CSGD ☔ 🌎 💻

This repository contains files to post-process precipitation ensemble forecasts using the CSGD method. The scripts are designed to be compatible with HOOPLA. That is, CSGD outputs can be used as input variables of the HOOPLA hydrological models.

The CSGD post-processor is based on a complex heteroscedastic, nonlinear regression model conceived to address the peculiarities of precipitation (e.g., its intermittent and highly skewed nature and its typically large forecast errors). This method yields full predictive probability distributions for precipitation accumulations based on ensemble model output statistics (EMOS) and censored, shifted gamma distributions.

HOOPLA is an automatic software that allows to carry out model calibration and obtain hydrological simulations and forecasts at different time steps.

The following gives a brief description of the individual files:

• FormatData.r: This code prepares the observations and precipitation forecasts in HOOPLA format to work directly on the CSGD functions. It reads the Matlab variables in Hydromet_obs & Ens_met_fcst folders.
• AuxiliaryFunctions.r: A collection of small functions (calculating ensemble statistics, data selection, etc.) used by the other R-scripts
• CSGD.r: This code is a simplified variant of the method proposed by Scheuerer and Hamill 2015b: Censored, Shifted Gamma Distributions (CSGD).
• CreateSchaakeRanks.r: This code selects historical trajectories by the method proposed by Scheuerer et al. 2017 and determines the rank order: Min. Divergence Schaake Shuffle (MDSS).

How does it work? 📝

🔵 When using HOOPLA data

Preliminary steps

1. Copy and paste the Hydromet_obs & Ens_met_fcst folders from HOOPLA into the RAW_DATA/time step folder (time step could be 3h or 24h). Folders path in HOOPLA:

   Hydromet_obs & Ens_met_fcst: HOOPLA-master/Data/time step/

2. Copy and paste the catchment_names.mat file into the RAW_DATA/time step folder. File path in HOOPLA:

   Catchments names: HOOPLA-master/Data/time step/Misc

Execute the functions in the following order:

1. AuxiliaryFunctions.r
2. FormatData.r
3. CSGD.r
4. CreateSchaakeRanks.r

🔴 without using HOOPLA data

Preliminary steps

You need a DATA_FOR_CSGD_timestep.Rdata file in the RAW_DATA/time step folder with the following variables:

• Pt_Obs = observations --> dim(days, nbLT,nBV)
• Pt_Fcast = Forecasts -> dim(days, nbmMet, nbLT, nBV)
• dates = a numeric date format of the post-processing period
• ts = time step
• nameC = list with catchments names
• nBV = No. of catchments
• nbLT = lead times
• nbmMet = No. of meteorological ensemble members
• month.string = c("Jan","Feb","Mar","Apr","Mai","Jun","Jul","Aug","Sep","Oct","Nov","Dec")
• years = list with the years of the post-processing period [e.g., c(2000,2001,2002,...etc)]
• nyrs = No. of years

Execute the functions in the following order:

1. AuxiliaryFunctions.r
2. CSGD.r
3. CreateSchaakeRanks.r

‼️ Before executing the functions, modify the forecasts/observations settings as desired. All parts that can be modified in the scripts are indicated ("THE ONLY PART TO MODIFY"):

• setwd("C:/Main/Folder/Path/")
• ts = time step
• nbmMet = No. of meteorological ensemble members
• nbLT = lead times
• StartDateFcast = first day of the forecast period
• FinDateFcast = last day of the forecast period
• StartDateObs = first day of the observation period
• FinDateObs = last day of the observation period
• HH = Define time window for observation dates. Equal 6 if ts= 3h and equal 0 if ts= 24h

Contact

Please, feel free to contact me if you find any bugs or have any questions. 😃 📧: [email protected]