1

I would like to perform a multidimensional interpolation on my data. My data has the following structure: Time, Latitude, Longitude, Speed, Course. At the moment I split the data into 4 separate time series, like: (Time, Latitude), (Time, Longitude) and so on. For each time series i am now calculating a cubic spline and linear interpolation. Afterwards i'm merging the calculated data, to one dataset.

However, I am afraid that due to a separate interpolation, some dependencies between the data will be ignored and the interpolation results will be worse than in a single multidimensional interpolation.

Therefore I would like to perform an additional multidimensional interpolation on the data. I would also like to calculate at least the multidimensional-linear and multidimensional-cubic interpolation. By entering a time point (e.g. x=23) I want to get the latitude, longitude, speed and course values.

Does anyone know some Python libraries that could be useful for my project? Big thank you in advance.

Perhaps necessary addition to the question: For me its unclear how i can create an interpolation where I only have to enter the timestamp to get the remaining data. In the example of scipy i have to enter A, B, C to get a value for D. But my question was whether it is possible to perform a interpolation, so i can enter A and get B, C, D.

Braveness
  • 49
  • 7
  • https://docs.scipy.org/doc/scipy-0.16.1/reference/generated/scipy.interpolate.RegularGridInterpolator.html – AnsFourtyTwo Sep 30 '19 at 15:08
  • 1
    Did you give a look at https://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html#multivariate-data-interpolation-griddata (2 min google search)? From what I get, you have all the tools available to perform this analysis on your own and quantify/qualify the differences between your two strategies. Anyways, where is your data ? your code? What are the data dependencies you are afraid of ? If there are known constraints, you may as well define your own interpolation model. – LoneWanderer Sep 30 '19 at 15:09
  • Of course, I found this library, too. But for me its unclear how i can create an interpolation where I only have to enter the timestamp to get the remaining data. In the example of scipy i have to enter A, B, C to get a value for D. But my question was whether it is possible to perform a interpolation, so i can enter A and get B, C, D. – Braveness Oct 01 '19 at 08:47
  • Unrelated, but from a physical point of view interpolating lat-lan values is far from trivial. Close to the equator, a minute of lat or long give more or less same distance. But when you reach the 60° of latitude, a minute of lat gives half the distance of a minute of long. Numerical methods always give numerical results. Whether the results are physically relevant is **the** question. – Serge Ballesta Oct 07 '19 at 09:18
  • But my tracks only has a length of about 15km, so you can probably neglect the influence of the earth's curvature again, can't you? I think the influence of the Earth's curvature would be bigger (and therefore more relevant) if the track would extend over several degrees on earth. – Braveness Oct 07 '19 at 09:28

0 Answers0