Exceedance frequencies for 2011 differ from reference dataset

[veenstrajelmer]

The overschrijdingsfrequenties derived within KWK are a combination of weibul-fit over measurement data for high frequencies and HYDRA-data for low-frequencies. The latter are still to be updated, differences are expected. However, the first ones (high frequencies) also differ from reference datasets. This is unexpected.

Some details:

• the examples in this issue focus on higher frequencies (5 times per year to once per 10 years)
• the Gele-boekje-2016 values correspond to the exceedance frequencies from “Kenmerkende waarden getijgebied 2011" (juli, 2013)
• therefore, the different values for KWK-python and Gele-boekje-2016 both cover the exceedance waterlevel data up to 1-1-2011

Some stations from south to north:

Vlissingen:
freq KWK-python Gele-boekje-2016
5	291	305
2	308	320
1	323	335
0.5	337	350
0.2	356	365
0.1	370	385

Hoek van Holland:
freq KWK-python Gele-boekje-2016
5	193	210
2	214	230
1	230	245
0.5	247	260
0.2	268	285
0.1	283	300

Den Helder:
freq KWK-python Gele-boekje-2016
5	149	165
2	172	190
1	192	210
0.5	210	230
0.2	233	255
0.1	249	275

Harlingen:
freq KWK-python Gele-boekje-2016
5	217	210
2	249	230
1	272	245
0.5	294	260
0.2	321	285
0.1	341	300

Schiermonnikoog:
freq KWK-python Gele-boekje-2016
5	219	225
2	248	250
1	271	275
0.5	292	295
0.2	317	320
0.1	334	340

The westerly stations seem to show a shift in the frequency table (row-wise), but the northerly stations not per se. The differences can be caused by:

• According to “Kenmerkdende waarden kustwateren en grote rivieren (2013)”, all values were rounded to 5 cm so that explains part of the differences
• the data could be different in both methods. However, the KWK-python calculation is based on all available extremes for each station. If I look at the availability in Missing/duplicated values in WATHTE timeseries and extremes Rijkswaterstaat/wm-ws-dl#39, it seems that at least for Hoek van Holland and Vlissingen all extremes are available. This also seems to be the case in the locally created check-figures with extremes. It could be that not all data was available for the old method, or some data was excluded.
• The method might be sligtly different. I assumed the method implemented within the kenmerkendewaarden project in 2020 by BD corresponded to what was used before, but this might not be the case. It would also not explain why some stations do come close to reference results.