Status and Perspectives on Using Radar Data in Hydrological and by yaoyufang

VIEWS: 51 PAGES: 80

									Status and Perspectives on Using Radar Data in
   Hydrological and NWP Models in Europe

             Edited by Silas Michaelides

                November 18, 2004
Contents

1   Introduction                                                                                      1

2   Austria (Robert Schatzl)                                                                          3
    2.1 Austrian weather radar net . . . . . . . . . . . . . . . . . . . . . . .                      3
    2.2 NWP-models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                        4
    2.3 Hydrological models . . . . . . . . . . . . . . . . . . . . . . . . . .                       4

3   Belgium (Laurent Delobbe)                                                                         5
    3.1 Radars in Belgium . . . . . . . . . . . . . . . . . . . . . . . . . . .                       5
    3.2 Use of radar data for hydrological applications . . . . . . . . . . . . .                     5
    3.3 Use of radar data in NWP models . . . . . . . . . . . . . . . . . . .                         7

4   Cyprus (Silas Michaelides)                                                                        8
    4.1 Radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .            .   .   .   .    8
    4.2 Rain-gauge Network . . . . . . . . . . . . . . . . . . . . . .               .   .   .   .    9
    4.3 Hydrological use of radar . . . . . . . . . . . . . . . . . . . .            .   .   .   .    9
        4.3.1 Radar Data Processing . . . . . . . . . . . . . . . . .                .   .   .   .    9
        4.3.2 Comparison between Radar Data and Rain-gauge Data                      .   .   .   .   10
        4.3.3 Comparison between Ground-based and TRMM radar                         .   .   .   .   10

5   Czech Republic (Milan Salek)                                                                     11
    5.1 Technical background . . . . . . . . . . . . . . . . . . . . . . . . . .                     11
    5.2 Utilization of the radar-based quantitative precipitation estimates in
        hydrological modelling . . . . . . . . . . . . . . . . . . . . . . . . .                     12

6   Denmark (Xianq-Yu Huang)                                                                         15

7   Finland (Bertel Vehvilänen)                                                                      16
    7.1 Kyrönjoki basin . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                    16
    7.2 Kemijoki basin . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                     17

8   France (Jean-Luc Chèze)                                                                          20
    8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . .   .   .   .   .   .   .   .   .   20
    8.2 Present status . . . . . . . . . . . . . . . . . . . . . .   .   .   .   .   .   .   .   .   20
    8.3 On-going projects . . . . . . . . . . . . . . . . . . . .    .   .   .   .   .   .   .   .   21
        8.3.1 Monitoring soil moisture and stream flow . . .          .   .   .   .   .   .   .   .   21
        8.3.2 Flood forecasting for medium size watersheds .         .   .   .   .   .   .   .   .   22


                                           i
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe                                         ii


            8.3.3     Flash flood awareness service . . . . . . . . . . . . . . . . .                                         23

9   Germany (Jan Handwerker and Christian Keil)                                                                              25
    9.1 The German radar network . . . . . . . . . . . . . . . . . . . . . . .                                               25
    9.2 Use of radar data for hydrological applications . . . . . . . . . . . . .                                            26
    9.3 Use of radar data for NWP . . . . . . . . . . . . . . . . . . . . . . .                                              26

10 Greece (Vassiliki Kotroni)                                                                                                27
   10.1 Weather radars in Greece . . . . . . . . . . . . . . . . . . . . . . . .                                             27
   10.2 NWP-models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                               27
   10.3 Hydrological models . . . . . . . . . . . . . . . . . . . . . . . . . .                                              28

11 Hungary (Akos Horvath)                                                                                                    29
   11.1 Introduction . . . . . . . . . . . . . . . . .               .   .   .   .   .   .   .   .   .   .   .   .   .   .   29
   11.2 The numerical model and objective analysis                   .   .   .   .   .   .   .   .   .   .   .   .   .   .   30
   11.3 Experiments and a case study . . . . . . . .                 .   .   .   .   .   .   .   .   .   .   .   .   .   .   31
   11.4 Summary . . . . . . . . . . . . . . . . . .                  .   .   .   .   .   .   .   .   .   .   .   .   .   .   34

12 Ireland (Michael Bruen)                                                                                                   35
   12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                           35
   12.2 Future Plans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                           35

13 Italy (Pier Paolo Alberoni, Mauro Tollardo)                                                                               36
   13.1 CIMA activities on hydrometeorological application of radar mea-
         sured fields: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                          36
   13.2 Adige River Flood Forecasting System . . . . . . . . . . . . . . . . .                                               37
   13.3 Use of weather radar data in Piemonte for hydrological risk management                                               38
   13.4 Use of weather radar data in Emilia Romagna for hydrological risk
         management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                            38

14 The Netherlands (Iwan Holleman)                                                                                           40
   14.1 Current status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                           40
   14.2 Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                           42

15 Norway (Uta Gjertsen)                                                                                                     44

16 Poland (Jan Szturc)                                                                                                       46
   16.1 Weather radar network in Poland POLRAD . . . . . . . . . . . . . .                                                   46

17 Portugal (Manuel Rosa Dias)                                                                                               50
   17.1 Use of radar data in hydrological models . . . . . . . . . . . . . . . .                                             50
   17.2 Use of radar data in NWP models . . . . . . . . . . . . . . . . . . .                                                51

18 Slovenia (Gregor Gregoric)                                                                                                52

19 Spain (Joan Bech)                                                                                                         54
   19.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                           54
   19.2 Radar and NWP . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                              54
   19.3 Radar and hydrology . . . . . . . . . . . . . . . . . . . . . . . . . .                                              57
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   iii


20 Sweden (Daniel Michelson)                                                           59
   20.1 Research use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .       60

21 Switzerland (Andrea Rossa)                                                          62
   21.1 Radar network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .      62
   21.2 Operational use of radar data in hydrological models . . . . . . . . .         62
   21.3 Future challenges and strategies . . . . . . . . . . . . . . . . . . . .       64

22 United Kingdom (Robert Moore)                                                       66
   22.1 Operational use of radar and NWP data in hydrological models . . . .           66
   22.2 Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .     68

Acknowledgements                                                                       69

Bibliography                                                                           70
Chapter 1

Introduction

This report is an assemblage of the contributions from all the countries participating
in Cost 717 (Use of radar observations in hydrological and NWP models). The report
was prepared to include all contributions from these countries and forms the basis for
drafting a respective Chapter in the Final Report of Cost 717. The countries whose
reports are presented in the following are:

   • Austria

   • Belgium

   • Cyprus

   • Czech Republic

   • Denmark

   • Finland

   • France

   • Germany

   • Greece

   • Hungary

   • Ireland

   • Italy

   • Netherlands

   • Norway

   • Poland

   • Portugal

   • Slovenia


                                          1
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   2


    • Spain

    • Sweden

    • Switzerland

    • United Kingdom

The name(s) of the contributors who have communicated a county’s report are noted,
respectively.
Chapter 2

Austria (Robert Schatzl)

2.1    Austrian weather radar net
Austria operates four weather radars located at Vienna Airport (Schwechat), Feld-
kirchen (near Salzburg), Zirbitzkogel (near Klagenfurt) and Patscherkofel (near
Innsbruck) all of them being C-band doppler radars operated by Austro Control
(Air traffic control). Automatic calibration in the reflectivity domain is performed
by ingesting a known signal into the receive channel of radar and by subsequent
adjustment of the radar constants. This check is carried out regularly (e.g. ever two
weeks) semi-automatically. An inspection by the maintenance team is performed once
per month including manual calibration of the radar.

    Special attention is paid in the correct clutter treatment due to the mountainous
surrounding of the radars. Beside the build in Doppler based clutter removal a spe-
cial clutter removal algorithm has been implemented by TU-Graz based on statistical
methods. This process is carried out in the following stages:

   • Stage 1 - Featurisor: Calculates a number of parameters (= features) from the
     base data: For each pixel the parameters are calculated from the Z, V and W
     moments, using images ranging over one or more days. This causes the history
     of each particular pixel to be taken into account.

   • Stage 2 - Discriminator: The single parameters are classified according to certain
     rules to get a single clutter probability value p for each pixel.

   • Stage 3 - Thresholder: Each clutter-probability is compared with a threshold to
     achieve a clutter yes/no decision.

   • Stage 4 - Elimination: All clutter pixels are replaced by an estimate (the median
     value) of the surrounding pixels.

   • Stage 5 - Plausibility Checks: The last stage performs a number of plausibility
     checks for each pixel. Again a number of parameters (like the intensity height
     gradient etc.) are calculated and compared with thresholds. Pixels which exceed
     the threshold are eliminated (replaced by an estimate based on the surrounding
     pixels).

                                         3
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe      4


2.2      NWP-models
In 2003, the Austrian Met service (Zentralanstalt fur Meteorologie und Geodynamik,
ZAMG) began to evaluate weather radar data in quantity and to process these data
(in contrast to the qualitative use through the meteorologist for storm warning etc.).
At the moment, a nowcasting algorithm is in operational use, which performs an
extrapolation of the precipitation amount of the next two hours from the observed
shift of the radar echos. In this algorithm, no data of NWP - models are used.

    At the moment, a system is developed (end product around 2005), which combines
radar data and precipitation data from rain gauges with forecasts from models. The
system is called INCA (Integrated Nowcasting through Comprehensive Analysis) and
includes the registration of the present precipitation status using all data available and
based on this an improved NWP - forecast for the next hours. It is planned to perform
the synthesis radar data - model outside the NWP - model, that means a correction of
the NWP forecast. The system is working with high spatial and temporal resolution (1
km and 15 minutes).

    In a second step (about 2006) it is planned to include radar data directly in the
starting conditions of the NWP - model (in this case ALADIN).


2.3      Hydrological models
At the moment, there is no operational use of radar data in hydrological models
in Austria. In the past, some research was done especially in smaller catchments,
where radar data where used to evaluate flood events caused to the lack of observed
precipitation data.

    Furthermore, there was also some research in the comparison between radar
data and rain gauge data (Koeck, 1998; Schatzl, 2001) especially for Western Styrian
catchments, where it was tried to find an algorithm (also for operational use) for the
calibration of radar data.

    At the moment and in the next future, some hydrographical services of the
Austrian federal states are or intend to use radar data for the qualitative estimation of
rainfall events.

    In Styria, it is planned for the next future to use radar data in an existing rainfall -
runoff model for the Western Styrian catchments (Kainach, Lassnitz and Sulm), where
the density of the rain gauge net is also very high. Some results of this project can be
expected in the next year.
Chapter 3

Belgium (Laurent Delobbe)

3.1    Radars in Belgium
Two weather radars are operational in Belgium: the radar of Wideumont which is
operated by the Royal Meteorological Institute of Belgium (RMI) and the radar of
Zaventem operated by Belgocontrol, the Belgian public company ensuring the air
traffic safety. The radar of Wideumont was installed in 2001 in the South of Belgium
near the borders with France and Luxembourg. It is a C-band Doppler radar with a
Magnetron transmitter (Gematronik METEOR 500 C). The new radar of Zaventem
near Brussels has been operational since 2003. It is a C-band Doppler radar with a
Klystron transmitter (RADTEC/SIGMET).

     The installation of a third radar is planned near the coast. This project is a joint
initiative of the Royal Meteorological Institute of Belgium and the hydrological service
of the Flemish Region (AWZ). Finally, the hydrological service of the Walloon Region
and the RMI are involved in the project "Radar du Nord" led by Météo-France. The
project is supported by the European INTERREG IIIA program. The aim is to install
a new weather radar in the north of France near the Belgian border and to strengthen
the exchanges of radar and gauge data between the three partners.


3.2    Use of radar data for hydrological applications
Radar observations are operationally used as input for hydrological modelling in
the framework of a new flood forecasting system called "Operational Basin Model-
Demer" (OBM-Demer). The system has been operational since April 2003. It is
operated by the division water of the Ministry of Flanders - Aminal. The Demer basin
is 2275 km2 large and is situated in the Southeast of Flanders. A composite radar
image generated at RMI and covering France, the United Kingdom, the Netherlands,
Luxembourg and Belgium is transmitted every 15 minutes. The forecast system is
mainly based on the FloodWorks and HYRAD softwares (Wallingford Software and
Centre for Ecology and Hydrology - UK).

   Radar observations from Wideumont and Zaventem are qualitatively used by the
hydrological service of the Walloon Region (MET/SETHY) for the monitoring of


                                           5
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   6




Figure 3.1: precipitation accumulation per river catchment: comparison between the
estimates from the radars of Wideumont and Zaventem

precipitation over the Walloon river catchments, e.g. the Meuse catchment. Radar
products used by the hydrologists are instantaneous images, 1h and 24h precipita-
tion accumulation images, and 1h and 24h accumulations per river catchment. Many
catchments are covered by both radars which allows the comparison of the two radar
estimates (Figure 3.1). The MET/SETHY operates an automatic rain gauge network
of about 100 stations distributed over an hydrological area of 16842 km2 . Such a dense
network offers the opportunity to efficiently merge rain gauge and radar data. RMI is
developing for the MET/SETHY a software tool for the joined analysis of gauge and
radar data. This ongoing project is a first step towards a real time adjustment of radar
precipitation data.
    As far as research activities are concerned, RMI is involved in a close collaboration
with the Hydrology and Quantitative Water Management Group of the Wageningen
University concerning the hydrometeorology of the Ardennes region. The aim of the
project is to investigate whether an improved assessment of the space-time structure
of precipitation, as can be obtained with the radar of Wideumont, in combination with
an innovative approach towards modelling the rainfall runoff process will lead to an
improved understanding of the hydrometeorology of Ardennes catchments. In the
framework of this project, correction procedures based on volume reflectivity data,
such as attenuation and VPR corrections, will be implemented and tested. In the same
line of research, several Belgian partners are involved in the Floodsite project of the
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   7


EC Sixth Framework Programme. The project includes pilot studies on flash flood
forecasting in four application sites. The Ardennes Area is one of the pilot application
sites. The Ardennes pilot study is led by the Wageningen University and includes
groups from The Netherlands, Belgium and Luxembourg. The radar of Wideumont
provides a very good coverage of most of the Ardennes area and the contribution of
weather radar to the forecasting of flash flood events will be an important aspect of the
Ardennes pilot study.


3.3      Use of radar data in NWP models
RMI participates in the Aladin consortium. A Belgian version of the Aladin model
is operationally run for the regional weather forecasts and RMI contributes to the
research activities of the consortium. There is currently no assimilation of radar
data in the Aladin model and no research activities in this field are carried out in
Belgium. The use of radar precipitation data is limited to the verification of the Aladin
precipitation forecast for a number of selected episodes.

     There is presently a growing interest in assimilating radar wind data in NWP
models. Radar wind profiles are extracted from Doppler radial velocity data for
the two Belgian radars. Wind profiles are generated every 15 minutes for the radar
of Wideumont and 5 minutes for the radar of Zaventem. The quality of the wind
profiles is evaluated within the European CWINDE project (COST Wind Initiative
for a Network Demonstration in Europe). The profiles are sent to the CWINDE data
hub run by the UK Met Office for the Eumetnet WINPROF programme. Monthly
statistics on the data quality are generated by comparing the data with the Met Office’s
numerical model output. A future assimilation of Belgian radar wind profiles in
European NWP models can be considered.
Chapter 4

Cyprus (Silas Michaelides)

4.1    Radar
In 1995, the Cyprus Meteorological Service was equipped with an Enterprise Elec-
tronics Corporation (EEC, U.S.A) Doppler radar, which was originally designed for
use in now-casting at the Weather Forecasting Office at Larnaka Airport. Since its
installation, this radar is used exclusively by the Weather Forecasters operationally
and especially, in issuing warnings for Thunderstorms and related hazardous weather
phenomena, mainly to the aviation operators and authorities but also to a number of
other users, such as mariners, fishermen, the public, etc.

     The radar is installed on the North-western mountainous region of the island, near
Kykkos, a Medieval monastery, hence the name Kykkos given to the radar site. The
radar is a C-band one with Doppler capabilities. The information from the radar site
is transferred to the weather forecasting unit at Larnaka Airport, on the south-eastern
coast of the island, via a combination of UHF and land-line configuration.

    The radar images can be processed locally, both at Kykkos and at Larnaka airport,
by using EDGE, a purpose specific software, provided by EEC, recently upgraded.
Although the operator at Larnaka Airport has a lot of freedom in using the radar in
a surveillance (real time) mode and view images from the radar site, there is also the
possibility to store volumes of radar data, according to pre-defined scan strategies.
However, few if any of these facilities have been used in the past, the main reason
being the lack of any access to the digitally stored information. This information is by
default stored in a binary format, proprietary to the manufacturing company, EEC.

     The radar hardware is supplemented with two Unix based computers with tape
archiving facilities. As part of the COST 717 activities, the radar’s computer network
has been expanded with two additional computers (one PC and one Unix based Work-
station) and the whole system is now connected to the internet, via a dial-up connec-
tion. As a result of these upgrades, it is now possible to access and process the radar’s
digital data, remotely or from other computers on the network, without disturbing the
operational use of the system by the Weather Forecasters.



                                           8
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   9


4.2      Rain-gauge Network
Cyprus Meteorological Service operates a large number of daily rainfall measuring
stations. Many of these stations are operated by trained part-time observers and some
are operated by professional staff. A total of 146 stations are currently operational
and measure daily rainfall, thus providing a very good coverage of the Government
controlled area. All of the currently operating 146 rain gauges on the island, and the
radar site itself, have been georeferenced.


4.3      Hydrological use of radar
The radar can be used in order to improve flood detecting and forecasting because of
its ability to provide high spatial and temporal resolution rainfall estimates. Its data
can be used as numerical input to a hydrological model to make hydrological forecasts
more accurate.

     Weather radar gives detailed information concerning precipitation distribution over
rain-gauge measurement. The comparison of ground-based radar data, space radar
data, in situ data (rain-gauge measurements) and numerical model data are important
in validating the measurements of these different methods used in measuring precip-
itation fields. Processing and procurement of good calibration data is a pre-requisite,
therefore strategies to improve and standardize calibrations should be developed. Since
Cyprus Meteorological Service became a member of the VOLTAIRE project, some rel-
ative work has been done concerning the numerical representation of the radar data and
calibration subjects.

4.3.1     Radar Data Processing
The radar application software, namely EDGE, is able to store binary volume files
that contain the respective information, as determined by the operator. The operator
can apply any possible scan strategy and store combinations of products in pictorial
form but also as part of the volume files. However, until recently, this procedure for
data storing rendered the system unavailable for any quantitative work. It’s necessary
to transform the original binary volume files into an easily accessible format; so it has
been provided by the company a basic software for transforming the binary digital
data into ASCII-dBZ format in a very straightforward way.

    Because sampling (start-end azimuth, spacing of rays etc.) varies from scan to
scan, each radar volume scan had to be interpolated to a fixed Polar grid (prior to their
conversion to the fixed Cartesian grid). As part of the effort to further post-process the
radar data, a software was developed which performs such a conversion from the bulk
ASCII-dBZ archive into Polar (x-y) coordinate system that could be easily used in a
GIS (Geographic Information System) application software.

    Data quality control is essential for comparing and using rainfall data in hydro-
logical models. An objective of VOLTAIRE is to improve the data quality of ground
radar data by reducing or correcting the measurement errors e.g. in mountainous ter-
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   10


rain. Three different clutter algorithms - Cluttermap, "Segment size" algorithm, and
"Texture based" algorithm - were chosen to minimate or eliminate ground clutter in
2D radar data (see Golz et al. (2004)).

4.3.2     Comparison between Radar Data and Rain-gauge Data
A comparison between radar data and rain gauge data was performed using the daily
rainfall. The reflectivity Z obtained by the radar is converted into a rain rate parameter
                                             Z
R through the following formula, R= 3 ( 316 )2 where R is measured in mm/h and
Z is measured in mm6 m-3. One daily map needs to be calculated so that it will
be compared with the daily rainfall measurements. This daily map is achieved by
averaging all maps of the day and then multiplying by 24 so that the values of the final
map will be measured in mm/day.

    Rain values over rain gauge sites are determined using GIS (Geographic Informa-
tion System) application software. These data are compared with rain gauge measure-
ments provided by the Cyprus Meteorological Service. For precision reasons, stations
strongly affected by ground clutter are excluded.

4.3.3     Comparison between Ground-based and TRMM radar
Cyprus is the only European country within the TRMM orbit. Therefore it provides an
excellent opportunity for comparison between the precipitation estimate techniques.
The TRMM’s data collection centre provides "hdf" format files, which contain infor-
mation about the whole orbit of the satellite. So it is necessary to define the area upon
which the comparison will be executed and then the map will be built up.
   Radar and TRMM maps must be comparable. Obviously, maps must deal with the
same area and must also have the same resolution. In fact, radar maps have a resolution
equal to 0.01◦ , ten times more TRMM resolution, so it is necessary to aggregate pixels.

    A method for evaluating offsets between the maps of the two radars consists in
through analysis of the surface extensions over predefined thresholds. Assuming that
precipitations can rotate and/or move but cannot change their intensity, it is possible to
evaluate the trend of the pixel number beyond predefined thresholds. Thresholds start
from 10 dBZ and end at 52 dBZ with steps of 3 dBZ. Excluding pixels covered by the
mask (a discarded area which has effects of ground clutter and beam occultation), for
every threshold the number of pixel is calculated. Then the median value maps are
created and compared with the TRMM maps.
Chapter 5

Czech Republic (Milan Salek)

5.1    Technical background
On the Czech territory there are two weather radars which are operated by the Czech
                                                              c
Hydrometeorological Institute (CHMI, see [Novák and Krᡠmar (2002)]). They
serve mainly for detection of precipitation-related weather phenomena, qualitative
nowcasting, wind measurement, quantitative precipitation estimation (QPE) and
atmospheric research. The radar reflectivity product used for precipitation estimate is
(pseudo)CAPPI 2km and for the precipitation estimates the VPR correction procedure
                                                          c
is optionally applied (for more details see [Novák and Krᡠmar (2001)]).

    The radar-based precipitation estimates for 1-hour, 6-hour and 24-hour accumula-
tions are routinely adjusted by Mean Field Bias (MFB, also called Adjustment Factor,
AF) which is calculated from time moving window of at least three ’precipitating’
days until the total average precipitation depth exceeds predefined threshold. Then
the MFB-adjusted estimate is combined with available raingauge measurements by
a simplified procedure of D.-J. Seo (see Fulton et al., 1998; Seo, 1998). Another
procedure of the Czech Institute of Atmospheric Physics combining the adjusted
radar and raingauge measurement is nowadays being tested on daily estimates, too.
The purpose of the Multisensor QPE system is NWP model verification, use in
hydrological information systems including hydrological modelling and information
support of general warning service of the Czech Hydrometeorological Institute (Šálek,
2000; Šálek et al., 2004).

    The CHMI is calibrating, testing and deploying set of hydrological models for
rural areas. They are intended to serve as an important component of the hydrological
forecasting system with special emphasis on flood warning service. Regarding the
precipitation input, the designers and operators still rely mostly on classical raingauge
measurements and for areal QPE they use the Thiessen’s polygon method.




                                           11
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   12


5.2      Utilization of the radar-based quantitative precipitation
         estimates in hydrological modelling
Utilization of weather radar in hydrology at the Czech Hydrometeorological Institute
began in the second half of nineties when the radar-based precipitation estimates
were monitored by the hydrological warning service. Only since the beginning of
new millenium started the use of radar data in hydrological models. The radar-based
hourly estimates were first used for calibration of the hydrological models; they
served mainly for hourly distribution of the 24-hour precipitation accumulation that
were recorded by the climatological network of manually operated raingauges.

    The situation changed during 2002 as new radar precipitation estimate processing
and the multisensor precipitation estimation on new 1x1 km grid was put in operation
at the CHMI. Then it was also technically more feasible to compute direct QPE for
predefined areas. It has to be noted that new attractive presentation of the radar infor-
mation (including radar-based QPE) called JSPrecipView, based on JavaScript/PHP
and available on the internal network of CHMI, was much more persuading (see also
Šálek et al. (2004)). It was one of the reasons why the hydrological community at the
CHMI got more interested in utilizing this new kind of information.

    First operational use of ’radar-influenced’ QPE in hydrological modelling was
the computation of the hourly estimates at the site of the planned and/or installed
raingauges. These values, averaged for the particular radar pixel and eight neighboring
areal elements, were derived from the combination of the MFB-adjusted radar estimate
and available raingauge measurements and served mainly as an auxiliary or alternative
value in case of missing measurement of the particular gauge. As the original radar
and MFB-adjusted radar estimates were available at the gauge’s location, too, it was
also possible to assess the quality of the particular rauingauge measurements.

    After successful tests of ’direct’ areal QPE in hydrological simulations in 2002,
new precipitation inputs from the combined radar-raingauge (merged) estimates were
prepared and tested, especially for hydrological model Hydrog that is used mainly in
the eastern part of the Czech Republic. For the sake of simple implementation and
use, the hourly merged radar-raingauge mean areal QPEs were first computed for
Thiessen’s polygons designed from the positions of planned/installed raingauges. The
system has been in use since autumn 2003 (see Fig. 5.1 and Fig. 5.2).

    Since the areal distribution of precipitation based on Thiessen’s polygons is too
artificial, some test have been performed using mean areal precipitation computed for
subcatchments of the Svitava River basin of area 1118 km2 , which is mostly in rural
area. The tests were performed using gage-only, MFB-adjusted radar and merged
estimates on three variants of average subcatchment size 108, 80 and 59 km2 . The test
                                 r
are summarized in (Šálek and Bˇezková, 2004) and showed the feasibility of using the
combined and/or radar areal QPE. It was confirmed that the radar provides valuable
information especially in convective precipitation but in these cases it is recommended
to use rather small areas for computation of mean precipitation (not more than 100
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   13




Figure 5.1: Areas of the Thiessen’s polygons for which the areal precipitation esti-
mates are computed (in the eastern part of the Czech Republic), along with the radar
sites and the maximum range of the CHMI radars. The polygons southwest of the
radar Skalky are used for Svratka and Svitava catchments, the polygons to the east of
                                    c
the radar Skalky are utilized for Beˇ va catchment.


km2 ).

   It should be stressed that the use of radar data is possible also because the Czech
weather radar network is well maintained and the two radars in the country have very
good radar horizons and the visibility of the meteorological targets is good.

    Since the radar-raingauge merged QPE shows good results, the areal estimates
will be used also in hydrological modelling for other river basins. Further testing will
be performed with the various configurations of areas for which the areal QPE are
computed. The work on the algorithm of the multisensor precipitation algorithm will
also continue in cooperation with the Institution of Atmospheric Physics (see e.g.,
Sokol et al., 2002). The possibility to use quantitative precipitation nowcasting will be
examined, too.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   14




Figure 5.2: Example of a discharge simulation with the raingauge-only input (upper
image) and radar-raingauge combined areal precipitation estimate (bottom image) at
Svitava river catchment at Bilovice, 5 October 2003. The smooth curve is the simula-
tion, the dashed line is the measured discharge.
Chapter 6

Denmark (Xianq-Yu Huang)

The DMI contribution to COST717 is mainly through the HIRLAM collaboration, in
which SMHI and FMI have made a major progress in assimilating VAD and radial
wind data using the HIRLAM variational data assimilation system. DMI has been one
of the key institutes for the development and maintenance of the variational system
and for the software component handling the wind profiler data and radar data.




                                        15
Chapter 7

Finland (Bertel Vehvilänen)

Data of weather radars operated by Finnish Meteorological Institute (FMI) have
been in test use at two basins in the hydrological forecasting system WSFS
(www.environmet.fi/ waterforecast) at SYKE. Due to the large potential improvement
in short-term flood forecasting with radar data the use of data will be implemented to
cover whole country during this year.

    The hydrological model used for hydrological forecasting is a conceptual water
balance model. The meteorological input is precipitation and mean air temperature.
Discharge, water level and snow line observations are used to update the hydrological
model before forecasting. The experiences gathered during the operational test use
with radar data are described here.


7.1    Kyrönjoki basin
SYKE began to co-operate with FMI during 1998 to use radar precipitation data in
flood forecasting (Vehviläinen et al., 2002) in the Kyrönjoki basin (4000 km2 ) in
western Finland (Fig 7.1). At the basin there are only three real-time precipitation
gauges available. The weather radar at Ikaalinen was 50-150 km from the basin. The
radar is outside the catchment, which causes some deterioration of the results. The
hydrological forecasting model is calibrated with rain-gauge data, which may be a
disadvantage when radar precipitation data are used.

   The main results and experience thus far are:

   1. Average correction terms for radar precipitation can been estimated through the
      water balance simulation of WSFS. Precipitation correction term estimation is
      a standard procedure in the updating of hydrological model simulation. These
      correction terms are used then in hydrological forecasting. Nowadays FMI cor-
      rects radar precipitation data also with an advanced profiling method (Pohjola
      and Koistinen, 2002), which increase the accuracy of radar data.

   2. During summer, practically no differences were observed in the hydrological
      forecasts made by the two model versions; one using radar and one gauge pre-
      cipitation as input.

                                         16
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   17


   3. Weather radar gives more realistic rainfall distribution estimates, but so far this
      has not increased the accuracy of hydrological forecasts. This is partly due to
      the updating of the hydrological model against real-time discharge data, which
      masks improvements of better precipitation distribution.

   4. In summer, rain gauge measurements can be replaced by radar precipitation
      measurements, especially when real time discharge measurements are available
      for updating of hydrological model, without deteriorating forecast accuracy.

   5. In winter the simulation of water equivalent of snow using radar data is not ac-
      curate enough (Vehviläinen et al., 2002). Seasonal snowfall correction terms im-
      prove results considerably. In winter the precipitation correction was estimated
      against areal snow water equivalent based on snow line measurements.

   6. The simulation time-step in the hydrological model is one day. This is too long
      to get full benefit from radar precipitation data. The time-step will be shortened
      in the future.

   7. Soil moisture deficit error after dry rainless period, affects strongly on flood
      forecasting accuracy. The error in soil moisture deficit in millimetres can be
      half or even more compared to the precipitation event. The error of soil moisture
      originates mostly from errors in evapotranspitation simulation and is difficult to
      correct without soil moisture observations, which are not available.

    The radar data used in this application was mostly uncorrected radar data. Still, it
was clearly observed that even uncorrected radar data could be used in hydrological
forecasting in summer. This is important, if the number of rain gauges is reduced. FMI
is developing physical event-based precipitation correction methods (Saltikoff et al.
(2000) and Pohjola and Koistinen (2002)) and better corrected radar data are used in
operational forecasting already now.


7.2      Kemijoki basin
The Kemijoki basin, 50 000 km2 , in northern Finland has a weather radar in the
middle of the basin at Luosto (Fig. 7.1). The 150 km measurement radius of the radar
covers almost the whole basin.

    The hydrological model used is semi-distributed divided into 500 sub-basins
with mean area of 100 km2 . The areal precipitation and snow accumulation are
calculated on 25 km2 grid, but other processes (soil moisture changes, sub- surface
and groundwater flow) are calculated with sub-basin division. For each grid of 25 km2
precipitation is calculated from the nearest precipitation stations weighted by square
of the inverse distance between the grid and precipitation station.

    Two spatial precipitation values are available: one based on gauge observations
and one on radar precipitation. The more reliable estimate is chosen, or a weighted
sum of the values is used. The weight is based on the reliability of the precipitation
values decided earlier or if obvious errors can be detected case by case. The reliability
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   18


of the precipitation observation from gauge and radar depends mostly on the distance
from the target area. The other aspects, which affect on the reliability of the radar data
are form of precipitation (water/snow), amount of the precipitation.




Figure 7.1: The Kyrönjoki basin is at western coast and Kemijoki basin at the North.
The sites of corresponding weather radars at Ikaalinen and at Luosto with effective
measurement radius for rainfall (outer) and snowfall (inner).

    The procedure for estimating the daily precipitation for each sub-basin is follow-
ing:

   1. The precipitation for each sub-basin is estimated hourly from automatic and
      standard precipitation gauge observations and radar measurements.

   2. Hourly sub-basin precipitation is calculated based on the reliability estimate
      from gauge and radar data.

   3. Daily precipitation is calculated as a sum of hourly values.

    The reliability of the weather radar data depends on the amount of observed
precipitation. The weather radar might miss the precipitation, mainly when the clouds
are too low and too distant. It is less likely that the weather radar overestimates the
precipitation.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   19


    The main use of the hydrological model is to forecast floods and to give early
warning of floods. Therefore it is less dangerous to overestimate the possibility of a
flood than to underestimate it. Based on this the radar observation, which estimates no
precipitation, is considered to be less reliable than observation with some precipitation.
Especially observations, which indicate that there is a high amount of precipitation,
are always used even if the other source of observations disagrees. The radar data is
corrected by FMI with profile correction method (Pohjola and Koistinen, 2002).
Chapter 8

France (Jean-Luc Chèze)

8.1    Introduction
Although this report is focused on the use of radar data in hydrological models, it is
important to mention the qualitative use of radar images by operational hydrology. For
example, all the french flood warning services are equipped since many years with ter-
minals for visualization of radar images. But, coming back to a more quantitative use
of radar data, an operational application runned by a flood warning service is described
and also many on-going projects that deal with basins of different scales.


8.2    Present status
The DIREN Midi-Pyrénées (French acronym for Regional Direction for Environment
in Midi-Pyrénées) is in charge of flood warning activity on the Upper Garonne river
Basin. To hasten advances in the operational use of radar imagery for flood warning
application, a cooperative project started in 1995 by the DIREN Midi-Pyrénées and
Météo-France and the Laboratoire d’Aérologie within the framework of GISELE
(French acronym of Groupement Inter-organismes pour l’Evaluation des Lames
d’Eau). The aim of this project was to built and test a real-time system devoted to
runoff forecasting and flood warning, especially on the small upstream watersheds.

    The GISELE real-time hydrological system can be briefly described as follow :
radar data are processed on the Toulouse radar site itself by a dedicated computer.
The images of hourly rainfall amount are transmitted to the Flood Warning Service
(DIREN Midi-Pyrénées) also located in Toulouse, where they enter the local net-
work to be used as input for the rainfall-runoff modelling or to be visualized on a
VALERHY terminal (VALERHY is a system developed by Météo-France for the
display of meteorological information for hydrological users).

    Images of rain accumulations are computed from images generated every 5minutes
for 3 elevations (0.8, 1.4 and 2.2◦ ).The influence of the vertical reflectivity profile on
radar measurement can be corrected by the analysis of the radar signal at successive
elevations. The correction is performed in real-time since summer 1998.



                                          20
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   21


    The VALERHY system is used to display images of rainfall accumulations;
additional functions concerning the display of other meteorological information
(raingauge, satellite imagery, ...). The software computes the areal rainfall amounts
for different sub-basins by integrating the images of rainfall accumulations.

    Areal rainfall amounts for elementary watersheds are computed either from the
rainfall accumulations derived from radar data, or from the data of the raingauges
available on the watersheds.

    The results are issued from a runoff forecasting system based on a multi-model
procedure (SOPHIE). This procedure allows the comparison of the results of the
forecasts from either raingauge data or radar data.

    The SOPHIE software is based on a multi-model procedure. A forecast is
performed for a given station, using several models taking into account differ-
ent input data and in different ways. Each model contributes to the final result,
called weighted forecast, with a weight that is a function of errors registered for the
near past. So, the forecast fitting process takes into account errors due to bad estimates.

    Now, it is planned to implement the calculation of images of rain accumulations
on the new french radar computer (CASTOR 2) and to start the generalization of this
hydrological system to other flood warning services.


8.3      On-going projects
8.3.1     Monitoring soil moisture and stream flow
Based on the soil-vegetation-atmosphere transfer scheme ISBA originally developed
at Météo-France for numerical weather prediction a large scale hydrological model
has been designed. It is in essence an assimilation suite able to compute, in near real
time(i.e with a lag of one or two days) the evolution of soil moisture and stream flow
for major French rivers. This system is currently operational at Météo-France. The
system is of course mainly driven by precipitations which are retrieved from conven-
tional surface rain gauges. These point measurements are then interpolated using a
rather sophisticated scheme called SAFRAN. The spatial resolution obtained, 8km, is
however too coarse for some applications such as forest fire monitoring in some hilly
areas or pre-alert to floods over small watersheds. Also, in cases of strong precipita-
tions, some ground stations can be out of service (see comments associated to figure
8.1). Accordingly, attempts have been made to drive the system using radar estimates
of precipitations. A new analysis scheme has been designed which automatically se-
lects the radar information for areas of strong precipitations and surface rain gauges
estimations for stratiform, low intensity rainfall. The figure below shows the improve-
ment in streamflow prediction obtained using this method for the flow event in the
lower Rhtne valley in December 2003.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   22




Figure 8.1: Stream flow in Beaucaire (last gauging station on the Rhtne ) from Novem-
ber 1 to December 31. In red, best estimate available off-line running the model with
all surface information available. In blue, what would be obtained if the real time rain
gauges system was optimal (all gauges reporting). Yellow, estimation obtained in real
time using radar+rain gauges available as described above. Green, real time simulation
using the rain gauges that actually reported during the period. ( Morel et al, personal
communication)


8.3.2     Flood forecasting for medium size watersheds
In September 2003 the flood alert system in France has been reorganised. The new
organisation, with a central facility called SCHAPI ( French acronym for Service
Central d’Hydrometeorologie et d’Appui à la Prévision des Inondations) which
will promote the use of modern methods of flood forecasting in the operational
environment offers new perspectives. The use of weather radars in particular will be
more intensive. SCHAPI is currently testing several prediction methods.

    One of them was originally developed in a pilot project of the French ministry of
research called PACTES. A flood scenario is simulated, by "chaining" rainfall predic-
tion, runoff generation as computed by an hydrological model, stream flow, and finally
determination of water height and flood extent in the watershed. This kind of sim-
ulation chain can be run in two modes (using in fact different simulation models to
account for processing time constraints):

    • off-line mode: to assess flood scenarios and their impacts

    • on-line mode: for real-time forecasting, based on rainfall observations.

    Finally, the output of these "simulated flood maps" is combined with vulnerability
and infrastructure data to generate "information products" which can be directly in-
terpreted and used by the operational actors of the flood management process. This
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   23


approach based on "physical modelling" of the flood process can to some extent cir-
cumvent the difficulties associated with more empirical methods:
    • it is applicable in watersheds where historical data is scarce. The cost however
      is the large amont of data, starting with the digital elevation model, necessary to
      describe the watershed properly.
    • it is able to handle "extreme" events and in particular torrential floods for which
      historical records are not available.
    The information on precipitations is mostly based on radar estimates, in the
"off-line mode" a data base of extreme precipitation events available both in the study
area and in areas of similar climate. In some cases observed rainfall field are used and
the prediction potential of the system is only due to the concentration time of the flow
in the watershed. Alternatively, forecasted rain field are introduced in the system. For
small watersheds in sloppy areas this may turn out to be the only possibility to obtain
enough lead time. The forecasting method used at the moment for precipitations is
simply an "intelligent extrapolation" of the trajectories of the rain cells (so called
2PIR method). In both cases the radar is calibrated against surface measurements on
a monthly basis using the HYDRAM method. The anticipation possible using 2PIR
ranges from 30 minutes to 90 minutes.

    Another rather innovative project, called ESPADA has been developed specifically
for the city of Nimes which suffered badly from floods in 1988. ESPADA is a join
project between Météo-France and 2 consulting firms. The approach is the same as
above and again ,based on radar information and local rain gauges. Run off generation
is estimated using 2 separates models, one for the urban part for which the flow con-
centration is extremely rapid and one for the rural part which is complex because of
the karstic nature of the terrain. The use of the 2PIR method for precipitations now-
casting is planned. The method has been fine tuned to cope with large, quasi-stationary
precipitation systems. It has been shown instead that smaller scale targets have to be
used as traces. Interestingly, it has been shown that a standard extrapolation technique
won’t work very well here and that the targets from which the extrapolation field is
estimated have to be selected very carefully.

8.3.3     Flash flood awareness service
A regional flood warning system has been designed at the regional forecasting centre
of Meteo-France in Aix-en-Provence. The Aix-en-Provence centre is in charge of
weather forecasting for South Eastern France and Corsica where most of the strong
precipitations occur The objective is to give an appreciation of the risk in a given
situation for a region trying to describe the situation at small scale (i.e. sub-basin
scale). The method is not intended for accurate forecasting of precipitations over
a particular watershed. The forecaster gets every 15 minutes the precipitations
observed by radar and also the precipitations forecasted until the end of the hour. The
precipitations fields are considered as objects, which can be modified by the forecaster.

   At watershed scale a computation of run-off generation is performed using the
SCS (US Soil Conservation Service) methods which transforms precipitations into
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   24


run-off using only 2 parameters to describe the watershed and which takes into
account the moisture deficit.

     A "specific run-off" is then computed at the (1km)2 scale and compared to an ex-
isting climatology of the same quantity obtained using the same SCS method adapted
to the conditions of south eastern France by CEMAGREF. A risk intensity map of the
region is then produced . The method is currently tested both at Aix-en-Provence and
at the head quarters of Civil Security in Valabre.
Chapter 9

Germany (Jan Handwerker and
Christian Keil)

9.1    The German radar network
The German weather service (DWD) operates one of the densest radar networks
worldwide consisting of 16 C-Band Doppler radars. All radars are operated with the
same scan strategy, consisting mainly of a volume scan with 18 elevations vom 37◦
down to 0.5◦ with a maximum range of 128 km and an additional 5 elevation scan
from 0.5◦ up to 4.5◦ with a range of 230 km. These 23 elevations are repeated every
15 min. From this volume scan, local reflectivity products as well as national and
international composite products of radar reflectivity are generated.

    Embeded in this scan strategy, a one elevation scan (the so called “precipitation
scan”) is repeated every 5 min. This scan uses the lowest unshielded elevation for
each 45◦ sector and is used to determine the local precipitation intensity with a spatial
resolution of 1 km × 1 km, based on the Z/R relation Z = 256.0R1.42 .

    The radars are maintained on a biannual basis, including a regular calibration.

    Recently, DWD installed new computers on the radar sites. The capacities of the
old IT hardware limited the development of sophisticated routines. Thus, the rain
intensities are calculated from the reflectivity values without any correction, e.g. due
to vertical profile, attenuation, etc.

     In cooperation with the “Länderarbeitsgemeinschaft Wasser” (federal state work-
ing group water) the DWD initiated the project “Routine Procedure for the Online
Calibration of Radar Precipitation Data with the Help of Automated Ground Precip-
itation Stations (Radolan)”, which aims at the improvement of the radar derived rain
intensities. Main topics of this project are improved clutter correction, determination
of Z/R relations depending on time and location, online calibration on the basis of daily
to hourly accumulated precipitation, comparison with “ground truth” measurements,
and shortening the repetition rate of the precipitation scan. Finally, an online-software
shall be developed to support the use of quantitative radar data for hydrological pur-


                                           25
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   26


poses.


9.2      Use of radar data for hydrological applications
Hydrological services are within the responsibilities of the federal states in Germany.
Thus, the structures and tools are heterogeneous. Depending on the local orography,
the relevance of hydrological modeling varies as well.

    In some states, regular flow predictions are calculated operationally on a daily
basis (e.g. Baden-Wuerttemberg, Bavaria). In case of strong precipitation they
are repeated hourly. In other states predictions are only performed during critical
situations (e.g. Rhineland-Palatinate).

    The calculations are based on flow gauge and ground precipitation sensors. Pre-
cipitation predictions are taken from Lokalmodell (up to 48 hours) and Globalmodell
(up to 172 hours). Radar data are not used operationally yet.

    At present, radar data are distributed in real time to the states flow prediction cen-
tres for test purposes. These data consist of hourly Radolan calibrated precipitation
measurements with a 1 km × 1 km spatial resolution. The usefulness of these data
will be investigated offline after the next flash flood case. If the data will provide use-
ful additional information compared to the current system, it is planned to expand the
models to make use of this information.


9.3      Use of radar data for NWP
The situation at NWP is comparable to that at hydrological applications: There is no
operational use of radar data yet, but it is planned to make use of it in the near future
for validation as well as for assimiliation.

    Validation of model output will be based on the radar simulation model RSM
(Haase and Crewel, 2000). This tool calculates reflectivity fields based on the output
of NWP. These reflectivity values will be compared to the measured data from the
radar network.

   Until 2006 a latent heat nudging scheme (Klink and Stephan, 2004) shall be
implemented to assimilate 2D radar composite data into the Lokalmodell.
Chapter 10

Greece (Vassiliki Kotroni)

10.1     Weather radars in Greece
Operational network
The Greek operational radar network includes:

   1. two S-band radars, partially digitised and installed in the Northern part of Greece
      (Thessaloniki and Larissa) that are operated by the Greek Agricultural Insurance
      Organisation for its hail protection program.

   2. two non-digitized C-band radars installed during the 1970’s in Athens and in
      western Greece (Andravida) operated by the Hellenic National Meteorological
      Service for weather surveillance.

   3. a newly installed (spring 2004) C-band radar, installed in Egina island (located
      in the Saronic Gulf, 20-30 km south of Athens). The radar is mainly used for
      weather surveillance of the highly populated area of Athens and nowcasting
      purposes.

The Hellenic National Meteorological Service will enhance the radar network in the
near future. Indeed a project with a budget of 8000000 Euros is under way that will
include updating and digitisation of the four old radars (two S-band and two C-band)
as well as the installation of three new C-band radars.

   Research network

   • An X-band dual polarisation mobile radar operated by the National Observatory
     of Athens for research purposes including quantitative precipitation forecast val-
     idation and assimilation in high resolution NWP models (radar characteristics:
     9.3 GHz, capable of simultaneous transmission of horizontal and vertical polar-
     ization, with an antenna system of 8 feet diameter and 3 dB beamwidth).


10.2     NWP-models
The X-band dual polarisation mobile radar has operated along with a video disdrom-
eter and a number of raingauges during winter 2004 in the weastern part of Greece

                                           27
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   28


with the aim to (a) to validate radar measurements against raingauge and disdrometer
data, (b) use the radar measurements for high resolution NWP validation and (c) to
study the impact of assimilation of radar measurements on the high resolution precip-
itation forecasts provided by regional scale weather forecast models. Currently, the
assimilation effort is being carried out. Namely, a simple method of adjustment of
the humidity fields that are used as initial conditions for the high resolution (5km x
5km, and 2 km x 2 km grid increment) weather forecasts is applied and evaluated. The
method is evaluated with the use of BOLAM (hydrostatic) and MM5 (non-hydrostatic)
models that are run operationally by the National Observatoty of Athens since 1999
(http://www.noa.gr/forecast). The humidity adjustment method has been already eval-
uated for sixteen cases with heavy and widespread precipitation over Greece, based on
the used of satellite data over a wider area and it was proved that it was able to provide
improved precipitation forecasts mainly for medium and high precipitation amounts
(Lagouvardos and Kotroni, 2004).


10.3       Hydrological models
At the moment, there is no operational use of radar data in hydrological models in
Greece.
Chapter 11

Hungary (Akos Horvath)

11.1     Introduction
Thunderstorms and theirs related phenomena can cause serious problems in the
Carpathian basin (Horvath and Geresdi, 2001). In the operative practice of the
Hungarian Meteorological Service (HMS) there are two approach of thunderstorm
prediction. The first is the synoptic approach when meteorologists using general
meteorological information (instability indexes, numerical models, soundings, etc.)
indicate the probability of the thunderstorms for smaller or larger regions for a given
time period ("this afternoon"). The second approach is the nowcasting of severe
convective activity, using radar, satellite and other real time data. The nowcasting
system of HMS, named MEANDER has been developed for this reason: it uses all
real-time data and calculates advections of thunderstorms (Horvath and Geresdi,
2003). This system is able to forecast of 60% of severe thunderstorms 1 hour ahead.
The remaining 40% comes from "unexpected" convections: from cases where the
evolution of phenomena is stronger than theirs advection. The first problem is to
forecast the place and time of the first appearance of thunderstorms, which can
determinate the weather of the next few hours. The second problem is the evolution
of thunderstorms: pure replacement of convective cells can cause unacceptable
errors beyond 30-40 minutes. A possible solution of this problem could be the "fast
responding model run": to run a numerical model when the first significant appearance
of convective activity takes place in such a way, that strong trigger effects indicates
the real place of existing thunderstorms in the objective analysis. It gives a chance,
that the model develops thunderstorms on the right place and makes better predictions
than linear replacement.

   The problem of radar data assimilation for NWP is a fast developing field, this
works is supported by COST 717 action (COST, 2003; Macpherson, 2000; Gregoric,
2001)

    In this experiment the PSU/NCAR meso-scale model, MM5 was chosen (Grell
et al., 1994). This model has well developed cloud physic and planetary boundary
layer schemes and the applied high resolution (3 km) horizontal grid allows to avoid
using cumulus parameterization schemes but consider direct convection. The nudg-


                                          29
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   30


ing capability of the model allows to do nudging procedure which partly useful for
decreasing spin up time of the model and partly it helps to create longer time period
perturbations to indicate thunderstorms in the objective analysis.


11.2       The numerical model and objective analysis
The MM5 model is applied in two steps. A larger domain model is used to run for a
longer term (36 hours) which provides boundary conditions for the meso-scale version
of MM5 (figure 11.1).




      Figure 11.1: The larger and the smaller domain of the applied MM5 model.

    The horizontal resolution of the large model is 10 km , the small scale model
is 3 km. The small scale model uses ETA planetary boundary model (Janic, 1994),
Reisner mixed-phase explicit moisture scheme (Reisner and Rasmussen, 1998), and a
rapid radiative transfer model for radiation scheme (Mlawer and Brown, 1997). The
surface scheme is a five layer soil model (Dudhia, 1996).

    The model uses four dimensional data assimilation using nudging technique. Two
hours analysis nudging is applied. The objective analyses at 0 hour, +1 hour and +2
hours model time are made by the MEANDER system analysis segment. The radar
data are taken into account in the objective analysis. Model experiments show that
vertical profiles of equivalent potential temperature (EPT) of well developed thunder-
storms have characteristic profiles, they can be considered as conservative values in
the middle troposphere. This experience is confirmed by some cross section of EPT
field, calculated by MM5 (figure 11.2).
    The basic idea is that on grid points, where radar data indicates thunderstorms,
profiles of EPT are considered constant values. Supposing that the relative humidity
is 100% in thunderstorms, the modified temperature profile can be recalculated as
a function of the pressure profile. This modified humidity and temperature values
represent thunderstorms in the objective analysis. Obviously it is important question
how to determinate EPT anomalies in the thunderstorms. In this experiments EPT
of thunderstorms are calculated from the lowest 1000 m humidity and temperature
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   31




Figure 11.2: A vertical cross section of EPT at a case of severe thunderstorms (upper
figure). The lower figure shows EPT field at the surface, and the arrow indicates the
direction of the cross section.

values which were taken from the objective analysis of the MEANDER system.

    The above described method works only for convective precipitating systems, it is
not proper for stratiform clouds. To distinguish thunderstorms, only grid points with
radar reflectivity higher then 40 dBz were considered.


11.3       Experiments and a case study
10 case studies were made for different convective situations. These studies show
that introductions of thunderstorms can be used to upgrade short term (3-5 hours)
forecasts. Most successful cases are prefrontal unstable situations and isobar-less
synoptic patterns. At cases of unsuccessful objective analysis the effect of radar
nudging goes down rapidly in time, soon after the end of the nudging period. There
are some cases when the enhanced nudging forcing may help to catch the strong
convective phenomena, despite the poor objective analysis. The following case study
is a good example for this case.

   In Jun 9, 2002 at 18:00 UTC a squall line hit the middle part of Hungary. There
were two supercells in this system which caused tornadoes in Budapest and in
Kecskemet (100 km southeast of Budapest). These cells were indicated by the radar
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   32




Figure 11.3: Radar image of 09.06.2002. 18:30 UTC. Arrows show supercells close
to Budapest and Kecskemet.




 Figure 11.4: Forecast of hourly accumulated precipitation for 19 UTC, 09.06.2002.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   33


network of HMS (figure 11.3).

    Large scale MM5 (and other models like ECMWF) did not predicted any
convective activity in this region. A model run of small scale model (started at 15
UTC) with two hours of analysis nudging did not generate convection, either. When
radar data were introduced in the initial time (at 15 UTC) the convection appeared but
died out in 1 hour. When radar data were introduced in two hours analysis nudging,
thunderstorms remained active (figure 11.4).

   When the nudging coefficients of temperature and humidity were screw up to the
magnitude of Coriolis forece, one of thunderstorms (which hit Kecskemet) showed
supercell signature on the wind field (figure 11.5).




Figure 11.5: Forecast of hourly accumlated precipitation and wind of 925 hPa for 19
UTC, 09.06.2002.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   34


11.4       Summary
Case studies show that using radar data as trigger effects in numerical model may
help the model to place the convective events to the "right" place. It seems, that four
dimensional nudging assimilation method also helps to develop severe convective
phenomena like supercells. It is important to start the model in the very beginning
of the convective activity and doing as long nudging term as possible because it may
help the model to breed real convection and decrease not realistic ones.

   This work connected to, and used experiences of COST 717 activity and supported
by Hungarian Scientific Research Fund, under grant T043010.
Chapter 12

Ireland (Michael Bruen)

12.1     Introduction
At the moment there are three radars operational on the island of Ireland. Two in
the Republic of Ireland (Dublin and Shannon) operated by Met Eireann and one near
Belfast, operated by the Met Office. The data from the Dublin and Shannon radars
are only operationally used qualitatively by Met Eireann. However, radar data may be
consulted in a quantitative way after major events. Data from the Dublin radar is used
in the Carpe Diem research project.


12.2     Future Plans
There are no immediate future plans for extending the use of radar. However, at the
moment, the Irish Government is currently producing a flood strategy (in draft form at
the moment) and flood warning will be identified as an essential part of this strategy.
Once the strategy is agreed, and the position of warning vis-a-vis protection estab-
lished, it is likely that pre-implementation studies will be commissioned. It is likely
that radar and NWP will be identified as essential ingredients of a robust warning sys-
tem. However, major decisions on whether such a warning system should be operated
at National or local levels will only be taken at that stage.




                                          35
Chapter 13

Italy (Pier Paolo Alberoni, Mauro
Tollardo)

A number of different applications and developments are running in country. Most
of them are based at regional level, following the framework that share hydrological
responsibility based on catchments area. We present here some of them.


13.1    CIMA activities on hydrometeorological application of
        radar measured fields:
One of the CIMA main task is the study of the precipitation fields and the interaction
between rainfall and runoff measurements. These studies are needed to implement
and improve the existing hydrological models, to design civil protection systems and
to enhance the understanding of the physical mechanisms underlying the precipitation
process.

   From an hydrological point of view radar measurement are used to:
   • Study of small-scale statistical properties of rainfall for better understanding the
     interaction between the scales of the precipitation and those of the hydrological
     models.
   • Compare the river discharge estimates obtained by using rain-gage and radar
     measurements as input for rainfall-runoff models.
   • Study the behaviour of rainfall-runoff models under different kind of radar-
     measured rainfall fields.
   The radar measurements are used also for meteo-hydrological application such as:
   • Improving the rainfall downscaling models by studying the Fourier power-
     spectra of radar measured rain rate fields under different synoptic conditions.
   • Validation of the rainfall downscaling models.
   • Comparison between the statistical properties of the rain cells observed with
     radar and generated by high resolution limited area model (Lokal Model) simu-
     lations.

                                          36
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   37


    • Study of the shape of rain cells.


13.2       Adige River Flood Forecasting System
Acquisition of an optimal precipitation estimation product (GRID estimate at 1
km grid size over the upper Adige river basin). The optimal precipitation product
is based on merging of radar and raingauge data. Radar data will be processed through
a Structured Algorithm System (SAS).

    The SAS will integrate the following key features:

    • radar visibility in rugged topography,

    • identification of the impact of the 3D structure of the atmosphere and type of
      precipitation on the measurement,

    • correction of the effects of attenuation by precipitation,

    • capability to merge data coming from radar and raingauge data,

    • control of the radar system stability using reference targets like ground detection.

    The system will be able to operate with or without ground measurement. It will be
extendable to the combined use of satellite data with high spatial and time resolution
(like METEOSAT Second Generation - MSG).

    Development of specific algorithms.
    The following processing techniques will be developed:

    • Identification of the type of rainfall using radar image analysis differentiating
      zones associated to the different driving mechanisms of precipitation develop-
      ment.

    • Use of ground clutters to check the radar stability and analysis of the pulse to
      pulse variability of radar reflectivity for echo discrimination.

    • Automatic quality control for real time quantitative use of radar detection.

    Development and test of the concept of Flash Flood Guidance (FFG)
    FFG is the approximate threshold basin-average rainfall depth over a given
duration that would cause a small stream to begin flooding. One conservative measure
of flooding flow, which will be implemented into the algorithm, is the bank-full
discharge. The FFG rainfall amounts depend on the soil moisture state: wet soils
are characterized by relatively low FFG (and therefore high flood threat). FFG is
computed daily for a number of sections of the river network covering the operation
domain. The FFG rainfall depth will be computed using the hydrological models
mentioned above. The rainfall volume is assumed to be uniformly distributed over the
time period and over the catchment area. Comparison of these thresholds with radar
estimated rainfall provides real-time information on flood threat.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   38


    The proposed basic algorithm will use radar-derived rainfall estimates over differ-
ent duration length (30min, 1hr, 3hr, 6hr), consistent with the duration of the FFG. The
procedure for the determination of FFG rainfall depth will be part of a GIS software.
It will compute FFG-values for each catchment, and display the results on the stream
network of the region.


13.3       Use of weather radar data in Piemonte for hydrological
           risk management
ARPA Piemonte manages two multiparametric weather radars (Bric della Croce,
Torino, Monte Settepani, Savona) and about 350 ground stations. In addition to these
measurement facilities an hydrological model is run operationally. Raw polar radar
data are processed in order to obtain reliable rainfall estimates which can be used
jointly with rain gauge measurements and as input to the hydrological model. In
the area surveyed by radar, North-western Italy, complex orography and variability
of beam propagation conditions affect the observation of weather. For this reason
several processing techniques must be applied to radar data in order to suppress non
meteorological echoes (clutter) and correct for the vertical variation of reflectivity.
For hydrological risk management, with a focus on flash floods, radar derived
rainfall fields are operationally used as input for FEST, a distributed, physically
based hydrological model. The model is used either for assessing the rainfall-runoff
transformation and for flood wave routing. Runoff is calculated for each cell using
the SCS-CN model and it is routed along the slopes and channels in the whole river
network using the Muskingum-Cunge model. The structure of the model allows to
obtain the simulated hydrograph at any cross-section. No ad hoc calibration is needed
(Mancini et al. (2000), Ravazzani et al. (2002)). The input to the model is real-time
provided by the composite (Bric and Settepani radar) precipitation map over the
last hour. If rain is revealed from radar map, the model is run hourly. Hydrological
forecast are based on the hypothesis that future rain is null.

     It is now under investigation the opportunity to ingest in the model radar precipi-
tation maps corrected using the assessment factor calculated with rain gauges data and
filtered with a Kalman filter (see poster P0043 ERAD 2004 Conference). A short term
radar forecast (1-3 hours) will also be used to improve the hydrological forecast.


13.4       Use of weather radar data in Emilia Romagna for hy-
           drological risk management
ARPA Emilia Romagna manages two multiparametric weather radars (San pietro,
Capofiume Bologna, Gattatico Reggio Emilia) and a network of ground stations. In
addition to these measurement facilities hydrological models are run operationally
over some of the catchments within the regional domain. Raw polar radar data are
processed in order to obtain reliable rainfall estimates taking into account propagation
effects, interaction with the orography, anaomalous propagation. As presented at the
ERAD2004 (COST 717 Final Seminar) all these techniques ae used in order to obtain
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   39


a quality flag for radar data. Till now we are still in the process to define this type of
index and we plan to move toward an operational use of quality flag in the next years.
Radar data are combined with rain gauge measurements in order to obtain an input
field to the hydrological model.

    Radar data are used in the framework of Reno river flood warning system. They
are combined to gauges with a methodology developed within the MUSIC EU project
(refer to the appropriate section of this report). The system is now under operational
evaluation, plan for the future include the extensive use of radar rainfall data in other
catchments.
Chapter 14

The Netherlands (Iwan Holleman)

14.1     Current status
The KNMI radars perform a 4-elevation reflectivity scan every 5 minutes. From these
scans pseudoCAPPI images are produced with a target height of 800 m above antenna
level. Ground clutter is removed from the pseudoCAPPI images using the stepwise
procedure described previously. The reflectivity values are converted to rainfall in-
tensities using a fixed Z-R relationship, Z = 200R1.6 . In this way, 3-hour accumu-
lations are produced every hour and a 24-hour accumulation is produced at 08 UTC.
KNMI maintains a dense homogeneous network of about 325 volunteers who report
the amount of accumulated precipitation daily at 08 UTC. The density of these cli-
matological stations is about one station every 100 km2 . In addition, an independent
network of 33 automatic precipitation (AWS) gauges which report every 10 minutes is
maintained. From these 10-minute accumulations, 24-hour accumulations have been
calculated from 08-08 UTC which will be used as independent verification data.
     The radar-gauge adjustment algorithm is adapted from the scheme used opera-
tionally at FMI (Koistinen and Puhakka, 1981) and at the Baltex Radar Data Centre
(Michelson et al., 2000). The ratio between the radar (R) and the gauge (G) accumula-
tion in dB is used for the analyzes, because it is a Gaussian distributed quantity and it
has a greater correlation length than the precipitation field itself. First a range adjust-
ment is applied for each radar which removes a possible bias and a range dependent
bias due to VPR effects:
                                      R
                            F (r) ≡     = a + b · r + c · r2                        (14.1)
                                      G
In figure 14.1, the ratio F for the radar in De Bilt and the climatological stations is
plotted as a function of range for a single day. At short ranges the ratio is close to
0 dB indicating that the calibration of the radar is not too far off. At ranges longer than
roughly 100 km, the underestimation by the radar of the accumulated precipitation is
increasing. The target height of the pseudoCAPPI is below the horizon for distances
larger than 100 km, and data from just the lowest elevation has to be taken. Due to
overshooting of precipitation and gradients in the vertical profile of reflectivity, the
amount of precipitation is underestimated by the radar at long ranges. The solid line is
used to range adjust the radar accumulations.
    A spatial adjustment is performed on the composite of the range adjusted radar

                                            40
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe      41




Figure 14.1: This figure shows the ratio F for the radar in De Bilt and the climatolog-
ical stations is as a function of range for 8 November 2002.


accumulations using the Barnes objective analysis scheme. The field of the F ratios is
spatial interpolated using an inverse distance weighting method:
                                             N              2         2
                                             n=0 Fn · exp[−dn (i, j)/σ ]
                            F (i, j) =         N
                                                                                       (14.2)
                                                          2         2
                                               n=0 exp[−dn (i, j)/σ ]

where N is the number of radar-gauge ratios (Fn ), dn (i, j) is the distance between the
gauge location and the image pixel (i,j)σ is a scaling parameter. When the denominator
of F (i, j) is smaller than a certain threshold value it is set to this threshold value. In
this way the adjustment field is zero in regions with no precipitation gauges.
    A verification of the adjusted radar accumulations has been performed over a 7
month period in 2002. The radar accumulations have been adjusted using the 325
climatological stations and the adjusted radar accumulations have been verified using
the 33 independent AWS stations. The daily bias of the adjusted and unadjusted radar
accumulations as obtained from the verification are shown in figure 14.2. The inde-
pendent verification clearly shows that the gauge adjustment effectively reduces the
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   42




Figure 14.2: The bias of the adjusted and unadjusted radar accumulations as obtained
from the verification against the AWS stations.


daily bias of the radar accumulations. A bias larger than 0.5 mm is only occasion-
ally observed for the adjusted radar accumulations. In addition, the standard deviation
(not shown) of the adjusted accumulations is never larger than that of the unadjusted
accumulations, moreover it is often reduced by a factor of two or more.


14.2       Perspectives
Currently, adjusted 24-hour radar accumulations (using method described in previ-
ous section) and unadjusted 3-hour radar accumulations are delivered to commercial
service providers and water boards. In addition, KNMI has developed an automated
warning system for severe weather and/or precipitation in collaboration with about
50% of the water boards in the Netherlands. In this project “Severe weather for water
boards”, each water board can define a so-called warning profile based on past and
future precipitation. Using (adjusted) radar accumulations and precipitation forecasts
from the Hirlam NWP model, automatic warnings are issued when the pre-defined
criteria are exceeded for the area managed by a participating water board. Within this
project there is a demand for adjusted/optimized 3-hour accumulations. First of all, the
possibilities of physical corrections for the bright band and vertical profile of reflec-
tivity will be investigated (Holleman, 2004). In the near future, the water boards will
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   43


start supplying hourly gauge data to KNMI that are going to be used for adjustment of
the radar accumulations. It is expected that the majority of the water boards will join
this project and that thus an almost national warning system for severe hydrological
conditions will be established.
Chapter 15

Norway (Uta Gjertsen)

In Norway, quantitative precipitation estimates from weather radar, have not yet been
put into operational hydrological use. For the qualitative monitoring of precipitation,
radar reflectivity data is transferred from the Norwegian Meteorological Institute
(met.no) to the Norwegian Water Resources and Energy Directorate (NVE) in real
time. The operational flood forecasting at NVE provides forecasts of discharge twice
a week for the entire country, and issues flood warnings when necessary. The service
is based upon the forecasted discharge (daily values) from the Swedish conceptual
rainfall-runoff model HBV (Bergstrm and Forsman, 1973). The HBV model falls in
the class of lumped (or semi distributed) conceptual models which are adequately
suited to simulate the discharge at the outlet of a catchment, given long representative
data series of temperature, precipitation, and discharge for calibration. However, such
a fortunate situation with regards to data is rare.

    Increased demands on the accuracy of hydrological monitoring and forecasting,
for flood forecasting, water supply and hydropower management, make it necessary to
investigate the potential of radar precipitation data. NVE and met.no plan to explore
the possibilities of using the semi-distributed structure to employ the distributed
information as provided by weather radar data. At met.no, areal precipitation
estimates are derived routinely for the Norwegian radars. These are today used by
some hydropower companies.

    Radar precipitation data distributed on rain, sleet and snow is produced opera-
tionally at met.no. Information on the water phase of precipitation is obtained from
the temperature and humidity data of synoptic stations ((Gjertsen et al., 2004)). It
is planned to use the information on snow accumulation for avalanche forecasts and
forecasts of snowmelt and hydropower production potential.

    The weather forecasters at met.no use mainly radar reflectivity data to estimate the
location and intensity of precipitation areas and storms. In met.no’s meteorological
workstation DIANA, the forecasters have also access to gauge adjusted quantitative
precipitation estimates from radar. Radar precipitation is available both for single
sites and composites. Also accumulated precipitation and a product showing the water
phase of precipitation are integrated in DIANA. In the near future it is planned to add


                                          44
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe    45


radar wind information to DIANA.

     As a part of met.no’s collaboration with the UK Met Office, first steps towards
model verification using radar precipitation information have been taken (Forbes
et al., 2004). It is planned to continue this activity. An example is shown in Figure 15.1.

    Is planned to implement a radar-based nowcasting system in 2005. A precipitation
forecast based on extrapolated radar precipitation is mainly demanded by the Norwe-
gian Road Authorities in the planning of winter snow clearance. In a next step, the
radar data and high resolution NWP model will be integrated for improved short term
precipitation forecasting. Currently there is no activity on assimilation of radar data
into NWP models.




Figure 15.1: Comparison of radar-derived rain rate (top-left) with 1 hour accumulated
precipitation from the HIRLAM5 (top-right), UMHL (bottom-right) and UMMO (bot-
tom left) for 04 April 2004 at 07 UTC (T+7, T+13 for UMMO).from (Forbes et al,
2004)
Chapter 16

Poland (Jan Szturc)

16.1    Weather radar network in Poland POLRAD
The beginning of radar network in Poland is connected with aerological department of
Institute of Meteorology and Water Management (IMWM, Polish acronym is IMGW)
located in Legionowo near Warsaw (Warszawa). The first radar was British Decca
installed in 1964. It was replaced by Soviet MRL-2 in 1976 and then by MRL-5 in
1991.

    The last of them was modern radar which possibilities were developed by team
of aerological department. The main achievement was introduction of automatic
processing system. The system had been operated by IMWM by 2002.

     Frequency band                                                             C
     Doppler mode                                                             yes
     User software                                                    Rainbow 5
     Raw data types                                                     Z, V, W
     Raw data resolution (Bit)                             8 (Meteor 360AC: 4)
     Antenna                                                           parabolic
     Diameter (m)                                                             4.2
     Beam width (o)                                                           1.0
     Gain (dB)                                                                 45
     Polarization                                              linear horizontal
     Transmitter                     coaxial magnetron (Meteor 1500C: klystron)
     Wavelength (cm)                                                          5.3
     Frequency (GHz)                                                    5.4 - 5.8
     Pulse power (kW)                                                        250
     Pulse repetition frequency (Hz)                                  250 - 1200
Table 1:Parameters of all POLRAD network radars.System characteristics

    The modern weather radar network in Poland has been developed since 1995
when first Gematronik GmbH radar was installed on Ramza hill near Katowice (south
of Poland) covering mountainous area of country. It is a Meteor 360AC radar. The
similar radar has been running since 1998 in Pastewnik near Wrocaw, south of Poland


                                        46
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe                    47


too. In 2002 old MRL-5 radar in Legionowo was replaced with more modern Meteor
1500C one. The next radars were installed in 2003-2004.

    Finally the network, called POLRAD (POLish RADar network), operated by
IMWM, consists of eight radars (all produced by Gematronik), so the whole area of
the country is covered by weather radars ranges (see the map on the Fig. 16.1 and
geographical co-ordinates in the Table 2). They are C-Band, Doppler, magnetron or
klystron radars. Their parameters are summarised in Table 1.




    Figure 16.1: Example of composite map of the Polish radar network POLRAD

     Radar      Location Radar type                  Longitude        Latitude           Height     Since
     #                                                    (◦ N)           (◦ E)        (m.a.s.l.)
     1          Swidwin       500C                     53.790          15.831               146     2003
     2           Gdansk      1500C                     54.384          18.456               158     2003
     3            Poznan      500C                     52.410          16.832               123     2003
     4        Legionowo      1500C                     52.400          20.931               119     2002
     5         Pastewnik     360AC                     50.883          16.040               691     2000
     6            Ramza      360AC                     50.152          18.727               358     1995
     7        Brzuchania      500C                     50.394          20.080               453     2004
     8          Rzeszsw      1500C                     50.114          22.002               241     2003
Table 2. Location of POLRAD radars
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   48



    The radar products, available operationally, are in accordance with Rainbow 5
software. At present, radar data (both a composite map and products from individual
radar sites) are available via IMWM Intranet, public WWW page and also transmitted
by specialised system using coded parts of TV teletext. The composite map is pre-
sented in real-time mode at http://www.imgw.pl/wl/internet/aktualna/com_r_i.html.
The operational centre of Polish radar network, the Operational Radar Centre (ORC,
Polish acronym is RCO), is located at Warsaw in IMWM. The Centre consists of
three groups: (1) operational group which task is a continuous control of the network,
(2) technical service done by outsourcing, (3) research group. The ORC started
working in 2002. One of tasks of ORC is to take part in European co-operation in
weather radar data exchange. Poland is a member of European OPERA (Operational
Programme for the Exchange of weather RAdar information) programme since
1999, which is a continuation of GORN (Liaison Group for Operational European
Weather Radar Networking) programme (1992-1998). Furthermore, Polish radar
data is included into two international networks: (1) CERAD (Central Europe
RADar Network) for central Europe area, and (2) BALTRAD, i.e. BALTEX radar
network (Michelson et al., 2000). Research activities are focused on participation in
several research European projects. Workers of the ORC take part in the following
European programmes COST (Action 717), Interreg (RISK AWARE) and 5 FP
UE (MUSIC, FLOODRELIEF). Other scientific researches are conducted thanks
to State Committee for Scientific Research (Polish acronym is KBN) financial support.

    Radar data as a high-resolution input to hydrological models
  Technique                                    Spatial resolution  Temporal resolution
  Measured precipitation
  Rain gauges network of IMWM         Only point measurements,                   1 hour
                                        which one gauge is over (in telemetric network)
                                                        265 km2
  Weather radar network POLRAD                              1 km             10 minutes
  Processed precipitation
  Radar data processed by Nimrod                            4 km             10 minutes
  Estimation by NWP model HRM                      About 14 km                   1 hour
Table 3. Spatial and temporal resolution of measured and processed precipitation
input data to hydrological models.

    The IMWM operates an unified hydrological forecast system called System
of Hydrology (SH). The system is built for the whole country and constitutes an
important part of Polish flood defence structure. In general tasks of the System of
Hydrology (SH) are real-time operating, historical data processing, and scientific
activities. Hydrological modelling in the SH system is based on existing models such
as MIKE (in the south of Poland - mountainous regions), IHMS (in Central Poland),
and Baltic Sea Model. New hydrological concepts consist in subdivision of sub-basins
into Hydrological Response Units (HRUs), grid-based pre-processor, and greater
emphasis on methods suitable for continuous modelling. Next models of different
approaches (e.g. Saint Venant equations, Muskingum Cunge, Kinematic Wave etc.)
are being developed and included into the SH system. The system is planned to be
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   49


finished by the end of 2004.

    The precipitation input into the SH system will be real-time radar data after NIM-
ROD system processing. The Nimrod system of the Met Office (UK) is a radar-based
rainfall (1) analysis and (2) forecasting (nowcasting) system (Golding, 1998). The
NIMROD uses weather observations like raingauge stations data (telemetric and
standard), remote-sensing systems data (weather radars, satellites, lightning, etc.),
data from NWP model HRM licensed on German DWD model.

    All data are provided by IMWM. Satellite data are received from NOAA series
satellites which perform measurements through AVHRR and AMSU-B radiometers.
Lightning data are collected by a VAISALA’s system of lightning detection in Poland
named Safir.

    The radar data quality is controlled to remove spurious echoes such as clutters
and anaprops. Then data are subject to a vertical profile correction to correct for the
effects of range and bright band. Finally raingauge-based mean field bias correction
is applied (Weipert and Pierce, 2003). Temporal and spatial resolution of the products
is 10 minutes and 4 km respectively.

   For rainfall-runoff modelling quantitative precipitation nowcasting plays an
important role. The nowcasting methodology adopted in NIMROD is based on
merging of an extrapolated recent processed radar data with outputs from a mesoscale
NWP model. The forecasts are produced for 6 hours ahead.
Chapter 17

Portugal (Manuel Rosa Dias)

17.1     Use of radar data in hydrological models
In the last two decades, Institute of Meteorology (IM) has been carrying out R&TD
activities in the field of Radar Meteorology, namely in the scope of the measurement
of precipitation by radar, having in mind the applications in Hydrometeorology and
Hydrology. Several works have been carried out by IM and in co-operation with INAG
(Water Authority) and with both the classical and technical Universities of Lisbon.

    IM and INAG jointly participated in several programmes co-financed by the
European Commission, namely in the projects "Applications of Weather Radar for the
Alleviation of Climatic Hazards" (1986/1991), "Weather Radar and Storm and Flood
Hazards" (1991/1992) and "Storm, Flood and Radar Hydrology" (1993/95).

    Aiming at the implementation of hydrological forecasting models using radar
data, a number of equipment, software, tools and data sets were developed in the
scope of these projects, many of them innovative and leading to new practices at the
institutions involved.

    A small river basin named Alenquer, prone to flash floods and intensively
instrumented, was selected for studies of precipitation measurement by radar, as well
as for other hydrometeorological and hydrological studies and for testing different
hydrological models using radar data.

    A comprehensive data set was developed for that basin on appropriate processing
equipment and software packages. Event models were calibrated off-line for that
basin by using historical hydrometeorological data.

     Work has been performed by INAG, aiming at the implementation of a real-
time rainfall-runoff model, allowing to forecast the discharge at a level gauge
site in Alenquer basin. Use was made of the HEC-1 and HEC-2 packages in the
establishment of the parameters of the implemented model, which included a deter-
ministic component based on the unit hydrograph theory and a stochastic component
for the correction of the forecast errors based on a second order autoregressive process.


                                           50
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   51



    Nowadays, IM operates a network of two weather radars with the capability to
perform real-time radar/raingauge adjustment using the automatic weather stations
network also operated by IM.

    Raingauge adjusted radar estimates of rainfall are available, in real time, at IM
headquarters and meteorological centres across the country and can disseminated to
the interested external users also in real time.
    INAG is not now active in hydrological forecasting using radar data.

    However, at the academic level, the development of distributed hydrological mod-
els capable of taking advantage of the rainfall radar data has been pursued by some
investigators. Work has been oriented twofold in discretization: irregular elements
(mostly TIN but also other tessellations) and regular elements (raster models). Also
some consideration has been given to subsurface processes, mainly in connection with
agricultural plots with very good field instrumentation, either meteorological and veg-
etative or surface and subsurface water sensors.


17.2       Use of radar data in NWP models
IM has officially joined the ALADIN Project in April 1997 and since the beginning of
1999 is running operationally the ALADIN model twice a day.

    Radar reflectivity data assimilation is being developed simultaneously in the
ALADIN/AROME models at Meétéo-France. It has been decided to give priority to
reflectivity (either in single levels or volumetric data) and to postpone the development
of assimilation of Doppler radar wind data up to 2006.

    In France, it is planned that the data from operational radars will be used for
operational data assimilation in 2006-2012. In Portugal, it is not planned to use radar
data for data assimilation before 2010.

    In ALADIN/AROME models, 3D-Var may use radar data with a horizontal
resolution ranging from 2 to 10km and vertical resolution of the order of 300m.

    It is important to note that radar data cannot be assimilated without taking into
account the complete vertical profile of condensed species, temperature and humidity
in the troposphere. So, a 1D-Var approach will be used. This approach will need the
parameterisation of cloud microphysics.
Chapter 18

Slovenia (Gregor Gregoric)

COST-717 related activities in Slovenia were mainly divided into two efforts to
assimilate radar reflectivity measurements into a limited-area NWP model. The
first one is "radar–driven convective parameterization scheme" and the second is
implementation of latent heat nudging in operational NWP model.

    Radar–driven convective parameterization scheme could also be described as
a kind of nudging. The main idea is to replace some of the closure assumptions used
in the convective parameterization schemes (CPSs). For example, in Kain–Fritsch
CPS convective cloud radius near cloud base is assumed to be so large that convec-
tion eliminates all instability (that is, CAPE) from the model point in prescribed time.
That assumption leads to wide–spread parameterized convection in cases of instable
situations. The procedure is following:

   1. Analysis of radar data. A tool that identifies convective cells, position of their
      centers and their dimensions is applied. Modified version of TREC method
      (which correlates boxes of radar data with idealized, cone–shaped function) is
      used. Look–up tables for the modified CPS are produced.
   2. Numerical model is run with modified CPS. The CPS is forced to to check ex-
      istence of look–up tables. Convection is triggered only if radar data confirms
      existence of convective precipitation.

This method could also be described as a kind of "hard nudging" since cloud radii are
not forced toward observed values; they are replaced by observed values.

    The method was implemented in MM5 model (running in research mode in
University of Ljubljana) and studied in a number of well documented cases such
as MAP IOPs and was (during a short–term scientific mission) also compared to
another similar assimilation technique, radar–driven physical assimilation (developed
in the University of Bonn). The assimilation technique generally improves spatial
distribution of convective precipitation during the nudging period. After that period
problems with unrealistic model–resolved convection start to occur. Therefore this
method should be nudged more carefully into the model and possibly combined with
other nudging–type assimilation methods.


                                          52
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   53


     Implementation of Latent Heat Nudging in operational NWP model Latent
heat nudging (LHN) is a method of forcing NWP model with observed precipitation
rate. In this case the model is forced with heat released by observed precipitation. One
is not limited with profiles imposed by parameterization scheme as in case of physical
initialization. However since only two-dimensional observations are available some
kind of heating profile should be prescribed. There are two possibilities: idealized
profiles or scaled model profiles. The second choice was selected for implementation
in the ALADIN model.

    Despite its robustness this technique has generally positive impact on model
forecasts’ scores which lasts up to 30 hours ahead. Slovenia is quite challenging area
for any precipitation assimilation technique. Due to its position between Alps and
Mediterranean mesoscale cyclones and convective systems are frequent and various in
shape and origin.

    The assimilation chain was applied on daily model runs throughout year 2002.
Year 2002 was warmer than 61-90 average (as is appropriate for climate trends).
Precipitation amount in the western and southern part of Slovenia (which receive
more precipitation due to proximity of the coast and orography features) was larger
than average and less than average on the eastern part where precipitation depend on
larger scale processes. This makes year 2002 a suitable choice for tests of LHN runs.

    First results show that LHN improves model precipitation bias. It successfully
replaces precipitation systems from model’s envelope orography ridges where they
typically appear. However there are many cases of deterioration of the forecast (also)
due to known problems of unrealistic model resolved convection.
Chapter 19

Spain (Joan Bech)

19.1     Introduction
The use of weather radar observations in NWP and hydrological models has increased
significantly in Spain throughout the life time of the COST 717 action. This situation
is the result of a number of concurrent factors, which include a growing interest in
quantitative applications, an expansion in the number and type of radar-data users and
also the installation of new radar systems operated by different institutions which has
lead to a wider and unprecedented availability of radar products (Figure 19.1).

    The Spanish weather service (INM) operates since the early nineties a national
weather radar network designed to cover the peninsular Spanish territory and also
the Balearic and Canary Islands with fifteen units. The radars are C-band Ericsson
Doppler systems (Aguado et al., 1994). Originally, S-band radars were installed in the
Mediterranean area, characterised by torrential rainfall regime and, therefore, prone to
flash floods events. However, due to technical reasons, those radars were changed to
C-band at the end of the nineties.

     The University of Barcelona installed a Kavouras C-band Doppler radar in 1996
for research purposes. Later the system was transferred to the Catalan Meteorological
Service (SMC) which was building a regional network designed to cover the complex
topography of Catalonia (NE of Spain) with four radars, with similar technical char-
acteristics of the first unit (Bech et al., 2004). Another regional institution, Euskalmet,
has recently acquired a Gematronik dual polarization C-band radar in the Basque coun-
try (N of Spain) for weather surveillance and hydrological modelling (Aranda, 2002).
Another weather radar, a C-band EEC system, is used in intensive observation periods
in the NW of Spain to support hailfall studies by the University of Leon (Sánchez et al.,
1996; Fraile et al., 2001).


19.2     Radar and NWP
The use of radar observations, both reflectivity and velocity data, in order to improve
NWP model output has been an expanding topic in the last years by different Spanish
groups. During the EU funded project DARTH (Holt et al., 2002), the University of


                                           54
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   55




Figure 19.1: Spanish weather radars: INM (red), SMC (blue), EuskalMet (green) and
University of León (yellow). Solid symbols indicate the radar is operational and blank
symbols indicate planned. INM operates an additional radar in the Canary Islands.


Barcelona ran a series of simulations with the MASS model over the Scandinavian
area covered by the Nordrad weather radar network. Four different CAPPI levels of
reflectivity observations were used to enhance the characterisation of the humidity
field in the initialisation stage of the model (Bech et al., 1998; Codina et al., 1999).

    Furthermore, in the EU CarpeDiem project (Alberoni et al., 2002), techniques of
nudging and incremental analysis updates have been applied with the same simulation
system over the area of Catalonia. In this context, single level reflectivity observations
of SMC radars were considered and lead to improved quantitative precipitation
forecasts (Picanyol et al., 2004). Moreover, in the same project but a with a different
approach (using NWP data to improve the quality of radar data), the MASS model
output was also used to simulate the three-dimensional refractivity field in order
to study the radio propagation environment and its temporal evolution in radar
anomalous propagation events (Bech et al., 2003a; Bebbington et al., 2004).

    A field of intense work in the INM has been the use of velocity radar observations
VAD (Velocity Azimuth Display) for assimilation purposes (Salvador et al., 2004).
VAD products were already used to characterise and monitor the initial stages and
further development of convection and also for precipitation nowcasting purposes
(Martín et al., 2002; Conejo et al., 2003; Elizaga et al., 2004). The assimilation was
designed in the context of the HIRLAM consortium, following a three-dimensional
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   56




Figure 19.2: Madrid 1.2 km simulated CAPPI reflectivity with a HIRLAM model
resolution of 5 km

variational approach (Gustafsson et al., 2001; Lindskog et al., 2001), and was
implemented and tested successfully to be used with radars of the INM Doppler radar
network (Salvador and Navascués, 2004). Furthermore, a bias reduction scheme for
VAD profiles is being developed and VAD observation errors will be tuned with the
aim of the next operational assimilation of these wind profiles by the INM-HIRLAM
suite based on the present 3DVar and the future 4DVar assimilation systems.

    The use of the Radar Simulation Model (Haase and Fortelius, 2001) has also been
analysed by INM as a verification tool of the HIRLAM system (Sanz and García-
Moya, 2004). In particular, the Radar Simulation Model (RSM) has been used to
compare simulated radar reflectivity at different model levels and radar CAPPI prod-
ucts for several model resolutions (Figure 19.2). A parallel version of the RSM code
is expected to allow using this technique for routine verifications, difficult to perform
with the current RSM implementation because of its high cost in terms of computing
time.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   57


19.3       Radar and hydrology
In order to obtain radar quantitative precipitation estimates -an essential input for
hydrological models- a number of projects have been undertaken in Spain since the
mid 90s to implement or develop specific corrections to radar observations. This
research has ranged from fundamental and applied disdrometric studies about Z-R
relationships (Sempere-Torres et al., 1994; Cerro et al., 1997; Sempere-Torres et al.,
1998), to bright-band and precipitation type discrimination (Sánchez-Diezma et al.,
2000; Rigo and Llasat, 2004) or topographical blockage and anomalous propagation
(Bech et al., 2003b), just to mention a few.




Figure 19.3: View of the EHIMI tool showing a composite reflectivity image of the
SMC radar network

    However, the experiences of hydrological models using radar data have been
relatively limited compared to the widespread use of only traditional raingauge
networks (Corral et al., 2002; Sancho and Fleitz, 2002). The Applied Research Group
in Hydrometeorology (GRAHI) of the Polytechnical University of Catalonia has
been working actively to develop and implement operationally hydrological models
incorporating radar observations, in collaboration with both INM and SMC. A recent
project funded by the Catalan Water Agency (Sanchez-Diezma et al., 2002; Sempere-
Torres et al., 2004) consisted in the development of EHIMI, an integrated system to
correct, visualise and combine radar observations with raingauge data in order to feed
a hydrological model over a specific catchment near Barcelona city (Figure 19.3).
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   58


Further developments of the tool consider attenuation and VPR enhanced corrections
(Berenguer et al., 2002; Franco et al., 2002), operational real-time merging of radar
and gauge observations (Velasco-Forero et al., 2004) and the extension of the model
to a higher number of catchments.
Chapter 20

Sweden (Daniel Michelson)

Operational use of radar data in hydrological models
The hydrological forecasting system at SMHI is based on the semi-distributed con-
ceptual HBV model (e.g. Bergström, 1976; Lindström et al., 1997). The model was
originally developed to assist hydropower operations. The aim was to create a concep-
tual hydrological model with reasonable demands on computer facilities and calibra-
tion data. Therefore input data are designed to be as simple as possible, normally only
daily mean-values of temperature and precipitation are required. The original use for
hydrological forecasting has expanded to applications such as filling gaps in measured
time-series, simulation of stream-flow in ungauged rivers, design flood calculations
and water quality studies. Today, the model is the standard forecasting tool in Sweden
where some 50 catchments are calibrated for the national warning services. The HBV
model has been implemented in a software called Integrated Hydrological Modelling
System (IHMS) which includes a wide range of facilities for e.g. processing of input
data and presentation of results.
    For hydrological forecasting, the HBV model input in terms of precipitation and
temperature is provided by the High Resolution Limited Area Model (HIRLAM). At
SMHI, HIRLAM is run 4 times a day over a North Atlantic-European area at a hor-
izontal resolution of 44 km. Computations of surface pressure, wind, temperature,
humidity and cloud water are made at 40 vertical levels, from surface to 31 km. Inside
the 44 km model operates a nested model with 22 km resolution for Northern Europe.
The previous HIRLAM forecast is used as a background for 3-dimensional variational
analysis of observations from surface, upper air and potentially from space. Bound-
ary values are provided by the global model at European Centre for Medium-range
Weather Forecasts (ECMWF).
    Besides IHMS, other software has been developed for end users of hydrological
forecasts. WebHyPro is an Internet-based presentation system for real time data, pre-
sented on maps as values, isolines, tables or graphs. Besides the presentation of me-
teorological input and parameters computed by the HBV model (i.e., snow cover,
soil moisture content and runoff), radar and satellite images may be accessed and
displayed. The river simulator RISIM has been developed as a tool for training hy-
dropower staff to handle flooding events and other critical situations. RISIM is com-
bined with the HBV model to simulate and display inflow, water-levels in rivers and


                                          59
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   60


lakes, and planned discharge from reservoirs.
    There are twelve Ericsson C-band Doppler weather radars in Sweden covering the
country almost completely. Five of these are owned by SMHI and seven are owned
by the Swedish Armed Forces (SAF), and all together they constitute the Swedish
Radar Network (SWERAD). The maximum operational ranges in the Doppler and
non-Doppler modes are 120 km and 240 km, respectively. Data from all radars are
made available to the Nordic Weather Radar Network (NORDRAD). At present radar
data are not operationally applied for hydrological purposes, but ongoing research is
evaluating the potential of radar to improve hydrological forecasting in Swedish catch-
ments.


20.1       Research use
Two research projects are performed in the frame of 5th Framework Programme of
European Union. They are CARPE DIEM (Critical Assessment of Available Radar
Precipitation Estimation Techniques and Development of Innovative Approaches for
Environmental Management) 2002-2004 and ELDAS (Development of a European
Land Data Assimilation System to predict Floods and Droughts) 2002-2004.
    In the CARPE DIEM project, areal precipitation estimates from different sources
have been compared (Olsson et al., 2004). Besides radar, the sources included rain
gauges, an NWP model and different versions of a mesoscale analysis system. The
radar data used in the study were treated by a regional gauge adjustment procedure
(e.g. Michelson, 2003). The main comparison was performed using data for 2002
from catchment Gimån, located in central Sweden and of approximately 4300 km2
size. For most of the period the radar data compared well with data from the gauge-
derived sources, but during a few months the radar produced unrealistically high values
over a part of the catchment, possibly owing to temporal malfunctioning. In a second
phase the different areal precipitation estimates were used to drive the hydrological
HBV model, set up for the catchment. The resulting runoff was compared with the ob-
served, both for the entire catchment and for a smaller sub-catchment. Generally, the
accuracy of the radar-generated runoff was similar to the gauge-generated (Fig. 20.1).
The inhomogeneities found in the radar-estimated areal precipitation did not signif-
icantly deteriorate the generated runoff as they occurred in a period with low-flow
conditions and further affected only a remote part of the catchment. Other research
within CARPE DIEM includes methods for local gauge adjustment of radar data for
hydrological forecasting.
    In the ELDAS project the role of SMHI is to investigate the potential use of radar
precipitation for improving real-time hydrological forecasting in critical flooding situ-
ations. The focus of the study is on case studies of two recent major flooding events
in Sweden. One occurred in catchment Gimån in summer 2000 and the other in catch-
ment Glafsfjorden in autumn 2000. The objective is to assess whether the high tem-
poral and spatial resolution of radar data would have improved the forecasting and
possibly affected decisions during rescue operations.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   61




Figure 20.1: Runoff from catchment Gimån during 2002: observed and simulated
using precipitation estimated from rain gauges and radar, respectively.
Chapter 21

Switzerland (Andrea Rossa)

21.1     Radar network
MeteoSwiss operates since 1993-95 three meteorological radar in Switzerland,
respectively located in the northern, the southern and the south-western part of
the country. The composite image of the three radars covers the whole country
(see Figure 21.1). All of the three locations are equipped with C-band Doppler
weather radars (e.g. METEOR 360 AC by Gematronik, Germany) and operates with
the data processing system from Lassen Research (U.S.A.). MeteoSwiss issues a
number of products (for a full reference of the MeteoSwiss radar related products
see http://www.meteoswiss.ch/en/Science/Radar/IndexRadar.shtml), the most suitable
of which for hydrological application is RAIN, available starting from 1998. Prior
to 1998 other products with a less resolved intensity scale are available. RAIN
represents the best-estimate of the precipitation at the ground level. It accounts for the
radar visibility, the terrain topography, and the mean vertical reflectivity distribution
(Z-profile) to estimate the precipitation at ground level. The data taken from the last 6
measurements (each obtained from a full volume scan) lead to this product, containing
a moving average of the precipitation rate during the previous 30 minutes, updated
every 5min.

    The rainfall intensity scale of the RAIN product contains 16 equal steps on a log-
arithmic scale between 0.16 and 100 mm/h. The original data of RAIN are highly
resolved in space with a pixel size of 1x1 km2 , having the corresponding radar station
in the centre of the picture.


21.2     Operational use of radar data in hydrological models
Although MeteoSwiss has developed a considerable experience concerning rainfall
measurements by means of radar, the complex topography of Switzerland makes a
direct use of such measurements more complicated than in countries where the radar
visibility is optimal and the orographic influence on precipitation plays a negligible
role. This has been indeed the major obstacle to the operational implementation of
radar measurements into basin scale hydrological modelling.



                                           62
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   63


    To date there is no operational forecasting procedure which officially implements
quantitative precipitation estimates obtained from radar measurements. A few agen-
cies responsible of the flood forecasting service at the cantonal or federal level receive
the radar images and visualize them to integrate qualitatively the quantitative point
measurements obtained from raingages. This is the case, for instance of the the flood
forecasting service for the Swiss part of the Rhein river, which is operated by the Fed-
eral Office for Water and Geology. The flood warning procedure makes use of radar
images only to visualize and track qualitatively the spatial extent of storm events dur-
ing warning periods, and to assess the need of an intensification of the forecasting runs.

    In addition, a few research projects investigated the potential of operational use of
radar in flood forecasting, whereas no significant scientific contribution is available
with respect to the use of radar measurements for flood frequency analyses based on
rainfall-runoff modelling.

    Two EU-funded research projects have mainly addressed the opera-
tional use of radar precipitation estimates in flood forecasting.         These are
the project RAPHAEL (Runoff and Atmospheric Processes for flood HAzard
forEcasting and controL, Project n◦ ENV4-CT97-0552) and the project MUSIC
(http://www.geomin.unibo.it/orgv/hydro/music/) (MUltiple-Sensor Precipitation
Measurements, Integration, Calibration and Flood Forecasting, Project n◦ EVK1-CT-
2000-00058).

    Both projects targeted southern Switzerland (the Ticino river basin and its
tributaries, ∼3000 km2 ), where the climatic conditions, closer to those of mediter-
ranean climates, and the highly variable nature of precipitation in space, also due
to the orographic influence, suggest as appropriate the use of radar measurements.
Conversely, the pronounced orography is particularly challenging for the use of radar,
as part of the river basins are not seen properly by the radar.

    The results from RAPHAEL pointed out that the radar rainfall estimates were
generally underestimating the ground truth as estimated by means of a relatively dense
raingage network. Specifically, the streamflow simulation produced using the radar
estimates generally understimated the observed flows, with only one exception, but
were correct as for what the hydrograph time dynamics is concerned. This highlights
a basic (and known) problem of radars in capturing the correct amount of precipitation
in areas where the orographic complexity plays a significant role, thus suggesting
the need for developing future alternative strategies to achieve reliable quantitative
precipitation estimates distributed in space. A further limitation may also come
from the smoothing of the largest intensities due to the logarithmic scale used by
MeteoSwiss to describe rainfall intensity in their operational products.

    The results from the MUSIC project were limited to a smaller area, namely the
Maggia river basin (∼1000 km2 ). The innovative aspect of this project is to combine
the radar precipitation estimates with the observations provided by the raingages within
a Bayesian framework. The combined estimates provide a rainfall field, which merges
the spatial distribution observed by the radar with the more accurate observation of
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe     64


rainfall intensity obatined by raingages. The streamflow estimates computed using
the combined rainfall field outperform those obtained using the radar only, as well as
those obtained using the individual raingages (both in the case of Thiessen polygons
based spatial redistribution and in that of of block-kriged redistribution of raingages
estimates). This suggests a viable way to operational use of radar observations in
streamflow modelling, pointing out once more the importance of the correct estimate
of the spatial distribution of rainfall. In this respect, the results also indicate a way to
further investigate the potential of radar devices in hydrological modelling, as otulined
in the next section.


21.3       Future challenges and strategies
The results of the research projects that addressed the operational use of radar mea-
surements pointed out once more two important (and known) facts. Radar estimates
are appropriate to describe the spatial precipitation pattern, but often fail to provide
an accurate quantitative estimate in regions where the topography is particularly
enhanced and represents an obstacle to the visibility of radars.

    This circumstances indicate that there may be a need to complement the existing
large scale radar installations with (low cost) local weather radars. These can provide
estimates on much smaller areas, thus being less affected by visibility problems, thus
representing a complementary information to both raingages (still necessary for cal-
ibration purposes) and large scale radars. A research initiative may start in the next
year in the south-western part of Switzerland by investigating the performance of an
X-Band radar located in the mountainous region of Valais. A combined network of
large scale radars, local radars and raingages may represent on the short- and mid-term
the best compromise between accuracy and costs, especially if complemented by al-
gortihms that can integrate the measurements provided by the different sensors also
minimizing the variance of the estimate errors.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe     65




Figure   21.1:      Radar     composite    image   over     Switzerland                (from
http://www.meteoswiss.ch/en/Science/Radar/IndexRadar.shtml)
Chapter 22

United Kingdom (Robert Moore)

22.1     Operational use of radar and NWP data in hydrological
         models
Weather radar products are disseminated from the Met Office via ftp to the En-
vironment Agency’s Data Distribution Server in Leeds where CEH’s HYRAD
(HYdrological weather RADar) system provides Client/Server facilities to display the
products for 200 users throughout the 8 EA regions. The displays support animated
replay of rainfall fields, accumulation displays, and displays of hyetographs in total
and cumulative form over pixel or catchment areas. The main Met Office radar
products received by the EA are the single site radar products (1/2/5 km resolution),
quality controlled composites over the UK, and forecasts from the Nimrod (5 km)
and Gandolf (2 km) systems, the latter using different techniques in stratiform and
convective rain. Forecasts are for a 6 hour lead time at a 15 minute time-interval and
taper towards the NWP rainfall forecast with increasing lead-time. The radar data can
be further processed by Hyrad to remove residual anomalies, merged with raingauge
data if available, and forecasts constructed on a 1 km grid using a local advection
with smoothing technique. All products can be processed to derive catchment average
time-series rainfall made available via a generic interface to Flood Forecasting
and Modelling Systems (FFMSs) in use by the EA. NWP rainfall and temperature
forecasts out to 1 1 days ahead are also received from the Met Office for visualisation
                   2
within Hyrad and to pass to FFMSs. Other spatio-temporal products received include
precipitation type and a radar quality indicator. Receipt of MOSES (Met Office
Surface Exchange Scheme) products, including soil moisture and runoff fields on a
5 km grid, is planned for late 2004. These products are derived using data from the
Nimrod radar composite as one source of input.

    The flood forecasting systems and the set of models they use vary with each EA
region. Thames and Northeast region use CEH’s River Flow Forecasting System
(RFFS) Information Control Algorithm for automated forecast construction. This
generic, configurable system can be set up to use any combination of models in
a network arrangement reflecting that of the river basin. In Thames region it is
configured within the EA’s CASCADE system to use both the EA’s TCM (Thames
Catchment Model) rainfall-runoff model and CEH’s PDM (Probability Distributed


                                         66
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   67


Model) and the results can be compared before issuing a warning. It can be configured
to use radar (Nimrod, Gandolf, or possibly a Hyrad merged radar/raingauge estimate)
or raingauge data, and a logical switch is provided to toggle off radar when it is judged
less reliable than raingauges. The configuration in Northeast region encompasses the
Ouse river network in Yorkshire, the Tees and other Northumbria rivers. The network
of models utilise CEH’s PACK snowmelt model, the PDM rainfall-runoff model,
the KW channel flow routing model, the RFFS HYDRO hydrodynamic river model
for the tidal Ouse (developed from the NWS DWOPER code), and the RFFS ISIS
hydrodynamic model for the tidal Tees (a HR/Halcrow and CEH development).

    The Midlands region has developed its own forecast construction environment and
employs conceptual water balance models for rainfall-runoff modelling (the Midlands
Catchment Runoff Model) and channel flow routing (the DODO model). It employs
raingauge estimates of rainfall as input up to the present time, and radar rainfall
forecasts of hourly totals out to 6 hours obtained from the Nimrod product. Whilst
the Thames, Northeast and Midlands regions employ models based on continuity and
mass balance principles, simple transfer function models feature in the systems in
Northwest and Southwest regions supplied by the Bristol University group (previously
at Salford) and aim to make direct use of radar data.

    Southern region does not enjoy good radar coverage at present and only makes
limited use of radar data in flood forecasting models. Radar-based flood warning
is under development in Anglian region as part of a system interfacing a new Flow
Forecasting and Modelling System to Hyrad. Welsh region of the EA uses weather
radar only as an informal aid to flood warning.

    In Scotland, an operational trial of the use of weather radar in flood forecasting is
ongoing by the Scottish Environment Protection Agency (SEPA). This is employing
the White Cart Flood Forecasting System which provides flood warnings for the
White Cart Water draining through the southern part of Glasgow. The system employs
Hyrad and the RFFS in FloodWorks form and receives a Met Office data feed from
the Corse Hill radar. In Northern Ireland, the Castor Bay radar is only used informally
in support of flood warning and data are not automatically linked to models for flood
forecasting.

    In general, throughout the UK there is caution in the automated use of weather
radar data as input to hydrological models for real-time flood forecasting. This is due to
reasons concerning the lack of consistent quality in time and space of the weather radar
estimates of rainfall. Up to the time the forecast is made, more reliance is often placed
on information from the raingauge network, although this may be of poor quality at
times due to low gauge density and high spatial variability of rain. For future times,
however, radar rainfall forecasts are seen as the best available source of information
for extending the lead time of flood forecasts from rainfall-runoff models. Some use of
synoptic forecasts of rain, supported by NWP outputs, is made for higher lead times.
                                                                                   1
The recent automated receipt, via Hyrad, of NWP rainfall forecasts out to 1 2 days
offers further opportunities to extend the lead time-of flood warnings, at least at an
indicative first-alert level.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   68


22.2       Research
There are active research programmes within the Met Office, CEH (within the
CEH/Met Office Joint Centre for Hydro-Meteorological Research at Wallingford),
CLRC (RAL) and universities (Bristol, Essex, Imperial College, Lancaster, Newcas-
tle, Reading, Salford). Research topics include improved radar rainfall estimation
(methods and new radar technology), probabilistic rainfall and flood forecasting, grid-
based rainfall-runoff modelling, soil moisture estimation, and stochastic space-time
models of rainfall fields for design flood estimation by continuous catchment model
simulation.

   The results of the Natural Environment Research Council HYREX (HYdrological
Radar EXperiment) Special Topic Research Programme are published in (Moore and
Hall, 2000).
Acknowledgements

I wish to thank all the contributors who provided the information for this Working
Document. I also wish to thank Mr. Petros Constantinides for helping with the editing.




                                         69
Bibliography

Aguado, F., J. L. Camacho, E. Gutiérrez, J. M. Gutiérrez, and F. Pérez, Cost 75;
  weather radar systems., in International Seminar, pages 29–35, Brussel, 1994, EUR
  16013 EN. Edited by Collier.

Alberoni, P. P., E. Todini, M. Chandra, M. Lindskog, J. Koistinen, D. Bebbington,
  B. Codina, V. Levizzani, M. Bruen, N. Gustafsson, D. Zrnic, A. Rossa, and P. Bur-
  lando, Carpe diem: Eu project, in Second European Conference on Radar Meteo-
  rology, volume 11, pages 363–369, 2002, EUR 16013 EN. Edited by Collier.

Aranda, J. A., Personal comunication, euskalmet, Vitoria -Gasteiz 01010, c/ Donostia,
  San Sebastian, Spain, 2002.

Bebbington, D., A. Rae, J. Bech, B. Codina, and M. Picanyol, Modelling of weather
  radar echoes from anomalous propagation using a hybrid parabolic equation method
  and thermodynamic fields derived from a numerical weather prediction model, in
  European Conference on Radar in Meteorology and Hydrology (ERAD) - COST
  717 Final Seminar, page 29, 2004, Abstracts Book, Copernicus GmbH (c).

Bech, J., B. Codina, and J. Lorente, Forecasting weather radar propagation conditions
  using nwp mesoscale model data, in 25th EWGLAM - 10th SRNWP / COST 717
  meeting, page 5, Lisbon, 2003a.

Bech, J., B. Codina, J. Lorente, and D. Bebbington, The sensitivity of single polariza-
  tion weather radar beam blockage correction to variability in the vertical refractivity
  gradient, J. Atmos. and Oceanic Technol., 20, 845–855, 2003b.

Bech, J., B. Codina, A. Sairouni, J. Vidal, D. Michelson, and J. Lorente, Initialisation
  of a nwp mesoscale model using satellite and radar data, in DARTH Project Final
  Seminar, page 15, 1998.

Bech, J., E. Vilaclara, N. Pineda, T. Rigo, J. López, F. O’Hora, J. Lorente, D. Sempere,
  and F. X. Fábregas, The weather radar network of the catalan meteorological service:
  description and applications, in European Conference on Radar in Meteorology and
  Hydrology (ERAD) - COST 717 Final Seminar, volume 2, pages 416–420, 2004,
  ISBN 3-936586-29-2.

Berenguer, M., G. W. Lee, D. S. Torres, and I. Zawadzki, A variational method for
  attenuation correction of radar signal. in: H. russchenberg (editors), in ERAD, ISBN:
  3-936586-04-7, pages 11–16, 2002.


                                           70
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   71


Bergström, S., Development and application of a conceptual runoff model for scan-
  dinavian catchments, Technical report, Reports Hydrology and Oceanography 7,
  SMHI, 1976.

Bergstrm, S. and A. Forsman, Development of a conceptual deterministic rainfall-
  runoff model, Nordic Hydrology, 4, 147–170, 1973.

Cerro, C., B. Codina, J. Bech, and J. Lorente, Modeling raindrop size distribution and
  z-r relationships in the western mediterranean area, Journal of Applied Meteorology,
  36, 1470–1479, 1997.

Codina, B., J. Vidal, J. Bech, D. Michelson, J. More, and J. Lorente, Humidity initial-
  ization with satellite and radar data: a case study, in Geophysical Research Abstracts
  24th General Assembly European Geophysical Society, Hydrology Oceans and At-
  mosphere, ISSN:1029-70006, OA10, Numerical Weather Prediction, volume 1, 2,
  page 428, 1999.

Conejo, S., F. Elizaga, F. Martín, and I. S. Ambrosio, Use andapplication of vad (veloc-
  ity azimuth display) data in a case of severe convection in catalonia, in 5th Plinius
  Conference on Mediterranean Storms, Corsica (France), 2003.

Corral, C., M. Berenguer, D. S. Torres, and I. Escaler, Evaluation of a conceptual
  distributed rainfall-runoff model in the besos catchment in catalunya using radar
  information, in In: H. Russchenberg (Editors), Use of Radar Observations in Hy-
  drological and NWP Models. ERAD, pages 409–415, 2002, Delft. ISBN: 3-936586-
  04-7.

COST, Quality and assimilation of radar data for nwp, Technical Report EUR 20600,
 COST 717, 2003.

Dudhia, J., A multi-layer soil temperature model for mm5, in The 6th PSU/NCAR
  MM5 Users‘ Workshop Boulder Colorado, pages 49–50, 1996.

Einfalt, T. and C. Golz, Examples for the effect of using radar data correction proce-
  dures, in 1st VOLTAIRE Workshop, pages 11–15, Barcelona, Spain, 2003.

Elizaga, F., F. Martín, I. S. Ambrosio, O. Carretero, and S. Conejo, Development of
  radar-based applications at inm, Technical report, STAP - Analysis and Forecasting
  Techniques Department, INM - Instituto Nacional de Meteorologia (Spanish Met.
  Service), 2004.

Forbes, R., D. Bjrge, M. Homleid, U. Gjertsen, T. E. Nordeng, O. Vignes, and
  V. degaard, Assessment of the unified model for high resolution nwp over norway
  during the period jan-apr 2004, Technical report, met.no research report, 2004, un-
  der preparation.

Fraile, R., A. Castro, J. L. Sánchez, J. L. Marcos, and L. López, Noteworthy c-band
  radar parameters of storms on hail days in northwestern spain, Atmospheric Re-
  search, 59-60, 41–61, 2001.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   72


Franco, M., D. S. Torres, R. Sánchez-Diezma, and H. Andrieu, A methodology to
  identify the vertical profile of reflectivy from radar scans and to estimate the rain-
  rate at ground at different distances, in Use of Radar Observations in Hydrological
  and NWP Models. ERAD, edited by I. H. R. (Editors), pages 299–304, 2002, Delft.
  ISBN: 3-936586-04-7.

Fulton, R., J. Breidenbach, D. J. Seo, D. Miller, and T. O’Bannon, The wsr-88d rainfall
  algorithm, Wea. Forecasting, 13, 377–395, 1998.

Gabella, M., J. Joss, and G. Perona, Optimizing quantitative precipitation estimates
  using a non-coherent and a coherent radar operating on the same area, J. of Geophys.
  Res., 105, 2237–2245, 2000.

Gabella, M. and S. Michaelides, Radar adjusted rain fields and comparison with trmm
  data: a preliminary report on a selected 2002 event in cyprus, in 1st VOLTAIRE
  Workshop, pages 19–26, Barcelona, Spain, 2003.

Gjertsen, U., J. I. Dahl, V. degaard, and A. Kristensen, Combined use of radar and
  temperature measurements, in Annalen der Meteorologie 40, SIRWEC 2004, 12th
  International Road Weather Conference, Bingen, Germany, 2004.

Golding, B. W., Nimrod: a system for generating automated very short range forecasts,
  Meteorol. Appl., 5, 1–16, 1998.

Golz, C. and T. Einfalt, A review of rainfall data quality control methods, in 1st
  VOLTAIRE Workshop, pages 8–10, Barcelona, Spain, 2003.

Golz, C., T. Einfalt, and S. Michaelides, Quality control of rainfall measurements in
  cyprus, in 7th Panhellenic (International) Conference of Meteorology, Climatology
  and Atmospheric Physics, 28-30 September 2004, 2004, in press.

Gregoric, G., Diagnostics of deep convection using radar data and meso-scale numeri-
  cal model, in 30th Conference on Radar Meteorology, pages 177–178, AMS, 2001.

Grell, G. A., J. Dudhia, and D. R. Stauffer, A description of fifth-generation penn
  state/ncar meso-scale model (mm5), Technical Report NCAR/TN-397+IA, NCAR
  Technical Note, 1994.

Gustafsson, N., L. Berre, S. Hörnquist, X.-Y. Huang, M. Lindskog, B. Navascués,
  K. S. Mogensen, and S. Thorsteinsson, Three-dimensional variational data assimi-
  lation for a limited area model. part i: General formulation and the background error
  constraint, Tellus, 53A, 425–446, 2001.

Haase, G. and S. Crewel, Simulation of radar reflectivities using a mesoscale weather
  forecast model, Water Resources Research, 36, 2221–2231, 2000.

Haase, G. and C. Fortelius, Simulation of radar reflectivities using hirlam forecasts,
  Technical report, HIRLAM Tech. Rep., 2001.

Heise, E. and G. Rivin, Precipitation analysis and prediction. final report on the dwd-
  contribution to the eu-project “an european flood forecasting system”, Forschung
  und Entwicklung, Arbeitsergebnisse 80, DWD, 2004, 38 pp.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   73


Holleman, I., Vpr adjustment using a dual-cappi technique, in ERAD Publication se-
  ries, 2, pages 25–30, 2004.

Holt, A. R., P. P. Alberoni, J. Bech, J. Bebbington, M. Chandra, B. Codina, C. G. Col-
  lier, P. J. Hardaker, D. Michelson, and M. Monai, Darth - development of advanced
  radar techniques for hydrometeorology, in 2nd MITCH Workshop (Mitigation of
  Climate Induced Natural Hazards), Barcelona, 10-12 June 2002, 2002.

Horvath, A. and I. Geresdi, Severe convective storms and associated phenomena in
  hungary, Atmospheric research, 56, 127–146, 2001.

Horvath, A. and I. Geresdi, Severe storms and nowcasting in the carpathian basin,
  Atmospheric research, 67-68, 319–332, 2003.

Janic, I., The step-mountain eta coordinate model, Mon.Wea.Rev., 118, 1429–1443,
  1994.

Joss, J., M. Gabella, and G. Perona, Needs and expectations after the 1st voltaire
  workshop, upscaling and consequences of non-homogeneous beam filling, in 1st
  VOLTAIRE Workshop, pages 1–6, Barcelona, Spain, 2003.

Klink, S. and K. Stephan, Assimilation of radar data in the lm at dwd, Technical Re-
  port 4, COSMO Newsletter, 2004.

Koeck, K., National and international weather radar networks and their calibration by
  use if raingauge data for quantitative use, Technical report, Graz, 1998.

Koistinen, J. and T. Puhakka, An improved spatial gauge-radar adjustment technique,
  in 20th conference on Radar Meteorology, pages 179–186, AMS, 1981.

Lagouvardos, K. and V. Kotroni, Improvement of high resolution weather forecasts
  through humidity adjustment, based on satellite data, Quarterly Journal Royal Me-
  teorological Society (submitted), 2004.

Lindskog, M., N. Gustafsson, B. Navascués, K. S. Mogensen, X.-Y. Huang, X. Yang,
  U. Andræ, L. Berre, S. Thorsteinsson, and J. Rantakokko, Three-dimensional vari-
  ational data assimilation for a limited area model. part ii: Observation handling and
  assimilation experiments, Tellus, 53A, 447–468, 2001.

Lindström, G., B. Johansson, M. Persson, M. Gardelin, and S. Bergström, Develop-
  ment and test of the distributed hbv-96 hydrological model, J. Hydrol., 201, 272–
  288, 1997.

Macpherson, B., Radar data in nwp models, an outline of cost-717 working group 3,
 Phys. and Chem, Earth (B), 25, 1225–1228, 2000.

Mancini, M., N. Montaldo, and R. Rosso, Effetti di laminazione di un sistema di invasi
 artificiali nel bacino del fiume toce, Technical report, l’Acqua, 2000.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   74


Martín, F., I. S. Ambrosio, O. Carretero, and F. Elizaga, Radar based operational tools
 for heavy rain nowcasting: regional and national perspectives, in Mitigation of Cli-
 mate Induced Natural Hazards (MITCH) Workshop II: Advances in Flood Fore-
 casting, Barcelona (Spain), 10-12 June 2002, 2002, Flood Warning and Emergency
 Management.

Michelson, D. B., Quality Control of Weather Radar Data for Quantitative Applica-
 tion, Ph.D. thesis, Telford Institute of Environmental Systems, School of Environ-
 ment and Life Sciences, University of Salford, Greater Manchester, United King-
 dom, 2003.

Michelson, D. B., T. Andersson, J. Koistinen, C. G. Collier, J. Riedl, J. Szrurc,
 U. Gjertsen, A. Nielsen, and S. Overgaard, Baltex radar data centre products and
 their methodologies, Report RMK No. 90, Swedish Meteorological and Hydrologi-
 cal Institute (SMHI), 2000.

Mlawer, E. J. and P. D. Brown, Radiative transfer for inhomogeneous atmosphere,
 J.Geophys.Res., 102, 16663–16682, 1997.

Moore, R. J. and M. J. Hall, Hyrex: the hydrological radar experiment, Hydrology and
 Earth System Sciences, 4, 681, 2000.

                    c
Novák, P. and J. Krᡠmar, Vertical reflectivity profles in the czech weather radar net-
  work, in 30th International AMS Conference on Radar Meteorology, pages 179–
  186, Munich, Germany, 2001, poster presentation.

                    c
Novák, P. and J. Krᡠmar, Www page, version of october 2004, Technical report,
  CHMI Radar Department, 2002.

Olsson, J., B. Johansson, and S. Fogelberg, Comparison of areal precipitation esti-
  mates and their performance in runoff modelling: a case study for a central swedish
  catchment, Technical report, CARPE DIEM project report, SMHI, 2004.

Picanyol, M., J. R. Miró, B. Codina, and X. S. de Buruaga, A comparison experi-
  ment between nudging and incremental analysis updating in a mesoscale model,
  Geophysical Research Abstracts, 6, 07442, 2004.

Pohjola, H. and J. Koistinen, Diagnostics of vertical reflectivity profiles at the radar
  sites, in ERAD Publication series, pages 233–237, Delft University of Technology,
  2002.

Rahimi, A. R., G. J. G. Upton, and A. R. Holt, Dual-frequency links - a complement
  to gauges and radar for the measurement of rain, J. Hydrology, 288, 3–12, 2004.

Ravazzani, G., N. Montaldo, and M. Mancini, Modellistica idrologica distribuita per
  il caso di studio del bacino del fiume toce, in Atti del 28◦ Convegno di Idraulica e
  Costruzioni Idrauliche, Potenza, Italia, volume 1, pages 341–348, 2002.

Reisner, J. and J. Rasmussen, Explicit forecasting of supercooled liquid water in winter
  storms using the mm5 meso-scale model, Quart.J.Roy.Meteor.Soc., 124B, 1071–
  1107, 1998.
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   75


Rigo, T. and M. C. Llasat, A methodology for the classification of convective structures
  using meteorological radar: Application to heavy rainfall events on the mediter-
  ranean coast of the iberian peninsula, Natural Hazards and Earth System Sciences,
  4, 59–68, 2004.

Robbins, G. L., M. Grum, and C. G. Collier, Combining and comparing rainfall ob-
  serving systems in hydrological modelling of a rural catchment, in 3rd European
  conference on Radar Meteorology, pages xxx–yyy, 2004.

Saltikoff, E., J. Koistinen, and H. Hohti, Experience of real time spatial adjustment
  of the z-r relation according hydrometeors, Phys. Chem. Earth (B), 25, 1017–1020,
  2000.

Salvador, C. and B. Navascués, Vad experiments using the inm weather radar network,
  Technical Report 45, HIRLAM Newsletter, 2004.

Salvador, C., I. Sanz, B. Navascués, and J. A. García-Moya, Usage of weather radar
  network observations in nwp at inm, Technical report, Servicio de Modelización
  Numérica del Tiempo. Instituto Nacional de Meteorología, 2004.

Sánchez, J. L., R. Fraile, J. L. de la Madrid, M. T. de la Fuente, P. Rodríguez, and
  A. Castro, Crop damage: the hail size factor, J. Appl. Meteorol, 35, 1535– 1541,
  1996.

Sanchez-Diezma, R., D. Sempere-Torres, J. Bech, and E. Velasco, Development of a
  hydrometeorological flood warning system (ehimi) based on radar data, in On-line
  abstracts of Use of Radar Observations in Hydrological and NWP Models, 2002.

Sánchez-Diezma, R., I. Zawadzki, and D. Sempere-Torres, Identification of the bright
  band through the analysis of volumetric radar data, Journal of Geophysical Re-
  search, 105 (D2), 2225–2236, 2000.

Sancho, O. G. and J. Fleitz, Saih flood warning system and emergency management
  in the jucar river (spain): the case study of 22-24 october 2000, in Mitigation of
  Climate Induced Natural Hazards (MITCH) Workshop II: Advances in Flood Fore-
  casting, page 15, Barcelona (Spain), 10-12 June 2002, 2002, Flood Warning and
  Emergency Management.

Sanz, I. and J. A. García-Moya, Using rsm to validate mesoscale hirlam at inm, in
  Procceedings of HIRLAM and SRNWP Workshop on Verification Methods, 2004,
  De Bilt.

Schatzl, R., Skalenabhangiger vergleich zwischen wetterradardaten und nieder-
  schlagsmessungen, Dissertation, Graz, 2001.

Sempere-Torres, D., C. Corral, R. Sánchez-Diezma, M. Berenguer, C. Velasco,
  M. Franco, X. Llort, E. Velasco, and J. Pastor, Are radar rainfall estimates ready
  for hydrological applications? some reflections from the experience at. catalunya,
  in European Conference on Radar in Meteorology and Hydrology (ERAD) - COST
  717 Final Seminar. Abstracts Book, page 30, 2004, Copernicus GmbH (c).
Status and Perspectives on Using Radar Data in Hydrological and NWP Models in Europe   76


Sempere-Torres, D., J. M. Porra, and J. D. Creutin, Experimental evidence of a gen-
  eral description for raindrop size distribution properties, Journal of Geophysical
  Research, 103 D2, 1785–1797, 1998.

Sempere-Torres, D., J. M. P. Rovira, and J. D. Creutin, A general formulation for rain
  drop size distribution, Journal of Applied Meteorology, 12, 1494–1502, 1994.

Seo, D. J., Real-time estimation of rainfall fields using radar rainfall and rain gage
  data, J. Hydrol., 208, 37–52, 1998.
              ˇ c
Sokol, Z., D. Rezᡠová, and P. Pešice, Radar and raingauge estimates and hourly sums
  of precipitation for river basins in the czech republic, in ERAD Publication series,
  volume 1, pages 286–290, 2002.

Vehviläinen, B., O. Niemi, J. Koskela, and J. Koistinen, The use of doppler radar in
  the kyrönjoki watershed model, Suomen ympäristökeskuksen moniste 248, Vesistö-
  mallijärjestelmän kaukokartoitustietojen käytön ja prosessimallien kehittäminen,
  Helsinki, 2002.

Velasco-Forero, C., D. Sempere-Torres, R. Sánchez-Diezma, E. Cassiraga, and J. J.
  Gómez-Hernández, A non-parametric methodology to merge raingauges and radar
  by kriging: Sensitivity to errors in radar measurements, in Proceedings of ERAD
  (2004), pages 21–24, 2004, Copernicus GmbH.

Šálek, M., The radar and raingauge merge precipitation estimate of daily rainfall - first
  results in the czech republic, Phys. and Chem. of the Earth, 25, 977–979, 2000.

                  r
Šálek, M. and L. Bˇezková, Utilization of radar-based precipitation estimates in hydro-
  logical models in the czech republic, in ERAD Publication series, volume 2, 2004,
  in press.

Šálek, M., P. Novák, and D. J. Seo, Operational application of combined radar and
  raingauges precipitation estimation at the chmi, in ERAD Publication series, vol-
  ume 2, 2004, in press.

Weipert, A. and C. Pierce, Multi sensor supported flood detection and monitoring sys-
 tem for poland, in 31st Conference on Radar Meteorology (Volume II), pages 718–
 721, AMS, 2003.

								
To top