[x] maybe also plot slotgem 2021 for hw/lw (and write to csv)
[ ] discuss method with WS/SLM (like broken linear model or not)
[ ] make functions accept mean_array df instead of values/height df >> align with SLM
[ ] nodal cycle is not in same phase for hw/lw/wl for HOEKVHLD (and probably others), this is not physically correct, might be due to different data coverages
[ ] consider adding slr package as dependency for models or duplicate code?
[ ] when using linear+nodal, the slotgemiddelde varies depending on the projection year. One wants to compare consecutive slotgemiddelden and see increase (or not) if they depend on a nodal cycle this becomes unclear. Therefore, a simple linear fit might be more useful.
[ ] consider using only 10 years for (linear) fit. This way, increase in SLR will also be accounted for after a few years. However, it will be impacted by nodal cycle still, so maybe use last 20 years (then also for havengetallen). Although, maybe not all stations have 20+ years of data. And SLR response will be slower.
[ ] According to FB, one needs at least 1.5 times the nodal cycle period in order to not be affected by nodal tide, also when fitting a linear line. In general, meteo-effects will probably strongly influence a trendline over a shorter period.
[ ] discuss with AB, how are current slotgemiddelden (in getijcomponents) derived?
Todo:
mean_arraydf instead of values/height df >> align with SLM