Professional Documents
Culture Documents
143
ABSTRACT
The southeastern coast of Brazil is frequently affected by meteorological disturbances such as cold fronts,
which are sometimes associated with intense extratropical cyclones. These disturbances cause oscillations on
the sea surface, generating low-frequency motions. The relationship of these meteorologically driven forces
in low frequency to the storm-surge event is investigated in this work. A method to predict coastal sea level
variations related to meteorological events that use a neural network model (NNM) is presented here.
Pressure and wind values from NCEP–NCAR reanalysis data and tide gauge time series from the Cananéia
reference station in São Paulo State, Brazil, were used to analyze the relationship between these variables
and to use them as input to the model. Meteorological influences in the sea level fluctuations can be verified
by filtering the astronomical tide frequencies for periods lower than tidal cycles (periods higher than 24 h).
Thus, a low-pass filter was applied in the tide gauge and meteorological time series for periods lower than
tides to identify more readily the interactions between coastal sea level response and atmospheric-driven
forces. Statistical analyses on time and frequency domain were used. Maxima correlations and coherence
between the low-frequency sea level and meteorological series could be defined using the time lag of the
NNM input variables. The model was tested for 6-, 12-, 18-, and 24-hourly forecasts, and the results were
compared with filtered sea level values. The results show that this model is able to capture the effects of
atmospheric and oceanic interactions. It can be considered to be an efficient model for predicting the
nontidal residuals and can effectively complement the standard constant harmonic analysis model. A case
study of a storm that impacted coastal areas of southeastern Brazil in March 1998 was analyzed and
indicates that the neural network model can be effectively utilized in the Cananéia region.
1. Introduction tions is very important not only for marine services but
also for the protection of coastal residents, for moni-
Oscillations in sea level due to meteorological driving
toring the changes in marine ecosystems, and for de-
forces related to wind and pressure occur at different
signing and constructing onshore and offshore struc-
scales and frequencies in all coastal regions. Some
tures. Interactions between meteorological (atmo-
countries are affected by flooding, with serious damage.
spheric pressure, wind, sea surface temperature) and
Therefore, knowledge about the sea level height varia-
oceanic (salinity and deep sea) variables affect the
regular tides and modify the sea level conditions in
coastal regions, mainly in restricted waters such as bays.
Corresponding author address: Marilia M. F. de Oliveira, Insti- Tropical cyclones and extratropical storms are the
tuto Alberto Luiz Coimbra de Pós-Graduacxão e Pesquisa de En-
genharia—COPPE, UFRJ, Caixa Postal 68501, CEP 21945-970,
main cause of storm surges that can produce damage
Rio de Janeiro, RJ, Brazil. through high waves and sprawling water over large
E-mail: marilia@coc.ufrj.br coastal areas in a single storm. The principal factors
DOI: 10.1175/2008JAMC1907.1
involved in the generation and modification of storm the astronomical arguments using local constituents de-
surge are the action of wind stresses on the surface termined by previous analysis and do not take into ac-
water, reduction of atmospheric pressure (inverted ba- count meteorological influences. Thus, the observed
rometer effect), waves and swells in the shallow water and predicted values of the sea level are normally dif-
area, coastline configuration, and bathymetry (Pore ferent. Numerical models developed to predict surges
1964). are still considered to be insufficient because of the
Frontal systems associated with extratropical cy- complexity of the nonlinear processes involved. These
clones introduce variations that are observed in the tide models require a large amount of tidal and meteoro-
records and sometimes result in storm surges (Pugh logical data, collecting many factors such as central
1987, 2005). These storms differ with respect to size and pressure, speed of the cyclone, rainfall, and coastal to-
intensity, and the associated storm surges have charac- pography (Lee 2006).
teristics consistent with these differences. The different Nowadays, the neural network model (NNM) has
characteristics of these types of storms in particular been widely applied to modeling nonlinear dynamic
concern wind speeds and spatial scale of the storm systems using time series that translate the physical re-
(Storch and Worth 2008). lations between the input variables (predictors) and the
In the South Atlantic Ocean, along the Brazilian phenomenon that will be modeled (predictand). Elsner
coastline, there are few tide gauge records with long and Tsonis (1992) developed some methods for making
series to analyze and predict surge events. Characteris- short-term predictions of nonlinear time series data us-
tics of the meteorological tide variations along the ing a neural network model. These authors discuss the
southeast coast of Brazil have been studied by Marone implications of these methods for the study of weather
and Camargo (1994). Castro and Lee (1995) presented and climate.
a study of sea level fluctuations due to wind-driven The NNM has some important characteristics such as
forces in the southeast continental shelf. Ribeiro (1997) generalization, parallelism, nonlinearity, adaptability,
investigated a surge caused by the passage of a cyclone and robustness (Haykin 2001). These models have been
along the Rio de Janeiro coastline that raised the sea used in some fields of science and engineering. Szto-
level 0.60 m above the mean sea level datum, causing bryn (2003) applied NNM in hydrological forecasting
damage to coastal communities along the Guanabara where the variation of water level is only wind gener-
Bay. Netto and Lana (1997) studied the superficial sedi- ated. The results were successfully compared with ob-
ment characteristics of tidal flats in Paranaguá Bay. De served sea level and other routine methods. Lee (2006)
Mesquita (2008a,b) verified a similar behavior of the applied an NNM for forecasting storm surge in Taiwan
mean sea level oscillations along this area of the Bra- related to the passage of three typhoons over the re-
zilian coastline. Mantovanelli et al. (2004) verified the gion. The results indicate that NNM is efficiently ca-
tidal velocity and duration as a determinant of water pable of learning and predicting from these events.
transport and residual flow in Paranaguá Bay estuary. Tseng et al. (2007) used a typhoon-surge forecasting
Dalazoana et al. (2005) studied the mean sea level model developed with a backpropagation neural net-
variations using longer tide gauge temporal series from work in coastal northeastern Taiwan. To determine the
Cananéia and Fiscal Island (state of Rio de Janeiro) better forecasting model, four models were applied and
tide gauge and satellite altimetry to establish analysis tested under different compositions of the input vari-
methods applicable to Brazilian vertical datum regions. ables. For coastal and harbor engineering applications,
The classical method of harmonic analysis is used to Chang and Lin (2006) simulated tides at multipoints
predict the astronomical tides. Tidal curves appear as considering tide-generating forces. The NNM proposed
periodic oscillations and can be described in terms of is applicable for multipoint tidal prediction in which the
amplitude, period, or frequency. The harmonic analysis tidal type is similar to that of the original point.
is based in tidal variations represented by N harmonic The southeast coast of Brazil is sufficiently affected
constituents of the tide (Doodson and Warburg 1944). by cold fronts over 3–5-day periods. An important
Normally, 365 days of hourly data at a point are needed event that sometimes occurs because of the combina-
to extract the constituents with adequate separation of tion of tides and surges is the rising of the sea level with
closely spaced constituents using the least squares waves that reached the coastline. There are few NNM
method. These constituents can then be used to provide applications to predict the variability of sea level along
reliable predictions for future tides at the respective the Brazilian coastline that are focused on the surge
point (Franco 1981). events. The relationships describing the response of the
Predictions for reference stations are prepared from coastal sea level to the influence of cold fronts were
JANUARY 2009 DE OLIVEIRA ET AL. 145
FIG. 2. Map of the study area with the localization of Cananéia Estuary and the tide gauge station.
(1991) have verified two regions of persistent cyclogen- nomenon great climatic disturbances occur in this re-
esis over South America; one over the San Matias Gulf gion, leading to abundant rain.
in Argentina (42.58S, 62.58W) and another over Uru- ENSO strongly influences rainfall patterns mainly in
guay (31.58S, 558W). The climate is subtropical humid, southern Brazil that tend to be above the median from
and during the El Niño–South Oscillation (ENSO) phe-
FIG. 3. Tidal cycles at Cananéia station in March 1998 by the FIG. 4. Observed and predicted tide at Cananéia station in
M2, S2, O1, and K1 constituents, including shallow-water terms. March 1998.
JANUARY 2009 DE OLIVEIRA ET AL. 147
FIG. 5. Map showing the shape of the southeastern coastal region of Brazil and the local-
ization of the tide gauge station and reanalysis points (chart modified from one provided by
CHM).
November to February (Rao and Rada 1990). The El do Brasil (CHM) during this period were also used.
Niño event in 1997–98 was considered, by some mea- Figure 5 shows the location of the Cananéia Estuary
sures, to be the strongest on record, causing major cli- tide gauge and the reanalysis grid points.
matic impacts around the world (McPhaden 1999).
During the 1997–98 El Niño event, the sea level in the
4. Statistical analysis
Cananéia region was significantly affected by a storm
surge, which is analyzed in this work. The atmospheric pressure, wind, and tide gauge time
series were analyzed statistically by estimating the cen-
ter of the distribution (mean and median), variances,
3. Meteorological and sea level dataset
standard deviation, asymmetry, and kurtosis. From the
Hourly sea level records for the period 1997–98 were percentiles analyses, a few outliers in the sea level
obtained from the tide gauge station installed at record could be identified from box plots. They were
Cananéia Estuary at latitude 258019S and longitude replaced by the average values between the previous
478559W. The equipment has been operated and main- and following hourly data. Before fitting, both series
tained by the Instituto Oceanográfico/Universidade de were used for the period from January 1997 to Decem-
São Paulo, and data were available (http://ilikai.soest. ber 1998 to study the coastal sea level response related
hawaii.edu/UHSLC/jasl.html) through the Joint Ar- to the meteorological conditions as well as the behavior
chive for Sea Level program of the University of Ha- of the coastal sea level in this Brazilian region.
waii Sea Level Center. Atmospheric pressure and wind
a. Filtering data using a low-pass filter
from the National Centers for Environmental Predic-
tion–National Center for Atmospheric Research re- This study was focused on the oscillations in sea level
analysis data (Kistler et al. 2001) at 0000, 0600, 1200, caused by frequency lower than astronomically driven
and 1800 UTC between 258009 and 278309S and from forces that is related to the passage of frontal systems
the coastline up to 458009W at 2.58 3 2.58 and 1.9058 3 that have periods around 3–5 days. Tides and inertial
1.8758 grid points—latitude and longitude, respec- motions usually cause a high-frequency noise in sea
tively—were used for studying the local influences level records used to analyze low-frequency motion in
caused by the passage of cold fronts. Meteorological the ocean (Thompson 1983). To eliminate diurnal and
analysis and forecasts from the forecast daily bulletins shorter-period tide oscillations from the input dataset,
transmitted by the Centro de Hidrografia da Marinha the Thompson low-pass filter, a symmetric digital filter,
148 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 48
was used. This filter is defined by the following expres- tion, and, to pass low frequencies correctly, a constraint
sions: is considered (Thompson 1983):
n
Yt 5 å
n
8
1 w , V1
<
>
pðw V1 Þ
LðwÞ 5 1 2 1 1 cos
= for V1 , w , V2 ; ð3Þ
>
: V2 V1
0 V2 , w
where V1 and V2 are cutoff frequencies chosen within a Spectral analyses of filtered meteorological data and
definite range. The cutoff frequencies used in this study filtered sea level records were carried out, using fast
were V1 5 6.48 h21 and V2 5 11.28 h21, with periods of Fourier transform. Cross-spectral analyses were ob-
56.25 and 32.14 h, respectively. The Thompson filter tained to identify the frequency characteristics of the
uses 15 harmonic components and local inertial fre- local and remote meteorological events that have an
quency (Coriolis parameter: 2V sinu, where u is lati- influence on the variation of the sea level at the
tude) to calculate the weights that will be used for fil- Cananéia tide gauge station. The coherence between
tering by convolution of the hourly dataset. the peaks of the meteorological and sea level time se-
Hourly observed sea level records were then filtered ries was analyzed to verify the linear correlation be-
to remove the oscillations or noises related to tidal fre- tween the components of the bivaried process.
quencies. For the reanalysis dataset, the same filter was
used, considering the 6-hourly intervals. After filtering,
the hourly sea level series was replaced at 6-hourly in- 5. Neural network modeling
tervals as the reanalysis data so that both datasets could
The progress of neurobiology allowed several re-
be compared for the same time interval and frequencies.
searchers to develop this model to emulate the cerebral
capacity of learning in the attempt to solve problems of
b. Series analysis in the time and frequency domain a complex nature. It has the capability of abstraction in
Storm surge is usually considered to be driven by two representing the characteristics of the phenomenon
processes: the extreme wind stress and atmospheric through the information from a large database. To de-
pressure. Therefore, cross correlations among the fil- termine the best linear approach to a dataset, Rumel-
tered sea level, atmospheric pressure, and zonal and hart et al. (1986) has developed the backpropagation
meridional wind stresses were calculated. The zonal learning algorithm, which is widely used in the mul-
wind stress (zws) and meridional wind stress (mws) tiple-layers model.
were calculated using the following equations: An NNM receives a set of inputs Xi that is multiplied
by a weight Wi and added, consisting of a linear com-
T x 5 rCd jW jU and ð4Þ bination. It is expressed as
FIG. 6. Oscillations of observed and filtered sea level time series from Cananéia tide gauge
station using the Thompson low-pass filter.
Y k 5 uðuk 1 bk Þ; ð7Þ
FIG. 8. Fourier analysis of the sea level variation (meteorological tide) related to
meteorologically driven forces at Cananéia Estuary: (a) pressure, (b) zws, and (c) mws.
surface according to some stochastic criteria (Wasser- technique, given only a training set (Cigizoglu and Alp
man 1993); the generalized regression neural network 2006); and NNM feed forward–MLP, which used the
(GRNN), a method for estimating the joint probability supervised learning and backpropagation algorithm.
density function of x and y as in the standard regression All samples were used with intervals of 6 h (LT)
JANUARY 2009 DE OLIVEIRA ET AL. 151
FIG. 12. Scatterplots of simulated and filtered sea level data in the training stage for 1997
and the validation stage for 1998.
sphere at the Cananéia tide gauge station. This model shown in Fig. 13. The results obtained for 24-hourly
presented the best performance, with a correlation co- time-lag simulations of approximately 0.83 for the cor-
efficient around 0.99 for the 6-hourly time-lag simula- relation coefficient suggest that this model could still be
tion, and it can be efficient to forecast storm surge, as used for forecasting the low-frequency sea level at this
FIG. 13. Comparison between observed and predicted coastal sea level with both models. It
is verified that the HM undervalues the prediction for a storm surge on 26–28 Mar 1998.
154 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 48
time lag with good performance. Forecasting for peri- Gan, M. A., and V. B. Rao, 1991: Surface cyclogenesis over South
ods larger than 24-hours could be improved by consid- America. Mon. Wea. Rev., 119, 1293–1302.
Haykin, S., 2001: Redes Neurais: Princı´pios e Prática (Neural Net-
ering hydrodynamic variables such as river discharges.
works: Principle and Practice). 2nd ed. Bookman, 900 pp.
The results indicate that the NNM can also be useful Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Re-
as a complement to the standard harmonic model and analysis: Monthly means CD-ROM and documentation. Bull.
thus to improve the sea level forecast. Also, the pro- Amer. Meteor. Soc., 82, 247–267.
posed NNM for predicting the surge level can be fur- Lee, T. L., 2006: Neural network prediction of a storm surge.
Ocean Eng., 33, 483–494.
ther applied to other locations along the Brazilian coast
Mantovanelli, A., and Coauthors, 2004: Combined tidal velocity
or in other sites in the world. In addition, this NNM and duration asymmetries as a determinant of water trans-
could be developed in conjunction with a numerical port and residual flow in Paranaguá Bay estuary. Estuarine
ocean model (e.g., Princeton Ocean Model) to improve Coastal Shelf Sci., 59, 523–537.
forecasting water levels at the key locations. Marone, E., and R. Camargo, 1994: Marés meteorológicas no lito-
ral do estado do Paraná: o evento de 18 de agosto de 1993
Acknowledgments. We are thankful to CHM for sup- (Meteorological tides in the coast of the state of Paraná: The
event of 18 August 1993). Nerı´tica, 8, 1–2.
plying the tide gauge records and the meteorological McPhaden, M., 1999: El Niño: The child prodigy of 1997–98. Na-
information. Appreciation and thanks are also given to ture, 398, 559–562.
three anonymous reviewers for their constructive com- Netto, S. A., and P. C. Lana, 1997: Influence of Spartina Alterni-
ments and suggestions to improve the manuscript. flora on superficial sediment characteristics of tidal flats in
Paranaguá Bay (South-eastern, Brazil). Estuarine Coastal Shelf
Sci., 44, 641–648.
REFERENCES Paiva, A. M., 1993: Estudo das variac xões do nı́vel do mar em
Angulo, R. J., and G. C. Lessa, 1997: The Brazilian sea-level Arraial do Cabo-RJ (Study of sea level variations in Arraial
curves: A critical review emphasis on curves from Paranaguá do Cabo, Rio de Janeiro). Instituto Alberto Luiz Coimbra de
and Cananéia regions. Mar. Geol., 140, 141–166. Pós-Graduacxão e Pesquisa de Engenharia Research Rep.,
Besnard, W., 1950: General aspects about the Cananéia-Iguape COPPE, Universidade Federal do Rio de Janeiro.
region-1. Inst. Paulista Oceanogr., 1, 9–26. Pore, N. A., 1964: The relation of wind and pressure to extratrop-
Castro, B. M., and T. N. Lee, 1995: Wind-forced sea level vari- ical storm surge at Atlantic City. J. Appl. Meteor., 3, 155–163.
ability on the southeast Brazilian shelf. J. Geophys. Res., 100, Pugh, D. T., 1987: Tides, Surges and Mean Sea Level. John Wiley
16 045–16 056. and Sons, 472 pp.
Chang, H.-K., and L.-C. Lin, 2006: Multi-point tidal prediction ——, 2005: Changing Sea Levels: Effects of Tides, Weather and
using artificial neural network with tide-generating forces. Climate. Cambridge University Press, 265 pp.
Coastal Eng., 53, 857–864. Rao, V. B., and K. Rada, 1990: Characteristics of rainfall over
Cigizoglu, H. R., and M. Alp, 2006: Generalized regression neural Brazil: Annual variations and connections with the southern
network in modelling river sediment yield. Adv. Eng. Soft- oscillation. Theor. Appl. Climatol., 42, 81–91.
ware, 37, 63–68. Ribeiro, C. E. P., 1997: Uma nova técnica adaptativa para análise
Dalazoana, R., R. T. Luz, and S. R. C. de Freitas, 2005: Mean sea direcional de ondas (A new adaptive technique for direction-
level studies from tide gauge and satellite altimetry time se- al analysis of waves). Anais do II Encontro de Ondas e Marés.
ries looking for the integration of Brazilian Network to Instituto de Estudos do Mar Almirante Paulo Moreira, 182–
SIRGAS. Rev. Bras. Cartogr., 57, 140–153. 193.
de Mesquita, A. R., cited 2008a: Marés, Circulac xão e Nı́vel do mar Rumelhart, D. E., G. E. Hinton, and R. J. Willians, 1986: Learn-
na Costa Sudeste do Brasil (Tides, circulation and sea level in ing internal representations by error propagation, parallel
the southeastern coast of Brazil). [Available online at http:// distributed processing. Exploration in the Microstructures of
www.mares.io.usp.br/sudeste/sudeste.html.] Cognition, Vol. 1, D. E. Rumelhart and J. L. McClelland,
——, cited 2008b: Sea level variations along the Brazilian coast: A Eds., MIT Press, 318–362.
short review. Brazilian Symp. on Sandy Beaches. [Available Seluchi, M. E., 1995: Diagnóstico y prognóstico de situaciones
online at http://www.mares.io.usp.br/praias/praias.html.] sinópticas conducentes a ciclogênesis sobre el este de Su-
Doodson, A. T., and H. D. Warburg, 1944: Manual de Mare´s do damerica (Diagnostics and prognostics of synoptic situations
Almirantado (Manual of Tides of the Admiralty). 2nd ed. conducive to cyclogenesis on the east of South America).
Brazilian Navy, 372 pp. Geofis. Int., 34, 171–186.
Elsner, J. B., and A. A. Tsonis, 1992: Nonlinear prediction, chaos, Shaeffer-Novelli, Y., H. S. L. Mesquita, and G. Cintrón-Molero,
and noise. Bull. Amer. Meteor. Soc., 73, 49–60. 1990: The Cananéia lagoon estuarine system. Estuaries, 13,
Filippo, A. M., 2003: Variabilidade do nı́vel do mar em func xão de 193–203.
eventos meteorológicos de baixa freqüência (Variability of SMA, 1990: Macrozoning of the Cananéia-Iguape estuary:
sea level as a function of meteorological events of low fre- Coastal management planning. Secretaria do Meio Ambiente
quency). Ph.D. thesis, Fuminense Federal University, Ni- do Estado de São Paulo, 41 pp.
terói, Rio de Janeiro State, , Brazil, 100 pp. ——, 1996: Regulation of the Cananéia-Iguape-Peruı́be-
Franco, A. S., 1981: Tides: Fundamentals Analysis and Prediction. environment protected area (EPA): Action planning. Insti-
IPT, 232 pp. tuto Brasileiro do Meio Ambiente e dos Recursos Naturais
Fu, L.-M., 1994: Neural Networks in Computer Intelligence. Renováveis (IBAMA), 64 pp.
McGraw-Hill, 460 pp. Stech, J. L., and J. A. Lorenzzetti, 1992: The response of the South
JANUARY 2009 DE OLIVEIRA ET AL. 155
Brazil Bight to the passage of wintertime cold fronts. J. Geo- féricas locais em São Francisco do Sul-SC (Meteorological
phys. Res., 97, 9507–9520. tides and local atmospheric forcing in São Francisco do Sul,
Storch, H., and K. Worth, 2008: Storm surges, perspectives and Santa Catarina). M.S. thesis, Federal University of Santa
options. Sustainability Sci., 3, 33–43. Catarina, Florianópolis, Santa Catarina State, Brazil, 115 pp.
Sztobryn, M., 2003: Forecast of storm surge by means of artificial Tseng, C. M., C. D. Jan, J. S. Wang, and C. M. Wang, 2007: Ap-
neural network. J. Sea Res., 49, 317–322. plication of artificial neural network in typhoon surge fore-
Thompson, R. O. R. Y., 1983: Low-pass filters to suppress inertial casting. Ocean Eng., 34, 1757–1768.
and tidal frequencies. J. Phys. Oceanogr., 13, 1077–1083. Wasserman, P. D., 1993: Advanced Methods in Neural Computing.
Truccolo, E. C., 1998: Maré meteorológica e forc xantes atmos- John Wiley and Sons, 250 pp.