Computer-based flood forecasting in Australia by yiq68006


									Hydrological forecasting ~ Prévisions hydrologiques (Proceedings of the Oxford Symposium,
April 1980; Actes du Colloque d'Oxford, avril 1980): IAHS-AISH Publ. no. 129.

Computer-based flood forecasting in Australia

A. J. H A L L and J. F . E L L I O T T   Head Office, Bureau of Meteorology, Australia

Abstract. This paper outlines the planned automation and improvement of forecasting services
in the Australian Bureau of Meteorology with the installation of a network of minicomputer
systems, termed the Automated Regional Operations System (AROS), in the main Regional Fore-
casting Centres. Flood forecasting will be a special application of these computers and a detailed
description of the proposed computer-based system to replace the existing manual system is
presented. Advantages of the AROS Flood Forecasting System (AFFS) are discussed and it is
considered that the system will be highly cost effective.

Prévision des crues basée sur l'emploi de l'ordinateur en Australie
Résumé. Cette note décrit le plan d'automatisation et d'amélioration des services de prévision
dans le Bureau de la Météorologie avec mise en place d'un réseau de mini-calculateurs dans les
principaux centres régionaux de prévision. Ce système sera designé sous le nom de Système
Régional et Opérationnel Automatisé (SROA). La prévision d'inondation sera un des services
effectués sur les mini-calculateurs. Une description du système de prévision qui remplacera le
système actuel, non-automatisé, est présentée. Les avantages du système de SROA de prévision
d'inondation sont discutés et il sera démontré que la système serait très économique.

Flood forecasting services in Australia are provided by the Bureau of Meteorology, the
national meteorological service which is a division of the Australian Department of
Science and the Environment. Flood forecasting is carried out by the Bureau in the
more flood-prone eastern states of Queensland, New South Wales, Victoria and
Tasmania. Forecasts are made at the state level by the Bureau's Regional Forecasting
Centres located in the capital city of each state. Each of the more populated and
developed states currently has up to 32 river basins for which qualitative flood warn-
ings are provided. Quantitative flood forecasting systems are available for 25 river
basins in one state. Where required, quantitative estimates of flood height are made
using mathematical models for up to 10 forecast points within a basin.
    As part of a planned automation and improvement of forecasting services in the
Bureau, a network of minicomputer systems, termed the Automated Regional Opera-
tions System (AROS), is proposed for installation in each of the Regional Forecasting
Centres. The equipment of the prototype system is to be installed in the Victorian
Regional Forecasting Centre, Melbourne, in 1980. The main function of these
computers is to handle the large amount of data required for the Bureau's weather fore-
casting services. The computers will present these data and the numerical weather
prediction output from a larger central computer to the forecaster, and enable the
texts of forecasts to be prepared directly on the computer for automatic dissemination.
Flood forecasting will be a special application of the computers, enabling a range of
mathematical modelling procedures, more complex than the existing manual or hand/
programmable calculator techniques, to be used. A number of the component sub-
systems of AROS will be common to both weather and flood forecasting and will
encompass both; e.g. data collection, verification, handling within the computer
(storage, retrieval, display, etc.) and forecast message composition and dissemination.
58    A. J. Hall and J. F. Elliott
Four phases can be identified in the current flood warning organization: routine
assessment, flood precautionary, flood alert and flood forecast phases.
    (1) Routine assessment phase: this covers the receipt and logging of daily rainfall
and river height reports, computation of average basin rainfall and basin moisture
indices to allow initial loss estimates to be made for all basins, and the receipt and
logging of any special heavy rainfall or flood warning river height reports. This infor-
mation is summarized for the duty forecasters, and used in conjunction with 24 and
48-h quantitative precipitation forecasts (QPF) to determine the necessity to enter a
flood precautionary phase. Daily bulletins of river height readings and rainfall are also
    (2) Flood precautionary phase: the issue of confidential advice t o local authorities;
 this may require a brief modelling study to assess the sensitivity of the basin to possible
 rain amounts (QPF).
    (3) Flood alert phase: the issue of initial flood alert warnings to the public in a
qualitative form; requests for additional rainfall readings at 1,3, 6-h intervals as
 appropriate for each basin; tentative or provisional flood modelling to assess the
possible degree of flooding (minor, moderate or major); preparation of concise sum-
maries and presentation of rainfalls and river heights as received; plotting of observed
and forecast hydrographs and flood stage profiles to decide whether the flood fore-
casting phase is to be entered.
    (4) Flood forecasting phase: this involves full scale modelling with a combination of
observed and forecast rainfalls to determine forecasts of future river height, issuing of
quantitative flood warnings to the public, and logging of incoming data (1, 3 and 6 h)
and presentation as tables and maps for forecaster's assessment. Following cancellation
of the forecasting phase, model 'carryover' values are stored and the data collected,
received and generated during the flood is archived.

The AROS minicomputer system is designed to be fully reliable and able to maintain
operations 24-h a day. In addition, a batch-processing system for development work
and other non-routine processing will be available. Duplication of many of the hard-
ware components and the ability to switch peripheral devices and communication lines
between processors is proposed to ensure a high degree of operational reliability, with
ease and rapidity of recovery from fault situations being of prime importance.
    The system will interface to the Bureau's existing modes of meteorological and
hydrological observation networks, mostly in a variety of coded formats as low speed
communication lines via Telecom, Australia's public telegram network (TRESS), Telex,
the Department of Transport's Aeronautical Fixed Telecommunications Network
(AFTN) and to the Bureau's private wire network. A separate facsimile network is also
maintained to carry weather satellite pictures and graphic products to the Regional
Office from the National Meteorological Analysis Centre (NMAC) at the Head Office.
    AROS is to interface with the new Telephone Observation Collections System
(TOCS) currently (August 1979) under tender. Briefly, TOCS will make use of the
switched telephone network to receive coded observations from the Bureau's observing
points and will be controlled by a microprocessor-based central station which will be
linked to the minicomputer system by a 9600 bits per second synchronous or asynch-
ronous data link. TOCS is primarily designed to reduce the expensive telegram and
voice telephone communication costs and to reduce transcription errors.
                                    Computer-based flood forecasting in Australia         59
   Under the supervision of an operator at a visual display unit (VDU), the mini-
computer will maintain control of all communication lines and interfaces. It will
receive, present for correction as necessary, decode, and store on disc all observations.
Commands entered via VDUs permit access to such facilities as:
   (1) display and modification of observations on the VDU screen;
   (2) display of observations in various tabular and graphical forms on a printer/
   (3) automatic preparation of charts containing plotted observations t o be analysed;
   (4) automatic computation of objective forecasting aids;
   (5) text editing to enable preparation of forecasts and warnings on a VDU screen,
   (6) automatic distribution of forecasts and warnings to preset, but modifiable,
Special software routines will alert forecasters of the arrival of observations of a critical
nature. Various other situations requiring attention, such as failure of an item of hard-
ware, will generate appropriate alarms. Other proposed facilities include:
   (1) ability to enter manually produced charts into the computer for further proces-
sing or for automatic distribution via an interface to a facsimile network;
   (2) ability to display on a video screen a large number of graphic products such as
contour charts, satellite cloud pictures, and observational plots;
   (3) ability to output graphic products via a facsimile digital-to-analogue interface,
   (4) possibly, the ability to digitize and accept data from a local radar display and
analogue facsimile channels.
   The possible system configuration is shown in Fig. 1. Two minicomputers of a
minimum of 192K 8-bit bytes addressable memory expandable to at least 256K 8-bit
bytes will be used. At least three independent and identical disc drives will be used,
simultaneously accessible from both computers, with a total storage capacity equiva-
lent to at least 80 million 8-bit bytes. Two identical printer/plotters using non-impact
printing techniques are required for producing hard copy of both alphanumeric and
graphic data. These will be switchable between computers and have a paper width of
at least 280 mm (11 in.) with a resolution of at least four dots per millimetre (100 dots
per inch).
   The graphics display sub-system may be switched between computers, and be
capable of displaying at least four individually selectable overlaid images such as
geographic outline, cloud satellite picture, contour chart, and selected observational
plots. A graphic tablet digitizer with an active digitizing area of at least 510 x 760 mm
(20 x 30 in.), and a resolution of 0.25 mm (0.01 in.) is also under consideration.

The AROS Flood Forecasting Sub-system (AFFS) is designed around the existing
manual system and will automate as many of the functions as possible, commensurate
with increasing the efficiency and effectiveness of the flood forecasting service. The
AFFS is to tie in with the other sub-systems of AROS and some common components
will be used, e.g. data collection, verification and storage, message formulation and
dissemination. AROS components which present observed data; e.g. maps, tables,
graphs, etc, will be used by both AFFS and meteorological forecasting.
   Figure 2 shows the AFFS components. These components are not necessarily
separately identifiable, but may be integrated with other components of AROS. The
function of each component is:
60   A. J. Hall and J. F. Elliott

                                             2 PRINTER/
                                                              / " T v C U s WITH \   |             LE
                                                          \   V -BOAR„           Y, "DNS"

                                                               CARD READER



                                      —     DIGITIZING

                                                                 FACSIMILE                 FACSIMILE

                                                                INTERFACE       "*   "*    NETWORK

                                                                               RADAR INPUT

                                                                   TELEX                    TFI FX
                                                                INTERFACE                   NETWORK
                                        I   POSSIBLE
                                        1 PAPER TAPE
                                        i   PUNCHES (2)

                                                              TELEPHONE OBSERVATION
                                                              COLLECTION SYSTEM

                                                              PUBLIC TELEGRAM NETWORK(TRESSj

                                                              AERONAUTICAL FIXED
                                                              TELECOMMUNICATIONS NETWORK

                                                          - BUREAU'S PRIVATE WIRE NETWORK

FIGURE 1.   Configuration of the Automated Regional Operations System.

FIGURE 2.   Components of the AROS Flood Forecasting Sub-system.
                                  Computer-based flood forecasting in Australia       61
    (\)Data input - RAWFILE: that portion of the total information received by
AROS which is relevant to AFFS; e.g. rainfall, river height, outflow rates and storage
levels from dams, QPFs — usually at 3 or 6-h intervals during floods, and daily maxi-
mum and minimum temperature for evaporation estimates (for details of data volumes
see Hall and Elliott, 1979).
    (2) Data edit — DATEDIT: quality monitoring of the incoming data, adjusting
where necessary, e.g. missing rainfall data, interpolation to produce regular time period
data, conversion of stage heights to discharge where required, etc.
    (3) Working data file - BASFILE: contains all the real time working data required
for operational use. The type of data collected will depend upon which of the various
phases is in operation; e.g. daily data for monitoring, 3 to 6-h data for forecasting for
durations up to 5—10 days for small to medium basins, and daily data for durations up
to two to three months for some of the long, slow flowing inland rivers. The contents
of BASFILE will be periodically archived onto permanent storage.
    (4) Data presentation - DATDISP and MAPPER: prepares all the tabulations,
listings, plots and basin representation of data to aid the forecaster.
    (5) Arealaveraging - A VERAGE: calculates areal averages of BASFILE data where
required; e.g. 1, 3 or 6-h rainfalls.
    (6) Routine catchment assessment DATSUM: prepares a summary of information
required to assess the current status of the basins (moisture indices, initial loss poten-
tial, etc.) and the likely consequences of any expected rainfall.
    (7) River forecast - RBFCST: controls the detailed hydrological modelling required
for preparing a forecast for a basin. Because of the wide range of basins and different
climates to be handled by this system, the hydrological modelling component will
need to be flexible. Existing manual models include lumped unitgraph—empirical loss
models and graphical—nomogram techniques. Where required these will be replaced
with a continuous soil moisture accounting model, most likely the Sacramento model
(Burnash et al., 1973). Where possible a number of generalized models of each of the
separate processes of basin modelling will be arranged to suit each particular situation.
It is ultimately intended that RBFCST be used to interpret a control vector defining
the combination and arrangement of modules for each basin and to operate the model
so defined.
    (8) Modelling data file - MODFILE: contains all the data used and generated by
the operational forecast program RBFCST, e.g. forecast hydrographs, hyetographs to
each sub-basin including forecast rainfall, and model parameters which vary between
    (9) Work data file — WORKFILE: used as temporary disc storage area as required
for RBFCST operations.
   (10) Permanent catchment data file — HYDCHR : contains all permanent characteris-
tics of each basin used in the transformation of the basic data and operation of the
models; e.g. rating tables, storage—discharge tables for flood routing, critical or key
heights, sub-basin areas, river lengths, areal rainfall station weights, model parameters,
profiles of past flood events for comparison, basin maps, boundaries, station locations,

   (1) A much faster and more accurate examination of river systems is possible. As
more rivers become potential flood hazards the general overview required by the fore-
caster becomes the most difficult task. The computer can be used to sort out the most
pertinent information to be presented to the forecaster, thereby reducing the present
level of mental strain.
62    A. j . Hall and J. F. Elliott
    (2) Data handling is minimized since all observational data received are immediately
accessible and are presented in the most pertinent format to the forecaster. Similarly,
all basin parameters can be recalled and presented as required.
    (3) All routine calculations (e.g. daily basin assessment) are carried o u t automatically.
    (4) The computer allows the use of more complex models in the forecasting systems
which lead to the following advantages:
       (a) Continuous soil moisture accounting models (e.g. Sacramento model) provide
    a more comprehensive monitoring of the basin moisture status and provide a com-
    plete forecast of all components of the flood hydrograph.
       (b) A more objective forecast can be made as the forecaster can quickly assess
    the consequences of a wider range of forecast rainfalls.
       (c) Modelling is not necessarily restricted to a constant time interval.
       (d) The availability of a wide range of modelling procedures will allow the most
    appropriate routine to be selected for each application.
       (e) The use of theoretically sounder models should lead to greater confidence
    when predicting extreme events, and allow the effect of physical changes within
    the basin to be anticipated.
    (5) The availability of a computer will hasten the development of new forecasting
systems and allow for more comprehensive post-flood performance analysis.
    (6) In addition to flood forecasting, AFFS could provide a more comprehensive
range of streamflow forecasts to assist various water management authorities (e.g.
drought forecasts, reservoir management decisions), and as an advisory service to
industry on such matters as timing of pollutant releases into natural water courses, etc.

The introduction of computer-based flood forecasting systems will improve the quality
and efficiency of flood forecasting services in Australia. These operations are planned
on a state basis and some of the complexities of developing a computer-based flood
forecasting service for a large region have been presented. These involve data collec-
tion on a large scale, real-time hydrologie modelling with an emphasis on flexibility,
and forecast dissemination to the public. The forecasting system must also be geared
to seek and receive public response to its forecast, including updating and correcting,
and act accordingly. The advantages of AFFS in handling these problems have been
   The brief description of AROS and AFFS shows that a number of the components
of AROS are common to both the flood forecasting function and the meteorological
forecasts produced by a national weather service, and therefore the system is expected
to be highly cost-effective.

Acknowledgements.      This paper is published with the permission of the Director of Meteorology,

Burnash, R. J. C, Ferrai, R. L. and McGuire, R. A. (1973) A generalized streamflow simulation
       system: conceptual modelling for digital computers. US Department of Commerce, National
       Weather Service, and State of California, Department of Water Resources, Sacramento,
Hall, A. J. and Elliott, J. F. (1979) Development of a computer based flood forecasting system in
       Australia. International Symposium on Logistics and Benefits of Using Mathematical Models
       of Hydrologie and Water Resources Systems, Pisa, Italy, October 1978: Pergamon Press (to
       be published).

To top