mva 0.0.2

A package to correct the bias of forecasts/hindcasts. Read the documentation at https://github.com/gouthamnaveen/mva

Documentation

A python package to adjust the bias of probabilistic forecasts/hindcasts using "Mean and Variance Adjustment" method.

Read documentation at https://github.com/gouthamnaveen/mva

References:

[1] Torralba, V., Doblas-Reyes, F. J., MacLeod, D., Christel, I. & Davis, M. Seasonal Climate Prediction: A New Source of Information for the Management of Wind Energy Resources. Journal of Applied Meteorology and Climatology 56, 1231–1247 (2017).

[2] Manzanas, R. et al. Bias adjustment and ensemble recalibration methods for seasonal forecasting: a comprehensive intercomparison using the C3S dataset. Clim Dyn 53, 1287–1305 (2019).

Installation:

pip install mva

Parameters:

hindcast: numpy.ndarray

The hindcast (or training) data. Kindly maintain the shape of the array as (years/samples,time,ensemble-members,grid-points).

observation: numpy.ndarray

The truth or observations corresponding to the hindcast. Kindly maintain the shape of the array as (years/samples,time,grid-points).

forecast: numpy.ndarray, optional

The forecast (or test) data. Kindly maintain the shape of the array as (time,ensemble-members,grid-points).

Note: Kindly respect the array shapes even if the computation is done for one time/grid point/ensemble member.

Methods:

adjust_hindcast():

This method corrects the bias of the hindcast using hindcast of the remaining years in the set (i.e., leave-one-out approach) and the corresponding observations.

Returns:

bias_adjusted_hindcast (Note: It has the same shape as the hindcast)

adjust_forecast():

This method corrects the bias of the forecast using hindcast and the corresponding observations. This method works only when the forecast parameter is given.

Returns:

bias_adjusted_forecast (Note: It has the same shape as the forecast)

Demonstration:

import numpy as np
import mva.mva as mva

Let's imagine that we have loaded the data of hindcast, forecast, and observation.

Example - 1

In [1]: hcast.shape
Out[1]: (20,46,10,6)
In [2]: fcast.shape
Out[2]: (46,50,6)
In [3]: obs.shape
Out[3]: (20,46,6)
In [4]: bc = mva(hcast,obs,fcast)
In [5]: ad_hcast = bc.adjust_hindcast()
In [6]: ad_hcast.shape
Out[6]: (20,46,10,6)
In [7]: ad_fcast = bc.adjust_forecast()
In [8]: ad_fcast.shape
Out[8]: (46,50,6)

Example - 2

In [1]: hcast.shape
Out[1]: (20,46,10,6)
In [2]: fcast.shape
Out[2]: (48,50,6)
In [3]: obs.shape
Out[3]: (20,46,6)
In [4]: ad_hcast = mva(hcast,obs,fcast).adjust_hindcast()
In [5]: ad_hcast.shape
Out[5]: (20,46,10,6)
In [6]: ad_fcast = mva(hcast,obs,fcast).adjust_forecast()
Out[6]: Please respect the array shapes and try again!