necessary.

A Flash Flood Index (FFI) will be developed to enhance the information provided by the FFMP. Parameters that could potentially be used in the FFI will be researched to determine which factors significantly affect flash flooding. These parameters include basin area, hydraulic roughness of the land surface, hydraulic conductivity of the soil, and antecedent moisture conditions. The FFI would be validated on geographically diverse basins using a distributed-parameter physically based flash flood model (FFM) and observed stream discharge. The WSR-88D has improved considerably the detection and warning of severe storms. The WSR-88D processing system has built-in detection for features related to tornadoes and precursors such as circulation. Operationally in the NWS, no such detection tool exists for flooding except very general flash flood guidance threshold values based on antecedent rainfall. The skill or ability of the proposed FFI is to predict which basins are likely to flood. Using the WSR-88D network of radars, the proposed physically based FFI will be a next-generation tool for monitoring, prediction, and warning for the NWS reducing societal costs associated with this flood hazards.

Flash Flood Algorithm Development

The validity of radar data for various historic storm events over the Blue River test basin in Oklahoma will be checked at NSSL with raingage measurements. Radar data verified as reasonably accurate will then be used to calibrate the hydrologic model for the Blue River.

Improved Hydrologic Prediction: Oak Creek Watershed Pilot Project

The overall goal of this proposed project at NSSL is to demonstrate the utility of the distributed hydrologic approach to prediction of flash floods, water supply, and reservoir operations in the subject watersheds administered by the Salt River Project (SRP). Improved forecasts of rainfall-runoff and flash flood prediction have important consequences to the SRP. The proposed hydrologic modeling system relies on the WSR-88D radar precipitation processing system (PPS), rain gages, river gages, and a physically-based distributed parameter watershed model, called OUFEA. The proposed project is a pilot project to demonstrate the feasibility of distributed parameter hydrologic modeling coupled with the WSR-88D radar. This model will be applied to the 233 square mile Oak Creek watershed located near Sedona, Arizona (southwest of Flagstaff). Through other funding, improvements to the WSR-88D PPS are being developed as a part of the WDSS-II, a real-time processor of radar data for the WSR-88D radar. Though precipitation errors cause one of the largest uncertainties, the PPS offers unprecedented fine-resolution estimates of precipitation for hydrologic prediction.

Multi-Sensor Estimation of Precipitation

Statistical analysis of the STORM-WAVE rainfall data will continue at NSSL. Work will also continue on stratifying lightning and satellite data, based on raingage observations.  A database will be established that contains co-located raingage, radar, satellite, lightning, and environmental data for STORM-WAVE rainfall events. This statistical work will then be applied to the database to develop a Neural Network model and a physically based algorithm (which includes identification of bright-band and other precipitation types) that can be tested in the real-time.

An Evaluation of Algorithms that Infer Precipitation Type near the Ground

During the next year this research will be completed at NSSL.  Data will be obtained from datasets at NSSL and the evaluation procedure will be determined.  Research results will be presented at a Canadian conference and a manuscript will be prepared for submission to an American Meteorological Society journal.

SCAN Implementation and Testing

Testing of SCAN will continue at the Sterling and Tulsa forecast offices and will be observed at NSSL.

CODE Development

Version 1.0 of CODE is expected to be available at NSSL for a beta test in December 1999.  An updated version is anticipated in spring 2000.

Explicit Modeling of Convection in the Terminal Area

Forecasts for other severe weather events beside May 3 will be evaluated at CAPS using a verification procedure with and without the phase-shifting technique. Cases will be chosen for which 3-km forecasts already exist. In addition, radar reflectivity will be used to evaluate forecasts rather than vertically integrated condensate (VIC). This is to alleviate uncertainties associated with attempting to interpret moisture variable densities from reflectivity data for different hydrometeor types and also allows for use of the radar data observations in their native form. Additionally, Level III (NIDS) data will be incorporated, which has already undergone quality control checks and provides a composite from multiple radars.

Improvements in Land Use Specification in MM5-PLACE

Work will continue at NSSL on tests of the MM5-PLACE model to determine the importance of the individual land use parameters.

The Morning Convection Project

Mesoscale convective systems often live through the night and affect the Great Plains during morning hours.  Usually these systems die out during the late morning, but not always.  Forecasters are thus faced with a decision as to whether to forecast dissipation or maintenance.  Little is known about what controls the evolution of these systems; therefore, this project will strive to increase understanding of this evolution and develop techniques to improve forecasts of them.

This project has gone on informally for several years and involves collaboration between personnel at NSSL, the forecast offices at Norman and Dodge City, and at OU.  Recently a proposal was funded through COMET’s NWS Cooperative Projects opportunity that will support work for three years beginning in January 2000.  The primary benefit of support will be a CIMMS graduate student who will perform research for the project as a master’s thesis research topic.  A climatology of these systems over a six year period is planned that will include investigating the wind and stability structure ahead of such systems.  For the past three summers NWS personnel have logged comments at a special web site on the forecast and evolution of real-time systems.  A preliminary climatology as part of a 1998 REU program has found that these systems occur quite frequently and exhibit  preferred directions of system motion, along with a tendency toward declining in intensity or dissipation.


Progress – FY 99

‘Al Moubarak’ and Moroccan Precipitation

Over the last five years, there has been a strong collaborative effort between CIMMS and the Kingdom of Morocco to increase our understanding of the interannual-to-decadal variability of Moroccan winter precipitation, particularly within the context of the North Atlantic climate system (‘Al Moubarak’). This effort has also included the wider global climate system (tropical Pacific sea surface temperatures) as it relates to the late rainy season. The project was motivated by a predominance of extremely poor Moroccan winter precipitation seasons since the late 1970s. This understanding is being used to develop a seasonal precipitation prediction capability for Morocco. This work has been done in direct collaboration with the Moroccan Direction de la Météorologie Nationale (DMN) with the financial support of the U.S. Agency for International Development, including through DMN scientists being in residence at CIMMS.

During this fiscal year, the development and issuance to Moroccan government officials of “Experimental Precipitation Predictions for Morocco” continued, with the prediction for 1998-99 verifying extremely well. Also, a book chapter, “Climate Variability in Northern Africa: Understanding Droughts in the Sahel and the Maghreb”, was published in “Beyond El Nino: Decadal and Interdecadal Climate Variability” (Antonio Navarra, Ed., Springer Verlag, 1999). This chapter features research on the climate variability of the Sahel and Maghreb. Another publication prepared was “Towards the Seasonal Prediction of Moroccan Precipitation and its implications for Water Resources Management” for the proceedings of the Abidjan’98 Conference held at Abidjan, Cote d’Ivoire, in November 1998.

The study of the relationship between weather systems and Moroccan precipitation patterns within the North Atlantic climate system also continues. A major objective of this phase is to enhance our understanding of other meteorological processes affecting Moroccan precipitation. To that end, separate studies are being done to examine which North Atlantic Oscillation (NAO) periods had significant or little control over Moroccan precipitation patterns:

Weather Systems Analysis

For the documentation of the weather systems, a cyclone-tracking algorithm was obtained from Colorado State University and it has been used to track the intensity and the path of weather systems in the North Atlantic since 1958 using four per day NCEP/NCAR reanalysis data. From this algorithm, the following ten variables have been outputted with daily resolution on a 2.5 by 2.5 degree grid from 1958 through 1997: total cyclone events, number of cyclone days, total system events, number of system days, cyclone central pressure, cyclone intensity, cyclone translation velocity (all cyclones and no-stationary as well), max deepening, and cyclogenesis and cyclolysis events. Software has been developed to plot the storm tracks of each storm over a month’s time. On these plots, an identification number and the time of each observation of position document the storms. The sea level pressure anomaly maps using the NCEP reanalysis data have been produced for the 17 months identified as extreme precipitation cases for one of the NAO control and non-control categories for both coastal and central Morocco.

These results have been binned in boxes that span 30° W-0° in longitude and either 1.0, 2.0 or 2.5° in latitude. These two basic datasets were used to create seven different “running domain” data sets. The motivation behind these data sets is to have smoother latitudinal profiles of the variables mentioned above. The running domain data sets differ in the number of bins that are included for each latitude estimate of a variable and in the latitude resolution of the estimates. As an illustration, if the “running domain” is moved one degree for every estimation, then the latitudinal resolution of the estimates will be one degree. The latitude estimate is made at the mean latitude of the area covered by the bins. With each of these datasets, the annual cycle of the latitudinal profiles of each of these variables has been plotted. Monthly latitudinal profile plots were produced showing the composite profile for each of the four categories chosen to show NAO control and non-control. Each of these composite profiles also has climatology as a background. The running domain dataset with 10 bins (at 2.0° /bin) per latitudinal estimate yielded the best results.

Rainfall Event Analysis

An eighteen year (1979-1996) daily Moroccan rainfall dataset was obtained for 25 reasonably well distributed stations. A monthly dataset was created for each station to quantify the rainfall events. The parameters included the number of rainfall events/month, the total precipitation from all of these events, the mean precipitation/event, and the mean length/event. This monthly dataset is being used in conjunction with the monthly cyclone data described above to analyze the relationship between weather systems and Moroccan precipitation patterns. Annual cycles for each of these parameters at each station (which were grouped by region) were plotted. Contour plots were made showing the spatial scale of the monthly climatologies. Composites of these variables for the four categories representing NAO control and non-control were done.

Predictability of East African Rains

Diagnosis and predictability studies of rainfall in east Africa continued at CIMMS. A paper was submitted to the Journal of Climate entitled ‘East African rainfall and tropical circulation/convection on intraseasonal to interannual timescales’. The current research results improve understanding of the tropical atmospheric dynamics associated with the regional rainfall and increase confidence in prediction systems; identifying circulation systems that need monitoring in real-time, including early warning (10-15 days) risk on extended wet spells. Further new results were presented at the European Geophysical Society Annual meeting in April 1999. These included the finding that after a rainfall event over East Africa during the Short Rains (October-December), the disturbance, particularly apparent in the 200 hPa divergence, often propagates into the Indian Ocean and intensifies into structures that resemble the Madden-Julian Oscillation (MJO). Some wet spells over east Africa can be traced back westward through equatorial Africa and into the tropical Atlantic Ocean. Such intraseasonal disturbances seem to be much more common during equatorial/tropical south Atlantic warm events. This appears to lay the basis for a connection between the state of the equatorial Atlantic, east African rainfall events and possible MJO developments in the Indian Ocean.

Teleconnection structures across the Indian Ocean were also studied for March, April and May. The ENSO convection anomaly pole over the western Pacific during March and April is weaker and further east relative to that found in October-December. This weakness and location seem to lead to an absence of atmospheric teleconnection between ENSO and Indian Ocean circulations into east Africa. However, in May, the western Pacific pole of ENSO convection anomalies is once again stronger and further east, and again teleconnects across the Indian Ocean with opposite convection pole centered over east Africa/Western Indian Ocean. These results shed light on why ENSO connections to east African rainfall are weaker in the Long Rains of March-May.

Contribution to International Climate Outlook Forums

Scientists at CIMMS again contributed strongly to Climate Outlook Forums in West, East and Southern Africa (and one in the Caribbean). Verification of all the first round of African forums was supervised by CIMMS scientists and results were presented at the forums, demonstrating that the levels of skill achieved by all forum map forecast products was well above the level expected by a random forecast strategy, and quite consistent with the levels of skill achieved by forecast models (empirical and numerical) on past data. CIMMS scientists continued to contribute to further regional training workshops in Africa for National Meteorological Service personnel. These included a further workshop for West Africa and first workshops for East and Southern Africa, such that almost every National Meteorological Service south of the Sahara now has a member of staff conversant with regional climate prediction and a set of seasonal rainfall prediction models for regions within their country based on sea-surface temperature predictors. The training included verification methods as well as prediction methods.

Regional Climate Impacts: Information and Forecasting for Users – The Case of Malawi

As part of the Malawi Environmental Management Project (a World Bank sponsored initiative), scientists at CIMMS again visited the Malawi Meteorological Department headquarters to continue to develop capacity for user-tailored regional climate prediction systems. Meetings were held with users in Malawi including the Tea Association, the National Insurance Company, the Regional Water Board, and World Vision International. Impacts of climate anomalies on their activities were discussed and pilot climate services, incorporating climate predictions and real-time monitoring through the season, were defined for each of these customers. This experiment with the commercial sector provides the perfect complement to the RANET project aimed at the small-scale farmer.

Variability of West African Disturbance Lines

A large set of historical daily rainfall data from the West African Sahel is being used by a CIMMS graduate student to document the role that variations in the characteristics of individual rainfall disturbances have played in the long-term decline and the large interannual variability of seasonal rainfall totals in that region. The acquisition of daily rainfall data through 1998 is allowing this analysis to be extended to the most recent years possible. In collaboration with the TAMSAT group at the University of Reading, satellite rainfall estimation techniques have been applied to verify the results obtained from the raingage data alone, and have shown good agreement with those results in the years when data from both sources are available. Some of this work was presented at the Abidjan ’98 conference in Abidjan, Côte d’Ivoire, in November 1998 and at the WAMAP Workshop in Dakar, Senegal in June 1999.

Summertime Moisture Budget over the Midwestern U.S.

Land-atmosphere interactions are central to our environment. The hydrological cycle is a manifestation of this interaction with the surface fluxes of precipitation and evapotranspiration being balanced by the large-scale surface fluxes of the overlying atmosphere. The hydrological cycle over the Midwest during its growing season is particularly important due to its strong ties to the crop status for this world-renowned, unirrigated agricultural region and its variability. In FY98, to augment our study of the summer hydrological cycle over the Midwest, scientists at CIMMS acquired and analyzed the solar radiation data over our study area. This was done to quantify our theories about the role of solar radiation forcing of evapotranspiration (and recycling) over different timescales during wet and dry summers. Early in FY99, an article was submitted to Science, “Role of Local Evapotranspiration for Growing Season Precipitation and Crop Yields over the Midwestern United States”.

Expansion and Analysis of the Comprehensive Pacific Rainfall Data Base

Taylor’s Atlas of rainfall data has been combined with the Comprehensive Pacific Rainfall Database (CPRDB), totaling 72 stations of combined data, in work that has been performed at OU. An exploratory data analysis was done on this dataset of monthly rainfall. Data from these 72 station records extends into 1997 and goes back in most instances prior to 1950, and in a few instances prior to 1900. We first investigated simple least-squares regression analyses with the data for 20 year time periods (a minimum of 20 years was chosen due to the 2-5 year ENSO cycle), a 50 year period (1947-1996), and for long station records with a sufficient amount of data. Trend results for the 72 stations were also analyzed spatially. We also explored techniques in filtering (mostly in the frequency domain with relatively long, unbroken datasets) to better assess rainfall trends. Also explored were Neural Network techniques for time series prediction and cluster analysis.

The Taylor’s Atlas data was also digitized for distribution to users. Basic data quality checking was performed. A more thorough quality check of the data is underway.

We are also currently performing a thorough quality check of the CPRDB rainfall data. This rainfall data is daily and goes back to 1971 for 359 stations. Included in the quality checking analysis is:

Neighboring station data with significant discrepancies.
High/low percentage of zero rainfall amounts
Unusually high rainfall amounts
Comparison of rainfall to ENSO events and season
Identification of possible rainfall errors due to the specific instrument used.

We have also recently completed assembling a database of Pacific hurricane data, which will be used to possibly check unusually high rainfall amounts (as well as suspicious low/zero amounts) during a tropical storm.

The Surface Reference Data Center of the Global Precipitation Climatology Project

The Surface Reference Data Center (SRDC) has been successfully transferred to the Environmental Verification and Analysis Center (EVAC) at OU. Transfer was initiated with the help of Dr. Alan McNab during July 1998. The importance of validation of GPCP satellite rainfall products necessitated a sharpening of the role of SRDC. As a consequence of this need a meeting among GPCP principals and SRDC personnel was held during the January 1999 meeting of the American Meteorological Society in Dallas, Texas. An important result of this meeting was the recognition of the need for a medium through which comments and suggestions could be solicited from the research community at large as to how the SRDC might meet the community’s verification requirements. It was decided to experiment with a newsletter and to develop a web page to include a bulletin board. Both the newsletter and the web page are designed to publish researcher’s comments and preliminary research related to GPCP activities, particularly verification activities. An added benefit of the newsletter is an increased awareness of GPCP activities among the research community at large.

During the January meeting, specific objectives were set out for the SRDC to accomplish before May 1999. These objectives were:

Conduct validation exercises using Dr. George Huffman’s daily satellite rainfall product over Oklahoma using the Oklahoma Mesonetwork;
Produce and publish the ‘Validator’ newsletter;
Set up a data conduit for U.S. cooperative network data from the National Climate Data Center (NCDC) to the SRDC;
Solicit suggestions for the functionality of the SRDC from GPCP data producers, and
Start up routine validation of GPCP products.

To date, the SRDC has met and exceeded these tasks on time. While accomplishing these objectives we discovered several problem areas which, when dealt with, will considerably streamline our future operations. We would like to suggest that a standard data archive format be established for GPCP satellite algorithm output. This will facilitate their access by SRDC personnel. We would also like to request our non-U.S. GPCP participants to ask their respective local data archive organizations if sample raingauge data sets could be made available for the SRDC. It is important to note the entire data sets are not required, only samples from relatively dense networks. This will allow the verification of GPCP products while not overburdening the local data organizations with the task of extracting large data sets. Given the current funding for the SRDC it has also become quite apparent that the SRDC will only be able to conduct verification of select GPCP products. However, we are making every effort to produce, and place on-line, reference data and their associated error statistics. Thus, individual researchers should be able to access and download reference data meeting their error and scale criteria and conduct verification exercises. Another approach is on-line comparisons of satellite and surface observations. However, since validation efforts are frequently very specific to the algorithm being tested, it may be difficult to automate such a process.

Another problem that has become obvious is the time offset between surface raingauge and satellite rainfall estimates at the daily scale. Several possible remedies to this problem include interpolation of reference data and/or the satellite estimates, production of satellite estimates using image data corresponding to the surface time of observation and the use of hourly raingauge estimates to match the daily satellite observations.

Standard error estimates have been constructed for 1 x 1-degree boxes over Oklahoma. In addition the signal to noise ratio has also been constructed for each box. These statistics will allow researchers to pick and choose rainfall box averages according to their particular error requirements. These data are on the web and are freely available.

Water Vapor Measurements and Related Field Studies

The temperature and relative humidity calibration chamber that was purchased by the ARM Program is now in operation at the SGP CART Central Facility through the efforts of scientists at CIMMS. The chamber has proven useful in testing sensors used by the program and has detected instrumentation errors that otherwise would have gone undetected or would have required manufacturer testing to find. A calibration schedule has been developed for the field test sensors used by ARM technicians and these sensors are now calibrated every two weeks in the chamber. This will ensure that accurate field tests are performed, with the ultimate goal of improving data quality. Calibration of individual radiosondes is now also feasible using the calibration chamber. Sonde data ingest equipment is available at the Central Facility and will be utilized for these experiments. Additional details regarding the magnitude of the sonde relative humidity contamination are now available and a carefully constructed test procedure will be implemented to examine if surface calibrations can improve sonde measurements.

A contamination correction developed by sonde vendor Vaisala has been applied to radiosondes launched by the ARM Program in 1997 and 1998. Over 1,000 sondes were launched from the Central Facility in 1998, and the magnitude of the correction and its impact on precipitable water vapor and the radiative fluxes are being examined in collaboration with NCAR scientists. Preliminary results indicate a large seasonal variability in the magnitude of the correction (weaker in the winter and fall and larger in the spring and summer). Surface longwave fluxes were modified by as much as 7 W/m² in the spring and summer. The relative impact of the water vapor correction on the radiative fluxes appears to be larger when the correction to precipitable water vapor is smaller. These results are similar to those obtained in the re-analysis of TOGA COARE sondes, except that the seasonal variability was not evident. Seasonal variability could have important implications from a climatic point of view and will be examined further.

Chilled mirror dew point measurements at the Central Facility sonde launching system are being collected by the ARM site data system and are examined routinely for accuracy. Moisture measurements from two different instruments are being compared to assess the ability of the chilled mirror to make long-term unattended measurements. Results to date indicate that mirror cleaning is required approximately every two to four weeks. This does not represent a significant limitation if the systems are deployed in locations where such maintenance is feasible. However, it is unlikely that these systems would be appropriate for use in totally remote situations. The chilled mirror systems appear quite capable of making the measurements for which they were designed, namely NIST traceable observations during IOPs and other short-term experiments.

Investigation of Temporal and Spatial Variations of Broadband Surface Albedo across the ARM Southern Great Plains Area

In the fall and winter of 1998, 21 ARM SGP CART extended facilities (EF) deploying the SIRS (Solar and Infrared Radiation Station) platform were visited. Vegetation characteristics were noted for each EF. Since local albedo calculations will be used for climatological analysis and comparison with satellite albedo estimates, the vegetation at a site must be representative of an appropriately large area surrounding the measurement location. Therefore, each EF was assigned a rating based on this representativeness. Of the 21 sites that were rated, six sites were determined to have a high degree of representativeness, and will be studied further.

Soil Moisture from the Oklahoma ‘Moistnet’

Coordination with the USDA/ARS and Oklahoma Mesonet soil moisture measurement networks for similarity in calibration and validation procedures is ongoing at OU. This effort was facilitated by an “Oklahoma Soil Moisture Summit” held at the USDA/ARS Grazinglands Research Laboratory in March 1999. Delivery of ARM SGP soil moisture data to GCIP under present contracts has ended.

ARM Southern Great Plains Site Observations of the Smoke Pall Associated with the 1998 Central American Fires

Drought-stricken areas of Central America and Mexico were victimized in 1998 by forest and brush fires that burned out of control during much of the first half of the year. Wind trajectories at various times during the episode helped transport smoke from these fires over the Gulf of Mexico and into portions of the U.S. Visibilities were greatly reduced during these favorable flow periods from New Mexico to south Florida and northward to Wisconsin as a result of this smoke and haze. Public health advisories and public information statements were issued by agencies such as the NWS in May in Oklahoma. This event was also detected by the unique array of instrumentation deployed at ARM’s SGP CART and by sensors of the Oklahoma Department of Environmental Quality/Air Quality Division. Observations from these measurement devices suggested elevated levels of aerosol loading and ozone concentration over the CART during May 1998 when flow conditions were favorable for the transport of the Central American smoke pall into Oklahoma and Kansas.

Trajectories ending at the CART Central Facility in May indicated that May 13-15 and May 17-19 should have been particularly good days for observing the smoke pall in Oklahoma and Kansas. Indeed, analyses from the Aerosol Observing System, Raman lidar, Solar and Infrared Radiation Stations, Multi-Filter Rotating Shadowband Radiometer, Cimel Sunphotometer, Micropulse Lidar, a University of Utah Polarization Diversity Lidar, and condensation nuclei counters on the North Dakota Citation aircraft showed elevated levels of aerosols. Additionally, ozone monitors of the Oklahoma Air Quality Division showed concentrations on May 11 in excess of EPA clean air limits of 0.080 parts per million, although the weather conditions that day (not excessively hot or stagnant) were not conducive to producing ozone, suggesting transport into the region. Interestingly, the smoke pall intrusion of May 13-15 was generally limited to the lowest two kilometers of the atmosphere, while the later intrusion, which occurred after a strong cold front had cleaned the atmosphere late on May 15, showed elevated aerosols up to six kilometers. This multi-collaborator observational study, coordinated at CIMMS, particularly showcased a new capability for retrieving aerosol extinction profiles from Raman lidar data.


Work is underway at NSSL in two areas. A new “monsoon” index has been developed that identifies better the region of southwest North America that experiences a climatically distinct summer rainy period. The index keys on both the precipitation during July and August and on the lack of precipitation during the prior two months. The characteristic dry period has not historically been incorporated into such indices. This new index illustrates that the summer monsoon is most distinct from northwestern Mexico and southeastern Arizona west-northwestward into the California deserts and extreme southern Nevada. Prior indices have indicated the dominant monsoon affects to extend from northwest Mexico across southeastern Arizona and most of New Mexico.

The synoptic patterns that actually lead to the inverse correlation of summer precipitation between the central Plains and northwestern Mexico are being investigated for several years where the wet/dry and dry/wet signals were very strong for climate zones in eastern Nebraska and southeastern Arizona.

Plans – FY 00

‘Al Moubarak’ and Moroccan Precipitation

Development and issuance of ” Experimental Precipitation Prediction Statements” (EPPS) by scientists at CIMMS for each season will continue. Throughout the period, we will draw on the work based on “Characterization and Seasonal Behavior of Al Moubarak” to issue such EPPS. A main objective is to substantially increase the EPPS skill so as to yield results that would be of improved practical assistance to several critical Moroccan sectors (e.g., hydraulics, agriculture).

Capabilities will be developed to use the ECHAM4 model in a real time prediction mode in support of the preparation of EPPS, and continue using ARPEGE-Climat model in a real time prediction mode. Consistent with the original 1994 proposal, this project component will be developed through the conducting of additional ECHAM4 experiments motivated exclusively by the desire to understand the variability of Al Moubarak and Moroccan precipitation. We expect a substantial improvement in the EPPS skill using ECHAM4 in a real time prediction mode.

Capabilities of ECHAM4 and ARPEGE-Climat models to predict general patterns of Moroccan precipitation for different time-scales will be assessed. The objective of such work is to maximize the EPPS skill for a given Moroccan region and season.

Statistical studies will be started on the characteristics and predictability of agriculturally important start (October-November) of the precipitation season. The primary results are expected to substantially expand our existing Al Moubarak-based statistical and dynamical framework for Moroccan precipitation prediction, including the precipitation season, when Al Moubarak control can be weak.

We will continue our investigation of the relationship between the weather systems and Moroccan precipitation patterns. This investigation will focus on the interannual variability of the weather systems and rainfall event data, the distribution of the rainfall event data when it is stratified by different cyclone parameters and the reasons behind the NAO-control and non-control through documentation of our selected case studies.

Plans for this year also include publishing at least two journal articles on past research from the Al Moubarak project and participation in the AGU Chapman conference on the North Atlantic Oscillation. Dr. P.J. Lamb agreed to serve as a member of the Program Committee.

Predictability of East African Rains

Research studies on east African rainfall variability and predictability will continue at CIMMS, including a University of Nairobi Ph.D. Dissertation by Charles Mutai. The aforementioned findings will be further developed to form hypotheses for relationships between rainfall and atmospheric moisture flux/sea surface temperature anomalies in both the Indo-Pacific and tropical Atlantic regions, particularly in October-November.

Contribution to International Climate Outlook Forums

The contribution of CIMMS scientists to the climate outlook forums is now expected to be made more through e-mail and Internet communication. The major progress made in the region in the last two years makes this feasible. More advanced training will be undertaken at a workshop to be held in October/November 1999 at the University of Oklahoma for African and Central/South American participants. This training will cover more details of the science of climate prediction, and consider more closely regional climate impacts and their relation to such features as dry spell lengths. CIMMS scientists will also play a role in developing rural applications of regional climate predictions through the USAID-funded RANET project with ACMAD, that is being routed through CIMMS. This project will make wind-up radio technology available to villages for receipt of meteorological information relevant to agriculture. This will provide the impetus for the development of tailored products for the rural community.

Regional Climate Impacts: Information and Forecasting for Users – The Case of Malawi

Scientists at CIMMS will assist Malawi Meteorological Service in developing regional forecast and monitoring products for pilot customers during the 1999/2000 rainy season.

Variability of West African Disturbance Lines

The completion of the update to the West African rainfall analysis by a CIMMS graduate student will allow for the submission of a manuscript to an appropriate journal in the coming year.

The Surface Reference Data Center of the Global Precipitation Climatology Project

It is becoming increasingly obvious that, as the atoll raingauge data are becoming widely utilized, a comprehensive quality control system needs to be implemented at the SRDC. This has already started and we hope to have an operational system established by the end of August. Work is continuing on the comparison of the mesonet averages with the GPCP product. In addition, work that was started using Pacific data has been put off until the quality control system is established.

West African Monsoon

North American Precipitation Variability

Water Vapor Measurements and Related Field Studies

A collaborative project with the Oklahoma Mesonet and CIMMS has begun to examine the effects of pyrgeometer dome heating on calculated longwave radiative fluxes. An Eppley Precision Infrared Radiometer (PIR) has been modified to incorporate three thermistors to measure the dome temperature instead of the usual single thermistor. There has been speculation that the temperature variation across the PIR dome could result in significant errors in the calculated longwave flux. Data are now being analyzed. Preliminary results indicate that the temperature variation across the PIR dome is less than 0.5° C. This translates into an error in the calculated infrared flux of approximately 7 W/m². Ventilation of the PIR sensor may reduce the temperature error to less than 0.3° C, with an associated infrared flux error of less than 5 W/m².

Investigation of Temporal and Spatial Variations of Broadband Surface Albedo across the ARM Southern Great Plains Area

An albedo study has started at OU using the six sites found to be the most representative of their surroundings. Data analysis of the sites is still in a preliminary stage. A computer program is being written to locally organize and process netCDF-formatted data. An investigation of SIROS and SIRS data quality for 1997-1998 has also commenced.

ARM Southern Great Plains Site Observations of the Smoke Pall Associated with the 1998 Central American Fires

A manuscript will be submitted by CIMMS to the Bulletin of the American Meteorological Society in fall 1999 on the observational aspects of this work. The paper will emphasize the unique observational capabilities of the SGP CART site and its unlimited potential. More work is planned to study the nature and distribution of the aerosols observed during the smoke event.


During the coming year work will continue at NSSL on the development of a monsoon index and preliminary results will be presented at the NOAA Climate Diagnostics Workshop in November 1999.


Progress – FY 99

El Nino and its Impacts on Society

Agricultural Production and Economic Decision Models

Plans – FY 00

Engineering Research Center for Natural Hazard and Disaster Research

Floods are the most devastating of all weather-related hazards in the U.S. On average over the past 30 years, 139 lives are lost each year due to flooding, with the death toll in recent years rising to over 200 per year. In Texas alone, 28 people died in floods during 1998. Though often drawing more media attention, lightning, tornadoes, and hurricanes caused an average of only 87, 82, and 27 lives lost per year, respectively. During the same 30-year period, property damage amounted to $1 billion annually with recent figures averaging over $2 billion per year due to floods.

Considering that

Flooding is a $2 billion-per-year problem (considering damages and lost revenue)
Big-government solutions are no longer feasible or even available
Historically, the record shows increasing numbers of extreme storm events, such as the Mississippi floods of 1993
Development in urban areas has increased the flooding impact of storms
Few alternatives exist for alleviating flood damages in the built environment; and
The NRC and NSTC both agree that flooding disasters is a national priority,

an NSF Engineering Research Center (ERC) has been proposed at OU that would develop the only feasible alternative left– an improved information system for stakeholders to make their own decisions.

Specific ERC goals are to

Develop flood prediction, warning and design methodologies using radar estimates of rainfall for river basins and urban watersheds,
Involve industry/practitioners in specification of the engineered system, and provide needed education them in the potential uses of this technology, and
Develop the engineered system capitalizing on recent developments in radar and mesonet technology, atmospheric/hydrologic modeling, and information management techniques.

The strategic plan of the proposed ERC will focus on an investment in the development of technology to reduce societal costs of weather-related disasters. The engineered system consisting of a System for Information, Measurements, and Modeling, will be referred to as SIMM. The end-to-end engineered system will begin with radar/mesonet rainfall and other meteorological measurements ingested into atmospheric/hydrologic models to predict floods, and will filter customizing and delivering disaster information to stakeholders.

Modern Meteorology and the Insurance Industry

Agricultural Production and Seasonal Climate Prediction Models


Progress – FY 99

Open Systems Radar Data Acquisition (ORDA)

The objective of this NSSL work is to design and develop signal processing and automatic calibration algorithms for the WSR-88D radar. Activities during the past year have included analyzing and documenting signal processing functions currently operational in the WSR-88D legacy system, and designing, implementing, and testing the basic set of signal processing functions on the new ORDA. The latter included most of the signal processing functions performed in the legacy Hardwired Signal Processor (HSP) and Programmable Signal Processor (PSP). The code was developed for the new WSR-88D host computer running the real-time operating system VxWorks (WindRiver Systems) and the array of signal processors (SHARCs) running on the Mercury Computer Systems operating system (MCOS).

Radar Quality Improvements

This work at NSSL investigates and implements procedures to improve the quality of WSR-88D radar data. Activities have included the investigation of regression filters as ground clutter filters in the context of variable PRT, and comparison of the performance of regression filters used as ground cancellers with the performance of WSR-88D ground clutter filter algorithms.  A paper with these findings was published in the Journal of Atmospheric and Oceanic Technology. An investigation was begun of an efficient technique to reduce estimation errors in the processing of polarimetric radar data, including the estimation of spectral moments. Very promising results have been obtained to date.

AWIPS-Based Severe Weather Training

A milestone of the modernization of the NWS was reached in 1999 when for the first year all NWS offices were able integrate AWIPS capabilities into real-time forecasting and warning operations. In severe weather operations, the development of AWIPS involves a paradigm shift from a single sensor approach (primarily radar-based) to an integrated sensor approach (radar, satellite, lightning and other data). In order to respond to the high demand for severe weather training because of both the AWIPS paradigm and hardware/software platform shifts, OSF/OTB personnel, in cooperation with COMET, have been (1) developing a limited Displaced Real-Time (DRT) capability in AWIPS which has been used to simulate “live” warning exercises in decision making workshops, (2) converting baseline radar training to AWIPS format, and (3) developing display and archive capabilities for AWIPS applications development.

Severe Weather Training Applications

To help reduce the false-alarm rate and increase the probability of detection and lead times for severe weather warnings, development continues at OSF/OTB on how to most effectively integrate current observational data into the warning process. Important activities have included:

Participating in developing current tornado warning guidance for operational forecasters,
Evaluating the WSR-88D’s ability to detect tornadogenesis features and the limitations of its viewing angle,
Developing more web-based tutorials on issuing warnings for pulse storms, downburst prediction, and other types of severe weather,
Developing joint radar and satellite methods for determining thunderstorm characteristics such as updraft evolution and the thunderstorm equilibrium level,
Documenting the importance of boundary and updraft evolution as seen by radar and satellite on significant events such as the Jarrell, TX, tornado on May 27, 1997, and
Investigating mesocyclone and tornado evolution and algorithm performance for close-range events such as the May 3, 1999 tornado outbreak in Oklahoma.

Analysis of Tornado Detection Algorithm Performance

This OSF Applications Branch research seeks to determine the frequency of TVS (Tornado Vortex Signature) and ETVS (Elevated TVS) false alarms when storms pass directly over a WSR-88D.  A random selection of 20 cases from NSSL Level II archives was used for this analysis.  All cases have at least one occurrence of a tornado, recorded within or slighter greater than 25 km of the radar vicinity.  The analysis compared the NSSL Tornado Detection Algorithm (TDA) performance in WATADS 10.0 with respect to the deactivation of WATADS 9.0’s Build 10 dealiasing algorithm. Results were tabulated, and it was observed that TDA performance in WATADS 10.0 is more sensitive to detections. But, the tradeoff is a higher false alarm rate (FAR) for ETVS of 9.38% and TVS of 10.36%, as compared to WATADS 9.0 FARs for ETVS of 3.83% and TVS of 3.31%.  In addition, the WATADS 10.0 dealiasing algorithm enhanced its former WATADS 9.0 dealiasing algorithm by filling ‘good’ gates/bins to accurately reflect a shear.

Scoring Detection Algorithms

The OSF Applications Branch is coding, debugging, and developing a data structure leading to an automated process for scoring output of Doppler radar algorithms.  The Scoring Algorithm Detections program was written in the C programming language.  The program provides comparative scoring and will be used to evaluate candidate algorithms for inclusion in the WSR-88D program.  This is an important step in developing a structured algorithm improvement process.  During the prototyping stage, candidate WSR-88D severe weather algorithms developed through technology transfer by the NSSL will be compared with baseline WSR-88D algorithms.  Algorithm data files are obtained from WATADS Version 10.1.  The new program compares the number of hits and misses of each selected case, and calculates several established scoring parameters (POD, FAR, CSI, and HSS).

Generation of Terrain Hybrid Scan and Occultation Data Files in Preparation for the Snow Accumulation Algorithm

In preparation for implementation of the Snow Accumulation Algorithm (SAA), it was necessary for the OSF Applications Branch to generate site-specific terrain hybrid scan and occultation data files.  These files account for and remove local terrain that otherwise would cause unrealistically high returns at low elevation angles.  We used UNIX batch processing to deal with such large volume of data sets.  All data files were annotated, resized, and compressed in the Graphics Interchange Format for future reference.  During the process, we discovered ImageMagick (vice XView) for ease of automating various image file manipulations.  (ImageMagick is a freeware from the World Wide Web and has a robust collection of tools and libraries to read, write, and manipulate an image in any of the more popular image formats.)

Investigation of WSR-88D Level II Data:  Severe Weather Event Detection

This work by the OSF Applications Branch is follow-on to a previous similar project and began late in the fiscal year.  Well-truthed Level II data were mined for structure in the velocity and spectrum width fields. While this approach has been tried previously, the robust Neural Network technique was applied with limited success and remains a work in progress.

The use of CODE at the WSR-88D OSF

The planned transition from legacy hardware and software to a more open environment offers numerous opportunities for improvement to the WSR-88D radar system.  One such improvement is CODE.  Still under development, CODE is envisioned as a process for facilitating introduction of new and improved algorithms and products into the WSR-88D system.  A recent survey indicates that by year 2000 there could be as many as 60 algorithms vying for inclusion in the WSR-88D baseline.  In addition, more than 70 NWS forecast offices, universities, and DoD organizations are using the WSR-88D Algorithm Testing and Display System (WATADS) to improve existing algorithms through post-event studies.  In addition, CODE promises to open the algorithm development process to more researchers.

The objectives of CODE center on standardization, reuse of software, ease of including new software in the baseline, ease of data transfer, and common functionality.  CODE will be a software system and a development environment running on a workstation separate from the Open System Radar Product Generator (ORPG.)  It will behave much like the current WATADS with notable additions, namely, the ability to add an algorithm to the baseline, enhance existing algorithms and products, and develop new algorithms with minimal software re-engineering using modular common functionality.  With CODE, it will be possible to test algorithms using archived or real-time data.  When applications are created in CODE, both the developer and the OSF will be able to run the applications on a common system during initial testing and validation.  Applications thus introduced can be integrated into the ORPG much faster because the need for re-engineering is limited. Work locally involves OSF and NSSL scientists.

Polarimetric Radar Developments

Regular weather observations were conducted at NSSL using NSSL’s Cimarron dual-polarization radar. Two-dimensional video disdrometer data obtained during the MEaPRS campaign have been examined and compared with the corresponding radar polarimetric data. The comparison revealed the capability of the dual-polarization radar to roughly estimate median size of raindrops and to discriminate between rain types characterized by different drop size distributions. A new scheme for automatic identification of different precipitation types with polarimetric radar has been developed and tested. The results of the classification for four storms have been verified and confirmed using raingage data, 2D video disdrometer data, and ground reports of hail.

A scheme for automatic hydrometeor classification was refined. In situ aircraft data for the case of May 17, 1995 were analyzed in detail to further improve the existing version of the classification algorithm.

The NCAR S-POL radar data collected during the 1998 TEFLUN experiment in Florida have been examined and compared with similar data sets obtained with the Cimarron radar in central Oklahoma.

A relation between negative Kdp signatures and tilted crystals with high degree of common alignment in electrically charged zones of the cloud was further clarified using the data from the National Lightning Detection Network and new polarimetric variables from the NCAR S-POL radar.

Development of a Multiple Pulse Repetition Frequency Dealiasing Algorithm (MPDA)

Scientists at NSSL should finish a final version of the MPDA at the end of 1999. The MPDA is a new WSR-88D scanning strategy that combines multiple scans of data into a single product that reduces range folded data and velocity dealiasing errors. Several new Velocity Coverage Patterns (VCPs) have been developed based on the results of the 1998 studies. Data have been collected with these VCPs and the MPDA is currently being modified to further enhance its performance.

Scale Filtering of Radar Data

Two ways to speed up the identification of large-scale features in radar images have been developed at NSSL using 1) a Fast Fourier Transform (FFT) 2) a computationally efficient spatial filtering technique.  The FFT work has been accepted by the Journal of Atmospheric and Oceanic Technology and the spatial filtering technique has been submitted to IEEE Transactions on Image Processing.

Developing a V-notch Detection Algorithm in Satellite Images

It has been determined at NSSL that in the absence of context, there is very little to distinguish small V-notches from any other spot in satellite images. Work is therefore in progress to determine context from the satellite images.

Bounded Weak Echo Region Algorithm

The algorithm to search for bounded weak echo regions (BWER) has been changed at NSSL in a top-down manner and a genetic algorithm has been developed to automatically tune the BWER algorithm for use by a Neural Network for vortex detection.

County Warning Area Algorithm

NWS personnel are sometimes deluged with radar information on storms within their County Warning Area (CWA). Some offices have several WSR-88Ds available for inspection of storms. In warning operations it can become overwhelming to sort through all the data and chose the radar with the “best” view of the storm. Scientists at NSSL are developing the Multi-Radar Ingest and Analysis System (MuRIAS) to provide a means of ingesting and sorting the WSR-88Ds across a given CWA. This algorithm ranks storms based on severity and alerts the meteorologist to the particular radar it found that provides the best view.

Cell and Area Tracking

The goal of this NSSL project is to integrate the Storm Cell Identification and Tracking (SCIT) algorithm and the Scale Separation and Correlation Tracking (SS/CT) algorithm into a single multi-scale precipitation tracking and forecast package.  Code for the SS/CT algorithm was obtained from MIT/Lincoln Laboratories (LL) and implemented at NSSL.  This involved modifying the code to allow it to ingest NSSL’s RADS formatted data.  Initially, the algorithm was ingesting 0.5 degree base reflectivity, but later this was changed to composite reflectivity and then to VIL.  The MIT/LL version of the software uses VIL as the input field for tracking precipitation.  Output from the MIT/LL version was also obtained from MIT/LL to compare the NSSL version with the MIT/LL version to ensure that the NSSL version was implemented properly.  The MIT/LL output was converted from its native format to NSSL’s RADS format to facilitate a comparison.  This comparison is underway.

Clear-Air Adjoint Method Wind Retrievals from the WSR-88D Radar

A Clear-Air Adjoint-Method (CAAM) of wind retrieval from WSR-88D radars was completed at NSSL.  The CAAM is utilized primarily in clear-air or precipitation-free regimes using WSR-88D radars for several reasons:  1) to reduce the complexity of the equation set used in the retrieval process; 2) because past wind retrieval studies have primarily examined precipitation events only; and 3) retrieval studies from WSR-88D radars have only just recently been carried out.  CAAM is based upon the equation set used in Xu’s simple adjoint wind-retrieval method.  This technique uses predictive equations for reflectivity and radial wind while utilizing a cost function containing least-squared differences of observed and estimated variables including radial wind, reflectivity, time-mean radial wind, and divergence.  Because of CAAM’s application to precipitation-free events using WSR-88D radars, additional improvements over Xu’s simple adjoint system were needed.

These improvements were examined using both idealized and real-data cases. Within the idealized experiments, various configurations of atmospheric structures, including constant wind flow and a propagating boundary regime, were investigated.  The idealized cases revealed the dependence of accurate retrieval solution on the existence of gradients within the observation fields.  Whenever the gradients were of insufficient magnitude across any portion of the radar domain, then the non-zero first guess field and the utilization of two fitting terms (time-mean radial wind and divergence terms) within the cost function became very important in producing accurate retrievals.  This importance was also underscored in the real-data cases as well and points to the fact that regions of diffuse reflectivity gradients can exist often in precipitation-free environments.

Overall, results indicated a strong importance on the first-guess provided for the retrieval technique.  Instead of using a zero first-guess as in past retrieval studies, an initial wind field was provided using either the  Velocity-Azimuth Display (VAD) or Volume-Velocity Processing (VVP) techniques.  In addition, the use of divergence calculations from the VVP method improved the accuracy of CAAM.  Model produced divergence calculations were constrained toward the VVP-provided divergence calculations through the cost function.


Studies have been completed at NSSL on the advantages of employing data from multiple WSR-88D radars to determine better the vertical structures of convective echoes. The results are important since the trends of radar-measured cell characteristics with time (i.e., the trends plots that are used in NWS operations) can be extremely non-representative due to the large uncertainties in feature heights that are unavoidable with the fixed volume scan modes used for the WSR-88D. These studies have been completed and reported in the formal literature.

SWAT-V Severe Weather Algorithms for the WSR-88D

The Severe Weather Warning Applications and Technology Transfer – Vortex Focus Group (SWAT-V) team members at NSSL continued their primary mission of developing severe weather applications primarily for the WSR-88D radar, with an emphasis on mesocyclone and tornado vortex signature applications, and transferring technology and knowledge to the NWS and FAA. Activities and projects are listed below:

Severe Storms Analysis Package (SSAP)

Development, maintenance and enhancement continued for the Severe Storms Analysis Package (SSAP), including the following meteorological algorithms that have been evaluated off-line and in real-time NWS forecast office operational tests:

Tornado Warning Applications:
Mesocyclone Detection Algorithm (MDA)
Tornado Detection Algorithm (TDA)
Vortex Detection and Diagnosis Algorithm (VDDA)

Severe Storm Warning Applications:

Storm Cell Identification and Tracking (SCIT) algorithm
Hail Detection Algorithm (HDA)
Damaging Downburst Prediction and Detection Algorithm (DDPDA)

Applications related to both:

Near-Storm Environmental (NSE) algorithm

A manuscript entitled “Evaluating the performance of WSR-88D severe storm detection algorithms” was published in the AMS journal Weather and Forecasting June1998 issue.

Mesocyclone Detection Algorithm (MDA)

NSSL continued to test the Mesocyclone Detection Algorithm (MDA) on an ever-expanding database consisting of a variety of tornadic and non-tornadic supercell cases. The MDA allows for the detection of storm-scale vortices of various sizes and strengths, and classifies them into a number of different vortex types (including Mesocyclone, Low-Topped Mesocyclone, etc.). Trends of vortex attributes are also computed. The database now contains over 58 individual storm event days from a variety of locations across the country. The database includes over 400 tornadoes with over 390 hours of radar data. This represents 18 more events analyzed since FY97.

With each detection, a vortex is also diagnosed using a Neural Network to determine the probability of associated tornadoes or severe weather.

A Bounded Weak Echo Region (BWER) algorithm was developed in FY98 to detect and classify (using probabilistic “confidence” factors) weak-echo vaults within severe thunderstorms using WSR-88D reflectivity data. The BWER data as been integrated with the MDA (and TDA) data with 43 of the 58 cases. These combined MDA/TDA/BWER data were used to develop a new NN, and to develop statistical analyses for the 1999 NWS Tornado Warning Guidance documentation.

Enhancements made to the MDA include the addition of Near-Storm Environmental (NSE) algorithm parameters (as derived from RUC mesoscale model grids). Over 100 NSE parameters have been incorporated into the MDA (and TDA). Work is currently underway with a statistical analysis of the data. These data will also be used to train a new NN.

SWAT-V has been developing a special Web-based case study collection of a variety of storm types collected nationwide. This web page depicts a wide range of the types of tornadic storms that have been observed with the WSR-88Ds, provides detailed discussions, images, and algorithm output evaluations.

The NSSL spent a great amount of effort to add tracking and trending functionality to the WSR-88D Mesocyclone Algorithm. It is planned to add this functionality in the System for Convective Analysis and Nowcasting (SCAN) deployed at the Tulsa forecast office.

A manuscript entitled “The National Severe Storms Laboratory Mesocyclone Detection Algorithm for the WSR-88D” was published in the AMS journal Weather and Forecasting June 1998 issue.

Tornado Detection Algorithm (TDA)

In cooperation with the OSF/Applications Branch, the Tornado Detection Algorithm (TDA) was implemented into the WSR-88D Build 10 Radar Products Generator (RPG) in the fall of 1998. The majority of the TDA work was spent on testing the TDA and preparing it for this implementation.

A manuscript entitled “The National Severe Storms Laboratory Tornado Detection Algorithm” was published within the AMS journal Weather and Forecasting June 1998 issue.

Vortex Detection and Diagnosis Algorithm (VDDA)

SWAT-V developed some of the groundwork to merge the NSSL TDA and MDA into a single algorithm called the Vortex Detection and Diagnosis Algorithm (VDDA). The VDDA will consist of some techniques from both the MDA and TDA, as well as some new techniques. These techniques will also include the integration of data from other sensors.

The reasons for the development of the VDDA are several-fold. First, from the analysis of many supercell cases in the past few years, we have discovered that there are a variety of storm-scale vortices that can be tornadic in supercells, and they range in size from TVS-like to mesocyclone. Second, we would like to share all the analysis techniques of the two vortex-detection algorithms (MDA and TDA), such as the vertical and time association techniques. Also, the integration of data from other radar-based algorithms and other sensors (such as NSE data) will provide a more thorough analysis of the vortices.

We have also developed a database of over 16,000 simulated WSR-88D data consisting of vortices of varying, sizes, strength, and range from the radar. The MDA and TDA were initially tested on these data to determine the strengths and weaknesses of each algorithm, and to help develop new two-dimensional feature detection and diagnosis techniques. We have also developed a data set of gust fronts, mesocyclones with embedded TVSs, and mesocyclone with rear-flank downdrafts for testing. Testing will also incorporate the new detection and diagnosis techniques described earlier.

Unfortunately, the testing of the simulated database did not reach maturity, as there were no easy ways to display the advanced algorithm output using existing NSSL radar algorithm display software. Advanced display software (WDSS- II) will be available in FY00 or FY01.

Neural Network (NN) and Statistical Analyses

Three Neural Networks (NN) were developed – one for MDA, one for TDA, and another for circulations detected by MDA and TDA, jointly. All three NNs had two output nodes, corresponding to the probability of tornado, and the probability of damaging wind, respectively. These were trained and validated on a data set consisting of 29 storm days, and as many as 56 attributes. All three NNs had sufficiently matured for incorporation into SSAP. Reliability diagrams showed that all the probabilities produced by the NNs were highly reliable.

A new data set then became available for analysis. It consisted of 43 storm days, incorporates BWER attributes, and has as many as 85 attributes. As a result of the introduction of BWER attributes it was necessary to develop 6 different NNs (each with two output nodes, for the probability of tornado, and damaging wind):

NN_I: for circulations detected by MDA, only.
NN_II: for circulations detected by TDA, only.
NN_III: for circulations detected by MDA and TDA, jointly.
NN_IV: for circulations detected by MDA and BWER, jointly.
NN_V: for circulations detected by TDA and BWER, jointly.
NN_VI: for circulations detected by MDA, TDA, and BWER, jointly.

Upon preprocessing the data, it was found that if all of the available attributes are input into the NN, some of the resulting NNs tend to overfit the data. This occurs when the sample size is small in comparison to the number of parameters of the NN, the latter being proportional to the number of input nodes. As a result, the optimal NNs do not necessarily require all of the attributes as inputs. The reliability diagrams indicated that the produced probabilities are completely reliable within the statistical error bars.

The aforementioned 43-day data set was utilized for a determination of the best predictors of tornadoes. In spite of the numerous subtleties in the analysis, it was found that it is possible to isolate a few of the many predictors as being the best predictors. Several methods were employed; each designed to address a different facet of the problem.

A NN has also been developed for diagnosing hail. It is based on the Hail Detection Algorithm and is designed to predict the size of hail. A similar NN is being developed for producing a posterior probability of hail.

Tornado Warning Guidance

There was no update to the original 1997 NWS Tornado Warning Guidance document made for the 1998 convective season. However, during the latter part of 1998, scientists at NSSL and the OSF worked together to generate new, supplemental Tornado Warning Guidance for the NWS for the 1999 convective season (TWG99). The TWG99 is based on an update of our latest ideas about and understanding of tornado prediction. A majority of the new guidance comes from an extensive analysis of WSR-88D data using the NSSL experimental vortex detection algorithms. Numerous radar-based parameters that measure different aspects of thunderstorm vortices were evaluated to determine how well they discriminate between tornadic and non-tornadic storms.

New guidance documentation included sections on 1) qualitative general guidance information gleaned from the latest basic and applied research experience, and 2) supplemental information for the Build 10 Tornado Detection Algorithm (TDA).

Near-Storm Environment (NSE) Algorithm

The goal of the Near-Storm Environment (NSE) algorithm is to provide to the NSSL WSR-88D algorithms information concerning the environment of each storm cell, such as shear and stability parameters. Currently, NSE uses output from the Rapid Update Cycle (RUC) model to help determine the environment of storm cells.

During FY98, RUC-I data were gathered for those cases in the 43-case MDA/TDA data set where they were available. This represented 14 of the 43 cases, or about 31,000 of the 82,000 vortex detections. Also, a variety of new NSE parameters were developed within the NSE algorithm. Finally, a method to bi-linearly interpolate NSE gridded information to the location of the MDA and TDA centroids was computed.

The NSE algorithm and the MDA and TDA were run on the 14 cases to create a large data set of integration algorithm information. Preliminary results looking at bivariate distributions of the data suggest that only a few NSE variables offer some diagnostic value for determining whether or not an algorithm-detected storm-scale vortex is tornadic.

Vortex/Severe Weather Climatology

Since 1995, Level II data have been routinely collected at NWS and DOD WSR-88D sites and archived at the National Climatic Data Center. As a result, climatologies of radar observed signatures (hail, mesocyclones, tornadic vortex signatures [TVS], etc. using a relatively large Doppler radar data set are now possible. In 1998, the NSSL embarked upon a pilot study to ultimately determine a more reliable estimate of the frequency of these radar signatures and how often they are associated with tornadoes and other severe/hazardous weather. This pilot study incorporated the use of NSSL’s enhanced severe weather and tornado detection algorithms to identify the radar signatures. The Level II data used were collected at the Pittsburgh WSR-88D (KPBZ) in 1996. Approximately 120 Level II tapes were processed spanning 3 months of time using a SUN Sparc 20 workstation. Data from the St. Louis WSR-88D (KLSX) have also been processed by NSSL’s algorithms. In addition, the study incorporates the SPC Smooth Log as the source of reports of tornadoes and severe weather. Preliminary results from the study using the KPBZ data were presented in 1998 at both the AMS Severe Local Storms Conference in Minneapolis, MN and the National Weather Association in Oklahoma City. Concrete results are still pending as the statistical analysis continues. Plans include completion of the statistical analysis by March 1999 using the KPBZ data as well as the KLSX data and process/analyze the Level II data from other radars across the U.S.

WSR-88D Database

SWAT-V has continued to acquire the WSR-88D base data (Level II) for development and testing of the SSAP meteorological algorithms. Currently, the NSSL has the Level II data associated with over 3500 tornadoes (2000 more than in FY97), as well as over 25,000 of severe weather reports.

The NSSL, with support from the OSF, continues to collect Level II data for the purposes of various research needs and algorithm development and evaluation. Much of the data collected in 1998 included those associated with the MEaPRS campaign and significant tornado events. In addition to the Level II data, upper air data associated with the collected data, have also been collected and made available on via anonymous ftp or the World Wide Web. The NSSL also obtained the 1996 SPC Smooth Log that has been used to create the necessary ground truth files for all 1996 Level II data contained within the NSSL/OSF Level II database. Data inventories, soundings, and ground truth data can be found at the NSSL web site.

WDSS Testing

The NSSL deployed the WDSS for real-time testing at the Pleasant Hill, MO forecast office during the 1998 convective season. Although there was a general lack of convective storms compared to other years, feedback from the forecasters was positive regarding the use of WDSS in the warning decision making process. Of particular note, the forecasters continue to be impressed by the organization of information within the WDSS and the ease at which the information can be accessed. The Pleasant Hill environment was a unique test site for the WDSS since the WDSS was displayable on the AWIPS workstations as a complimentary warning decision tool. As a result, the staff was able to conduct sectored warning operations where forecast staff was assigned warning responsibilities for specific portions of the county warning area (CWA).

A winter version of the WDSS was deployed in the Reno NWS forecast office in November 1998 for testing of the USBR Snow Accumulation Algorithm (SAA). So far, the Reno staff has been impressed with the SAA performance as snow depth verification closely matches that of the SAA’s depth estimates. The WDSS is capable of displaying the SAA output in a similar fashion as the precipitation estimates. This test will continue through April 1999.

The NSSL also has deployed a prototype next-generation WDSS in the Jackson forecast office. This WDSS is capable of ingesting radar data from multiple WSR-88Ds. In addition, the next-generation WDSS consists of a mosaic product and a composite table consisting of information from all of NSSL’s severe weather and tornado detection algorithms for all radars. The focus of the next-generation WDSS is on the warning responsibilities of the NWS for their entire county warning area.

Plans – FY 00


NSSL will continue developing, testing, and documenting basic functions for the signal processor for the ORDA. We will also begin studying the implementation of test functions for the on-the-site failure diagnostics program, finish pre-prototype software for the Signal Processing System (SPS) on the ORDA, and re-evaluate automatic calibration algorithms for their implementation on the WSR-88D within the new open system platform.

Radar Quality Improvements

NSSL scientists will continue investigating efficient techniques to reduce estimation errors in the processing of polarimetric radar data, including the estimation of spectral moments. Considerations for the implementation of these techniques after the adoption of a digital receiver will be studied. We will also investigate techniques to mitigate ambiguities, especially in a dual receiver configuration planned for polarimetric measurements with the WSR-88D radar.

AWIPS-Based Severe Weather Training

The OSF/OTB will continue developing AWIPS-based training on radar usage and integration into NWS warning operations. Plans for 1999-2000 include:

Improving Displaced Real-Time (DRT) capability in AWIPS to be used in simulating “live” warning exercises in decision-making workshops,
Developing training on new functionality and changes to successive AWIPS software builds as they apply to severe weather operations,
Converting baseline radar training to AWIPS format, and
Developing AWIPS-based integrated sensor techniques for determining convective initiation and convective evolution.

Severe Weather Training Applications

The newly created NWS strategic plan for the next five years contains ambitious targets such as a 50 percent reduction in false alarm rate and a 50 percent increase in probability of detections for severe thunderstorms. Although the WSR-88D radar network has been the primary warning tool since the completion of the 1998 deployment, the large numbers of surprise severe weather events experienced regularly across the nation shows that these performance goals will be challenging to achieve. OSF/OTB scientists plan to work closely with field offices to share lessons learned from these challenging cases nationwide, and severe weather evaluation techniques will be reviewed and refined given new research and technology. Plans for 1999-2000 include:

Developing new tornado warning guidance
Creating an archive of challenging event web-based training modules,
Developing integrated sensor techniques using radar, satellite, lightning, and near-storm environment data,
Determining the strengths and weaknesses of the WSR-88D radar and algorithms by studying reflectivity and velocity data for more close-range severe storm events, such as the 3 May 1999 Oklahoma tornado outbreak, and
Investigating new approaches to discriminate between tornadic/non-tornadic supercells and severe/non-severe thunderstorms.

Analysis of Tornado Detection Algorithm Performance

The OSF/Applications Branch is accumulating case studies and plans to continue analyses during the next year.  Developing case studies is a time-consuming, manual process, and we have teamed with data base developers at the NSSL to share the burden.  In addition, we have approached the National Climatic Data Center (NCDC) for large-scale development of well-truthed case studies.

Scoring Detection Algorithms

The OSF/Applications Branch is in the early stages of this continuing work.  The WSR-88D system is undergoing major changes with the advent of Open Systems.  Open Systems will usher in CODE, designed to provide a programming environment for new algorithm development and algorithm improvement.  Under CODE, algorithms may be developed and presented for evaluation more quickly than in the past, so scoring and verification methods must be automated.

Generation of Terrain Hybrid Scan and Occultation Data Files in Preparation for the Snow Accumulation Algorithm

The OSF/Applications Branch continues to test the Snow Accumulation Algorithm at snow sites around the U.S. in preparation for implementation in Open Build 2 somewhere in the 2001 timeframe.  We are continuing to refine the occultation and hybrid scan files.

Investigation of WSR-88D Level II Data:  Severe Weather Event Detection

The OSF/Applications Branch plans further work on the use of Neural Networks for severe weather event detection in Level II data.

Investigation of WSR-88D Level II Data:  Precipitation Estimation

This OSF/Applications Branch work began in June 1999 and is similar to that reported immediately above.  The plan is to apply robust, multivariate statistical techniques to mine Level II data for structure that would lead indirectly to elucidation of drop size distribution and thus to estimation of rainfall rate.

The use of CODE at the WSR-88D OSF

Development of CODE with NWS Headquarters, the NSSL, and Mitretek continues.  CODE will become a vital part of the WSR-88D Program and is scheduled for deployment shortly after the turn of the century.

Polarimetric Radar Developments

Regular weather observations with the dual-polarization Cimarron radar will be continued at NSSL and combined with observations from the NSSL’s 2D-video disdrometer, recently bought from Joanneum Research company.

The potential for automatic classification of rain type prior to rainfall radar quantification will be assessed using existing Cimarron radar data and the Little Washita River raingage micronetwork.

Analysis of range effects on the quality of polarimetric rainfall estimation by radar will begin using Cimarron radar data and Oklahoma Mesonet data.

The NCAR S-POL radar data obtained in the “full-polarimetric mode” during the TRMM-LBA experiment in Brazil and the Mesoscale Alpine Project (MAP) in Italy in 1999 will be thoroughly examined in order to:

Compare polarimetric signatures of different precipitation types and clouds in different geographic regions of the world,
Investigate the information content of a full covariance scattering matrix and its use for better hydrometeor classification, and
Assess the possible inclusion of linear depolarization ratio and correlation coefficients between co-polar and cross-polar components of a radar return in the set of polarimetric variables to be measured by the polarimetrically upgraded version of the WSR-88D.

Development of MPDA

The MPDA design should be finished at NSSL by the end of 1999. It will then be turned over to software engineering for re-coding to Open System standards. Upon approval it will be delivered to the OSF for inclusion on the national WSR-88D system.

County Warning Area Algorithm

MuRIAS is currently being evaluated at NSSL. Some changes in the rule base that ranks the radars are likely. Once finished the algorithm will be included in the next generation WDSS-II.

Cell and Area Tracking

The comparison between the NSSL version and MIT/LL versions of the SS/CT algorithm will be completed.  It is expected at this point that the NSSL implementation of the SS/CT algorithm will produce results similar to those from the MIT/LL version.  Concepts to integrate the output from both the SCIT algorithm and the SS/CT algorithm will be investigated.  After an integrated display concept has been developed, the combined SCIT and SS/CT algorithms will be implemented into NSSL’s WDSS and tested in real-time at several WSR-88D sites during the year 2000 convective season.

Improved Mesoscale Model Initial Conditions Utilizing Radar-retrieved Wind Fields from the WSR-88D Network

Based on the accurate retrievals produced by the CAAM technique, an investigation will start at NSSL concerning the assimilation of the retrieved wind data within a numerical model.  Radar-retrieved wind information has only recently been used to modify model initial conditions and primarily, only for convective, storm-scale models.  This study investigates the ability to use retrieved wind information from multiple WSR-88D radars to improve simulations using the Penn State/NCAR (MM5) mesoscale model.  The CAAM technique is applied to the WSR-88D radar network across much of the central United States for a mesoscale convective system outflow event. Retrieved wind fields from several WSR-88D sites within a 2000 km x 2000 km domain are interpolated to the model’s lowest vertical levels, up to a height of 3500m.  CAAM is used to produce a horizontal wind field with a radius of 150 km from each WSR-88D radar.  Multiple wind fields are produced at 250-m vertical increments surrounding each radar site. Through a static initialization procedure, improvements within a simulation of a mesoscale convective system and its resultant outflow boundary will be examined as a result of adding retrieved wind information. Two different methods of producing model initial conditions will be implemented.  Model grid points up to 150 km in radius surrounding each radar site can be directly modified by retrieved results. However, this method can leave areas in the model initial condition unmodified in between radar sites.  Therefore, a second method of adjusting model conditions will be studied.  The second technique objectively analyzes radar-retrieved information from multiple WSR-88D sites to fill in regions initial on the model grid that are not directly observed by radar.

Tuning MIGFA for WSR-88D Radars: The 1998/1999 Sterling Test

The Machine Intelligent Gust Front Algorithm (MIGFA), initially developed for Federal Aviation Administration radar systems, is being tuned by MIT/LL for use in the WSR-88D radar system. The effort will benefit the three agencies (FAA, NWS, and DOD) supporting WSR-88D. The purpose of MIGFA is to provide NWS forecasters with automated detections of boundary layer convergence lines relevant to NWS forecast operations. The MIGFA detects boundaries and forecasts their future positions based on evidence from signatures identified in the WSR-88D base reflectivity and velocity fields.  The algorithm is being tested at the Sterling forecast office (1998-1999).  Using the data collected from this test, NSSL and NCAR are evaluating the MIGFA algorithm qualitatively to discern the character of boundaries detected well versus those detected poorly.  This feedback will help MIT/LL improve the MIGFA’s performance.

SWAT-V Severe Weather Algorithms for the WSR-88D

Plans at NSSL for the next fiscal year include:


The main focus in FY99 is to convert all the algorithms so output in “rapid update” mode. Currently, the algorithm information is processed after an entire volume scan is completed, and volume products are typically 5-6 minutes older than the lowest-altitude elevation scan in that volume. Rapid Update mode allows for algorithm information to be processed after each elevation scan, for more timely output of information to the warning forecasters.

Development has begun to convert the SSAP algorithms to C-language and for the future WSR-88D Open-RPG system. The SCIT algorithm and HDA algorithm conversions are in progress. TDA, MDA, and DDPDA will follow into FY99 and FY00.

More robust methods for vertical and time association will be developed in FY99. Also, terrain information will be integrated into the algorithms.


The NSSL MDA will be integrated into the WSR-88D system (Open-RPG) during FY00-FY01, and should be fielded nationwide at the end of that period. For FY99, work will be completed on the MDA in preparation for this inclusion into the WSR-88D system.

Beyond FY99, future work will include the continued development of a storm-scale Vortex Detection and Diagnosis Algorithm (VDDA) combining detection capabilities of both the MDA and TDA.


As the TDA is now part of the WSR-88D system, time will be spent in FY99 to provide consultation to the users of the algorithm via the OSF/Applications Branch and the OSF/OTB.


The VDDA work has been put on hold, pending the implementation of the NSSL MDA into the WSR-88D system. VDDA work will recommence again in FY00.

Neural Network and Statistical Analyses

A new NN will be trained that incorporates MDA, TDA, and Near-Storm Environment (NSE) algorithm input on 14 storm events (109 tornadoes). Statistical analyses of these data will also be performed.

Tornado Warning Guidance

As the amount of available data increases, and as the statistical methodology matures, further analysis will be performed to continue the offering of guidance for tornado warning for the 2000 or 2001 convective season. Primary emphasis will be on the development of situation-specific guidance based on variations in storm types as dictated by the near-storm environment.

NSE Algorithm

The integrated MDA/TDA/NSE data set will be statistically analyzed to determine its overall predictive value. Also, the data set will be used to train a new NN.

If the NSE data cannot provide predictive value on the scales of single thunderstorms and volume scans (5- or 6- minutes), we propose to use the NSE information on a more regional space and time scale, and to help with the automated adjustment of site-adaptable parameters for varying storm types (as dictated by the NSE).

Vortex/Severe Weather Climatology

We will continue the analysis of the Pittsburgh and St. Louis data sets. We will also begin the automatic transfer of NSSL algorithm information from all the WDSS sites so that we can start building automatic radar severe weather climatologies nationwide.

WSR-88D Database

The NSSL, with support from the OSF/Applications Branch, shall continue to acquire additional radar data.

WDSS Testing

The WDSS was deployed at the Tucson forecast office during the 1999 summer convective season with an emphasis on looking at downburst producing storms. A second WDSS is slated for St. Louis. For the 1999-20000 winter season, the Tucson WDSS will be deployed to Missoula, MT. Efforts will continue into 1999 to further develop and refine NSSL’s next-generation WDSS, WDSS-II.