The advent of dual polarisation radar has radically changed the nature of operational radar QPE. Use of multi-parameter classifiers for quality control and dual polarisation rain rate estimators (Brandes et al., 2002, 2003; Giangrande and Ryzhkov, 2008) are feeding improvements in the quality and reliability of radar products worldwide (Ryzhkov et al., 2005; Figueras I Ventura and Tabary, 2013; Helmert et al., 2014). However, the applicability of dual polarisation measurements directly to QPEs is still limited to the rain level. The assumptions underpinning the relationships betweenZDR, KDP and rain
rate are based on the strongly preferrential orientation behaviour of liquid rain drops, but equivalent constraints for more randomly oriented ice crystals, aggregates and melting particle mixtures have not yet been possible to derive. At high latitudes, this severely limits the applicability of dual polarisation rain rate estimators to what is often a only a small proportion of the radar sampling domain (eg figure 1.8). The use of single polarisation reflectivity measurements in and above the bright band for surface QPE, and the methods required to adjust these for VPR, remains an important area for future research.
There has already been significant investigation into the use of dual polarisation infor- mation for improving determination of stratiform VPRs (reviewed in chapter 2). Many papers have established skill inρhv,ZDR and LDR for identifying and locating the radar
bright band (eg Tabary et al., 2006; Boodoo et al., 2010; Hall et al., 2015). For sys- tems without access to freezing level information, using dual polarisation parameters to determine a domain-averaged melting layer height is a significant refinement of the cli- matological bright band corrections proposed in much of the literature, and provides for the first time the prospect of a VPR correction accurate enough for real time application (eg Tabary, 2007). However, little attention has been given to the use of microphysical information contained in dual polarisation measurements to determine profile character- istics on the more local scale. The demonstrable skill of dual polarisation parameters in hydrometeor classification raises the question as to whether this information could be combined with knowledge of microphysics underlying the VPR to improve the accuracy of local surface reflectivity estimates.
The pixel-by-pixel Kitchen et al. (1994) scheme applied in the UK provides a unique testbed for investigating the potential of dual polarisation to improve VPR classifica- tion and correction at the local scale. While stratiform VPRs are well-treated by this scheme, the identification and correction of non-bright band VPRs could benefit signifi- cantly from further research. Use of a reflectivity-based criterion for convective diagnosis is known to underdiagnose non-bright band conditions, meaning that potentially a sig-
nificant proportion of non-bright band cases are being corrected inappropriately. This implies widespread underestimation of rain rates in the very high impact situations for which accurate QPEs are most urgently required. Dual polarisation measurements have the potential to provide more reliable methods of identifying bright band (Smyth and Illingworth, 1998; Illingworth and Thompson, 2011), which could reduce the occurrence of inappropriate bright band corrections and the associated rain rate underestimation. Beyond identifying non-bright band conditions, recent observational literature (eg Delrieu et al., 2009; Kirstetter et al., 2013; Matrosov et al., 2016) suggests that the assumption that reflectivity is constant with height in all cases without bright band may be inaccu- rate. Once the issue of classification has been addressed, there is scope for improving the characterisation of non-bright band profiles within the Kitchen et al. (1994) framework, by developing new idealised local profiles for different types of VPR.
This thesis aims to apply new information from dual polarisation parameters to improving correction for VPR in the Met Office operational radar processing chain. High quality measurements from the upgraded radar network will be used to distinguish intelligently between different types of vertical profile, which are characterised using a large new dataset from the C-band research radar at Wardon Hill. The investigation focuses on the linear depolarisation ratio (LDR), which has been shown to respond particularly to the large melting snowflakes responsible for stratiform reflectivity bright bands (Smyth and Illingworth, 1998; Illingworth and Thompson, 2011).
This thesis is organised as follows. Chapter 2 reviews the current literature on VPR classification and correction schemes, and includes a detailed description of the Kitchen et al. (1994) approach on which this thesis builds. Chapter 3 provides details of the Wardon Hill radar, and describes the high resolution dual polarisation dataset collected to support this investigation. In chapter 4, the quantitative skill of LDR in distinguishing between different VPR types is investigated, and is compared to the skill of the current UK operational convective diagnosis criterion. Chapter 5 develops this result into an op- erationally feasible algorithm, and demonstrates the impact of LDR-based classification on QPE accuracy in a real time environment.
Having improved the real time classification of VPRs using dual polarisation measure- ments, the remainder of this thesis investigates refinements to the different profile shapes available for correction in different meteorological conditions. Chapters 6 and 7 exploit the reflectivity information from the high resolution profile dataset to suggest improve- ments to the idealised shapes used for VPR correction in the UK. In chapter 6 a new non-bright band VPR shape is proposed, with support from previous literature, and evaluated through both simulations and a real time implementation. Chapter 7 uses observations from stratiform VPRs in the Wardon Hill dataset to improve residual long range QPE bias through small changes to the Kitchen et al. (1994) idealised profile. Fi-
nally, chapter 8 summarises the outcomes of these investigations and suggests areas for future research and development.
Chapter 2
Existing approaches to VPR:
classification, determination and
correction
2.1
Introduction
Radar quantitative precipitation estimation (QPE) is achieved through the conversion of a reflectivity measurement aloft into a rain rate estimate at the ground (section 1.5). An important step in this process is the adjustment of a meteorological reflectivity from its value measured at height to an estimate of the value near the ground. This is known as correcting for the vertical profile of reflectivity, or VPR.
The VPR defines the variation in atmospheric reflectivity as a function of height above the ground surface. While there is little variation at low levels other than that caused by partial beam blocking, in areas where the radar beam samples above or close to the 0oC isotherm the difference between measured and surface reflectivities can exceed an order of magnitude. In high latitude climates, the ranges at which radar PPI measurements intersect the melting layer combined with broadening of the radar beam cause these effects to influence a significant proportion of the radar domain. VPR is therefore a significant, if not the most significant, source of error in high latitude radar QPEs. In this chapter the existing literature pertaining to VPR classification, determination and correction is reviewed. Section 2.2 describes the microphysical processes that occur in different types of precipitation, with reference to both observational and modelling studies, and how they lead to the general shape of the resulting profiles. This provides context for the many approaches to correcting for the VPR in PPIs which are reviewed
in section 2.3. These range from purely empirical ratio-based methods to highly complex linear parameterisations and probabilistic schemes, on a range of different spatial and temporal scales. Section 2.4 presents approaches to the classification of different pre- cipitation types, such as convection, which are needed as input to certain bright band and VPR correction schemes. Three operational correction schemes and their support- ing classification methods are discussed in detail and compared in section 2.5. Finally, section 2.6 summarises the main points of the review and revisits the motivations for the remainder of this thesis.