[All times are CET (UTC + 1h)]
Welcome to the course with pratical information and a brief introudction to your instructors.
You will learn some of the basics of working in R, including common data types and classes, installing and attaching packages, vectors and arrays and basic math operations.
You will be introduced to data frames and lists - two of the most useful classes for storing data in R. harp makes extensive use of both data frames and lists, so this lesson serves to provide a foundation for working with harp. You will be shown how to manipulate data in data frames and introduced to the concept of tidy data.
If you've had any problems installing harp, this session will give you an opportunity to get help from your instructors to install harp successfully.
You will be introduced to harp's main functions for reading external data:
read_grid()read_forecast()read_obs_file()read_obs()read_analysis()SQLite format
that enables fast access to the data.
You will be introduced to the harp point verification workflow, read > join > verify > plot, for deterministic forecasts and the functions needed to follow this workflow:
read_point_forecast()read_point_obs()join_to_fcst()det_verify()plot_point_verif()We will extend what we learned in the previous lesson to apply the same workflow to ensemble forecasts. You will also learn how to deal with lagged ensembles, and if time allows, multimodel ensembles. You will be introduced to the main ensemble verification function and functions for producing specific scores:
ens_verify()ens_spread_and_skill()ens_crps()ens_rank_histogram()ens_brier()ens_reliability()ens_roc()ens_value()
You will be introduced to the harpSpatial package, and shown how to run
spatial verifcation in harp using the verify_spatial()
function.
We will continue to explore the harpSpatial package and how to plot
spatial verification scores using the
plot_spatial_verif() function.
You will be shown how MetCoOp has implemented harp into their production chains to run operationally. You will also be shown how MetCoOp have set up a shinyserver to share verfication resuts internally.
You will be introcuded to the groupings argument to harp's point
verification functions. This enables you to group data together and get
verification scores for each group. This will include grouping by vertical
level to get scores for vertical profiles, grouping by valid time, grouping
by forecast cycle and by groups of weather stations.
You will be shown how to compute the statistical signficance of verification scores using bootstrapping, how to take account of serial correlations in the data and how to plot the results on score cards.
You will be shown some of harp's methods for plotting forecast data, as well as getting some tips on how use the ggplot2 package to make beautiful charts tailored to your own needs.
In the weeks after the course, we will run some live coding sessions where you can learn about the workflows we use in the real world, using real data with harp. You will have the opportunity to see how to incorporate harp functions into reusable scripts, how to build your own package that makes use of harp functions, and how to build functions that harp can use to read data in any format. We will also go through how to make automated reports and some much more advanced plotting techniques.