Title of the paper - JESUS G, Custom deployment of a Nowcast-forecast information system in coastal regions.pdf

Custom deployment of a Nowcast

forecast information system in

coastal regions

Gonçalo Jesus

João Gomes

Nuno Ribeiro

and Anabela Oliveira

Information Technology Unit

Hydraulics and Environment Division

Laboratorio Nacional de Engenharia Civil

Avenid

a do Brasil 101

1700

066 Lisboa

Tel

.: + 351 21 844 3885; Fax: + 35

21 844 3016

gjesus@lnec.pt; jlgomes@lnec.pt

;

nribeiro@lnec.pt

;

aoliveira@lnec.pt

Abstract

This paper presents a nowcast

forecast information system, tailored for coastal applications.

The information system is based on the custom deployment of a generic forecasting platform,

modified for short to long wav

e predictions, and implemented in a modular way to

accommodate future developments. It integrates a suite of high resolution numerical models,

prepared for application in high performance environments. The architectural and

technological details are presen

ted here, along with examples from three distinct applications

for the forecasting of short and long waves. Emphasis is given on the several phases of the

daily execution process of the system, on how the forecast models are integrated, as well as on

the

ecent improvement in the models outputs visualization. Future developments, towards an

early warning system for coupled circulation, and a coastal oil spill prevention and prediction

system, coupling remote sensing and numerical models, will be discussed,

along with several

technological issues.

Keywords:

real

time forecasting, spatial data, nowcast

forecast, coastal systems

visualization

tools

Introduction

1.1

Overview

The protection of important environmental and social assets in river and coastal regions

quires the early warning of potential hazards, such as floods, pollution events and tsunamis.

Timely hazard forecasting is an essential part of risk management for vulnerable communities,

providing the necessary information for population evacuation, pollu

tion protection resources

allocation and efficient emergency personnel management. Hydraulic forecasting information

systems have been under development for over three decades, addressing many problems and

spanning several areas, such as river floods, wind

and wave forecast and tidal prediction.

Forecast systems combine our ability to measure and to simulate the behavior of water bodies,

by integrating numerical models, monitoring networks and information technology systems, to

provide real

time, short

term

, predictions of the main physical and chemical parameters.

With the recent emergence of new, reliable and cost

effective automatic data acquisition and

transmission systems, and highly efficient numerical models (Baptista, 2006), most important

constraint

s for widespread operational use of hydraulic models in real

time forecasting have

been minimized. In particular, coastal nowcast

forecast systems have evolved from research

tools to operational tools for the management of harbors, marine resources and eme

rgency

operations, providing accurate and timely information on waves and currents conditions. A

decade ago predictions were restri

cted to physical quantities, but now are

being extended to

biochemical phenomena (Baptista 2006, González 2008).

Motivation

he ability to simulate and forecast the dynamics of estuarine and coastal zones is fundamental

to assess the social, ecological and economical impacts of both human interventions and

climate changes in these regions.

Due to the complexity of the processes

and the uncertainty

associated with the forcings in the water dynamics of coastal regions, developing those

computational nowcast

forecast systems requires a set of high performance and high spatial

resolution mathematical models and large computational re

sources.

The vast amount of daily

model results generated by forecasting models poses an additional challenge, as human

resources to validate and analyze these data are unavailable. One of the easiest and friendliest

ways of reaching a broad spectrum of st

akeholders, from the coastal managers to the avid

recreational user, is to produce images and animations integrating the products of the running

forecast models, taking advantage of the several visualization tool boxes, many of which

support GIS (Geographi

c Information System)

capacities.

CMOP (Center for Coastal Margin Observation & Prediction, U.S.A.) developed a real

time

generic hydrodynamic forecast system

–

RDFS, Rapid Deployment Forecast System

–

which

can be customized for any geographical location

(Baptista 2006). This system was firstly

adapted by CMOP and

LNEC

for the tidal hydrodynamics of Portuguese coast (Oliveira 2010),

through integration of

existing state

the

art numerical models applications (Rodrigues

2009).

This paper focus on the new

developments of the RDFS applied to the Portuguese coasts,

namely its extension to short wave propagation and its whole execution process, with

integration of important inputs as several sources of forcings, based on global forecasting

models and extrapola

tion of near

real time measurements. In the other end of the process,

important developments were made in the resultant visualization products.

Development of the information system

3.1

Overview

The current physical architecture of the RDFS is defined by sever

al rack mounted computer

servers with a Linux operating system logically clustered around a central file server. The

central file server provides archival storage for model outputs, access to model data, and tools

for managing the forecasts. In each comput

er server one or more forecasts, depending on the

server capacity, are running on a daily basis. Figure 1 shows how physical components interact

with each others, and where the files, web and database servers are positioned, and finally,

how the outside us

ers can connect to the RDFS through web browsers.

The whole RDFS functioning process runs on daily basis, in which the execution scripts start

the different phases of the process, like forecast execution, products generation and next run

preparation. These

scripts interact with the database in order to fetch the location of the

forecast systems and later to remotely transfer the products files to the central

node. The

interface is described in the section below,

and

Figure 1 shows that

this

is a web

based

pplication using Apache web

server, that also interacts with the database.

3.2

Technologies

In order to integrate all the components of the system, several technologies were used,

including Perl programming language. The RDFS backbone is constituted by a set o

f perl

scripts that are replicated to all the deployed forecast models. Those scripts prepare and launch

the forecasts, interacting with a PostgreSQL database server to get the specific forecast

information, as well as important data regarding points of ag

gregation for river inputs

consisting of real

time flows, temperatures and river climatologies. Because this is a daily

process, the main scripts are being executed from crontab

daemon

The preparation part of the process consists in setting up the environ

ment to the current status,

retrieving the referred data from the database server and changing the initial parameters of the

forecast, also saving the previous environments for future executions. Launching the forecast

involves the correct execution of the

forecast models, which may run in a clustered system,

and not actually be in the current computer server. After this process, another set of Perl scripts

used to process results, perform model and data comparisons and drive products generation,

using v

isualization tools such as VisTrails or the matplotlib library.

The end

user will visualize all the products and interact with the given results, using the RDFS

user interface (UI). This UI is basically a customized deploy of Drupal, a PHP

based Content

nagement System (CMS), that is used to access model metadata, status and products.

Figure 1. Physical architecture of the RDFS

3.3

Models application and data integration

The RDFS platform uses three open

source models, Wave Watch III (WW3) for wave

forecast

, and ELCIRC and SELFE for hydrodynamic runs. The inclusion of a wave model for

the near

coast wave simulations, SWAN, is underway. This set of models is capable of

coupling waves, tides, storm surges, river fluxes and winds, providing a forecast of sea le

vel

variations, currents, temperatures, salinit

y and waves for a target area.

WW3 (Tolman 2009) is a third generation wave model developed at NOAA/NCEP. ELCIRC

(Zhang 2004) is an unstructured

grid model that solves the free surface elevation, 3D water

vel

ocity, salinity and temperature, which represent mass conservation, momentum

conservation, and conservation of salt and heat. SELFE (Zhang 2008) is an unstructured

grid

model designed for the simulation of 3D baroclinic circulation across river

ocean sc

ales.

The original RDFS, developed at CMOP, implements hydrodynamic forecasting using models

SELFE or ELCIRC, based on wind, tidal and river flow forcings. The present implementation

extended this platform to short wave prediction, by integrating the wave

model WW3 and

forcings by global wind predictions of NOAA.

Two different setups can now be implemented in the RDFS for a defined region. One setup

uses a waves model (currently WW3, the integration of a coupled wave

circulation is

underway) while another s

etup uses a hydrodynamic model (SELFE or ELCIRC).

On both hydrodynamic models, ELCIRC and SELFE, two types of boundaries are considered:

oceanic, forced by the regional tide model of

(Fortunato

2002), by the temperature collected by

the buoys from Institu

to Hidrográfico da Marinha Portuguesa (IH), riverine forced by the river

flux and temperature from Sistema Nacional de Informação de Recursos Hídricos (SNIRH).

When data is not available (due to lack of measurements or forcing forecasting fail

ure),

climato

logy data is used.

In the wave propagation setup, to provide boundary condition for local applications of model

SWAN for any region of the Portuguese coast, two nested run

of WW3 are used, based on the

application of

(Dodet

2009). The first one covers the

entire North Atlantic Ocean and the

second one covers the Portuguese coast. Both runs are forced with wind data from the Global

Forecast System (GFS) from the NOAA.

3.4

Products

Custom

tailored products are developed automatically for each type of model fore

casting.

Depending on the complexity of the hydrodynamic simulations, just water levels and

velocities, or also salinity and temperature fields are produced for target area. For wave

simulations, significant wave heights and directions are provided. For al

l products, images and

animations for pre

defined zooming areas are automatically generated, providing detailed

representations of the simulated phenomena. The usage of selective coarsening of results

representation is underway to prevent difficulties in a

nalysing results due to an overly full

mapping of quantities.

Introducing

GIS visualization tool

4.1

RDFS UI improvements

Since the beginning of the RDFS backend development we

have

added a map server support

and

OpenGIS Web Mapping Service for

model resu

lts

visualization

. The main goals of the

integration of these

frameworks

are

to provide product

visualization

, including animations, and

model output query capabilities

to the system

users

. The UI also connect

to PostgreSQL

database server, and to PostGIS

module to retrieve the spatial components of the information.

4.2

GIS backend

A simplistic benchmark evaluating costs, usability and performance of the most known market

and community options determined that the open source Geoserver system fulfilled the

requ

irements

of the map server

. However, the products results

which are produced

daily

are

not

georeferenced by default, and

neither are their

format and appearance ready for deploy in a

map server.

Existing products

were

adapted

for the new UI also

using

ily

“run

once” scripts

written in

Perl, which

make use of

tools such as imagemagik among others. For the prototype

version herein presented, geotiff format was chosen in order to

use

all the previous generated

images.

The deployment in Geoserver of all the

products, and the ones that are produced on a daily

basis, is possible using PhantomJS and CasperJS (including jQuery library). This process,

repeated

automatically

as many times as products, simulates a user

interaction with a browser

to successfully dep

loy a geotiff source in Geoserver.

xisting

geotiffs are then served by

eos

erver as

WMS (Web Map Services)

one of

the format

supported by the RDFS UI

WebGIS tool

Future developments are expected on the backend side, regarding Geoserver

configurations f

or increasing performance, and especially changes in the spatial information

sources paradigm.

Figure 2. Screens of the new WebGIS prototype being developed, showing different visualization features

and products.

4.3

Building a WebGIS

With the goal

s of attractiveness, simplicity and detail, a visualization tool prototype was

designed and developed with Flex/Flash technology, over OpenScales framework. Although

this framework is on development for some years

now

, there is a lack of documentation and

examples that

make this WebGIS framework not

easily customizable.

The project is to build

modular, extensible

generic

and

customizable

WebGIS

, not only to support RDFS data

but

also

spatial data from other LNEC’ projects or partners.

he prototype congr

egates data from several WMS, and shows them to the users in an

intuitive

way, even if they

aren’t

familiarized with GIS software.

The

interface allows the user to better

manipulate existing products, for

instance,

using a slider

the users can

navigate

thr

ough

determined intervals of a model result simulation, as shown in Figure

the near future,

for

example,

it will be possible to query data on

the

fly, elaborate graphs and charts, and consult

real

time data gathered from field sensors.

Conclusions an

d future research

new

real

time forecast system

RDFS

was developed and applied

to all Portuguese coast

using

existing state

the

art numerical models related to the short and long wave prediction

of coastal zones, allowing the users to analyze and vi

sualize large amounts of scientific data

through animations and images representing the products of the forecast systems.

This

deployment runs two types of forecasts and is applied to three coastal areas.

Our vision is

to develop a multi

purpose, broader

nowcast

forecast

information

system

. The

present

RDFS is a first step towards a broader nowcast

forecast system that will incl

ude water

quality issues and

integration with satellite data products

, and an

early warning system for

coupled circulation due sto

rms surges, waves, tides and river floods, targeting the extended

application of the European Flood Framework Directi

ve to river to coastal domains.

Acknowledgements

This study was funded by FCT’s

PhD Grant (

SFRH/BD/82489/2011

)

and FCT’s

projects G

Cast (G

RID/GRI/81733/2006), Pac:man (PTDC/AAC

AMB/113469/2009), SPRES (Interreg

Atlantic Area Transnational Cooperation Programme 2007

2013 project SPRES

2011

1/168)

and FLAD (project CWOS). The authors thank Antonio Baptista, Paul Turner and Joseph

Zhang, fro

m CMOP, for the RDFS platform and the support for its adaptation, and the models

ELCIRC and SELFE. Thanks are also due to the WW3 development team for making the

model available and NOAA for the wind forecasts. Finally, real time forecasting cannot be

achi

eved without proper model validation. The authors thank Marta Rodrigues, André

Fortunato and Guillaume Dodet, from LNEC’s Estuarine and Coastal Division, and Xavier

Bertin, (CNRS), for the model’s applications.

References

Baptista, A.M.

(

2006

)

. CORIE: the

first decade of a coastal

margin collaborative observatory. Oceans'06,

Boston.

Dodet, G., X. Bertin and R. Taborda (2010).Wave climate variability in the North

East Atlantic Ocean

over the last six decades, Ocean Modelling, 31/3

4, 120

131.

Fortunato, A.B.

, L. Pinto, A. Oliveira and J.S. Ferreira (2002).

Tidally generated Shelf Waves off the

Western Iberian Coast, Cont. Shelf Research, 22/14, 1935

1950.

González, M., Ferrer, L., Uriarte, A., Urtizberea, A., Caballero, A.,

(

2008

)

Operational Oceanography

stem applied to the Prestige oil

spillage event, Journal of Marine Systems, 72 (1

4): 178

188.

Oliveira , A., A.B. Fortunato, A.M. Baptista , P.Turner , G. Jesus , M. Rodrigues, N. A. Ribeiro, A.

Mendes , J.P. Fernandes , A. Azevedo , N. Bruneau, G. Dodet

(2010) Sistema de previsão em tempo real

para a circulação estuarina e costeira, 1as Jornadas de Engenharia Hidrográfica, Instituto hidrográfico, 4

pp.

Rodrigues, M., A. Oliveira, H. Queiroga, A.B. Fortunato and Y.J. Zhang (2009) Three

Dimensional

Modeli

ng of the Lower Trophic Levels in the Ria de Aveiro (Portugal), Ecol.

Modell., 220, 1274

1290.

Tolman, H. L. (2009) User manual and system documentation of WAVEWATCH III version 3.14.

NOAA/NWS/NCEP/MMAB Technical Note 276, 194 p.

Zhang, Y.

L., Baptista, A.

M. and Myers, E.P. (2004) "A cross

scale model for 3D baroclinic circulation in

estuary

plume

shelf systems: I. Formulation and skill assessment". Cont. Shelf Res. 24, 2187

2214.

Zhang, Y., Baptista, A.M. (2008) SELFE: A semi

implicit Eulerian

–

Lagrangian f

inite

element model,

Ocean Modelling,

21/3

, 2008, 71