HarmoniQuA - State-of-the-art

HarmoniQuA - State-of-the-art
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-1
7. STATE-OF-THE-ART FOR FLOOD FORECASTING MODELLING
Gábor Bálint
Water Resources Research Centre, Hungary (VITUKI)
7.1
Definition of the flood forecasting domain
There are several different types of flooding. The analysis below focuses on
precipitation (both rainfall and snowmelt) induced floods regardless whether those occur
in localised or distributed forms or develop as fast response or slow gradual process. Ice
jam induced floods, mudflows, ravines and floods caused by the failure of hydraulic
structures are not directly addressed. Inundation generated by marine conditions is left
out from the scope of this domain.
Flood forecasting (FF) modelling is aimed at producing future estimates of a flood
related hydrological variable based on the present state and past behaviour of the
modelled catchment or river reach. As in case of any forecasting procedure, “real-time”
and “future” remain the key words and purely statistical predictions of flood frequency
fall out of scope of the given analysis. However it does not mean the exclusion of
probabilistic forecast, ‘ensemble’ forecast, which represents transient approaches based
both on assumptions and/or set of forecast of input variables on one side and on the
current state of the modelled system.
Distinction is made between flood detection, indicating the likelihood of flood
formation from hydro-meteorological analysis; flood forecasting, meaning quantitative
estimates of hydrological variables during forthcoming flood events; and flood warning
issued to the appropriate authorities, specialised users and to the public on the extent,
severity and timing of the flood or giving only qualitative estimates indicating the
possibility of exceedance of certain thresholds. No strict limits exist between the above
activities but attention is given only to the first two stages. Any modelling tool being
able to work in flood conditions can be considered. Consequently no strict distinction is
made between general purpose hydrological forecasting and flood forecasting.
The whole ranges of possible catchment size and possible lead time are covered by
flood forecasting: from very small (urban catchments few km2 large or hilly,
mountainous catchments covered by flash flood watch and warning – lead time of
forecast from 5 minutes up to 3-6 hours) through small and medium size river basins up
to large basins covering large regions in continental scale (over 10,000 km2 – lead time
of forecasting days or even months).
Flood forecasting is an operational, result-oriented activity and as such pays less
attention to the modelled system than to the output of the forecasting procedure. Outputs
usually can be peak river flow, peak stage or flood crest, flood flow and stage or water
level hydrographs, flood volume, and inundation depths over the floodplain.
An attempt to get a prior estimate of the listed variables may require only a single
data collection or modelling step, but often requires a whole chain of coupled or
integrated modules. Flood forecasting modelling may need observed meteorological
data, hydrometeorological analyses, meteorological forecast, rainfall-runoff modelling,
channel routing, hydraulic or hydrodynamic modelling, and updating or error correction
steps. Beside the above listed process-oriented elements representing analysis and
forecasting models the flood forecasting system includes or is closely linked to
observation and data transmission network, pre-processing module handling, data
ingestion, assimilation and post-processing module(s) occupied with forecast product
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-2
preparation, dissemination and to decision support systems. Consequently, the flood
forecasting system is rather the managing and integrating shell of the listed modules,
which also performs the role of the interface to the linked systems. Flood forecasting
modelling is rather the implementation of different modelling tools for a given
application than a completely autonomous domain.
Owing to its large practical importance FF modelling is crucial in any water related
decision-making. In the context of the HarmoniQuA project, it has close links with other
domains. First of all, precipitation-runoff modelling, since this is the most essential tool
that many forecast systems apply. Flood forecasting is also directly related to the
hydrodynamic domain, and flood forecasting is one of the fields where this type of
modelling is implemented. Floods play an important role in the wash out, leaching,
transport of sediment and pollutants. Therefore flood forecasting is the basis for realtime modelling of pollutant transport, and also of early warning for pollution incidents
during flood events. Links therefore also exist to the surface water quality domain.
7.2
Needs of QA guidelines
Generally, hydrological forecasting has a well-defined application value, and within
that field even more interest is focused on flood forecasting. Simple gauge relation
based flood forecasting dates back to ancient civilisations. The application of this simple
approach in graphical form first appeared in the flood forecasting system of the Seine
basin soon after the mid-XIX century, but in some places it is used until nowadays. The
development of complex FF models is historically a more recent development. One of
the first computerised rainfall runoff models that found practical application in flood
forecasting dates back to the 1960s (Sittner et al., 1969). Wide spread FF modelling and
application really started even later with the advent of the age of personal computers in
the late ‘70s and early ‘80s.
Flood forecasting for most countries is a task carried out by state or regional
environmental or water related agencies. Meteorological services always play a certain
role and flood forecasting typically is their direct responsibility in cases of joined hydrometeorological services or institutions. Local authorities, municipalities may also be
involved by their role prescribed by legislation or voluntarily recognising local interests
filling up the gaps or inadequacies of nation wide or regional systems.
Since the first flood forecast centres were established at the very beginning,
hydrological forecasts have been issued for a flood crest at a given point on a river or
stream. Each flood forecast was based on a pre-determined flood stage where damage
would occur in the reach surrounding the forecast point. These forecast products were
issued as single point values for an expected time when the crest might occur. The
watershed areas affecting these forecast points were generally large, on the order of
thousands of square kilometres. In later years, the number of hydrological forecast
points increased and the hydrological (rainfall-runoff) forecast areas have become
increasingly smaller, 500 to 1,000 km2 in the USA and most European countries (Braatz
et.al., 1997; Packman, Chapter 17). These tendencies may change general attitudes
towards flood forecasting systems, instead of looking at each of those as unique entities,
forecasting schemes can be treated rather as the realisation of fairly similar elements
requiring uniform Quality Assurance procedures to guarantee their proper functioning
As technologies have been advancing and the demands for more accurate and timely
flood forecasting has been increasing, the public, private and other state agency sectors
are insisting upon the expanded use of hydrological/meteorological analyses and
products for the flood management of water courses. To meet this need, services
responsible for hydrological forecasting are capitalising upon modernisation in remote
sensing, data automation and advanced hydrological hydrometeorological modelling.
Extreme flood events from 1993 until recently give impetus to apply more and more FF
modelling tools in real life conditions (Casale et al., 1996). Similar tendencies are
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-3
reported from the USA, where the National Weather Service Advanced Hydrologic
Prediction System (AHPS) intends to meet these objectives (Green et al., 1994, Braatz
et al., 1997).
The diversity and complexity of flood forecasting technologies have on the other
hand made it difficult to find overall satisfactory solutions to the implementation of
different FF models in operational practice. The gap between the community of model
developers and the user community in many European countries in the field of rainfall
runoff modelling reported by Perrin et al. (Chapter 5) is also valid for all flood
forecasting applications.
The above factors explain the almost non-existence of Quality Assurance guidelines
in the field of FF models. However given the wide range of existing hydrological
techniques, and operational tools to link and apply them, makes it feasible to select
different FF models. In this respect, coupling them into integrated, complex systems is
also an issue to be addressed together with the creation of interfaces to decision support
systems. All these steps are to be supported by QA guidelines to maintain good practice
to meet users needs towards flood forecasting or at least urging forecasters to clearly
state accuracy and other quality criteria their systems are able to pass.
Guidelines are needed to elaborate and implement procedures to validate model
outputs to prevent fals warnings. However natural inherent uncertainity of any any flood
forecast has to be recognised and evaluation of the flood warning system is part of the
planning of a decision support system (Haimes et al. 1996).
7.3
Discussion in scientific literature
Flood forecasting schemes may have the most diverse structure depending on
catchment size, response or concentration time and the availability of real-time input
data. The core of the forecasting system is often shifted from modelling tools to the
observation and data collection systems. In such cases the remotely sensed, telemetred
or otherwise observed and transmitted precipitation or upstream river flow value with or
without very simple transformation may generate an alarm or early warning.
Under similar conditions meteorological forecast, namely quantitative precipitation
forecast (QPF), occupies the same role and serves as basis for flood warning. More
complex nowcasting or very short time weather forecasting schemes may have similar
applications (e. g. Stallings, and Wenzel, 1995; HYDROMET 2001).
As mentioned above flood forecasting modelling includes process models including
updating or error correction modules and a number of steps representing the workflow
of the preparation of hydrological forecasting. The quality and timeliness of flood
forecasting is often decided by the efficiency of observation and data transmission
network and the possibilities of data assimilation. Also the presentation of forecast
results, the quality of forecast product and its timely transmission, dissemination decide
the overall value of the flood forecasting system together with effective use of the skills
of professional employees occupied with the job. The weight of flood forecasting
modelling should be judged in this context.
The proper selection of modelling tools for flood forecasting purposes is, beside the
type of flood, decided by the degree and development of the observation network and
telecommunication and data processing facilities; the length and quality of data records;
and not least the availability of qualified personnel to run and maintain the models on a
routine bases (WMO 1994; Boyle et al., 2001).
Data, and the quality of the data, are critical to any hydrological forecasting
procedure. Operational use of models is preceded and accompanied by developing,
implementing, and maintaining hydrological models and systems. The forecast models
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-4
used are developed and calibrated for specific rivers and watercourses based on
historical events. They are conditioned and constrained operationally using current
observations and, in the case of operational ensemble forecasts, with historical data as
well. Inaccurate, inconsistent, incomplete or insufficient data can cause significant
problems in the forecast process.
There are several high standard approaches for data analysis of both operational and
historical data, as well as for archiving and retrieving historical data. The data analysis
approaches are targeted at helping users to reduce uncertainties associated with the use
of data, and they include: automated double-mass analysis for inconsistency checks,
wavelet transform for nonstationary time-series analysis, cluster analysis for the study of
regionalisation, and robust outlier detection for spatial inconsistency checks (Haimes,
1979). These approaches are being integrated into our operational development and
forecast processes.
Traditional sources of historical data have to be used. A recent tendency is to capture
and archive operational data that have been lost in the past. Tools that provide new ways
to view and access historical data and inventories of historical data may support the
forecasting procedure. World Wide Web (WWW) facilities (Pan et al., 1998;
Vähviläinen, 2002) have been developed as simplified interfaces to traditional data
access utilities and a new interactive "browser" has been developed that allows users to
query and browse inventories of the historical data as well as view and retrieve the data
itself. The ability to query and view the data inventories greatly improves the user's
effectiveness in selecting data and the efficient use of forecasting models.
Quality control (QC) of input data
The purpose of quality control is to prevent "bad" data from being used in various
hydrological processes (calibration, modelling, forecasting, etc.). For quality control of
hydrometeorological data one performs a variety of checks, such as range checking,
spatial inconsistency checking, temporal inconsistency checking, internal consistency
checking, and multi-sensor inconsistency checking (Krajewski, 1986,1987). Real-time
quality control systems are to be designed to support forecasting operations. The
features of these systems include real-time access to operational data as it is captured
from the observation network, efficient and robust outlier detection, reanalysis,
integration of a variety of information for decision making, and multiple temporal scale
data handling (Bissell and Zimmerman, 1992). Real-time access and processing are
important because they support the forecaster's need to make decisions in short time
frames (one minute or less) as the data are arriving.
The proper strategy and methods to be used for spatial inconsistency checks is the
key issue. To satisfy the operational requirements, simple and robust algorithms are
needed to perform efficient and effective detection of suspect data, and then these
suspect data have to be reanalysed. Finally, in a proper system users can perform multisensor data comparison, and if possible, validate these outliers. Regionalisation,
elevation zones etc. are needed for the implementation of spatial inconsistency checks
that support efficient real-time quality control (Fovell et al., 1993). Cluster analysis tools
can be applied for regionalisation using different temporal scale data as input. The steps
of robust outlier detection are: Determine the median and mean absolute deviation
(MAD) of N stations in each climate region (step 1); Determine indices of different
percentile for each station (step 2). Madsen (1992) has implemented a similar approach
for daily precipitation quality control. Similar approaches including reanalysis and
estimation methods are also suggested (Miller et al., 1992).
Calibration of FF models
The general statements (Refsgaard and Henriksen, Chapter 3) regarding the
assessment, selection, calibration and running models are valid also for FF use. Fread et
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-5
al. (1995) suggest to combine procedures needed to process historical
hydrometeorological data and to estimate model parameters for a specific basin. The
models simulate snow accumulation and ablation, calculate runoff and the temporal
distribution of its delay from the basin to the basin outlet, and route streamflow through
reservoirs and channel systems. There are many modular systems used for hydrological
forecasting purposes that allow the hydrologist to select from a variety of models and to
configure them in a manner that is descriptive of the basin. As part of the calibration
procedure, for a particular basin, the simulated streamflow is statistically and visually
compared to the observed streamflow to determine the necessary model parameter
adjustments. The ideal model parameters are those with which the model simulated
streamflow most closely matches the observed streamflow.
Criteria
Flood forecasting is a well defined area of model application of (e.g. rainfall runoff),
but the choice of criteria is still largely user dependent. Different criteria listed by Perrin
et al., (Chapter 5) for rainfall runoff modelling, and those recommended by WMO
(1992) together with other tools for the simulated real-time intercomparison of
forecasting models may serve:
· different forms of model error (cumulative, quadratic, absolute);
· different target variables (streamflow or transformed values with root square or
logarithmic transformation);
· absolute or relative forms, for example the classical root mean square error or the
Nash and Sutcliffe (1970) criterion R2;
· multiple step ahead forecast statistics, such as the root mean square error of the
step ahead forecasts, applied for flood event wise or according to different lead
times of forecast;
· peak statistics: maximum forecasted peak, time of occurrence of the maximum
forecasted peak, differences in times and occurrence of the maximum forecasted
and actual peaks, forecast of the time when the flow will cross the threshold,
differences in forecasted and actual times of crossing the threshold;
· additional to the Nash – Sutcliffe criterion this one is compared with the criterion
calculated for the ‘naive’ forecast model given by “one step ahead forecast =
value now”. This description is valid for any existing lead time of the forecast.
The given criterion is sometimes referred as efficiency or persistence criterion
(Corradini 1986; Kitanidis and Bras, 1980);
· the statistics are calculated for model residual errors, i.e. errors without updating
to enable the separate evaluation of the performance of the process model and
that of the statistical updating.
Many prescribed procedures suggest to use linear or logarithmic graphs of calculated
and observed flood variables plotted against time (Glaudemans et al., 2002).
Real-time updating
Once the models have been calibrated for a basin, they can be used operationally with
real-time hydrometeorological data to forecast river flows and stages. Real-time
hydrological forecasting models consist in one or more “process models” or “simulation
models” which can be supplemented by a procedure of forecast updating. The process
model utilizes measured or estimated input data. The process model consists of a set of
equations that contain state variables and parameters. The process model output is
observable and is generally discharge or stage. Updating procedures consider the
prediction errors (differences between computed and measured discharges/stages) in
order to modify the model’s forecast and improve the model’s performance during
operational use. The difference between simulated and measured discharges up to the
time of forecast can be accounted for by errors induced by the model input data,
imperfect model structure, limitations in model calibration in terms of short data series,
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-6
the time dependent change of catchment characteristics, and errors in the discharge
hydrographs at the gauging station.
The real-time updating procedures differ from the techniques of periodical, historical
recalibration of models. The periodical recalibration of the model parameters may be
necessary as the characteristics of the catchment slowly change in time due to non
stationary influences.
Forecast updating procedures consist of approaches, which update one or more of the
following: input variables, parameters or output variables. They generally consist of
either completely automated methods or manually interactive ones (e.g. trial and error).
One of the most widely studied updating technique, that generated considerable
attention in the last decades is that of the Kalman filter (Refsgaard et al. 1983, SzöllősiNagy 1987, Refsgaard 1997). Despite the initial high expectations raised among
hydrologists concerning Kalman filter as an updating tool for flood forecasting
reservations concerning its superiority have emerged. Xiong and O’Connor (2002)
advocate the use of simpler approaches such as autoregressive (AR) schemes, and they
insist on the reasonable and proportional use of statistical updating. The function of
these procedures is to improve the performance of process models rather than hiding
their inadequacies.
One of the WMO (1992) intercomparison studies concludes that, it is valuable for
models to adopt automatic updating algorithms. This reduces the need for human
intervention and subsequently increases the operational nature of the forecast model.
This also increases the transferability of the procedures.
Ensemble forecast
Since the 1980s the idea to enable a hydrologist to make extended probabilistic
forecasts of streamflow and other hydrological variables (Day 1985) has been
developed. The assumption usually made is that historical meteorological data are
representative of possible future conditions and these data are used as input data to
hydrological models along with the current model states obtained from the forecast
component. A separate streamflow time series is simulated for each year of historical
data using the current conditions as the starting point for each simulation. The
streamflow time series can be analysed for peak flows, minimum flows, flow volumes,
etc., for any period in the future. A statistical analysis is performed using the values
obtained from each year's simulation to produce a probabilistic forecast for the
streamflow variable. This analysis can be repeated for different forecast periods and
additional streamflow variables of interest. Short-term quantitative forecasts of
precipitation and temperature can be blended with historical data to produce a more
realistic transition in meteorological conditions.
A number of attempts exist to make use of medium range weather forecasting
(ECMWF, DWD etc) for hydrological, especially flood forecasting purposes. 41
different perturbed realisations of future weather conditions (Todini 2000; EFFS) are
routed through forecasting models.
Interactive forecasting procedures
Similarly to ensemble forecast testing different scenarios, changing initial and
boundary conditions may serve in sake of holistic search of better forecast result. The
forecaster can interactively make changes to the parameters, data, or current conditions
used for hydrological simulation and quickly see the results of such changes. These
changes can be categorized into those affecting time series and those affecting a specific
hydrological model. A graphical user interface may support to perform the required
changes.
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-7
Product generation
The value of flood forecasting can be enhanced by proper presentation of forecast
results. As an example US National Weather Service River Forecasting System prepares
the following products: River Statement (RVS), Flood Statement (FLS), and Flood
Warning (FLW) (Office of Hydrology, 1996). When initiated, it compares observed and
forecast river stage data with threshold stages, and tracks the history of recently issued
products, and then determines a recommended product and the forecast points the
product should include. The forecaster can accept these recommendations and generate
the product using predefined templates that control the product format and content.
Alternatively, the forecaster can customize the product extensively - e.g., a different
product can be created with different forecast points included. Also, the forecaster can
select from the predefined templates for each section of the product, and thereby control
the precise wording and appearance of the product. A default set of predefined phrase
templates is provided, and each office is able to modify or add to these templates to
meet their local needs. In addition to providing the functions necessary to customize,
generate, and edit a product, RiverPro provides textual displays of information to
support the forecaster in the decision-making process for product issuance. The
forecaster can view tabular summaries of the stage data and reference data for stations,
and can review information about previously issued products, including the product
itself. After the product has been tailored as necessary, and reviewed, the product can be
issued to the appropriate data dissemination circuits.
Uncertainties and risk communication
Hydrological forecasting systems should provide information regarding the relative
uncertainty of hydrological variables (i.e., river stage and discharge). The increased
lead-time of forecast may greatly improve the capability of water facility and emergency
managers to take timely and effective actions that significantly mitigate the impact of
major floods, however the increased lead-time is usually associated with increased
uncertainty.
Interface to Decision support systems (DSS)
FF remains one of the most important field of hydrological model application and
potential difficulties identified by Cunge and Samuels (1996) should be considered in
case of forecasting usage, namely:
· lack of appreciation of the range of uncertainty in the forecast results;
· the temptation to believe every number that a computer produces;
· illusory visualisation of model results;
· possibility of using models outside their range of definition;
· unsatisfactory calibration of the model.
Trans-boundary rivers
Distribution of forecasting tasks may need special organisational and managerial
efforts even within national borders regulating inter agency collaboration, but transboundary rivers certainly need special approaches. Several major rivers cross or form
national or provincial borders. Thus flood forecasting in these river basins has the
additional complexity of requiring international co-ordination and co-operation. The
issues in forecasting trans-boundary rivers are not restricted to the major rivers as far as
administrative borders do not coincide with limits of drainage basins. The RIBAMOD
project identified some trans-national issues on flood management, among those many
are directly related to FF issues (Samuels, 1999):
· hydro-meteorological networks for flood forecasting;
· trans-boundary compilation of radar images for flood forecasting;
· exchange of flood forecast information between states.
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-8
Combination of elements with evident heterogeneity is the main issue flood
forecasting faces on trans-boundary rivers (Balint et al, 1990). Quality assurance should
tackle problems arising from the use of input data having different data formats,
accuracy, observation time and frequency. Inclusion of foreign forecast results is often
needed although they bring additional uncertainty into the system. European
standardisation of data exchange and forecasting approaches could deliver real benefit in
improving flood forecasting.
7.4
Existing guidelines
As underlined previously, flood forecasting for most of the countries is a task carried
out by state or regional environmental or water related agencies or
meteorological/hydrometeorological services. There is today a lack of guidelines in the
flood forecasting domain partly linked to this institutional set up.
In the UK case study (Packman, Chapter 17) it is reported that flood–forecasting
modelling in the UK is wholly an EA/SEPA function, separate modelling strategies have
developed in each region and even for different catchments within each region. Basic
rainfall-runoff models, real-time updating, data gathering procedures, and operating
systems all vary considerably. Models range from simple graphical relationships
between flow gauges, through transfer function models, to complex deterministic
models that include hydrodynamic river routing. Following recent flood events, the EA
is developing a more standardised, modular approach, recognising a role for models of
differing complexity, but including them as generic modules within a standard operating
shell. The approach is being implemented through large contracts to upgrade flood
forecasting hardware and software – initially in three EA regions. The Tender
documents comprise four stages:
Stage 1 Data abstraction - appropriate checking of hydrometric, catchment, and
channel data
Schematisation – appropriate division into sub-catchments, reaches, and
structures.
Outputs – preliminary model, programme of work, required accuracy/reliability
Stage 2 Review – test significance of model components, sufficiency of data,
consultation
Stage 3 Calibration – continuous data & events, automatic & manual method, physical
limits
Verification – to required accuracy. Recalibration & re-verification
Outputs – verified model, detailed verification report and data sets.
Stage 4 Test real-time operation and updating.
Overall, the use of standardised section headings emphasises the modularity of the
approach, but does not give a clear structure for presenting modelling issues and
guidelines.
In the USA the National Weather Service's (NWS) river forecasting centres (RFC)
work under uniform requirements. The NWS Office of Hydrology is currently
developing a subset of these applications, referred to as the WFO Hydrological Forecast
System (WHFS), that provides the WFO forecaster with the capability to issue warnings
of flood and flash flood events in real-time.
More guideline types of documents are available for flood warning purposes,
however in this domain flood forecasting occupies a minor place only. (US ACE 1988,
FCD Maricopa County 1997). The “Flood Control District of Maricopa County:
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-9
Guidelines for developing a comprehensive flood warning program” mostly deals with
the dissemination of warnings and emergency response measures. Forecast procedures
are hidden within the task of “flood threat recognition”. The main concern is the setting
of the system of observations: observers, automated gauges, radar and satellite data,
meteorological support. Hydrological models are dealt within the category of decision
aids.
Co-ordination of national and provincial hydrological forecasting agencies exist in
the Rhine and Danube basins (IKSR-CIPR 1995, ICPDR).
7.5 Conclusions and recommendations with respect to the further HarmoniQuA
work
At this stage of the work, it was not possible to identify general good practice
guidelines about flood forecasting modelling. The previous analysis shows however that
scientific studies and several works related to water issues made by hydrologists could
serve as a basis to build practical guidelines in the context of the HarmoniQuA project.
In such guidelines, one could distinguish several quality assurance procedures
concerning model selection, model evaluation (with a selection of adequate criteria),
model parameter determination and operational implementation.
It can also be mentioned that the review of previous and ongoing comparative
assessments of models WMO (1992), AFORISM (Todini, 1996), and EFFS could be of
help to define standard procedures for model evaluation.
Concerning criteria, links should be made with the BMW project that aims at
proposing criteria for model evaluation.
7.6
Acknowledgement
The present work was carried out within the Project ‘Harmonising Quality Assurance
in model based catchments and river basin management (HarmoniQuA)’, which is
partly funded by the EC Energy, Environment and Sustainable Development programme
(Contract EVK2-CT2001-00097). Comments by Péter Bartha are kindly acknowledged.
7.7
References
Balint, G.;Ambrus, S.; Bishop, R. (1990) Flood forecasts on transboundary rivers in Hungary with
parallels in Canada. Proceedings of the Joint AWRA-CWRA Conference on International and
Transboundary Water Resources Issues, Toronto 62-71
Bissell, V.C., and Zimmerman, J.R. (1992) A Data Quality Description Model For Hydrometeorological
Operations, Reprint Of The Ams Eighth International Conference On Interactive Information and
Processing Systems For Meteorology, Oceanography, and Hydrology, Atlanta, Georgia, 56-62,
January 5, 1992.
Boyle, D. P., Gupta, H. V., Sorooshian, S., Korn, V., Zhang, Z. and Smith, M. (2001) Toward improved
streamflow forecasts: value of semidistributed modeling. Water Resources Research, 37(11), 27492759.
Braatz, Dean T., Halquist, John B., Wavrin, Robert J., Ingram, John J., Feldt, J. John (1997) NWS
Hydrological Products and Services: Moving from the Traditional to the Technically Advanced, 13th
International Conference on Interative Information and Processing System for Meteorology,
Oceanography, and Hydrology, American Meteorological Society, Long Beach, California, J63-J69.
Casale R., Havnø K. and Samuels P. (Eds.) (1996) RIBAMOD River basin modelling management and
flood mitigation Concerted Action , Proceedings of the first expert meeting on Model Structure and
Decision Support, EUR 17456 EN, ISBN 92-827-9562-4
Corradini, C., Melone, F. and Ubertini, L. (1986). A semi-distributed adaptive model for real-time flood
forecasting. Water Resources Bulletin, 22(6), 1031-1038.
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-10
Cunge J. A. and Samuels P. (1996) Future modelling needs In: Casale R., Havnø K. and Samuels
P.RIBAMOD River basin modelling management and flood mitigation Concerted Action ,
Proceedings of the first expert meeting, EUR 17456 EN, ISBN 92-827-9562-4
Day, G.N., (1985) Extended streamflow forecasting using NWSRFS, Journal of Water Resources
Planning and Management, ASCE, Vol. III, No. 2, 157-170.
EFFS (2002) An European flood forecasting system. http://effs.wldelft.nl/
FCD (1997) Flood Control District of Maricopa County: Guidelines for developing a comprehensive
flood warning program. Phoenix, Arizona, January 1997
Fovell, R. G., and M. C. Fovell (1993) Climate zones of the conterminous United States defined using
cluster analysis, J. Climatol., 6, 2103-2135.
Fread, D.L., R.C. Shedd, G.F. Smith, R. Farnsworth, C.N. Hoffeditz, L.A. Wenzel, S.M. Wiele, J.A.
Smith, and G.N. Day (1995) Modernization in the National Weather Service River and Flood
Program, Weather and Forecasting, Vol. 10, No. 3, American Meteorological Society, Boston, MA.
Glaudemans, M.J. , J.M. Roe, P.S. Tilles (2002) The ingest quality control, and processing of
hydrometeorological data at National weather service field offices. 82th AMS Annual Meeting,
Interactive Symposium on the Advanced Weather Interactive Processing System, Orlando, Florida
January 13-17, 2002 Paper J9.1
Glaudemans, M.J. , R. A. Erb, E.B. Wells, J. Zimmerman, J.J. Hill-Maxwell, K.S. Mack (2002) Overview
and status of the hydrologic forecast system in the National Weather Service weather forecast offices.
82th AMS Annual Meeting, Interactive Symposium on the Advanced Weather Interactive Processing
System, Orlando, Florida January 13-17, 2002 Paper J7.6
Green, C., van der Veen, A., Wierstra, E. and Penning-Rowsell, E. C.(1994) Vulnerability refined:
Analysing full flood impacts. In: Penning-Rowsell, E. C. and Fordham, M. (Eds) Floods across
Europe: Hazard assessment , modelling, and management. Middlesex University Press, London
Haimes, Y.Y. (1979) The worth of streamflow data in water resources planning: computational results
Water Resources Research 15(6), 1335-1342
Haimes, Y.Y., Li, D. Tulsiani, V. Lambert, J.H. and R. Krzysztofowicz (1996) Risk –based evaluation of
flood warning and preparadness systems Vol. 1 Overview, Vol. 2 Technical, US Army Corps of
Engineers, IWR Report 96-R-25 and 96-R-26
HYDROMET (2001) The developmentof active on-line hydrological and meterological models to
minimize impact of flood http://www.cen.bris.ac.uk/civil/research/wemrc/hydromet/hydrometp1.htm
IKSR-CIPR International Committee for Protection of the Rhine (1995) Grundlagen und Strategie zum
Aktionsplan Hochwasser, IKSR-CIPR, Koblenz
Kitanidis, P. K. and Bras, R. L. (1980) Real-time forecasting with a conceptual hydrologic model. 2.
Application and results. Water Resources Research, 16(6), 1034-1044.
Krajewski, Witold F. (1986) Quality Control Of Hydrometorological Data, Preprints Of The Second
International Conference On Interactive Information and Processing Systems For Meteorology,
Oceanography, and Hydrology, Miami, Florida, January 13, 1986.
Krajewski, Witold F. (1987) Radar-Rainfall Data Quality Control By The Influence Function Method,
Water Resources Research, Vol. 23, No. 5
Madsen, H., (1992) Semi-automatic quality control daily precipitation measurements, the 5th Int. Meeting
on Stat. Climatology, Canada, 375-377.
Miller, P., and S. Benjamin (1992) A system for the hourly assimilation of surface observations in
mountainous and flat terrain, Monthly Weather Review, 120, 2342-2359.
Nash, J. E. and Sutcliffe, J. V. (1970) River flow forecasting through conceptual models. Part I - A
discussion of principles. Journal of Hydrology, 27(3), 282-290.
NWS (1994) The Great Flood of 1993, NOAA Natural Disaster Survey Report, U.S. Department of
Commerce, National Oceanic and Atmospheric Administration, National Weather Service, Office of
Hydrology, Silver Spring, MD.
Pan, J.J., Bonnin, G.M., Motl, R.L., and Friedman, H., (1998) Recent developments in data analysis,
quality control, and data browsing at the National Weather Service Office of Hydrology, 78th AMS
Annual Meeting, Special Symposium On Hydrology, Phoenix, Arizona, January 11-16, 1998.
Penning-Rowsell, E. C., and Peerbolte B.(1994) Concepts , policy and reaserch. In: Penning-Rowsell, E.
C. and Fordham, M. (Eds) Floods across Europe: Hazard assessment , modelling, and management.
Middlesex University Press, London (1-17)
RAPHAEL (2002) Runoff and atmospheric processes for flood hazard forecasting and control.
http://www.vista-geo.de/Raphael.htm
Refsgaard, J. C. (1997) Validation and intercomparison of different updating procedures for real-time
forecasting. Nordic Hydrology, 28, 65-84
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
7-11
Refsgaard, J. C. and Knudsen, J. (1996) Operational validation and intercomparison of different types of
hydrological models. Water Resources Research, 32(7), 2189-2202.
Refsgaard, J. C. Rojsberg, D., and Marussen, L.M. (1983) Application of the Kalman filter to real time
operation and uncertainty analysis in hydrological modelling. IUGG/IAHS Symposium Hamburg,
IAHS Publ. No. 147 (273-282)
Samuels, P.G. (1999) RIBAMOD, River Basin Modelling, Management and Flood Mitigation. HR
Wallingford Report SR 551
Sittner, W.T., Schauss, C.E. and Monroe (1969) Continuous hydrograph synthesis with an API-type
hydrologic model, Water Resources Res, 5, 1007-1022.
Stallings, E.A., and L.A. Wenzel (1995) Organization of the River and Flood Program in the National
Weather Service, Weather and Forecasting, Vol. 10, No. 3, American Meteorological Society,
Boston, MA.
Szöllősi Nagy, A. (1987) Hydrologic forecasts and warnings In: O. Starosolszky (Ed) Applied surface
hydrology, Water Resources Publications, Littleton, Colorado, Chapter 11 , 592-613
TELFLOOD (1999) Forecasting floods in urban areas downstream of steep catchment
http://www.ucd.ie/%7Eofrss/enginee/civilen/1056.html
Todini E. (1995) - AFORISM - A Comprehensive Forecasting System for Flood Risk Mitigation and
Control, Final Report of Contract EPOC-CT90-0023 (manuscript)
U.S. Army Corps of Engineers, Engineering Division (1996) U.S. Army Corps of Engineers Annual Flood
Damage Report to Congress for Fiscal Year 1995, USACE, Washington D.C.
U.S. Army Corps of Engineers, (1988) General guidelines for comprehensive flood warning /preparedness
studies USACE, Washington D.C.
Vehviläinen, B., Huttunen, M. (2002): The Finnish watershed simulation and forecasting system (WSFS).
In: 21st Conference of the Danubian Countries on the Hydrological forecasting and hydrological Bases
of Water Management, 2-6 September 2002, INMH, Bucharest, Romania, Proceedings CD Topic 1,
ISBN 973-0-02759-5
WMO (1992) Simulated real-time intercomparison of hydrological models. Operational Hydrology
Report n° 38, WMO n°779, World Meteorological Organization, Geneva, Switzerland.
WMO (1994) Guide to hydrological practices, Data acquistion and processing, analysis, forecasting and
other applications. WMO n°168, World Meteorological Organization, Geneva, Switzerland.
Xiong, L. and O'Connor, K. M. (2002) Comparison of four updating models for real-time river flow
forecasting. Hydrological Sciences Journal, 47(7), 621-639.
Balint, Flood Forecasting Modelling
HarmoniQuA – State-of-the-art Report on QA guidelines, October 2002
Balint, Flood Forecasting Modelling
7-12
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement