In JuGross/ensAR: Autoregressive postprocessing methods for ensemble forecasts

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "README-"
)

Overview

The ensAR package establishes functions for statistical postprocessing of ensemble forecasts for a Gaussian distributed weather quantity. Predictive mean and standard deviation are derived by an autoregressive time series modification of the ensemble members.

It is assumed that the required data frame consists of one-step ahead forecasts matched with the corresponding observation of the weather quantity in the sense that for each row the difference between forecast and observation is the forecast error. The package contains data frames Magdeburg and ListSylt comprising daily temperature observations at German stations Magdeburg and List auf Sylt matched with 24-h forecasts generated for the respective observation day.

Installation

The latest version can be installed from Github

# install.packages("devtools")
library(devtools)
install_github("JuGross/ensAR")

Example

# Predictive distribution for one forecast day at station Magdeburg
# using the default number (90 + 30) of training days.
library(ensAR)
ensembleAR(Magdeburg[1:(90+30+1), ], obs_col = 6, mem_col = 7:56)