Skip to main content
European Commission logo print header

Critical assessment of available radar precipitation estimation techniques and development of innovative approaches for environmental management

Deliverables

Software has been developed to extract wind-field information from two operational Doppler radars with overlapping coverage area. The three dimensional wind-field is extracted from data relating to a square box (of dimension up to 30km) not including the line joining the two radars, at several heights above the ground. In addition to the (unambiguous) radial winds measured by each radar, the algorithm includes the continuity assumption. Use of the software is potentially of social benefit in helping to provide warning of strong winds, and particularly of the occurrence of cells containing extreme winds in convective storms. The software is capable of real-time implementation, and should thus be of value for application to civil defence warning of natural storm hazard. Application is potentially immediate for the radars operated by the Emilia Romagna region in Italy. It would be equally applicable to other pairs of Doppler radars where the radar spacing is of the order of 80km. The results will also be of value in testing/verifying the output of numerical models.
The report provides a critical review of the application of polarization data in weather radar systems. Through most of the history of weather radar, the deployment of polarimetric systems has been confined to research rather than operational systems. Where polarization has been installed in operational systems, it is not clear that its capabilities have yet been properly explored. This review describes earlier work and presents a concise outline of the theory of electromagnetic polarization and its basic application in weather radar. The commonly used parameters are described and derived, while many examples are presented which demonstrate both the obvious advantages and technical limitations associated with their use. Several examples are taken from previously unpublished data obtained in earlier EU-funded projects, and illustrate potentials for C-band exploitation of polarization data. The report also presents an overview of system requirements and the performance impact of the principal system components. The emerging techniques involving transmission in a mixed receiver basis (e.g. the often misleadingly named simultaneous HV) are also discussed.
A novel de-aliasing algorithm for Doppler radar velocity data has been developed at SMHI (Swedish Meteorological and Hydrological Institute). It is an accurate and robust tool based on a linear wind model and designed to eliminate multiple folding. The innovation of the new technique is that it maps the measurements onto the surface of a torus. Unlike many other concepts, it does not depend on wind information from a nearby sounding (e.g., radiosonde or wind profiler) or from a numerical weather prediction (NWP) model. De-aliased volume radar data can be used in variational assimilation schemes for NWP models through the generation of so-called super-observations (see result 11224, same project). A super-observation is an intelligently generalized observation created through smoothing in space, based on high-resolution data. It includes a number of derived variables, which collectively serve to describe the characteristics of a given observation. At SMHI, a method for generation of radial wind super-observations through horizontal averaging in polar space of the de-aliased polar volume data has been implemented. The Nordic Weather Radar Network (NORDRAD) is a cooperation project between Finland, Norway, Sweden, and Estonia. Since radial winds from Finnish radars are strongly affected by aliasing problems (only single-PRF data are available for the lowest four elevation angles), the proposed de-folding algorithm was applied to Doppler winds measured with the Vantaa radar during the winter storm of December 4, 1999. A major outcome of this case study was that the quality of the super-observations benefited from the new de-aliasing method. Currently, a corresponding experiment for a representative summer and winter period including all Finnish radars is under way. In the near future, super-observations of de-aliased radial winds from this experiment will be applied to HIRLAM's (High Resolution Limited Area Model) variational assimilation scheme (see result 11223, same project).
Crucial for accurate flood forecasting is an accurate estimate of areal precipitation, its magnitude and distribution over a catchment. Such estimates can be obtained from a wide range of sources, including rain gauges, weather radars and NWP models. Each source has its own characteristics and advantages, but also error and bias levels. Often flood forecasting is based mainly on one of the sources. However, with today's widespread production and high level of access to (real-time) data from the various sources, the possibility to improve forecasting by utilising data from several sources is apparent. In order to evaluate the accuracy of the data from each source, and assess the benefits from their combined use, their characteristics need to be compared. In the present study, areal precipitation estimates from five sources - NWP model (HIRLAM), weather radar (RADAR), rain gauges (PTHBV) and two versions of a mesoscale analysis (MA11 and MA22) - over a specific catchment during a specific time period were compared. The main observations were the following: - The mesoscale analysis consistently generated the lowest amount of precipitation, leading to a difference of about 100 mm for total precipitation over the area in 2002. This may be due to partly that correction for observation losses is not performed, and partly an underestimation of rainfall amounts in synoptical observations. - The seasonal cycle of precipitation in HIRLAM differed somewhat from the other sources, although the overall pattern agreed fairly well. Notable were an overestimation of spring rainfall (due to many forecasted small rainfalls never observed) and an underestimation of summer rainfall (due to forecasted but underestimated high rainfalls). - Throughout the year and especially in autumn, the mean areal precipitation amounts generally follow the order PTHBV>MA11>MA22, which can be explained similarly to the first item above, whereas HIRLAM and RADAR are more variable. - The areal standard deviation in the HIRLAM data was low on a daily basis but high for the 2002 totals. Areal smoothing in the model may explain the low daily areal variability, whereas the high areal variability of annual totals may originate from orographically induced small but systematic areal variations. - In the RADAR data a distinct inhomogeneity was found, with precipitation amounts being consistently overestimated by ~8 mm/day over a region of 500-1000 km² (as estimated from the coarse, averaged grid) during October-December. The source of this problem has not been possible to identify during this study, we can only speculate that it is related to temporally improper functionality of the Ostersund radar, northwest of the study catchment. - Areal correlation decreased nearly linearly for all sources, with the highest correlation coefficients in HIRLAM and PTHBV, and the lowest in RADAR. Overall the data from HIRLAM and RADAR agreed reasonably well with the other, "gauge-derived" sources, but the present comparison highlighted some differences. In HIRLAM, the seasonal cycle differed somewhat. This difference is possibly of a systematic character, judging from the 2002 data, but longer series are required to verify it. The different tendencies of the areal variability in the HIRLAM fields on a short-term (daily; low variability) and a long-term (annual; high variability) basis, respectively, may deserve further investigation. The temporal and areal inhomogeneities found in the RADAR data also require further analysis to identify the source of the problem and to improve the applicability for, e.g., hydrological forecasting. The performance of the different precipitation sources for flood estimation and forecasting will be reported as the subsequent result of the project (nr. 14783).
We have implemented a real time VPR correction scheme to represent reflectivity at ground level. The correction is calculated to all ranges (0-250 km) in a network of 7 C-band Doppler radars. The basic principle is comparison of the beam-smoothed reflectivity aloft to the reflectivity at ground level. Both of them are estimated from the measured VPRs close to the radars (0-40 km) applying the known elevation angles, elevation angle of the horizon and beam width.
The objective of the radar superobservation generator is to render superobservations of radar radial wind velocity and reflectivity factor directly from polar volume data. The superobservation generation is an intelligent transformation from high resolution to low resolution polar volume geometries according to a number of constraints and with a number of derived variables. The polar-to-polar transformation is based on a strategy for improved polar to Cartesian transformation. The strategy insures that all input polar data within a horizontal search radius are interpolated when determining the value of a given output polar bin. To minimize spatial correlation of the super-observations, the horizontal search radius used to select high-resolution polar data in the interpolation is constrained to not exceed the half arc distance between two adjacent output azimuth gates. The arc distance will thus be shorter than the range bin spacing out to a significant range, causing the search radius to increase with range until it is greater than or equal to the half output bin spacing. To minimize vertical correlations of the super observations two scans are constrained not to overlap. This means that no vertical interpolation is conducted. Non-overlapping scans are iteratively determined such that, starting with the lowest scan, the closest non-overlapping scan is identified and added to the output volume; this scan then becomes the starting point and the closest non-overlapping scan to it is found. This continues until the output volume is complete, or until a highest allowed elevation angle is reached. The super-observation operator can be used to assimilate radar observations in numerical weather prediction models in order to improve the precipitation and wind forecasts.Case studies using Swedish radar measurements show promising results.
The observation operator developed for 4D-Var is essential for utilization of radar radial wind data when determining the best possible initial model state by variation techniques in Numerical Weather Prediction. The radar data contain valuable information on high spatial resolutions that cannot be obtained through other data types. The qualities of weather forecasts are heavily dependent on the quality of the initial state. Therefore, the observation operator may be a link to possibly improve weather forecasts. Early experiments have indicated some gain in forecast quality when utilizing radial wind data by applying the observation operator. However more extended studies are needed to confirm these promising results.
Vertical profiles of reflectivity (VPRs) are derived from the 3D polar volumes inside the range of 2 to 40 km from each doppler radar. Profiles are measured in 200 m thick layers at intervals of 15 minutes. Classification of VPR is based on actual freezing level height and profile's shape. The main use of VPR is the calculation of vertical reflectivity profile correction but at the same time calculations also produce useful statistics about VPR characteristics.

Searching for OpenAIRE data...

There was an error trying to search data from OpenAIRE

No results available