Savaztky Golay filtering problem on LAI (Leaf Area Index ) time series in Python

Question

I'm new to Python and I'm hoping someone can help me understand my problem.

In fact, I'm working on data (count, min, max, sdt) extracted from a one-year LAI (Leaf Index Area) time series. My aim is to perform smoothing on the mean of the maximum LAI of my time series for segment_id (plot), using linear interpolation while applying a Savitzky-Golay filter. The results I obtained, even after adjusting the window and polyorder parameters, were not satisfactory.

import pandas as pd
import matplotlib.pyplot as plt
from scipy.signal import savgol_filter
import numpy as np

# import data
file_path = '......../LAI.csv'
df = pd.read_csv(file_path).drop(['Unnamed: 0'], axis=1)

# convert to dataframe
df = pd.DataFrame(df)

# sort by segment_id
df = df.sort_values(by=['segment_id'], ascending=True)

# convert date to datetime
df['date'] = pd.to_datetime(df['date'], format = '%Y%m%d')
df['date'] = df['date'].dt.dayofyear

# calculate mean by each segment_id 

mean_max = df.groupby(['segment_id']).mean()

# Define segment_ids 

segment_ids = list(range(0,1)) # plot only 2 segment_ids
# Iterate over each segment_id in segment_ids
for segment_id in segment_ids:
    df_filtered = df[df['segment_id'] == segment_id].copy()
    
     # linear interpolation
    
    x = df_filtered['date']
    y = df_filtered['mean']

    # remove negative values max
    y = y.mask(y < 0)

    if y.isna().all():  # if all values are NaN, skip this segment_id
        continue

    xnew = np.linspace(40, 400)

    nan_indices = y.isna()
    yinterp = np.interp(xnew, x[~nan_indices], y[~nan_indices])
    
    # savgol filter
    yinterp1 = savgol_filter(yinterp, 15, 3, mode ='interp', cval = 1) # window size 15, polynomial order 3
  
    
   plt.figure(figsize=(10, 5))    
    plt.plot(x, y, 'o',label='data')
    plt.plot(xnew, yinterp1, '-', label='linear')
    plt.legend(loc='best')
    plt.xlabel('Day of year')
    plt.ylabel('Mean of Max LAI')
    plt.show()

Both images with and without smoothing:

I'd like to obtain a smoothed curve along the data points, as shown in the image, by adjusting the filter parameters:

My problem fixed it's liked to data sorting

I just changed this x = df_filtered['date'].sort_index() y = df_filtered['mean'].sort_index()/1000