Fitmodel

[ARS1986]

Hej Jag använd akvifärs preferensmetod som resultat visar lite konstigt. det visar Jag är osäker kanske något fel med ’model.plot_fitmethod(). Jag bifogar koden också för att underlätta felsökning om du har tid.
I resultat visar några punkter som ligger utanför 95% konfidensintervall men visar inga avvikelse röda cirkel.

[eriktoller]

Could you alos upload the model.plot_fitmethod() and plots over the invididual time series model.plot_wells()?

[eriktoller]

You could also just copy the code and paste the relevant part, e.g. you plotting section

[ARS1986]

I update with latest version av gwrefpy (0.4.0) and still have same problem.

[ARS1986]

here are my code: import pandas as pd
import gwrefpy as gr
import matplotlib.pyplot as plt
import matplotlib.dates as mdates

# Skapa en Model-objekt och lägg till brunnarna
model = gr.Model(name='Akviför referens model_SFAL')

# Lägg till brunnarna till modellen en och en
for well in obs_well_objects:
    model.add_well(well)
for well in ref_well_objects:
    model.add_well(well)

# Tilldela tidsseriedata till brunnarna
for well in model.wells:
    if well.name in Obs_wells:
        data = filtered_combined_df_obs[filtered_combined_df_obs['Measurement'] == well.name][['CombinedDate', 'Value']]
        well.timeseries = pd.Series(data['Value'].values, index=data['CombinedDate'])
    elif well.name in Ref_wells:
        data = filtered_combined_df_ref[filtered_combined_df_ref['Measurement'] == well.name][['CombinedDate', 'Value']]
        well.timeseries = pd.Series(data['Value'].values, index=data['CombinedDate'])

# Kontrollera om timeseries data har tilldelats resp. brunn
for well in model.wells:
    print(f"Timeseries data för brunn {well.name}:")
    print(well.timeseries.head())

cutoff = '2020-06-01'

# Se till att cutoff är ett datetime objekt
cutoff_datetime = pd.to_datetime(cutoff)

# Fit the wells (Linear regression)
model.best_fit(Obs_wells[4], offset='0D', tmax=cutoff_datetime)

# Plot all fit methods
model.plot_fitmethod(plot_separately=True, plot_style='scientific', color_style='color')
model.remove_fits_by_n(Obs_wells[4], 1)

fig, ax = model.plot_fits(plot_separately=True, show_initiation_period=True, plot_style='scientific', color_style='color', plot_ref_well=True, mark_outliers=True)
fig2, ax2 = model.plot_fits(fits=model.fits[0], plot_ref_well=False)

# Sätt x-axel gränser
start = mdates.date2num(pd.to_datetime(start_date))
end = mdates.date2num(pd.to_datetime(end_date))
ax2.set_xlim(start, end)

# Lägg till markeringar för tunnel
start_tunnel = pd.to_datetime("2020-06-01")
slut_tunnel = pd.to_datetime("2022-09-05")

# Lägg till markeringar för infiltration
infiltration_starts = pd.to_datetime(["2021-10-05", "2021-12-08", "2022-10-04", "2025-04-11"])
infiltration_ends = pd.to_datetime(["2021-10-12", "2022-02-04", "2022-12-12", "2025-09-01"])
ax2.axvline(start_tunnel, color='black', linestyle='dashed', linewidth=2, label='Start tunneldrivning')
ax2.axvline(slut_tunnel, color='black', linestyle='dashed', linewidth=2, label='Slut tunneldrivning')

# Lägg till axvspan för den första infiltration perioden med etikett
ax2.axvspan(infiltration_starts[0], infiltration_ends[0], color='lightblue', alpha=0.5, label='Infiltration')

# Lägg till resten av infiltration perioderna utan etiketter
for start, end in zip(infiltration_starts[1:], infiltration_ends[1:]):
    ax2.axvspan(start, end, color='lightblue', alpha=0.5)

start = mdates.date2num(pd.to_datetime('2015-06-03'))
end = mdates.date2num(pd.to_datetime('2025-11-26'))
ax2.set_xlim(start, end)

# Resize to fit better
fig2.set_size_inches(10, 6)
ax2.legend(loc='upper left', fontsize='small')
plt.show()
print("\n✓ Klart! Alla plottar visas.")

[eriktoller]

I cannot see anything that should be wrong with your code. It is, however, difficult as I cannot reproduce the plots.

Can you create a minimal, reproducible, example using some fictive data that you can share?
https://stackoverflow.com/help/minimal-reproducible-example

[ARS1986]

i try to create example ASAP and share the result. Thanks for your support

[eriktoller]

If possible for you and your client, you could export and upload the .gwref file.

[ARS1986]

I am not sure if i can due to sensitive of project and need permit. i will see if i can get permit to share gwref file.