Since the incidental discovery during World War II that radar could detect and track meteorological phenomena, it has been pivotal in developing modern meteorology. Thanks to its temporal and spatial resolution, which remains unmatched for remote observing systems, radar has helped advance our understanding of storms and precipitation.
Weather radar also aids in issuing severe weather warnings, and has saved countless lives over past decades. And with modern communications technology, even members of the general public can view radar data from their mobile devices in real time.
Given radar's fundamental importance to meteorology, scientists have focused significant attention to developing new radar technologies. Most recently, weather radar has leaped forward in the research, development, and operational use of dual-polarization radars.
Unlike conventional weather radars, which transmit and receive horizontally polarized electromagnetic wave pulses, dual-polarization weather radars transmit and receive orthogonally polarized electromagnetic wave pulses in both the horizontal and vertical planes.
A dual-polarization tower belonging to the National Weather Service's WSR-88D system. CREDIT: NOAA
Dual-polarization radars have been researched and developed since the 1950s. However, it was not until about a decade ago that scientists began to consider a national network of dual-polarization radars. Upgrades to the existing Weather Surveillance Radar–1988 Doppler radar (WSR-88D) network began in 2011.
Since those upgrades were completed in June 2013, US meteorologists have been operating dual-polarization radars and are collecting an unprecedented amount of data for meteorological research.
Dual-polarization measurements
The power scattered by an atmospheric target back toward a radar is quantified by the radar reflectivity factor Z. The power is higher for a wave whose polarization vector is aligned with the longer axis of the target. When nonspherical particles are sampled by dual-polarization radar, the amount of backscattered power is therefore different in the horizontal and vertical polarization channels. That difference, along with phase information of the returned waves, can be exploited to infer the size, shape, composition, and behavior of targets.
How dual-polarization radars gather information about the hydrometeors in a radar resolution volume. CREDIT: National Weather Service, National Oceanic and Atmospheric Administration
Traditional radar measurements typically yield Z, Doppler velocity, and spectrum width. With dual-polarization radar, many new measurements become available, of which the most commonly used are differential reflectivity (ZDR), co-polar cross-correlation coefficient (ρhv), differential phase shift (ΦDP), and specific differential phase shift (KDP).
- • ZDR is the logarithmic ratio of the returned power at horizontal and vertical polarization and is greater than zero for oblate particles (where the horizontal dimension exceeds the vertical dimension). As raindrops form and subsequently increase in size, they become increasingly oblate due to drag. Consequently, ZDR can be understood as a measure of median drop diameter within a volume. However, ZDR can also be used to gain shape information about ice-phase hydrometeors, which do not deform due to air resistance. (Hydrometeor is the generic term for raindrops, hail stones, snow flakes, and other particles of liquid or solid water in the atmosphere.) Randomly tumbling particles such as dry hailstones will appear spherical to the radar and produce a near zero ZDR. Ice crystals, which are highly oblate and have large positive values of ZDR, can align vertically in the presence of an electric field and exhibit negative ZDR values; they serve as a proxy to detect electric fields within convective storms. ZDR is also affected by hydrometeor density: For a given size and shape, hydrometeors with higher density produce larger ZDR than less dense hydrometeors do.
- • ρhv is a measure of similarity between the returned waves at horizontal and vertical polarization. Volumes of homogenous hydrometeors exhibit values of ρhv near unity. Lower values are observed in areas of mixed-phase hydrometeors, non-Rayleigh scattering, or non-meteorological targets, such as bats, birds, and insects.
- • ΦDP represents the difference in accumulated phase shift between the horizontally and vertically polarized waves as they travel away from and back toward the radar.Electromagnetic waves slow as they pass through a medium—an ensemble of hydrometeors, in meteorological situations. As raindrops grow and become more oblate, the horizontally polarized wave slows relative to the vertically polarized wave and a differential phase shift develops. KDP, the derivative of ΦDP with respect to range, is frequently used due to the difficulty of interpreting ΦDP, which accumulates along the entire path of the wave. KDP helps pinpoint areas where the ΦDP is increasing most rapidly, which is useful for locating areas of heavy rain and oblate particles. Because they are phase measurements, both ΦDP and KDP have the advantage of being independent of attenuation and less susceptible to calibration errors.
These are just the most frequently used variables. Other variables deal with depolarization and the respective rates of attenuation of the horizontally and vertically polarized waves.
Current operational dual-polarization applications
Meteorologists have already made great strides in applications that use dual-polarization data. The Hydrometeor Classification Algorithm (HCA), for example, classifies observed radar targets in a volume as various types of hydrometeors, ground clutter, or biological echoes. It can also be used to locate the height at which snow, hail, and other solid hydrometeors melt and turn into rain.
The HCA has seen great success, particularly for hail identification, and scientists continue to refine and improve it. One such development is the Hail Size Discrimination Algorithm, which aims to identify the maximum hail size within regions that the HCA has already identified as containing hail.
An example of results from a Hydrometeor Classification Algorithm (HCA) for the blizzard that struck the Northeast US on 8–9 November 2013. Radar reflectivity (left). HCA for surface precipitation type (right) showing areas of rain (red), wet snow (green), dry snow (light blue), and ice crystals (dark blue). Circles indicate user-submitted precipitation types (red - rain, blue - snow) from mPING, a crowdsourcing project for the public to submit precipitation type observations in real time. CREDIT: Terry Schuur, National Severe Storms Laboratory
One of the primary goals of dual-polarization radar is to improve quantitative precipitation estimation (QPE), which has important implications for flash flood warnings. Traditionally, rain rate R has been estimated from Z. The 10-cm wavelength of the WSR-88D network ensures that even the largest raindrops, which do not exceed 8 mm in diameter, are Rayleigh scatterers. In that regime, Z is proportional to the sixth power of diameter, resulting in a highly nonlinear relationship between particle size distributions (PSDs) and Z. Large errors in QPE are therefore possible. Moreover, the relationship between PSDs and Z varies in different climate regimes and different meteorological environments. To account for that variation, researchers need to develop many different Z–R relations.
To be accurate and useful, Z–R relations must be selected according to the general meteorological environment at a given time. Additionally, any hail present in the volume can introduce significant errors into Z-based QPE. Some polarimetric variables, such as KDP, are much less sensitive to variations in PSDs and are less susceptible to hail contamination than Z is. Polarimetric QPE uses a combination of Z, ZDR, and KDP based on the results of the HCA. When compared to rain gauge measurements, polarimetric QPE has shown significant improvements over traditional Z–R relations.
Meteorologists now also use dual-polarization radar in the process for issuing severe weather warnings. One example is the polarimetric tornado debris signature. Debris lofted by tornadoes is irregularly shaped, large with respect to the radar wavelength, and moves chaotically. Very low values of ρhv result. When coupled with a rotational signature in the Doppler velocity data, low ρhv can serve as remotely sensed confirmation of a tornado touchdown.
Thunderstorm updrafts, which contain a low concentration of very large raindrops, also exhibit a distinct polarimetric signature called the ZDR column. ZDR columns feature high values of ZDR that extend well above the environmental freezing level; they appear when a storm's updraft lofts slowly freezing supercooled raindrops upward. ZDR columns are therefore positively correlated with updraft strength.
The development of conditions that manifest ZDR columns also depends on atmospheric instability and on the concentration and characteristics of aerosols. Current research investigates those dependencies with the goal of extracting aerosol information, and this has important implications for climate.
In addition to these applications, scientists have made advances in winter weather identification, heavy rain detection and flash flood characterization, and in the detection and filtering of non-meteorological echoes.
Potential benefits for convective-storm modeling
At the forefront of research in dual-polarization radar is the use of data in convection-resolving models, which use grid spacing small enough to characterize individual thunderstorms and their internal processes. The models are expected to play an increasingly large operational role in the future through initiatives such as Warn-on-Forecast, which aims to enlist models to increase lead times for severe weather warnings.
For convection-resolving models with grid spacing on the order of 1 km, radar remains the only source of data that can supply the requisite temporal and spatial resolution. The assimilation into models of Doppler velocity is relatively well established and has been shown to have a positive impact on estimating the wind field within the modeled storm. However, modelers assimilating Z face inherent difficulties due to the nonlinear relationship between Z and predicted model variables. Generally, microphysics schemes in numerical weather prediction models predict only the mass mixing ratio q of various hydrometeor types.
Data assimilation seeks to optimally combine observations with an existing model background field. However, because Z is observed in the field whereas q is predicted in models, one must be mapped to the other to enable direct comparison. We can accomplish this in one of two ways: convert the model q to Z through an operator, or retrieve q from the observed Z.
Currently, the latter method is most widely used, and this requires so-called Z–q retrieval equations. Due to the highly nonlinear relation between the two, however, we have to make a number of simplifying assumptions to make the comparison. The validity of these assumptions varies in space and time.
For example, rain PSDs are assumed to follow a general form, despite the significant changes to the PSD that can occur due to dynamic and thermodynamic processes in the environment. Even given those limitations, the assimilation of Z and use of retrieval equations has been shown to have positive impacts on model forecasts of convection.
Polarimetry’s potential to dispense with some of these assumptions is just beginning to be explored. One way to study the relationship between model variables and polarimetric variables is through the use of models that incorporate spectral bin microphysics. Spectral bin microphysics treats each particle size bin explicitly, allowing for interactions between particles of different sizes and with complex and realistic microphysical processes.
In contrast, bulk microphysics schemes describe a hydrometeor's PSD with a single function and microphysical processes act on the entire distribution, with no interactions between particles of different sizes. Due to their high computational expense, spectral bin models have been generally confined to research, while bulk microphysics schemes, which are oversimplified but highly efficient, are used frequently in today’s operational models.
When coupled with a polarimetric radar operator, which explicitly and accurately calculates the polarimetric variables for simulated PSDs without many of the assumptions previously mentioned, spectral bin models can give us insight into how various real-world processes affect the polarimetric variables. The results can then be parameterized and used to improve on current retrieval equations.
Similar to the improvements made in QPE, we expect that variables such as KDP will achieve more accurate retrievals of rain mixing ratios and smaller average errors than can the use of Z alone. The feasibility of directly assimilating the polarimetric variables and the utility of polarimetric retrievals is currently uncertain and remains an active area of research.
Researchers are beginning to investigate other potential uses of dual-polarization radar for improving convection-resolving models. One possibility is using the HCA to improve the modeled hydrometeor classes within a storm. Currently, in order to select an appropriate retrieval relation or determine which model hydrometeor class receives the most of an assimilation update, we use the model background temperature (a simplifying assumption). But the HCA provides a real-time three-dimensional field of hydrometeor types that could be used to improve the classification of hydrometeors within a modeled storm.
Another promising area of research is the use of so-called polarimetric fingerprints. The fingerprints refer to the characteristic and distinct signatures seen in the polarimetric variables due to various microphysical processes such as evaporation, aggregation, or melting. In addition to helping forecasters identify such processes in real time, cloud modelers may use these fingerprints to evaluate the microphysics in their models.
Microphysics parameterization schemes, which serve to represent the effects of sub-grid scale cloud processes, are widely acknowledged to limit the successful prediction of convective storms. After converting model output to the polarimetric variables, any inability to reproduce the fingerprints will help reveal deficiencies in the model’s microphysics scheme and help improve them.
Polarimetric fingerprints may serve as more than validation for parameterization schemes. As mentioned above, reducing spin-up time for modeled thunderstorms and their attendant precipitation has long been a major research priority. One technique to meet this goal is diabatic initialization. The major source of energy for a thunderstorm's updraft is latent heat release due to condensation. However, due to the small spatial scales at which the heat is released, it is not generally captured in assimilated observations and so is not accurately represented in the model background field.
But if latent heat data from a storm are to be assimilated, an accompanying increase in moisture and latent heating must be included in the initial conditions. Failure to do so leads to increased spin-up time and underestimated storm strength. The modeled storm could even fail to sustain itself and fizzle out. In addition to latent heating within updrafts, cold pools that generate new convection arise from latent cooling from the evaporation of rain and melting of hail. The characteristics of these cold pools have been shown to relate to the formation of tornadoes, highlighting the critical nature of accurately representing diabatic processes in a convective model.
In the past, meteorologists calculated the missing latent heat forcing in the updraft from the difference between modeled and observed precipitation, and from the vertical distribution determined from climatology or an assumed profile. More advanced techniques have since been developed that use Z and have shown positive impacts on storm spin-up time. However, a number of limiting assumptions are still required. Polarimetric fingerprints associated with phase changes, which are often more prominent than changes seen in Z alone, might be able to improve the estimate of both the magnitude and spatial distribution of the diabatic heating field.
Although condensation is not directly observed by radar, its effects (such as the initial ZDR echoes associated with rain formation) could be used to indirectly quantify diabatic heating effects. Additionally, the use of ZDR columns should also prove beneficial. As proxies for updrafts in time and space, ZDR columns could be used to promote updrafts or suppress spurious ones within the model. These could then serve as locaters for zones where significant diabatic heating occurs.
Weather radar polarimetry is currently and will remain an active area of research. We are still discovering its potential uses, and still investigating remaining uncertainties pertaining to the optimal use of data. Increasing collaboration between the radar polarimetry and cloud modeling communities will prove mutually beneficial and essential to tackling new problems in the years ahead.
The upgrade to WSR-88D has made it the largest dual-polarization radar network in the world, but many other countries are beginning to upgrade their systems as the benefits of polarimetry are realized. Despite the challenges that remain, meteorologists have great optimism for dual-polarization radar’s potential to improve convective-scale modeling.
Jacob Carlin is a PhD student at the University of Oklahoma, and his research focuses on dual-polarization radar and its potential applications for convective-storm modeling.
