Forecast verification: a worked example

In this example we obtain data from the NCEP's CFSv2 seasonal forecasting system (dataset = "CFSv2_seasonal"). In particular, in this example we will retrieve maximum daily surface temperature (var = "tasmax") for boreal summer (JJA, season = 6:8) for a rectangular domain centered on the Iberian Peninsula and France (lonLim = c(-10,15) and latLim = c(35,50)), for the period 1981-2000 (years = 1981:2000), and considering the first 9 ensemble members (members = 1:9) and a lead-month 2 forecast 2 (leadMonth = 2).

We will illustrate the verification of these predictions data against the observational gridded datasets WATCH Forcing Dataset-ERA-Interim (WFDEI, dataset = "WFDEI"), also available via the ECOMS-UDG. To this aim, we will use the tools developed within the projects SPECS and EUPORIAS. In particular, we will use the verification routines available in the R package SpecsVerification (available on CRAN). However, instead of using them directly, we will use the user-friendly interface implemented in package easyVerification, via the wrapper function veriApply.

tx.forecast <- loadECOMS(dataset = "CFSv2_seasonal",
                         var = "tasmax",
                         members = 1:4,
                         lonLim = c(-10 ,15),
                         latLim = c(35, 50),
                         season = 6:8,
                         years = 1991:2000,
                         leadMonth = 2)

## [2016-05-12 12:56:18] Defining homogeneization parameters for variable "tasmax"
## [2016-05-12 12:56:18] Opening dataset...
## [2016-05-12 12:56:48] The dataset was successfuly opened
## [2016-05-12 12:56:48] Defining geo-location parameters
## [2016-05-12 12:56:49] Defining initialization time parameters
## [2016-05-12 12:56:51] Retrieving data subset ...
## [2016-05-12 13:07:02] Done

plotMeanGrid(tx.forecast, multi.member = TRUE)