International Institute for
Infrastructural, Hydraulic and Environmental Engineering,
IHE Delft, The Netherlands
Corresponding Address: P. O. Box
3015, 2601 DA Delft, The Netherlands
Telephone: 31-15-215 18 41/871
Fax: 31-15-212 29 21
Email: abebe@ihe.nl
Abstract: A number of urban areas have experienced severe flooding in the past. The problem has worsened in southern Europe in the past decades mainly due to urbanisation, which has lead to the paving of natural ground surfaces and the blocking of natural watercourses by the infrastructure. In some areas this has aggravated the situation by shortening the hydrological response time, often triggering flash floods within few hours after the occurrence of heavy precipitation. The short time between the occurrence of heavy rain and corresponding flooding makes it difficult to forecast, especially when the meteorological conditions are such that it is difficult to forecast the rainfall. This is especially the case for coastal cities in mountainous terrain.
In order to get sufficient lead time to alarm inhabitants and plan reactive measures, it has been found important to use advanced forecasting and monitoring techniques and a decision support system to assist authorities in their decision making.
This article contains a concise presentation of a decision support system (DSS) developed to assist public authorities plan emergency activities when flooding is expected and to give guidance for restoration activities after the flood subsides. The DSS is designed mainly to work for areas with steep topography. Such areas are known to experience flash floods only a few hours after the occurrence of a severe precipitation event. The target areas are Liguria Region in Italy and the Greater Athens catchment in Greece. Severe and devastating flooding has occurred at both sites several times in the past two decades.
The study shows that the problem of flood warning based on forecasts is one of great complexity and uncertainty. It is believed that results can be improved with better forecasting techniques, improving catchment models, with experience using the DSS, and increasing experience of the public about the meaning of flood warnings. The system developed can be easily applied in other areas with similar flooding problems.
Keywords: decision support system, urban flood warning, and telematics
Several regions in the world experienced flooding in the past. The problem is particularly pronounced in areas of steep and complex orography. The severity and rapidity of flooding has increased in Southern Europe in the past decades due to urbanisation causing higher runoff rates and rapid catchment responses. Several flooding episodes have occurred in the last decades in these areas. Flood forecasting and real time monitoring have been found necessary to take reactive measures such as issue flood warnings, plan emergency activities, and initiate restoration.
This paper presents the work done in the development of a decision support system for flood warning in urban areas. It is part of the work done in the project TELEFLEUR (TELEmatics assisted handling of FLood Emergencies in URban areas) which is sponsored by the European Union DG 13 under its Telematics applications program, Environment sector. The target areas are Liguria Region in Italy and the Greater Athens catchment in Greece. Severe and devastating flooding has been observed in both areas about 10 times in the past two decades.
The main objective of the project is the development of a comprehensive operational system for handling urban flood emergencies that synthesises cutting edge telematics technology with advanced forecasting of meteorology and hydrology encapsulated in a Decision Support System (DSS). This paper however covers mainly the development and functionality of the DSS prototype developed under the project. The decision support system is initially intended to do the following activities:
–Manage the dynamic information provided by telemetry,
–Feed these data, along with relevant static data to an array of modelling tools,
–Forecast flooding conditions, and
–Assist public authorities in decisions regarding emergency measures.
The Athens area is
part of one large catchment area of about 430km2 where as the Liguria
Region consists of several small catchments the largest of which, Entella, has
an area of 370km2. A user requirement analysis indicated that the two
areas where the DSS is to be implemented share a similar type of topographical
and meteorological characteristics. Both constitute urban areas situated on the
coast and are prone to flash flooding by runoff from mountainous areas. Flooding
occurs within a few hours of the occurrence of heavy precipitation, often too
short to take preparatory measures. What is more, the rainfall is difficult to
predict in that the meteorological conditions are such that storm clouds can
form within a very short space of time. It is impossible to issue flood warnings
based on hydrological simulations from observed precipitation. Therefore it is
believed that flood forecasts and thus warnings should entirely depend on forecast precipitation instead of observations directly. Warnings
depending largely on forecasts, however, involve a lot of uncertainty, which
opens wide the possibility for the issuance of false alarms. This in fact
complicates the problem further, causing potential social, economic and other
related issues.
However, it is also
true that, with the experience of a synoptic analyst and with growing knowledge
about the hydrological and hydrodynamic response of a catchment, the quality of
decisions made using the DSS will improve with time. This dictates that the DSS
should be loosely coupled (in software engineering terms) with the models so
that the hydrological/hydraulic model of a catchment can be modified or changed
without altering the DSS, and thus can be used at other areas with similar
catchment properties.
Precipitation is the main cause of urban flooding in both areas. A precipitation forecast is done using an array of meteorological models nested in one another. The first of these models is the GCM (global circulation model) that covers the Northern Hemisphere with a ground resolution of 100km grid. The forecast is made for the coming 72 hours. The model takes approximately 12 hours to run. Nested within this model there are two Limited Area Models (LAM), one nested within the other. The smaller one has a resolution of around 10Km x 10km.
The results from these models however, have to be interpreted by a synoptic analyst. This is because there is a high level of uncertainty in the results from the models due to the complexity of weather phenomena making it impossible to localise the events either in space or in time. The synoptic analyst therefore produces three kinds of forecast:
– Maximum 12 hours cumulative precipitation, 2000km2 domain
– Maximum 6 hours cumulative precipitation, 500km2 domain
– Maximum 3 hours cumulative precipitation, 100km2 domain
This results in multiple, equally likely scenarios that may be expected.
A hydrological model is used to predict runoff along the main streams in the catchment. Then a hydrodynamic model predicts the flow profiles in the main reaches of the channel network at regular time intervals.
Since the resolution of the above precipitation forecasts does not satisfy the data needs of the hydrological/hydraulic model used for the purpose of the DSS, it was decided that model simulations have to be carried out for all scenarios to determine the worst case. Again the question arises, what is the worst case scenario? Two criteria can be identified to identify the worst case scenario: the maximum inundation at a certain location, and a location inundated for the longest duration. One more time series is added to assist decision-makers: the envelope of all the scenarios! A post processor is used to produce these three time series at key points in a catchment/basin.
As
a means of converting the model predictions to verbal warning levels, the DSS
uses three threshold water levels (or discharges) to classify the degree of
flooding at a location:
If threshold 1 is exceeded then flood degree is MILD.
If threshold 2 is exceeded then flood degree is SIGNIFICANT.
If threshold 3 is exceeded then flood degree is SEVERE.
Since several equally likely scenarios forecasted by the synoptic analyst are simulated, the DSS uses the probabilities of exceeding the thresholds to deal with the uncertainty of the situation. For example, the probability of a SEVERE event at a location is the number of events in which threshold 3 is exceeded divided by the total number of equally likely scenarios.
The DSS prototype is formulated to have the operational structure shown in Figure 1. The figure shows the logical linkage between the DSS software and other components such as the models, GIS, database, public authorities, etc., and the flow of information involved in the process. In order to meet the above requirements, the DSS software is designed in such a way that it has two main systems (the warning system and restoration system) and a number of supplementary modules described below.
This system (see the screen dump in Figure 2) is responsible for the issuance of flood-warnings based on the model simulation results. It is based on the results of the hydrological/hydraulic simulations resulting from the array of meteorological models, or can also include other forecast systems such as radar technology. It allows visualisation of flood warnings on top of GIS maps.
This system (see the screen dump in Figure 3) is responsible for the instantiation of restoration activities after a flood event has subsided. It is based on the real time data obtained from the remote water level and precipitation monitoring stations via telemetric system. Data acquisitions can be made either on request or at regular time intervals. It allows the visualisation of real time data graphically, and subsequently generates an event log when the water level at the remote monitoring stations rises above or falls below defined threshold water levels.
These modules are used to visualise and communicate results.
The GIS module is used to visualise the relative location of the watercourses, roads, emergency aid locations, escape routes, etc. depending on the availability of digital maps containing these relevant information.
The communication module facilitates communication of flood alarms, flood statements and other information using three means of communication. Electronic mail is there to send flood statements and other necessary information from the command center to public authorities. File transfer protocol (FTP) is used to upload and download data and other information between the command center and other FTP Servers. Hypertext transfer protocol (HTTP) is used for automatic generation of day-to-day flood warning internet pages with a format that can be dumped to a web server via FTP and can be accessed by the general public and other interested groups.
The Database module
is internally used by the warning system for dumping day-to-day decisions of the
DSS into a standard Microsoft Access database. Its main (external) use, however,
is to allow the browsing of information about earlier warnings issued by the DSS.
It has also a provision to enter observed flood events as judged by experts into
the database so that it can be used for a later comparison. It makes use of
tables and standard queries to facilitate filtering records, and a calendar to
navigate through large data sets indexed by date.
Typical flood forecast starts with firing the array of meteorological models followed by a hydrologic/hydraulic simulation. Then the results are taken in by the DSS. The effective time available for flood warning depends on the lead time of the forecast which, for instance, is 72 hours for the case shown in Figure 4, and the time needed to execute the models and the DSS. Therefore, the effective forecast time depends on the kind of meteorological and hydrological/hydraulic models employed in the forecast.
The following operational procedure is proposed for a typical operation of the DSS:
– Meteorological models: GCM through LAM or other forecast techniques such as radar imagery.
– Graphical visualisation of precipitation forecasts for possible extreme events (intense precipitation forecast initiates further processes).
– Simulation using a hydrological/hydraulic model to forecast the water level hydrograph at various points.
– Comparison of all the water levels within the time series with the relevant water level thresholds to analyse the evolution of the flood through time within the forecast period.
– Determination of peak water levels within the forecast period and their respective arrival time.
– Graphical information display through the GUI of the DSS on top of a catchment map.
– Generation (composition) of flood statements using a built-in flood statement format and generating HTML code for possible access by the public via the Internet.
– Analysis of the results for the need of possible issue of warnings to the concerned authorities (police, fire, hospital), media, and/or general public.
– Dumping the forecasts into the standard database system for documentation.
– Issuance of warnings by available means of communication such as: (1) Telephone for immediate actions, (2) Faxing the generated flood statement, (3) Posting HTML page on an Internet server, (4) Sending warning statements to authorities via email, or (5) Posting files with flood statements to an FTP site.
– Watching the monitoring system to confirm the situation and to prepare for restoration activities.
The study shows that the problem of flood warning based on forecasts is one of great complexity and uncertainty. However it has to be done in order to save lives and property, even though there are consequences of possible false alarms that may cause loss of public trust in the system. It is believed that results can be improved with better forecasting techniques, improving catchment models, with experience using the DSS, and increased experience of people about the meaning and implications of flood warnings.
The use of the decision support system is definitely important to assist authorities in planning activities, archive day-to-day situations, monitor real-time events, automate routine activities, and eventually provide a better understanding and control over the situation. The system discussed in this report can be easily applied in other areas with similar flooding problems.
Acknowledgement
The Authors of this article would like to acknowledge the EU-DG13 for financing the project and the project partners for their contribution. These include: National Observatory of Athens, Universita di Genova Dipartimento di Fisica, Ingenieurburero Brandt-Gerdes-Sitzmann Wasserwirtschaft GmbH, G-Systems, Water Corporation of Athens, Regionne Liguria.
Fig. 1 Operational diagram of the DSS
Fig. 2 Screen dump of the warning system
Fig. 3 Screen dump of the warning system
Fig. 4 Operational sequence of models and the DSS