Sea level verification
opr/test model intercomparison 2014
NOOS models
Dansk version

Background

This is a storm surge model intercomparison study for the North Sea. We compare one calendar year of observed and forecasted high waters, using data provided by the NOOS community, i.e. institutes from countries bordering the North West European Shelf Seas.

The station table is collated from locations each participant considers most important, and totals 23 stations.

Observations are collected in real time, and quality controlled for spurious peaks. Data gaps due to transmission problems are filled in using forecasts of longer range, where possible. All data are time interpolated to 10 min interval.

We examine only the forecast with the shortest range available for each model. This depends on the analysis time step. Validation of forecasts of longer range (1,2,.. days) remains to be worked out.

North Sea - Baltic Sea ocean models
institute country analysis step hindcast forecast range forecast step observation step
bsh germany 6 hours 0 hours 84 hours 5 min 15 min
dmi denmark 6 hours 0 hours 120 hours 10 min 10 min
met.no (tidl. dnmi) norway 12 hours 18 hours 60 hours 60 min 60 min
ukmo britain 24 hours 0 hours 48 hours 60 min 15 min (07,22,37,52)
knmi netherlands 6 hours 6 hours 45 hours 10 min 10 min
knmikal netherlands 6 hours 6 hours 45 hours 10 min 10 min
mumm belgium 12 hours 6 hours 120 hours 10 min 10 min
fcoo denmark 6 hours 0 hours 54 hours 15 min 15 min

Notes:

knmikal : dutch model with kalman filtering
knmi(kal): No forecast for Aberdeen
bsh: No forecast for NorthShields
2008: No data for RoompotBuiten, IJmuiden, Delfzijl, Tregde or Stavanger

We compare one calendar year of observed and hindcasted high waters

Results - standard errors

mean error (bias)
error standard deviation
peak bias
abs. mean peak error (10 highest peaks).

All stations



Single stations (empty plot -> no data)

TorsmindeKyst Esbjerg Vidaa Wick Aberdeen Immingham Lowestoft NorthShields Sheerness Helgoland Cuxhaven Borkum Bremerhaven Husum HoekVanHolland Vlissingen RoompotBuiten IJmuiden DenHelder Delfzijl Tregde Stavanger Oostende

Results - peak errors

abs. mean % peak error (resultatkontrakt, 3 highest peaks).

23 stations

peak table

Single stations (empty plot -> no data)

TorsmindeKyst Esbjerg Vidaa Wick Aberdeen Immingham Lowestoft NorthShields Sheerness Helgoland Cuxhaven Borkum Bremerhaven Husum HoekVanHolland Vlissingen RoompotBuiten IJmuiden DenHelder Delfzijl Tregde Stavanger Oostende

For each model

bsh ukmo knmi knmikal dnmi dmi mumm fcoo

Several years


Jacob Woge Nielsen - 23 Jan 2015