Extreme sea levels in the Mediterranean Sea

1
Extreme sea levels in the Mediterranean Sea
Marta Marcos, Mikis Tsimplis, Andrew Shaw
This work has been partly funded by EC GOCE
036961, CIRCE
2
Objectives
1st ) Estimate the extreme distribution and map
the return levels from tide gauge data and from
hindcast data in the Mediterranean Sea and the
Atlantic Iberian coasts 2nd) Explore the
temporal changes in extremes and its association
with mean sea level

• Data
• Methodology
• Mapping of return levels
• Observations vs. Hindcast data
• Temporal variability

3
Tide gauge data
Collection of hourly tide gauge data 75 stations
lt 68 yrs
Thanks to B. Pérez , F. Raicich, IEO (Spain), IGN
(Spain), CNR(Italy), APT(Itally), SONEL (France),
ESEAS MedGLOSS and Greek Hydrographic Service
asked for money to provide with the data
(18/week/st)
4
Tide gauge data
Data Processing

• Tidal analysis
• Quality control datum shifts, outliers, time
  drifts

Split
Civitavecchia
5
Hindcast data
HIPOCAS data set The barotropic model HAMSOM is
forced by a downscaling of atmospheric pressure
and wind fields generated by the model REMO (from
a NCEP re-analysis) The result consists in 44
years (1958-2001) of atmospheric and sea level
data constitute a homogeneous, high resolution
data set.
Split
Alicante
Developed by Puertos del Estado (Spain)
6
Methodology
Estimation of return levels of sea level
extremes Fitting the extreme distribution to a
Generalized Pareto Distribution Joint
Probability method for tides and surges
Temporal variability of sea level
extremes Percentiles
7
Methodology
The tail of the distribution is fitted to a
Generalized Pareto Distribution
Example of cumulative distribution functions of
observations
Coruña
Málaga
8
Methodology
Joint probability tidesurges
Test of interaction tides-surges
Distribution of extreme events as a function of
the tidal phase of the main tidal component M2
9
Return levels
50-yr return levels of observations (5 extremes
per year)
160 cm
250 cm
lt60cm
Higher extremes are found in the Atlantic coasts
due to the presence of large tides
10
Return levels
50-yr return levels of tidal residuals
145 cm
Values become consistent between Atlantic and
Mediterranean sites when tides are removed
11
Return levels
50-yr return levels of hindcast data
120 cm
Hindcast data present the same spatial pattern
but lower return levels
12
Return levels
Differences in the return levels from tidal
residuals and hindcast data
Differences are a measure of the error of the
surge model respect to the observations
13
Return levels
50-yr return levels of hindcast data
70 cm
NS gradient
14
Return levels
Joint probability tidessurges
Consistent pattern with observed return
levels The advantage of the hindcast is the
length and consistency of the time series
15
Temporal variability
Do sea level extremes change in time? How do
they change in relation to MSL?
50th, 90th, 95th, 99th and 99.9th percentiles of
observed sea level
Coruña
Alicante
Dubrovnik
16
Temporal variability
Correlations with winter NAO
17
Temporal variability
Trends in sea level extremes
18
Conclusions

• A total of 75 tide gauge stations with hourly
  data have been used to estimate sea level
  extremes, covering periods of up to 68 yrs
• Most of them are short records (70 shorter than
  20 yrs)
• Largest return levels are observed in the
  Atlantic stations due to the tides. Within the
  Mediterranean the largest values are found in the
  Adriatic
• Changes in sea level extremes are consistent
  with changes in MSL
• The hindcast data presents the same spatial
  pattern as the tide gauges, but underestimates
  significantly the sea level extremes
• With appropriate scaling the 2D models with
  future weather conditions should be sufficient to
  study the changes in extremes in the
  Mediterranean
• In the Atlantic non-linear interaction with
  tides has been observed and thus these runs need
  to include the tides.
• Ongoing work flood duration the role of
  higher/lower sampling frequencies