将函数应用于分组熊猫数据框中的列并将输出作为新列返回

Question

I have some weather dataset consisting of multiple columns:

StationID, altitude, datetime, longitude, latitude, rainfall

I have multiple stations, which are identified by their respective IDs. The rainfall column has accumulated rainfall amounts. For example, for station X in 10 days, I could have (in mm/day):

station X, 0 0 0 1 5 6 6 8 8 15

For station Y, I could have

*station Y, 0 1 14 14 14 15 18 18 18 20

But what I need are intensity values, that is, the amount from one day minus the other. This would give me the following values for stations X and Y (the first value starts with 0),

station X, 0 0 0 1 4 1 0 2 0 7

station Y, 0 1 13 0 0 1 3 0 0 2

I created a function, which takes in a time series and computes this difference:

def intensity(ts):
    ts2 = [0]
    for i in range(0,len(ts[:-1])):
        ts2.append((ts[i+1]-ts[i]))
    return ts2

test = [1,2,3,4,5,10,10,10,20,25]
intensity(test)

Now, my question is: how can I apply this function to the 'rainfall' column in my dataframe for each station group, ie:

dfg = df.groupby('station')

and then assign the output to a new column in the dataframe (eg: 'rain_intensity' column)?

Answer 1

I think you need:

print (df.groupby('station')['rainfall'].apply(intensity))

But better is diff with replace NaN to 0 by fillna and then if necessary convert to int :

print (df.groupby('StationID')['rainfall'].diff().fillna(0))

Sample:

df = pd.DataFrame({'rainfall': [0, 0, 0 ,1, 5, 6, 6, 8, 8, 15, 0, 1, 14, 14, 14, 15, 18, 18, 18, 20],
'StationID': ['station X'] * 10 + ['station Y'] * 10})

print (df)
    StationID  rainfall
0   station X         0
1   station X         0
2   station X         0
3   station X         1
4   station X         5
5   station X         6
6   station X         6
7   station X         8
8   station X         8
9   station X        15
10  station Y         0
11  station Y         1
12  station Y        14
13  station Y        14
14  station Y        14
15  station Y        15
16  station Y        18
17  station Y        18
18  station Y        18
19  station Y        20

def intensity(ts):
    ts = ts.tolist()
    ts2 = [0]
    for i in range(0,len(ts[:-1])):
        ts2.append((ts[i+1]-ts[i]))
    return pd.Series(ts2)

df['diff1'] = df.groupby('StationID')['rainfall'].apply(intensity).reset_index(drop=True)
df['diff2'] = df.groupby('StationID')['rainfall'].diff().fillna(0).astype(int)

print (df)
    StationID  rainfall  diff1  diff2
0   station X         0      0      0
1   station X         0      0      0
2   station X         0      0      0
3   station X         1      1      1
4   station X         5      4      4
5   station X         6      1      1
6   station X         6      0      0
7   station X         8      2      2
8   station X         8      0      0
9   station X        15      7      7
10  station Y         0      0      0
11  station Y         1      1      1
12  station Y        14     13     13
13  station Y        14      0      0
14  station Y        14      0      0
15  station Y        15      1      1
16  station Y        18      3      3
17  station Y        18      0      0
18  station Y        18      0      0
19  station Y        20      2      2