Welcome to the harp Training course for 2022. harp is a set of R packages developed within the ACCORD consortium for analysis and verifcation of numerical weather prediction (NWP) data. The course will take you from a brief introduction to R through how to read in NWP data, harp data structures, point verification for both deterministic and ensemble forecasts, spatial verificaiton of deterministic forecasts, statistical significance for verification scores and plotting forecast data.
As you know the course is taking place online, instead of at FMI in Helsinki as originally planned, as a result of travel restrictions arising from the COVID-19 situation. The link to the bluejeans event is available on the ACCORD MQA wiki . We will do our best to make the training as interesting and engaging as we can. Each class will be 45 minutes to 1 hour and will mostly go through the functions necessary to perform each task. Questions should be asked via the chat function in bluejeans.
After the course we will continue to offer training on a weekly(ish) basis where you can watch one of the instructors go through the workflows required to undertake specific analyses using real data. These sessions will include going through the verification of some ensemble forecasts from start to finish, a case study of an extreme weather event, how to write functions that harp can use to ingest local data formats, how to build a package to adapt harp for local use, advanced plotting and more.
Before attending the course you should have an installation of R. You will also need a few system libraries:
libproj-dev]libeccodes-dev]libnetcdf-dev]remotes package in R. You will also need to install the Rgrib2
and ncdf4 packages to be able to read grib and NetCDF files.
> install.packages("remotes")
> library(remotes)
> install_github("harphub/harp")
> install_github("harphub/Rgrib2")
> install.packages("ncdf4")
If you run into any problems, especially with system libraries, try consulting the harp installation instructions. You may also find it useful to have your own account on Github, along with a Github PAT (Personal Access Token) - this can become especially important if you have problems installing harp and exceed your hourly download limit from Github. The online book Happy Git with R includes a very helpful section describing how to setup and install a Github PAT.
For most of the lessons, the instructors will be using the RStdudio IDE. RStudio provides an excellent environment in which to develop R projects, and has a lot of features that will help you along the way. We reccomend that you also follow along with the lessons using RStudio.
If you do not have access to R or RStudio, you could make use of RStudio Cloud. This will provide you with most of the tools you need to follow the course. The only drawbacks are that you won't be able to work with grib files, and the available RAM on the free tier is limited to 1Gb (although that should be sufficient).
Andrew is the main developer for harp and will be your lead instructor for the course. He is based at the Norwegian Meteorogical Institute and started contributing to harp in 2014. He has been responsible for its development from a "black-box" into a set of packages that enable interactive analysis of NWP data. He is an experienced R programmer and is keen to make harp a user friendly experience. He also works on improving the understanding and usabilty of ensemble forecasts through modern data visualization techniques.
Alex is the Godfather of harp. He instigated the idea following discussions at a HIRLAM / ALADIN all staff workshop about the lack a common tool for the verification of ensemble forecasts and spatial verification of high resolution forecasts. He is based at the Royal Meteorogical Institute of Belgium and has vast experience as an R programmer. He is the author of the meteogrid and Rgrib2 packages that harp is built around as well as the author and maintainer of the harpSpatial package.
Mikko is based at the Finnish Meteorogical Institute (FMI) and has been responsible for implemtenting harp operationally for MetCoOp. He has experience of setting up the necessary environments to run harp efficeiently and to share the results via harp's web app.
Carl is the Area Leader for Model Quality Assurance (MQA) for the ACCORD consortium. He works at FMI and oversees the development of harp as part of his role as Area Leader for MQA. Carl is taking care of the administration of the harp Training Course and will be able to answer any questions you have about accessing the course via bluejeans.