Improve gemiddeldgetij

[veenstrajelmer]

Todo:

• [ ] we now use 10y average of tidal components instead of 4y component file (with slotgem)
• [ ] add correction with slotgemiddelden (this sort of happened via the component file, which is not used anymore). KD 20-06-2024: Oorspronkelijk werden de waarden van gemiddeld hoogwater en laagwater over de laatste 10 jaar zoals gepubliceerd in Tienjarige Overzichten als standaardwaarden gebruikt. In de jaren 1930 was het tijverschil in de westelijke Waddenzee sterk toegenomen door de afsluiting van de Zuiderzee. Verder was men inmiddels zich bewust van trends (men sprak destijds nog vaak van bodemdaling i.p.v. zeespiegelstijging). Het slotgemiddelde was de aflezing van een trendlijn na wegfiltering van de cyclus. Later: aflezing van het lineaire deel van een gefitte functie die de som was van een rechte en een sinusoide met opgelegde periode en fase aan het eind van het laatste decennium. Teruggaan naar het gemiddelde over het laatste decennium lijkt me daarom geen verbetering. >> add correction with waterlevel slotgemiddelde only (not re-scale with slotgemiddelden of high/low water). This still requires more data than the function currently recieves (entire timeseries instead of last 10 years). But it is quite simple, just replace A0 in component sets for raw signals with slotgem (or does scaling overwrites the mean?)
• [ ] we are considering a linear fit for slotgemiddelden in https://github.com/Deltares-research/kenmerkendewaarden/issues/85. A linear fit is also easier to fit over just 10 years of data in case of havengetallen/gemgetij. That does not have to be super accurate, but it is mainly to convert these KW from representative for the middle of the period to representative for the end of the period >> 10y fits can go awry because of outliers, so might be unwise to correct with such a value.
• [ ] should also be possible to use reshape_signal with tP_goal only
• [ ] https://github.com/Deltares-research/kenmerkendewaarden/issues/29
• [ ] springneap kromme is based on "zuiveren harmonischen van S2 en M2". This seems not to be the case, the code below shows the shallow water components that are only dependent on M2/S2, which is more than the list in the 1991 document. In the past, this list resulted in a kromme that was not repetitive, is this still the case?
• [ ] deprecate debug argument for calc_gemiddeldgetij
• [ ] boi timeseries csv should have timedifference between harvt10 and hoekvhld, this difference is lost with conversion from datetimes to timedeltas. Maybe manually add 20 minute offset like was done before?

import hatyan
import numpy as np
import pandas as pd

const_list_year = hatyan.get_const_list_hatyan("year")
dummy,shallowrel,dummy = hatyan.foreman.get_foreman_shallowrelations()
bool_M2S2only = shallowrel[1].isin([1,2]) & shallowrel[3].isin(['M2','S2']) & shallowrel[5].isin(['M2','S2',np.nan]) & shallowrel.index.isin(const_list_year)
shallowdeps_M2S2 = shallowrel.loc[bool_M2S2only,:5]
list_harmonischen_M2S2_code = shallowdeps_M2S2.index.append(pd.Index(["M2","S2","MU2"]))
# print(list_harmonischen_M2S2_code)
#['2SM2', '3MS4', 'M4', 'MS4', 'S4', 'M6', '2MS6', '2SM6', 'M7', 'M8', '3MS8', 'M10', 'M12', 'SM', '3MS2', '4MS4', '4MS6', '2(MS)8', '4MS10', '3M2S10', '5MS12', '4M2S12']

list_harmonischen_M2S2_1991 = pd.Index(['SM','3MS2','MU2','M2','S2','2SM2','3MS4','M4','MS4','4MS6','M6','2MS6','M8','3MS8','M10','4MS10','M12','5MS12'])
# print(list_harmonischen_M2S2_1991)

# list_harmonischen_M2S2_1991[~pd.Series(list_harmonischen_M2S2_1991).isin(list_harmonischen_M2S2_code)]
print(list_harmonischen_M2S2_code[~pd.Series(list_harmonischen_M2S2_code).isin(list_harmonischen_M2S2_1991)])
# ['S4', '2SM6', 'M7', '4MS4', '2(MS)8', '3M2S10', '4M2S12']

So this would result in 7 different components. If we use all schureman components instead of the year-list, "S6" is also included. "M7" is not even, so should not be included.