spatial-temporal interpolation in python

Question

I am new in using Python so I need a help.

I have data in two DataFrame in four columns: latitude, longitude, datetime and temperature.

In the DataFrame df2 I have latitude, longitude, datetime and I need to interpolate temperature using data from df1.

I need to use coordinate and datetime data to interpolate and I don't know how to do that.

Example of DataFrame:

df1:

lat     |    lon   |      Datetime         | temp
---------------------------------------------------
15.13   |  38.52   |  2019-03-09 16:05:07  |   23

12.14   |  37.536  |  2019-03-15 09:50:07  |   22

13.215  |  39.86   |  2019-03-09 11:03:47  |   21

11.1214 |  38.536  |  2019-03-10 16:41:18  |   22

12.14   |  37.536  |  2019-03-09 06:15:27  |   19

df2:

lat     |     lon    |     Datetime           
---------------------------------------------
13.13   |   38.82    |   2019-03-06 04:05:07    
11.14   |   36.36152 |  2019-03-15 19:51:07      
10.214  |   39.123   |   2019-03-19 11:01:08    
12.14   |   37.536   |   2019-03-10 16:15:27

Which method or function I need to use?

HMReliable · Accepted Answer · 2019-06-03 21:07:05Z

2

The best way to deal with a temporal interpolation is to convert the time into total seconds from a reference point in the past. You could then interpolate all values as though they were floats.

Here are your input dataframes df1 and df2:

df1 = pd.DataFrame({'lat':[15.13,12.14,13.215,11.1214,12.14], 
              'lon': [38.52, 37.536,39.86,38.536,37.536],
              'Datetime': pd.to_datetime(['2019-03-09 16:05:07','2019-03-15 09:50:07','2019-03-09 11:03:47','2019-03-10 16:41:18','2019-03-09 06:15:27']),
              'temp':[23,22,21,22,19]})


df2 = pd.DataFrame({'lat':[13.13,11.14,10.214,12.14], 
              'lon': [38.82, 36.36152,39.123,37.536],
              'Datetime': pd.to_datetime(['2019-03-06 04:05:07 ','2019-03-15 19:51:07','2019-03-19 11:01:08','2019-03-10 16:15:27'])})

Here is how you could convert time to floats, based on seconds from a reference point in the past:

df1['seconds'] = df1.Datetime.apply(lambda x: (pd.to_datetime(x)-pd.to_datetime('2019-03-01 00:00:00')).total_seconds())
df2['seconds'] = df2.Datetime.apply(lambda x: (pd.to_datetime(x)-pd.to_datetime('2019-03-01 00:00:00')).total_seconds())

And finally, you can use an interpolate function from scipy or any other package to interpolate using the lat, lon and seconds columns (note that some of your points in df2 fall outside the range defined in df1, and you get nans as a result):

from scipy.interpolate import griddata

griddata((df1.loc[:,['lat','lon','seconds']].values),
         df1.iloc[:,3].values,
         (df2.iloc[:,[0,1,3]].values))

edited Jun 3, 2019 at 21:07

answered Jun 3, 2019 at 20:22

HMReliable

8935 silver badges11 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Mima Mima Over a year ago

Thanks for help. When you use method nearest you get every values, but when linear method is used you get nan value as results. Is there any way to interpolate that points which fall outside the range?

Collectives™ on Stack Overflow

spatial-temporal interpolation in python

1 Answer 1

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related