2
$\begingroup$

I have a table that looks like the image below:

enter image description here

It has a lot of missing Aerosol Optical Depth (AOD) data. To solve the only the spatial component of the problem, one can use a multivariate interpolation like inverse distance weighting, popularly known as IDW. However, will the IDW multivariate interpolation work for the temporal component too? And if not, could someone advise me on what to do in such a scenario.

$\endgroup$
4
  • 2
    $\begingroup$ Most useful interpolation methods are n-dimensional and can adapt to spatio-temporal data. I'd expect better results from kriging (aka Gaussian process) than from IDW, and there are still other methods besides, like radial basis functions. What kind of tools do you have at your disposal? Do you know any programming languages? $\endgroup$
    – Matt Hall
    Mar 3, 2022 at 19:03
  • $\begingroup$ I am comfortable with Python and R. Could you elaborate on why you expect kriging to yield better results than IDW? $\endgroup$ Mar 3, 2022 at 19:22
  • 1
    $\begingroup$ With kriging you can fit an appropriate kernel, eg better controlling (reducing, mainly) the rate a point's influence falls off, and there's support for things like anisotropic kernels. And I like the visualizations and uncertainty modeling you can do with stochastic results (IDW is deterministic). Check out GSTools for Python. $\endgroup$
    – Matt Hall
    Mar 4, 2022 at 16:47
  • $\begingroup$ Will try GSTools and let you know. Thanks for the information. $\endgroup$ Mar 5, 2022 at 4:39

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.