Weather Radar and Hydrology Edited by

Probabilistic rainfall warning system with an interactive user interface

JARMO KOISTINEN¹, HARRI HOHTI¹, JANNE KAUHANEN¹,
JUHA KILPINEN¹, VESA KURKI¹, TUOMO LAURI¹, ANTTI MÄKELĹ, PERTTI NURMI¹, PIRKKO PYLKKÖ¹, PEKKA ROSSI¹ & DMITRI MOISSEEV²

1 Finnish Meteorological Institute, PO Box 503, FI-00101 Helsinki, Finland

[e-mail]

2 University of Helsinki, Department of Physics, PO Box 64, FI-00014 Helsinki, Finland
Abstract A real-time 24/7 alert system is under development. It consists of gridded forecasts of the best estimate rainfall and exceedence probabilities of rainfall class thresholds over a continuous time range of 30 minutes to 5 days. Nowcasting up to 6 h employs a 51 ensemble member extrapolation of weather radar measurements together with lightning location and satellite data. From approximately 2 h to 2 days a Poor man’s Ensemble Prediction System (PEPS) will be used, employing the NWP models HIRLAM and AROME. The longest forecasts use ECMWF EPS data. The mixing of the ensemble sets is performed through mixing of accumulations with equal exceedence probabilities. Alert dissemination employs SMS messages via mobile phones. The interactive user interface facilitates free selection of alert sites and warning thresholds at any location in Finland.

Key words rainfall; probabilistic forecasting; radar; NWP; mobile services

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 400-406

Impact of small-scale rainfall uncertainty on urban discharge forecasts
A. Gires¹, D. Schertzer¹, I. Tchiguirinskaia¹, S. Lovejoy², C. Onof³, C. Maksimovic³ & N. Simoes^3,4

1 Université Paris-Est, Ecole des Ponts ParisTech, LEESU, 6-8 Av Blaise Pascal Cité Descartes,
Marne-la-Vallée, 77455 Cx2, France

[e-mail]

2 McGill University, Physics Department, Montreal, PQ, Canada

3 Imperial College London, Department of Civil and Environmental Engineering, UK

4 Department of Civil Engineering, University of Coimbra, Coimbra, Portugal
Abstract We used a multifractal characterization of two heavy rainfall events in the London area to quantify the uncertainty associated with the rainfall variability at scales smaller than the usual C-band radar resolution (1 km² × 5 min) and how it transfers to sewer discharge forecasts. The radar data are downscaled to a higher resolution with the help of a multifractal cascade whose exponent values correspond to the estimates obtained from the radar data. A hundred downscaled realizations are thus obtained and input into a semi-distributed urban hydrological model. Both probability distributions of the extremes are shown to follow a power-law, which corresponds to a rather high dispersion of the results, and therefore to a large uncertainty. We also discuss the relationship between the respective exponents. In conclusion, we emphasize the corresponding gain obtained by higher resolution radar data.

Key words multifractals; rainfall downscaling; urban hydrology; power law

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 409-414

Joint analysis of radar observation and surface hydrological effects during summer thunderstorm events
P. P. Alberoni, m. celano, r. foraci, a. fornasiero, a. morgillo & s. nanni

ARPA Servizio Idro-Meteo-Clima, Viale Silvani, 6 Bologna, Italy

[e-mail]
Abstract During summer 2010 a special project was carried out in Emilia-Romagna (north Italy) focused on the issuing of warning associated with severe weather effects on local territory (e.g. sewer systems, road, small catchments and urban hydrological problems). This project has been set up to understand limits and, hopefully to improve, actual capabilities in the operational issuing of warning procedure. The first result, as expected, was the separation into two main classes of situation which caused such types of problem. On the one hand we had weather events where the synoptic forcing was well defined and where numerical modelling was able to correctly forecast the occurrence of such events. On the other extreme we had weather events where numerical models fail to forecast due to a number of causes like: weak or wrong synoptic forcing, very localised intense precipitation events, thunderstorms. This work focuses on an analysis of some of the events included in the second category, to highlight the link between observations and local problems and to understand how radar observations can play an important role in warning emission for this type of event. To tackle this ambitious goal, radar QPE is analysed along with the geo-localisation of surface problems occurring during the event; further particular attention is paid to the time evolution of surface precipitation pattern. An analysis of Limited Area Model performance will be carried out to highlight if, and in which circumstances, a very high resolution run could improve forecasting capability in such an event.

Key words warning; severe events

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 415-420

Observations of hailstorms by X-band dual polarization radar
SHIN-ICHI SUZUKI, KOYURU IWANAMI, TAKESHI MAESAKA,
SHINGO SHIMIZU, NAMIKO SAKURAI & MASAYUKI MAKI

National Research Institute for Earth Science and Disaster Prevention, 3-1 Tennoudai, Tsukuba, Ibaraki 305-0005, Japan

[e-mail]
Abstract Weather spotters reported small hail associated with convective storms during 2008–2010 in the Kanto area, Japan, and several of the storms were observed by an X-band dual polarization radar located in Ebina city, near Yokohama, Japan. Observed reflectivity Z_H and polarimetric parameters (e.g. differential reflectivity Z_DR, the specific differential phase K_DP, and the correlation coefficient_HV) were analysed in terms of hail detection by radar. Attenuation correction was applied to Z_H and Z_DR using the self-consistent method, because this correction is essential for X-band radar data. In several reflectivity cores, the rainfall rate estimated from K_DP was smaller than that estimated from Z_H, especially in regions where Z_H > 50 dBZ. This finding indicates the presence of hail, because K_DP is insensitive to hailstones. In many cases, the occurrence of small (but non-zero) values of K_DP and small values of _HV (<0.9) indicate the presence of wet hail or a mixture of hail and rain. Z_DR was smaller than that expected from Z_H in the case of rain. These results are consistent with those from S-band radars, suggesting the potential for detecting hail by X-band dual polarization radar.

Key words hail; X-band; dual polarization

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 421-426.

Multifractal study of three storms with different dynamics over the Paris region
I. Tchiguirinskaia¹, D. Schertzer¹, C. T. Hoang¹ & S. Lovejoy²

1 Université Paris-Est Ecole des Ponts ParisTech LEESU, 6-8 Av Blaise Pascal Cité Descartes, Marne-la-Vallee,
77455 Cx2, France

[e-mail]

2 McGill University, Physics Department, Montreal, PQ, Canada
Abstract Research is now triggered by the permanent need to better relate the measured radar reflectivity to surface rainfall. Knowledge of flow structure within cloud formation systems and the associated convective–stratiform separation may provide useful information in this respect. We will first discuss how stochastic multifractals can handle the differences of scales and measurement densities of the raingauge and radar data; and help to merge information from these data. Mosaics from the Météo-France ARAMIS radar network are used that correspond to horizontal projections of radar rainfall estimates for a 1 km × 1 km × 5 min grid over France. In particular, three storm events with different dynamics over the Paris region were selected to illustrate the efficiency of the multifractal framework. In spite of the difficulty that usually the same precipitation field comprises both stratiform and convective formations, their respective scaling properties allow the deciphering and classification of the radar data.

Key words multifractals; convective-stratiform formations; rainfall extremes; power law

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 429-434

Weather radar and hydrology: a UK operational perspective
robert j. moore, STEVEN J. COLE & ALICE J. ROBSON

Centre for Ecology & Hydrology, Wallingford OX10 8BB, UK

[e-mail]
Abstract Weather radar forms an essential and integral tool for water management in the UK, especially for monitoring and warning of flooding: the main focus of this perspective paper. An overview is first given of the radar network and its associated rainfall data products used by the environment agencies responsible for flood defence. The Hyrad (HYdrological RADar) system is deployed to receive, visualise and analyse these products, and to further process them for use within flood forecasting systems. Regional systems employ networks of models configured to make forecasts at specific locations. Very recently, countrywide systems employing an area-wide G2G (Grid-to-Grid) hydrological model have been implemented. Both types of system, used operationally in a complementary way, are reviewed in relation to their use of, and demands for, weather radar-related data. Activity on implementing probabilistic approaches to flood forecasting which benefit from using radar in ensemble rainfall prediction is outlined, and future prospects discussed.

Key words weather radar; hydrology; rainfall; flood; forecasting; distributed hydrological model

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 435-440.

On the accuracy of the past, present, and future tools for flash flood prediction in the USA
JONATHAN J. GOURLEY¹, ZACHARY L. FLAMIG², YANG HONG² & KENNETH W. HOWARD¹

1 National Severe Storms Laboratory, 120 David L. Boren Blvd., 73072, Norman, Oklahoma, USA

[e-mail]

2 Atmospheric Radar Research Center, University of Oklahoma, 120 David L. Boren Blvd., 73072, Norman, Oklahoma, USA
Abstract The skill of the USA National Weather Service’s flash flood guidance tool has been quantified from 2006 to 2008 using a combination of flash flood observations from spotter reports, automated stream discharge measurements, and witness reports from the public. A 15-year radar-based rainfall archive was used to run a distributed hydrologic model, thus enabling the estimation of flood frequencies at every 4-km grid cell. Exceedences of these return period flows were considered as predictors of flash flooding, and were validated using the same aforementioned datasets to establish the skill of present flash flood guidance. Significant improvements were realised using the forward modelling approach. Given the advent of 1-km², 5-min radar rainfall observations, distributed hydrologic models, and increased computing power, all of which are commensurate with the scales of flash flooding, it is now possible to directly forecast the probability of flash flooding over the conterminous USA in real time.

Key words flash flood; radar; distributed hydrologic model

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 441-447

Real-time radar-rainfall estimation for hydrologic forecasting: a prototype system in Iowa, USA
Witold F. Krajewski, Ricardo Mantilla, Bong-Chul Seo, Luciana Cunha, Piotr Domaszczynski, Radoslaw Goska & Satpreet Singh

IIHR-Hydroscience & Engineering, The University of Iowa, Iowa City, Iowa 52242, USA

[e-mail]
Abstract The estimation procedure to generate rainfall maps in real-time consists of Level II radar volume data collection, quality checks of the acquired data, and rainfall estimation algorithms such as non-meteorological target detection, advection correction, Z-R conversion, and grid transformation. The rainfall intensity map that is generated using data from seven radars around the State of Iowa is updated at nominal 5-min intervals, and the accumulation map is produced based on 15-min, 1-h, and daily intervals. These rainfall products are fed into a physically-based flood forecasting model called CUENCAS that uses landscape decomposition into hillslopes and channel links. The authors present preliminary results of analysis done on real-time radar-rainfall products using raingauge data and hydrological simulations from flood events in 2008 and 2009. They also show how differences in rainfall forcing affect peak flow discharge.

Key words precipitation; radar-rainfall; flood forecasting

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 448-453.

Which QPE suits my catchment best?
M. Heistermann, D. KneiS & A. Bronstert

University of Potsdam, Institute for Earth and Environmental Sciences, 14476 Potsdam, Germany

[e-mail]
Abstract We often seek to identify from a set of available QPE products the one with the least error for a particular catchment. However, point-based verification approaches such as cross-validation do not inform the user about the spatial representativeness of the error. Instead, we can force a hydrological model with different QPE products and select the QPE which best reproduces the observed discharge. In order to reduce effects of model calibration on the outcome of such a “hydrological verification”, we propose a Monte Carlo-based approach. We applied this approach in a case study for two catchments in southeast Germany and found that hydrological verification and cross-validation can, in fact, usefully complement one another.

Key words weather radar; quantitative precipitation estimation; verification; rainfall–runoff modelling

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 454-459

Study on a real-time flood forecasting method for locally heavy rainfall with high-resolution X-band polarimetric radar information
MAKOTO KIMURA¹, YOSHINOBU KIDO² & EIICHI NAKAKITA²

1 Central Research Institute, Nihon Suido Consultants Co., Ltd., PO Box 163-1122, 6-22-1, Nishi-Shinjuku, Shinjuku-ku, Tokyo, Japan

[e-mail]

2 Disaster Prevention Research Institute, Kyoto University, PO Box 611-0011, Gokasyo, Uji Kyoto, Japan
Abstract In recent times locally heavy rainfall has occurred frequently in Japan and caused serious human accidents; hence the need for flood forecasting systems has increased to reduce inundation damage. However, flood forecasting that secures lead-time for evacuations is extremely difficult because conventional radars cannot adequately measure the rainfall. Under this circumstance, X-band polarimetric radars have been installed, and flood forecasting with a higher accuracy is expected. In order to develop and optimize a real-time flood forecasting method for locally heavy rainfalls in urban drainage areas, we have considered several flood forecasting models with various computational accuracies and loads. Furthermore, we have evaluated these models for comprehensive prediction accuracies through case studies in an actual basin using X-band radar information. As a result, the detailed flood forecasting model might not always have the highest accuracy and the proposed simplified model which can apply the latest rainfall information with lower computational loads was effective.

Key words real-time flood forecasting; X-band polarimetric radar; locally heavy rainfall; urban drainage areas

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 460-465.

A rainfall–runoff model and a French–Italian X-band radar network for flood forecasting in the southern Alps
D. ORGANDE¹, P. ARNAUD², E. MOREAU³, S. DISS², P. JAVELLE², J.-A. FINE¹ & J. TESTUD³

1 Hydris hydrologie, 5 Avenue du Grand Chêne, 34270 Saint-Mathieu-de-Tréviers, France

[e-mail]

2 IRSTEA, 3275 Route de Cézanne, CS 40061, 13182 Aix en Provence Cedex 5, France

3 Novimet, 41 bis Avenue de l’Europe, BP 264, 78140 Vélizy Villacoublay, France
Abstract The aim of the CRISTAL project (Gestion des CRues par l’Integration des Systèmes Transfrontaliers de prévision et de prévention des bassins versants Alpins) is to develop an operational flood forecasting system for catchments located in the French Southern Alps and Italian Piedmont, based on rainfall data from two dual-polarisation X-band radars. The study deals with the calibration and initialization of the rainfall–runoff model on gauged French catchments (45–461 km² in area) on the Siagne, Paillon and Roya rivers. The GRD conceptual rainfall–runoff model is calibrated in order to reproduce measured flow. The model initialization consists of establishing a calculation rule to define the value of the daily production parameter in relation to known variables (such as previous rainfall or evapotranspiration). Hydrological simulations of recent events measured by X-band radars are presented and compared with raingauge and water-level records.

Key words flood forecasting; X-band radar; rainfall–runoff model; calibration; initialization; French–Italian border

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 466-471

River flow simulations with polarimetric weather radar
M. A. RICO-RAMIREZ¹, v. n. bringi² & m. THURAI²

1 Department of Civil Engineering, University of Bristol, Queen’s Building, Bristol BS8 1TR, UK

[e-mail]

2 Department of Electrical Engineering, Colorado State University, Fort Collins, Colorado 80523-1373, USA
Abstract Polarimetric weather radars offer advantages over conventional radars such as removal of non-meteorological echoes, attenuation correction, hydrometeor identification and more accurate rainfall estimation in the rain region, all of which lead to an overall improvement in data quality and the subsequent improvement in rainfall estimation. However, how much of that improvement is translated into more accurate river flow simulations given the fact that hydrological models are also subject to uncertainties due to model parameters and model structure? This paper examines the use of radar rainfall estimations from an operational polarimetric C-band weather radar linked directly to a hydrological model for river flow simulations. Several rainfall events from the winter and summer seasons were considered for the analysis. Polarimetric rain-rate algorithms were developed for both seasons, based on several months of disdrometer data. The results are presented in terms of radar and raingauge comparisons (over a large raingauge network) as well as flow simulations.

Key words polarimetric radar; radar errors; attenuation; rainfall estimation; flood forecasting

Weather Radar and Hydrology

(Proceedings of a symposium held in Exeter, UK, April 2011) (IAHS Publ. 351, 2012), 472-477.

Using combined raingauge and high-resolution radar data in an operational flood forecast system in Flanders
INGE DE Jongh¹, Els quintelier² & kris cauwenberghs²

1 VMM, Elfjulistraat 43, B-9000 Gent, Belgium

[e-mail]

2 VMM, Koning Albert-II-laan 20, B-1000 Brussels, Belgium
Abstract Since 2007 the Flemish Environmental Agency has an operational flood forecast system for Flanders. On the website www.overstromingsvoorspeller.be the water manager, civil services and interested citizens can follow in real-time how the situation of the rivers is progressing and what is expected in the next 48 h. For the past 48 h this real-time system uses a pseudo-CAPPI high-resolution radar composite of three radars (Zaventem, Wideumont and Avesnois (France)) combined with real-time raingauge data from more than 40 raingauges spread over Flanders, Belgium. Every 15 minutes, catchment rainfalls are calculated from the updated radar images and the hydrological models are run. Because at some locations a simple data-assimilation technique is used in flood forecast construction, using real-time river flow measurements, the historical catchment rainfall is in fact mainly used to calculate the water balance in the soil. In addition, for the ungauged catchments in the region, the hindcast is built completely with modelling results using raingauge-adjusted radar rainfall as input. Radar rainfall data are known to better resolve the spatial rainfall pattern, in comparison with interpolated raingauge data. Therefore it is a very helpful tool in real-time hydrological forecasting. However, error propagation can make radar rainfall data sometimes spurious. The hydrological forecasts for small catchments are especially more sensitive to the accuracy of the radar rainfall input data. The combination of raingauge and radar data to correct the retrieved radar rainfall is therefore necessary. The influence of merging the raingauge and radar data on the performance of the hydrological forecast system is illustrated for some individual storm events.