Agenda
Each day will be split into a morning and afternoon session, with a tea / coffee break in the middle of each session.
For the most part, instruction will be “code-along” sessions with the instructor with an opportunity to go through some of your own work and / or problems in the second half of the week.
Day 1: Monday 4 March 2024
13:30 - 17:00
Introduction and Getting started
We will give a brief introduction to harp and R and
Basic reading of data
Andrew will introduce the read_forecast()
function and go through reading data in Grib, NetCDF and vfld formats. We will look at how to deal with complex lagged forecasting systems and how to transform gridded data to geographic point locations
Day 2: Tuesday 5 March 2024
09:00 - 12:30
Point Verification Workflow
Andrew will go through the workflow for verifying point forecasts. From basic preparation of the observations, through to reading the forecasts and computing and plotting the verification scores. The complexity will be increased by grouping the data with the groupings
argument and adding conditions to the verification
13:30 - 17:00
Building a Verification Script
Andrew will build on the morning’s work by converting our verification workflow into a script that could be run in a production environment. We will cover how to loop over different forecast parameters, how to deal with harp’s non-standard evaluation by embracing variables with {{ }}
and then discuss some real uses of harp point verification in an operational environment.
Day 3: Wednesday March 2024
09:00 - 12:30
Spatial Verification Workflow
Alex will take you through the workflow required to to do spatial verification for scores including the Fractions Skill Score (FSS) and SAL (Structure Amplitude and Location). The new HIRA score will also be introduced and sources of spatial data will be discussed.
Andrew will briefly go through the workflow for harp’s new ensemble fractions skill score implementation and its distribution and error components.
13:30 - 17:00
Plotting and Manipulating Spatial Data
Andrew will take you through how harp data can work with (probably R’s most popular package) ggplot2. We will have the opportunity to experiment with the harp ggplot geoms geom_georaster()
, geom_geocontour()
and geom_geocontour_filled()
and different methods to map data to colours. In addition, we will go through harp’s geographical transformation functions, how to define your own georeferenced grids and how to deal with cross-sections.
Day 4: Thursay 7 March 2024
09:00 - 12:30
Building a Function to Read External Observations
This session will be a case study of how to use harp’s function recognition system to use read_obs()
to read observations from an external source. We will use MET Norway’s Frost database, which can be queried using a REST API.
In addition, there will be time to discuss any issues that you are having with harp
13:30 - 17:00
Contributing to harp
This session will be a case study of how to contribute to harp. We will fork a harp repository, create a feature branch, modify a function, test it, merge with the develop branch and make a pull request. The plan is to modify harpPoint’s functionality for computing probabilities by adding a comparator argument, such that categorical scores can be computed for cases of less than, greater than, equal to or between thresholds.
In addition, there will be time to discuss any issues that you are having with harp
Day 5: 8 March 2024
09:00 - 12:30
Interoperability of harp with Python
Juanje will introduce some work done with Python to do Langrangian spatial verification of Python. We will discuss the prospects for integrating Python functionalities into harp
In addition, Andrew will talk about some ideas for future directions for harp, such as functions to return harp verification functions.
This will also be a final opportunity to get help with specific issues with harp and R.