Return full ensemble prediction data, in addition to mean

[bnubald]

• IceNet version: v0.2.8/v0.2.9_dev

Description

Would be nice to be able to optionally get the raw ensemble prediction data as a return to this function.

icenet/icenet/process/predict.py

Lines 63 to 89 in cb1cb78

	def get_prediction_data(root: object, name: object, date: object) -> tuple:
	"""
	:param root:
	:param name:
	:param date:
	:return:
	"""
	logging.info("Post-processing {}".format(date))
	glob_str = os.path.join(root, "results", "predict", name, "*",
	date.strftime("%Y_%m_%d.npy"))
	np_files = glob.glob(glob_str)
	if not len(np_files):
	logging.warning("No files found")
	return None
	data = [np.load(f) for f in np_files]
	data = np.array(data)
	ens_members = data.shape[0]
	logging.debug("Data read from disk: {} from: {}".format(
	data.shape, np_files))
	return np.stack([data.mean(axis=0), data.std(axis=0)],
	axis=-1).squeeze(), ens_members

This function reads the prediction result of different ensembles from individual numpy files and returns: (mean sea ice concentration, number of ensemble members)

Would like to update to (mean sea ice concentration, full ensemble sea ice concentration, number of ensemble members). Although, latter variable is somewhat redundant, if full ensemble data is returned.

Useful if wanting to process individual ensemble results rather than just the mean.