TMP BF02

Download as pdf or txt
Download as pdf or txt
You are on page 1of 49

NEURAL NETWORK MODEL TO PREDICT A STORM SURGE

Marilia M. F. de Oliveiraa*, Nelson Francisco F. Ebeckena, Jorge Luiz

Fernandes de Oliveirab Isimar de Azevedo Santosc

a
Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia -

COPPE, Universidade Federal do Rio de Janeiro – UFRJ, Rio de Janeiro, Brazil


b
Departamento de Geografia, Instituto de Geociências – IGEO, Universidade

Federal Fluminense - UFF, Niterói, Brazil


c
Departamento de Meteorologia, Instituto de Geociências – IGEO / UFRJ,

Rio de Janeiro, Brazil

____________________________

Corresponding author address: Marilia M. F. de Oliveira, Instituto Alberto Luiz

Coimbra de Pós-Graduação e Pesquisa de Engenharia - COPPE, UFRJ, Caixa

Postal: 68501, CEP: 21945-970, Rio de Janeiro, RJ, Brazil.

E-mail: [email protected]

1
ABSTRACT

The southeast 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 meteorological driven forces in low-

frequency to the storm surge event is investigated in this work. A methodology to

predict coastal sea level variations related to meteorological events that use a

neural network model (NNM) is present here. Pressure and wind values from

NCEP/NCAR reanalysis data and tide gauge time series from Cananéia reference

station at São Paulo State, Brazil were used to analyse the relationship between

these variables and to use them as input of 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 hours).

Thus, a low-pass filter was applied in the tide gauge and meteorological time series

for periods lower than tides in order to more readily identify 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 the time lag of the NNM input variables. It was tested for 6-12-18-24-

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 an efficient model for predicting the non-

tidal residuals and can effectively compliment the standard constant harmonic

2
analysis model (HM). A case study of a storm that impacted coastal areas of

southeast Brazil in March 1998 was analysed and indicates that the neural network

model can be effectively utilized in the Cananéia region.

Key words: Artificial neural network, coastal sea level variations, surge, storm,

surge, time series forecasting.

3
1. Introduction

Oscillations in sea level due to meteorological driving forces related to wind and

pressure occur at different scales and frequencies in all coastal regions. Some

countries are affected by flooding with serious damage. Therefore, knowledge

about the sea level height variations is very important not only for marine services

but also for the protection of coastal residents, for monitoring the changes in

marine ecosystems as well as to design and construct onshore and offshore

structures.

Interactions between meteorological (atmospheric pressure, wind, sea surface

temperature) and 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.

Tropical cyclones and extratropical storms are the mainly responsible by the

occurrence of storm surges that can produce damage due to high elevation and

sprawling water over large coastal area in a single storm. The principal factors

involved in the generation and modification of storm surge are the action of wind

stresses on the surface water, reduction of atmospheric pressure (inverted

barometer effect), waves and swells in the shallow water area, coastline

configuration and bathymetry (Pore, 1964).

Frontal systems associated with extratropical cyclones introduce variations that

are observed in the tide records, sometimes resulting in storm surges (Pugh, 1987,

2005). These storms differ with respect to size and intensity and the associated

storm surges have characteristics consistent with these differences. The different

4
characteristics of these types of storms, in particular concerning wind speeds and

spatial scale of the storm (Storch, 2008).

In the South Atlantic Ocean, along the Brazilian coastline, there are few tide

gauge records with long series to analyze and predict surge events. Characteristics

of the meteorological tide variations along the Southeast coast of Brazil have been

studied by Marone and Camargo (1994). Castro and Lee (1995) presented a study

about the sea level fluctuations due to the wind-driven forces in the southeast

continental shelf. Ribeiro (1997) investigated a surge caused by the passage of a

cyclone along the Rio de Janeiro coastline that raised the sea level 0.60 m above

the mean sea level datum, causing damage to coastal communities along the

Guanabara Bay. Netto and Lana (1997) studied the superficial sediment

characteristics of tidal flats in Paranaguá Bay. Mesquita (1997, 2000) verified a

similar behavior of the mean sea level oscillations along this area of the Brazilian

coastline. Mantovanelli (2004) verified the tidal velocity and duration as a

determinant of water transport and residual flow in Paranaguá Bay estuary.

Dalazoana et al. (2005) studied the mean sea level variations using longer tide

gauge temporal series from Cananéia and Fiscal Island (State of Rio de Janeiro)

tide gauge and satellite altimetry to establish analysis methods applicable to

Brazilian vertical datum region.

The classical method of harmonic analysis is used to predict the astronomical

tides. Tidal curves appear as periodic oscillations and can be described in terms of

amplitude, period or frequency. The harmonic analysis is based in tidal variations

represented by N harmonic constituents of the tide (Doodson, 1944). Normally, 365

days of hourly data at a point are needed to extract the constituents with adequate

5
separation of closely spaced constituents using the least squares method. These

constituents can then be used to provide reliable predictions for future tides at the

respective point (Franco, 1981).

Predictions for reference stations are prepared from the astronomical

arguments, using local constituents determined by previous analysis and do not

take into account meteorological influences. Thus, the observed and predicted

values of the sea level are normally different. Numerical model developed to

predict surges are still considered insufficient due to the complexity between the

non-linear processes involved. These models require a large amount of tidal and

meteorological data, collecting many factors as central pressure, speed of the

cyclone, rainfall and coastal topography (Lee, 2006).

Nowadays, the neural network model (NNM) has been widely applied to

modeling non-linear dynamic systems, using time series that translate the physical

relations between the input variables (predictors) and the phenomenon that will be

modeled (predictand). Elsner and Tsonis, 1992 developed some methodologies for

making short-term predictions of nonlinear time-series data, using a neural network

model. These authors discuss the implication of these methodologies in the studies

of weather and climate.

The NNM has some important characteristics such as generalization,

parallelism, non-linearity, adaptability, robustness and others (Haykin 2001). These

models have been used in some fields of science and engineering. Sztobryn

(2003) applied NNM in hydrological forecasting where the variation of water level is

only wind generated. The results were successfully compared with observed sea

6
level and others routine methods. Lee (2006) applied a NNM for forecasting storm

surge in Taiwan related to the passage of three typhoons over the region. The

results indicate that NNM is efficiently capable of learning and predicting from

these events. Tseng et al (2007) used a typhoon-surge forecasting model

developed with a backpropagation neural network in the coastal of north-eastern

Taiwan. To compare the better forecasting model, four models were applied and

tested under different compositions of the input variables. For coastal and harbor

engineering applications, Chang (2006) simulated tides at multi-points considering

tide-generating forces. The NNM proposed is applicable for multi-points tidal

prediction in which the tidal type is similar to that of the original point.

The southeast coast of Brazil is sufficiently affected by cold fronts over 3 – 5

day periods. An important event that sometimes occurs due to combination of tides

and surges is the rising of the sea level with waves that reached the coastline.

There are few NNM applications to predict the variability of sea level along the

Brazilian coastline focused on the surge events. The relationships describing the

response of the coastal sea level due to the influence of cold fronts was analysed

using cross-correlations and spectral density between the tide gauge series and

meteorological variables. Maxima values and time lags of both analyses were

proposed as inputs of the sea level forecast model.

This paper presents a methodology to predict the coastal sea level variations

and surge using a NNM.

Although drastic storm surge typically does not occur along the coastal waters

of Brazil, these events can cause some damage to coastal regions. A strong storm

7
surge occurred along the southeast coast of Brazil in March 1998 caused severe

flooding in these coastal areas, destroying some coastline constructs. Fig. 1 shows

the curves related to water level over 6-day period at the Cananéia tide gauge

station in São Paulo. The storm surge revealed can be compared with other series.

In this case, the curves are referenced water levels to Mean Low Water (MLW) that

is referenced to a fixed level or station datum (tide gauge benchmark near a

gauge, to which the gauge zero is referred) for the convenience of plotting with

mostly positive number. In this paper will be demonstrate the application of the

NNM for this study case.

[Fig. 1]

Operational forecasting of high sea levels (storm surges) might be important in

the southeast coastal of Brazil, where there are registered sea level variations

above the astronomical tide predictions that can consistently impact coastal zones

in this area.

The aim of this study is to develop an empirical prediction of storm surge by

determining the relationship of the wind and pressure fields to storm surge. This

proposed model can be used to as complement of the standard constant harmonic

model to improve the prediction of the sea level variations.

2. Study area

The study area lies within the Cananéia estuary (24° 50’ – 25° 05’ S / 47° 45’ –

48° 00’ W); southeast coastal region of Brazil in São Paulo State (Fig. 2). This

8
region is located on the continental shelf, which is wider than the shelf of the

northern coast. The average width and declivity, near Cananéia city in São

Paulo State is about 218 km and 46 cm/km, respectively (Filippo, 2003). The

isobaths are oriented from southwest to northeast, parallel to the coastline with 45º

northern direction (Trucullo, 1998). It has wide coastal plains, long beach barriers,

and large estuaries (Angulo and Lessa, 1997). The Cananéia Estuary is an

important biological reserve and contains federal and state Environmental

Protected Areas (SMA, 1990/1996). This estuarine system covers an area of 135

km2 and is surrounded by a large mangrove area with high concentrations of

nutrients (Besnard, 1950; Schaffer-Novelli et al., 1990).

[Fig. 2]

The astronomical tide pattern at Cananéia station is semidiurnal with the

greatest amplitude (H) for the constituents M2 - principal lunar and S2 - principal

solar. The diurnal constituents O1 - principal lunar and K1 - declinational luni-solar

are also present as well as the shallow water constituents, M3 and M4 (third and

quarter diurnal lunar, respectively), which indicate the influence of the propagation

of the tide wave in the continental shelf (Fig.3).

[Fig. 3]

Surges that are verified in the tide gauge records normally are related to the

same extreme events; passage of cold fronts over this region (Mesquita, 1997,

9
2000). Fig. 4 shows the maximum recorded at the Cananéia gauge (3.13 m) during

a storm on 25 – 28 March 1998.

[Fig. 4]

2.1. Climatology of the region

The South American continent is affected by both tropical and extratropical

regions weather systems. The most severe weather systems in South America are

cold fronts, intense extratropical cyclones near the east coast causing intense

winds, upper level cyclonic vortices (ULCV), in some cases responsible for

cyclogenesis and frontogenesis, South Atlantic Convergence Zone (SACZ), squall

lines, mesoscale convective complexes and the Low Level Jet (LLJ). This region is

influenced by persistent high-pressure over the South Atlantic Ocean which

enhances northeast flow across the area. This circulation is disturbed, periodically,

by the passage of frontal systems caused by migrating anticyclones that move

from the southwest across the northeast in the southeast coast of Brazil. In this

region is verified the presence of strong cyclogenesis activity (Gan e Rao, 1991;

Seluchi, 1995) associated with ULCV that reach through the South America west

coast causing instability in the east and northeast sector. Gan e Rao (1991) has

verified two regions of persistent cyclogenesis over South America; one over the

San Matias Golf in Argentina (42.5°S, 62.5°W) and another over Uruguai (31.5°S,

55°W). The climate is subtropical humid and during the El Niño–South Oscillation

(ENSO) phenomenon great climatic disturbances occur in this region, leading to

abundant rain.

10
ENSO strongly influences rainfall patterns mainly in southern Brazil that tends

to be above the median from November to February (Rao, 1990). The El Niño

event in 1997–1998 was considered, by some measures, the strongest on record,

causing major climatic impacts around the world (McPhaden, 1999). During 1997-

1998 El Niño event the sea level in the Cananéia region was significantly impacted

by a storm surge and was analyzed in this work.

3. Meteorological and sea level data set

Hourly sea level records for the period 1997-1998 were obtained from the tide

gauge station installed to Cananéia estuary at latitude 25° 01’ S and longitude 47°

55’ W. The equipment has been operated and maintained by Instituto

Oceanográfico / Universidade de São Paulo (IO-USP) and data is available at

http://ilikai.soest.hawaii.edu/UHSLC/jasl.html through The Joint Archive for Sea

Level (JASL) program of University of Hawaii Sea Level Center (UHSLC).

Atmospheric pressure and wind from the NCEP/NCAR reanalysis data (Kistler et

al., 2001) at 00, 06, 12 and 18 UTC between 25º 00' S and 27º 30' S and from the

coastline up to 45º 00' W at 2.5º x 2.5º and 1.817º x 1.817º grid points - latitude

and longitude, respectively were used for studying the local influences caused by

the passage of cold fronts. Meteorological analysis and forecasts from the Forecast

Daily Bulletins transmitted by Centro de Hidrografia da Marinha (CHM) during this

period were also used. Fig. 5 shows the location of the Cananéia Estuary tide

gauge and the reanalysis grid points.

[Fig. 5]

11
4. Statistical analysis

The atmospheric pressure, wind and tide gauge time series were analyzed

statistically by estimating the center of the distribution (mean and median),

variances, standard deviation, asymmetry and kurtosis. From the percentiles

analyses could be identified few outliers in the sea level record with the box plot

graphic. They were substituted with the average values between the previous and

the following hourly data. Before fitting, both series were used for the period from

January 1997 to December 1998 to study the coastal sea level response related to

the meteorological conditions as well as the behavior of the coastal sea level in this

Brazilian region.

4.1. Filtering data using a low-pass filter

The present study was focused on the oscillations in sea level caused by

frequency lower than astronomical driven forces related to the passage of frontal

systems which have periods around 3 to 5 days. Tides and inertial motions usually

cause a high-frequency noise in sea level records used to analyze low-frequency

motion in the ocean (Thompson, 1983). To eliminate diurnal and shorter-periods

tides oscillations from input data set, the Thompson low-pass filter, a symmetric

digital filter, was used. This filter is defined by the following expressions:

n
Yt = ∑ wk xt + k w − k = wk (1)
k =−n

where Yt is the filtered time series, w− k , wk are the symmetric weights (n = -120

to +120 with total of 241 weights, including zero) and xt + k are the input data.

12
The symmetry is imposed to preserve phase information and to correctly pass

low-frequencies, a constraint is considered (Thompson, 1983).

n
∑ wk = 1 (2)
k =− n

The digital filter response is near unity at low frequencies, but near zero at high

frequencies, mainly at inertial frequency (w = w0 ) and tidal frequencies (w= wt )


An example of a function that has a desirable shape for this filter is:

1 w < Ω1
  π (w − Ω1 ) 
L(w) = 1/ 21+ cos  for Ω1 < w < Ω 2 (3)
0  Ω 2 − Ω1 
 Ω2 < w

where Ω1 and Ω 2 are cut-off frequencies chosen within a definite range. The cut-

off frequencies used in this study were Ω 1 =6.4º.h-1 and Ω 2 =11.2º.h-1, with

periods of 56.25 and 32.14 h, respectively. The Thompson filter uses 15 harmonic

components and local inertial frequency (Coriolis parameter: 2Ω sin φ) to calculate

the weights that will be use for filtering by convolution of the hourly data set.

Hourly observed sea level records were then filtered to remove the oscillations

or noises related to tidal frequencies. For the reanalysis data set was used the

same filter, considering the 6-hourly intervals.

After filtering, the hourly sea level series was replaced at 6-hourly interval as

the reanalysis data and so, both data sets could be compared for the same time

interval and frequencies.

13
4.2. Series analysis in the time and frequency domain

Storm surge is usually considered to be driven by two processes: the extreme

wind stress and atmospheric pressure. Therefore, cross-correlations between the

filtered sea level, atmospheric pressure, zonal and meridional wind stresses were

calculated. The zonal wind stress (zws) and meridional wind stress (mws) were

calculated using the follows equations:

T x = ρC d W U (4)

T y = ρC d W V (5)

where:

ρ = 1.22 Kg .m −3 (air density);

W = intensity of the wind (m.s-1) calculated from zonal (U) and meridional (V) wind

components of the reanalysis data set; and

Cd = 1.1 + 0.053 W (coefficient of drag for the southeast Brazil coast, (Stech and

Lorenzzetti, 1992).

The units used for wind stress are N.m-2, where 1 hPa is equal to 102 N.m-2.

Spectral analyses of filtered meteorological data and filtered sea level records

were carried out, using Fast Fourier Transformed (FFT). Cross-spectral analyses

were obtained to identify the frequency characteristics of the local and remote

meteorological events that have an influence on the variation of the sea level at

Cananéia tide gauge station. The coherence between the peaks of the

meteorological and sea level time series was analyzed to verify the linear

correlation between the components of the bivaried process.

14
5. Neural network modelling (NNM)

The progress of neurobiology allowed several researchers to develop this

model to emulate the cerebral capacity of learning in the attempt to solve problems

of complex nature. It has the capability of abstraction in representing the

characteristics of the phenomenon through the information from a great data base.

In order to determine the best linear approach to a data set, Rumelhart et al.

(1986) has developed the back propagation learning algorithm, which is widely

used in multiple layers model.

A NNM receives a set of inputs (Xi) that is multiplied by a weight (Wi) and

added, consisting of a linear combination. It is expressed as:

U= ∑ = Xi Wi (6)
i

The back propagation learning is used for supervisioned learning with multilayer

feed forward networks. This algorithm repeats the application of a chain rule to

compute each weight in the model with respect to an error function. The topology

of a multiple layers perceptron (MLP) is specified by the number of layers and the

number of nodes per layers. The layers are denoted by the input, hidden, and

output layers.

A basic element of this model is the activation function (linear, logic and

sigmoid) that computes the activation level across the NNM. The output signal

is given by the equation below:

Yk = ϕ (uk + bk) (7)

15
where:

ϕ (uk) = activation function, and

bk = bias ( controller over the activation function).

The first step on NNM approach to design specific network architecture

includes a number of layers, each layer consisting of a specific number of neurons.

The size and structure of the model needs to be a match for the nature of the

problem. This stage is usually not well known and therefore it is not an easy task,

involving often a trial and errors approach according to the characteristics of the

application domain. The network is then subjected to the training stage. In that

phase, neurons apply an iterative process to the input variables to adjust the

weights of the NNM in order to optimally predict the data on which the training is

performed. In this way, the error for each output unit is calculated and used to

update the weights. Consequently, one could say that the model attained the

optimum architecture or find a fit to when the error between the desired output and

the target is reduced. The NNM needs three data sets of input such as training,

testing and validation series. After learning, another new data set is used to test or

verify the performance of the trained neural network. Training and test series are

used for calibrating the model. Validation series verifies the generalization of the

model comparing the output data (predictions) with the actual ones (Fu, 1994).

In this paper, different training methods were applied to find the best

performance: Radial Basis Functions (RBF), a network particularly adapted to

approximation function. The hidden layer is defined by radial basis functions and

the learning fits a non-linear surface accordingly some stochastic criteria

16
(Wasserman, 1993); Generalized Regression Neural Network (GRNN), a method

for estimating the joint probability density function (pdf) of x and y as in standard

regression technique, given only a training set (Cigizoglu, 2006); and NNM feed

forward - MLP. It was used the supervised learning and back propagation

algorithm.

All samples were used with intervals of 6-hourly (LT) between the observations

and this data set was selected in 50% for training, 25% for testing and 25% for

validation.

The input variables for the NNM training were atmospheric variables, filtered

sea level series of previous hours, and observed wind for the actual time

6. Results of the statistical analysis

The filtered records generated the time series of the sea level response in low-

frequency related with the meteorological systems which were used in the NNM

(Fig. 6).

[Fig. 6]

Figs. 7(a) to 7(c) show the peaks of the cross-correlation between the low

frequency sea level at the Cananéia gauge station and pressure, zws and mws

from 1 and 2 reanalysis grid points. The greatest values of the crossed correlations

for the sea level response and pressure, zws and mws components were 48h, 30h,

0h, respectively with percentages around 32%, 35% and 47%.

[Fig. 7(a)]

17
[Fig. 7(b)]

[Fig. 7(c)]

In order to verify the inter-correlations of the sea level response and

meteorological variables in the same frequency, values of cross-spectral densities

and coherence between the series were analyzed. It was found that peaks of

energy and high coherence for periods from 5 to 3 days were related to passages

of cold fronts over the region.

Fig. 8(a) and 8(b) show the maxima peaks around 2.9 days for pressure and

1.7 days for zws. Another peak is also verified around 4.6 days for pressure and

2.8 days for zws. In the Fig. 8(c) can be observed peaks for periods around 6.2 to

2.8 days for mws. These values indicate a correlation of sea level variation and

meteorological events such as cold fronts that can be identified in low-frequency

motions. Similar results were found by Paiva (1993); Castro and Lee (1995) in

respect to the effect of waves in the continental shelf of the Brazilian southeast

coastline. High coherence values around 75% also were verified between the

meteorological variables and the low-frequency sea level variation to periods from

5 to 3 days (Fig.9).

[Fig. 8(a)]

[Fig. 8(b)]

18
[Fig. 8(c)]

[Fig. 9]

7. NNM performances

The maximum values of cross-correlations described previously were used as

input model. Then, a time lag was considered with respect to the sea level

response and the meteorological variables. Autocorrelations of the low frequency

sea level and wind speed for the current time was also used. Therefore, pressure,

zws, mws, 18, 12, and 6-hourly filtered sea level and wind speed predictors) as

input vectors. The filtered sea level relating to 6-hourly after was used as output

variable (predictand).

The filtering data set from January 1997 to December 1998 for predicting sea

level variations were used and the results were compared with actual and predict

values. Table 1 shows the best performances of the NNM with the correlation

coefficients (r). The MLP with 7-14-1 layers produced the best results.

[Table 1]

The back propagation algorithm was used for the NNM training. The activation

function used in the hidden and output units was the hyperbolic tangent function.

The software Statistica Neural Networks for Windows was employed in this work.

19
Table 2 shows the correlation coefficients to the selected pairs for training,

testing and validation for 6, 12, 18 and 24-hourly simulations. In both the stages a

high correlation was observed.

[Table 2]

The MLP (7-14-1) for forecasting the sea level variations for 6-hourly time lags

presents accurate results. The performance of the NNM to forecast the sea level

variations was satisfactory enough (correlation = 99.9%) for 6-hourly time lags. Fig.

10 shows the comparison between NNM generalization (validation) to predict the

variations of the low frequency sea level and the target (filtering data set). It is

observed that the two curves are quite similar, being in accordance with the

statistic results shown in Table 2.

[Fig. 10]

Fig. 11 shows the evolution for training and validation of the MLP model to

reach the error convergence. Learning rate and momentum parameters affect the

speed of the convergence of the back propagation algorithm. The stopping criterion

was based on the error to be minimized to improve the performance of the

network. The model attained the best performance for 700 epochs in which the

training error is 0,008276 and validation error is 0,008531 with learning rate of 0.01

and momentum of 0.9. After 700 epochs the process was established as showed in

the Figure below.

20
[Fig. 11]

The scatter plots shown in Fig. 12 have small disparity illustrating that NNM has

a small error in learning stage than in the validation stage according to the

correlation coefficients present in Table 2. This is a common result in establishing a

NNM. Normally the simulation performance of predictors is evaluated by the root

mean square (RMS) or the square of correlation coefficient (R2) that is called

coefficient of determination. Small RMS and large R2 values indicate that the

simulation performance is good (Chang, 2006).

The left column of this figure shows the target and desired output simulated by

NNM in training stage for 1997. This column indicates that there is little disparity

between filtered and simulated values for 6 and 12-hourly training in which R2 is

0.9981 and 0.9803, respectively. The R2 values for 18 and 24-hourly are around

0.912 and 0.7767, respectively, showing that the NNM preserves the influences of

physical process such as pressure and wind in the sea level variations. The right

column shows the scatter plots for 1998 in the validation stage. Small differences

between the two stages are verified. The R2 values for 6, 12 and 18-hourly present

similar results with the learning stage. For 24-hourly forecasting, the R2 presented

values lower than for testing stage. It can be related to the correlation between the

predictors and predictand.

[Fig. 12]

21
Strong (~15 m.s-1) south-westerly (190-260 degrees) winds blowing during 3 to

5 days, over the ocean parallel the coastline, is the most conducive wind vector for

producing storm surge along the southeast Brazilian coast.

Fig. 13 shows the curves of the sea level variation related to the storm surge

occurred on 26-28 March 1998 in southeast coastal Brazil. The value of the peak

of the high water level on 26 March was 3.13 m and the predict tide with Harmonic

Model (HM) was 2.53 m. The difference between the maxima peaks was around

0.60 m, characterizing the occurrence of a surge in this region. The value predicted

by NNM was around 0.63 m. Therefore, the value obtained with both models (HM +

NNM) was around 3.16 m. It can also be verified in the Figure that some peaks of

the high water predicted with both models are above the observed sea level. The

values of the low water level are quite similar

[Fig. 13]

8. Conclusion

Conventional numerical model developed to predict surges are still considered

insufficient due to the complexity between the non-linear processes involved. In

this paper, an alternative methodology based on the structure of neural network

model to predict coastal sea level variations related to meteorological events was

proposed.

Pre-processing of the data series in the time and frequency domain allowed

defining the input of the neural network model. Maxima correlations in the physical

22
process could determine the time lag between the meteorological variables and the

sea level response.

The results indicate that the MLP architecture of the network developed in this

work could generalize satisfactorily the non-linear behavior of the sea level

fluctuations due to the interactions ocean-atmosphere at Cananéia tide gauge

station. This model presented the best performance with correlation coefficient

around 99% for 6-hourly time lag simulation and it can be efficient to forecast storm

surge according to the Figure 14. The results obtained for 24-hourly time lag

simulations around 83% of correlation coefficient (r) suggest that this model could

be still used for forecasting the low-frequency sea level to this time lag with good

performance. Forecasting for periods larger than 24-hourly could be improved,

considering hydrodynamic variables such as river discharges.

The results indicate that the NNM can also be useful as complement for the

standard harmonic model (HM) and thus to improve the sea level forecast.

The proposed NNM for predicting the surge level can be further applied to other

locations along the Brazilian coast or in others sites in the world. In addition, this

NNM could be developed in conjunction a numerical ocean model (e.g. Princeton

Ocean Model - POM) to improve forecasting water levels at the key locations

23
9. Acknowledgements

We are thankful to Centro de Hidrografia da Marinha do Brasil (CHM) for

supplying the tide gauge records and the meteorological information. Appreciation

and thanks are also given to three anonymous reviewers for their constructive

comments and suggestions to improve the manuscript.

10. References

Besnard, W., 1950: General aspects about the Cananéia-Iguape region-1. Inst.

Paulista Oceanogr., 1, 9-26.

Castro, B. M., and T. N. Lee, 1995: Wind-forced sea level variability on the

southeast Brazilian shelf. J. Geophys. Res., 100, 16045-16056.

Chang, H. -K., and L. -C. Lin, 2006: Multi-point tidal prediction using artificial neural

network with tide-generating forces. Coast. Eng., 53, 857-864.

Cigizoglu, H. R., and M. Alp, 2006: Generalized regression neural network in

modelling river sediment yield. Adv. Eng. Software, 37, 63-68.

Dalazoana, R., R. T. Luz, and S. R. C. de Freitas, 2005: Mean sea level studies

from tide gauge and satellite altimetry time series looking for the integration of

Brazilian Network to SIRGAS. Rev. Bras. Cartogr., 57, 140-153.

Doodson, A. T., and H. D. Warburg, 1944: Manual de Marés do Almirantado. Trad.

Alberto dos Santos Franco. 2 ed. Rio de Janeiro: Brazilian Navy, 372 pp.

El-Shazly, A. H., 2005: The Efficiency of Neural Networks to Model and Predict

Monthly Mean Sea Level from Short Spans Applied to Alexandria Tide Gauge.

24
Pharaohs to Geoinformatics, FIG Working Week and GSDI-8, Cairo, Egypt,

April 16-21.

Elsner, J. B., and Tsonis, A. A., 1992: Nonlinear Prediction, Chaos, and Noise.

Bull. Amer. Meteor. Soc., 73, 49-60.

Franco, A. S., 1981: Tides: fundamentals analysis and prediction, São Paulo, IPT,

232 pp.

Filippo, A. M., 2003: Variabilidade do nível do mar em função de eventos

meteorológicos de baixa freqüência. D. Sc. Thesis, Fuminense Federal

University, Niterói, Rio de Janeiro State, Brazil.

Fu, L.-M, 1994: Neural Networks in Computer Intelligence, New York, Mc Graw-

Hill, Inc., 460 pp.

Gan, M. A., and V. B. Rao, 1991: Surface Cyclogenesis over South America. Mon.

Wea. Rev., 119, 1293-1302.

Gill, A. E., 1982: Atmosphere-Ocean Dynamics, International Geophysics Series.

Academic Press, New York, 662 pp.

Haykin, S., 2001: Redes Neurais: princípios e prática. Trad. Paulo Martins Engel. 2

ed. Porto Alegre: Bookman, 900 pp.

Kistler, R., E. Kalnay, W. Collins, S. Saha, G. White, J. Woollen, M. Chelliah, W.

Ebisuzaki, M. Kanamitsu, V. Kousky, H. van der Dool, R. Jenne, and M. Fiorino,

2001: The NCEP/NCAR 50 Year Reanalysis: Monthly Means CD-ROM and

Documentation, Bull. Amer. Meteor. Soc., 82, 247-267.

Lee, T. l., 2006: Neural network prediction of a storm surge. Ocean Eng., 33, 483-

494.

25
Mantovanelli, A., E. Marone, E. T. Silva, L. F. Lautert, M. S. Klingenfuss, V. P.

Prata, M. A. Noemberg, B.A. Knopper, and R.J. Angulo, 2004: Combined tidal

velocity and duration asymmetries as a determinant of water transport and

residual flow in Paranaguá Bay estuary. Estuar. Coast. Shelf Sci., 59, 523-537.

Marone, E., and R. Camargo, 1994: Marés meteorológicas no litoral do estado do

Paraná: o evento de 18 de agosto de 1993. Nerítica, 8, 1-2.

McPhaden, M., 1999: The child prodigy of 1997 – 98. Nature 398, 559–562.

Mesquita, A. R. de, 1997: Marés, Circulação e Nível do mar na Costa Sudeste do

Brasil, www Page, http://www.mares.io.usp.br/agn/51/una/una.html [avaiable on

September 25. 2003].

Mesquita, A. R. de, 2000: “Sea Level Variations along the Brazilians Coast: a short

review”, Brazilian Symposium on Sandy Beaches, www Page,

http://www.mares.io.usp.br/praias/praias.html [available on September 19,

2003].

Paiva, A. M., 1983: Estudo das variações do nível do mar em Arraial do Cabo - RJ,

Relatório de pesquisa, FAPERJ – COPPE/UFRJ.

Pore, N. A., 1964: The relation of wind and pressure to extratropical storm surge at

Atlantic City. J. Appl. Meteor., 3, 155-163.

Pugh, D. T., 1987: Tides, Surges and Mean Sea Level, Great Britain, John Wiley &

Sons, 472 pp.

Pugh, D. T., 2005: Changing sea levels: Effects of Tides, Weather and Climate,

Cambridge University Press, 265 pp.

26
Rao, V. B., and K. Rada, 1990: Characteristics of rainfall over Brazil: annual

variations and connections with the southern oscillation. Theor. Appl. Climatol.,

42, 81–91.

Ribeiro, C. E. P., 1997: Uma nova técnica adaptativa para análise direcional de

ondas. In: Anais do II Encontro de Ondas e Marés – Instituto de Estudos do

Mar Almirante Paulo Moreira, Arraial do Cabo.

Rumelhart, D. E., G. E. Hinton, and R. J. Willians, 1986: Learning Internal

Representations by Error Propagation, Parallel Distributed Processing:

Exploration in the Microstructures of Cognition, MIT Press, 1, 318-362.

Seluchi, M. E., 1995: Diagnóstico y prognóstico de situaciones sinópticas

conducentes a ciclogênesis sobre el este de Sudamerica. Geofís. Inter., 34,

171-186.

Shaeffer-Novelli Y., H. S. L. Mesquita, and G. Cintrón-Molero, 1990: The Cananéia

lagoon estuarine system. São Paulo, Brazil. Estuaries, 193-203.

SMA, 1990: Macrozoning of the Cananéia-Iguape estuary: coastal management

planning - Secretaria do Meio Ambiente do Estado de São Paulo, 41 pp.

SMA, 1996: Regulation of the Cananéia-Iguape-Peruíbe-environment protected

area (EPA): action planning - São Paulo/Instituto Brasileiro do Meio Ambiente e

dos Recursos Naturais Renováveis (IBAMA), 64 pp.

Stech, J. L. and J. A. Lorenzzetti, 1992: The response of the South Brazil Bight to

the passage of wintertime cold fronts. J. Geophys. Res., 97, 9507-9520.

Storch, H., and K. Worth, 2008: Storm surges, perspectives and options. Sustain.

Sci., 3, in press.

27
Sztobryn, M., 2003: Forecast of storm surge by means of artificial neural network.

J. Sea Res., 49, 317-322.

Thompson, R. O. R. Y., 1983: Low-pass filters to suppress inertial and tidal

frequencies. J. Phys. Oceanogr., 13, 1077-1083.

Truccolo, E. C., 1998: Maré Meteorológica e Forçantes Atmosféricas Locais em

São Francisco do Sul-SC. M. Sc. Thesis, Federal University of Santa Catarina,

Florianópolis, Santa Catarina State, Brazil,

Wasserman, P. D., 1993: Advanced Methods in Neural Computing, John Wiley &

Sons, Inc, New York, NY, USA, 250 pp.

28
• List of captions figure

Fig. 1: Storm surge, observed and predict water level at Cananéia tide gauge

station on March, 1998. Storm tide can be verified through the observed tide curve.

Fig. 2: Map of the study area with the localization of Cananéia estuary and the tide

gauge station.

Fig. 3: Tidal cycles at Cananéia station on March, 1998 by the M2, S2, O1, K1

constituents, including shallow water terms.

Fig. 4: Observed and predict tide at Cananéia station on March, 1998

Fig. 5: Map showing the shape of the Southeastern coastal region of Brazil and the

localization of the tide gauge station and reanalysis points (Chart modified-source:

CHM).

Fig. 6: Oscillations of observed and filtered sea level time series from Cananéia

tide gauge station using the Thompson low-pass filter.

Fig. 7(a) to (c): Peaks of the cross-correlation between the meteorological

variables and low-frequency sea level response, (a) pressure (Point 1), (b) zws

(Point 2), (c) mws (Point 2). The k-lags for the points 1 and 2 are 6 hourly. Then,

the x - values must be multiplied by 6.

Fig. 8(a) to (c): Fourier analysis of the sea level variation (meteorological tide)

related with meteorological driven forces at Cananéia estuary, (a) pressure, (b)

zws and (c) mws.

Fig. 9: Coherence between the meteorological variables and low - frequency sea

level at Cananéia estuary.

29
Fig. 10: Oscillations of the low-frequency sea level (target) and simulated curve by

the MLP.

Fig. 11: Performance of the MLP model for training and validation stages, showing

the convergence of the error and the established of the process.

Fig. 12: Scatter plots of simulated and filtered sea level data in training stage for

1997and validation stage for 1998.

Fig. 13: Comparison between observed coastal sea level and predicted with both

models. It is verified that the HM undervalues the prediction for a storm surge on

26-28 March, 1998.

30
Fig. 1: Storm surge, observed and predict water level at Cananéia tide gauge

station on March, 1998. Storm tide can be verified through the observed tide

curve.
Fig. 2: Map of the study area with the localization of Cananéia estuary and the

tide gauge station.


Fig. 3: Predicting tide at Cananéia station on March, 1998 with all remaining

176 constituents, including diurnal and shallow water terms.


Fig. 4: Observed and predict tide at Cananéia station on March 1998.
Fig. 5: Map showing the shape of the Southeastern coastal region of Brazil and

the localization of the tide gauge station and reanalysis points (Chart modified-

source: CHM).
Fig. 6: Oscillations of observed and filtered sea level time series from Cananéia

tide gauge station using the Thompson low-pass filter.


(a)

Fig. 7
(b)

Fig. 7
(c)

Fig. 7(a) to (c): Peaks of the cross-correlation between the meteorological

variables and low-frequency sea level response, (a) pressure (Point 1), (b) zws

(Point 2), (c) mws (Point 2). The k-lags for the points 1 and 2 are 6 hourly.

Then, the x - values must be multiplied by 6.


Fig. 8(a)
Fig. 8(b)
Fig. 8(c)

Fig. 8(a) to (c): Fourier analysis of the sea level variation (meteorological tide)

related with meteorological driven forces at Cananéia estuary, (a) pressure, (b)

zws and (c) mws.


Fig. 9: Coherence between the meteorological variables and low - frequency

sea level at Cananéia estuary.


Fig. 10: Oscillations of the low-frequency sea level (target) and simulated curve

by the MLP.
Fig. 11: Performance of the MLP model for training and validation stages,

showing the convergence of the error and the established of the process.
Fig. 12: Scatter plots of simulated and filtered sea level data in training stage for

1997 and validation stage for 1998.


Fig. 13: Comparison between observed coastal sea level and predicted with

both models. It is verified that the HM undervalues the prediction for a storm

surge on 26-28 March 1998.


Table 1: Correlation coefficients (r)
NNM Training (%) Validation(%)
RBF 95.36 93.63
GRNN 98.18 97.65
MLP (7-9-1) 95.73 93.83
MLP (7-11-1) 95.81 94.17
MLP (7-14-1) 99.90 99.87
Table 2: Correlation coefficients (r) for 6, 12, 18 and 24 hourly simulations
Time lag (h) Training Testing Validation
(%) (%) (%)
6 99.91 99.88 99.87
12 98.98 98.75 98.44
18 95.49 94.32 93.64
24 88.19 85.06 83.08

You might also like