Heppi output README. 
Michael Angus 2021. 

The following data repository provides the code used in the HEPPI project for WCSSP India. 

Included are the following: 

nc files:

the IMD observed data in latitude x longitude x time (129x135x334). There are 334 total timesteps across two monsoon seasons. 

the original forecast (NCMRWF_orig_forecast.nc) in latitude x longitude x ensemble member x time (129x135x23x334).

the first postprocessed forecast, using the univariate quantile mapping approach (NCMRWF_UQM_forecast.nc) in the same dimensions.

the second postprocessed forecast, using the Ensemble Member Output Statistics approach

(NCMRWF_EMOS_forecast.nc) in the same dimensions. 


Code for postprocessing the forecast: The EMOS script used, written in R (example_EMOS_fit.R)
				      The univariate quantile mapping, generalised as a function for which any of the methods tested can be applied (generalized_QM.m)
				      A zip file containing five python functions, used for preporocessing and apply the Multivariate Quantile Mapping approach (MQM_WCSSP.zip)


an excel spreadsheet with 9 test locations I used in the project. Also including the lat/lon and the grid cell on the IMD grid. 

M-files used for validation, including the figures presented in the final report:
verif_hitmiss.m a collection of validation measures based on the hit/miss fraction of the postpocessed forecasts
rank_histograms.m calculation of ranked histograms (see final report for details)
ver_rd.m Calculations of reliability diagrams, as discussed in the feedback to the final report.