[Iss 441] Shear by Time of Day bug in _apply()

[rach185]

The bw.Shear._apply() function uses integer months to index the shear coefficients. The shear coefficients matrix is labelled with calendar months and begins on the first full month of coverage in the input timeseries, i.e. does not always have the first column as January. This means that the shear coefficients applied to a particular month are not necessarily consistent.

Currently the bw.Shear._apply() function is updated to reference the coefficient matrix using calendar month names. Alternatively the coefficient matrix could be re-ordered to always begin in January - although may not be applicable for timeseries shorter than 1 year.

Sensitivity checks on this fix are within the range of a 0.00%-0.13% impact on wind speeds shear up 60m, with the average impact being 0.04% change in sheared wind speed. However, these sensitivity checks were not extensive.

[stephenholleran]

Hi @rach185,

I think we should get to the source of the problem which seems to be

we should order this resulting list so the first item is 1 (if there is a full year of data).

Then, we could use the same method for pulling the unique list of months instead of just 0 to 12 counting.

This should hopefully fix the issue?

[rach185]

I have updated the apply function so that it iterates through the sorted list of unique months, I've enumerated through this list to keep the j variable and preserve how df_wspds is called.

This update also allows users to calculate the shear by month when less than 1 year of data is available.

[rach185]

Reported bug when the coverage is too low in a month to calculate shear: column created for shear values but they cannot be filled -> conclusion, shear by TOD can only be used when all calendar months have good enough coverage. Error handling will be included.