As the IMPROVER product owner I would like IMPROVER to have the capability to compute the expected wind direction from multiple forecasts (rather than just from the control).
The purpose of this ticket is to build on the work done in IMPRO-476 to introduce the complex numbers approach into a plugin.
Usually blending between ensembles is done via a probabilistic approach (probability of temperature exceeding a threshold) - however this code blends by simply taking an average of the wind direction.
1) Implement the complex number wind direction averaging code from IMPRO-476.
2) Then implement code from IMPRO-476 to calculate the confidence values (r-values and CJs "average distance from central point"). These confidence values are an intrinsic componenent of 1) and will take approximately 15 minutes to include in the code. What we actually do with these confidence values will be the subject of ticket #491.
Related issues (follow on work):
Develop a CLI interface for the plugin and CLI tests. #521
Investigate resolution of low confidence values. #491
Implement results of 491 into operational code. #495
Incorporate plugin into suite. #492
Incorporate complex numbers into neighbourhooding code. #493
Acceptance criteria:
Introduce the complex numbers approach from method 3 to calculate the average wind direction in the IMPRO-476 notebook into a plugin.
Create unit tests.
Plot output to demonstrate the code functionality.
As the IMPROVER product owner I would like IMPROVER to have the capability to compute the expected wind direction from multiple forecasts (rather than just from the control).
The purpose of this ticket is to build on the work done in IMPRO-476 to introduce the complex numbers approach into a plugin. Usually blending between ensembles is done via a probabilistic approach (probability of temperature exceeding a threshold) - however this code blends by simply taking an average of the wind direction.
1) Implement the complex number wind direction averaging code from IMPRO-476. 2) Then implement code from IMPRO-476 to calculate the confidence values (r-values and CJs "average distance from central point"). These confidence values are an intrinsic componenent of 1) and will take approximately 15 minutes to include in the code. What we actually do with these confidence values will be the subject of ticket #491.
Related issues (follow on work):
Acceptance criteria: