Skip to content

GitLab

Merged
Created by edsmall@edsmallContributor

ARGODEV-150: Convert find_besthist

Jira Issue

ARGODEV-150 (https://jira.ceh.ac.uk/browse/ARGODEV-150)

Python Implementation

The algorithm has remained exactly the same, and will be outlined in the Matlab Implementation section below

I have split the function into four functions:

  • Calculate whether a point lies inside the chosen ellipse
  • Calculate the potential vorticity at a given latitude and depth
  • Calculate how strong correlated two points are
  • Bring all these functions together, along with some data manipulation, to find the best historical data

Testing

I have written a grand total of 19 tests for the above. They all currently pass

Old Matlab Implementation

  1. First, check whether the historical data lies within the chosen ellipse. If some of it does, select it.
  2. If we have more data than we need, then select 1/3 of our data randomly. Remove this data from the leftover data
  3. Find which data is most strongly correlated with out float data in the large ellipse/time frame. Select the best data from this pool for another 1/3, and remove it from the left over data
  4. Find which data is most strongly correlated with our float data in the small ellipse/time frame. Select the best data from this pool for our final 1/3.
  5. String all this data together and return it.

Our returned value should be the indices of the best historical data to use for our analysis.