# Plan for Griff Rees' work in UA 2.0

Griff will join the UA 2.0 (aka Science of Cities and Regions) project from 4th of September 2023 until the end of March 2024.

He will work on clim-recal, which is a bias correction tool for climate projection data initially developed under the DyME-CHH project and now adopted by the UA 2.0 project. REG has been working on this library for more than 6 months, alongside a PDRA. Two partners (Met Office and LCAT) are interested in using the library to generate data and for assessing different bias correction methods.

The two main project goals until March is to deliver v1.0 of the clim-recal tool to the Met Office and to generate a a set of bias corrected datasets for the whole of the UK, which will be used by both partners.

Griff's work will focus on these two goals. I have listed detailed objectives for Griff below, with the caveat that there is still a bit of uncertainty around project direction, especially after December. Any changes to the plan due to change of project direction will be communicated.

It is also important to mention that Griff is currently holding off medical treatment in the hope of participating in a study. This may have bearing on his work, and we will plan around that if necessary.

### Objectives:

- By end of September:
1. Finalise the clim-recal data pre-processing pipeline which:
- Crops input data to specific geographical areas
- Converts data to 360-day calendars and
- Groups data for calibration/validation/testing.
2. NOTE: This objective is removed from the list due to Griff's extra leave in September. <s>~~Use clim-recal to generate bias-corrected climate projections for: ~~
- ~~Three cities (London, Manchester, Glasgow).
- ~~Four of the thirteen independent runs in the input data. ~~
- ~~Three variables (rainfall, tasmax, tasmin) ~~
- ~~At least one bias correction algorithm e.g., Empirical Quantile Mapping, linear scaling, Scaled Distribution Mapping, Quantile delta mapping, Delta method.~~
- ~~Years 2020 to 2080~~</s>
3. Write two simple end-to-end tests for the two of the main python scripts in the pipeline.

- By second week of October:
- Plan of objectives for October and November
- Process for reaching that together (Griff and Greg)
- Update James when finalised

- By end of October:
- Add unit tests for functions under the the `load_data` folder and end-to-end tests for scripts
- Add GitHub Actions CI and automated lint/black
- Improve the repo documentation
- Support work on finalising assessment notebook:
- Assess bias corrected output using different utility measures (descriptive statistics, scoring on test data set).
- Use 1980-2010 for calibration and 2010-2020 for assessment.
- Use 1980-2010 for calibration and 2010-2020 for assessment.

- By end of November:
- Finalise assessment notebook
- Set up RStudio server instance for December workshop
- <s>Add and debug at least one more debiasing algorithm to clim-recal and generate bias corrected data using it. </s>
- <s>Add new algorithm to assessment notebook.</s>
- Set up automated docs generation
- Design/plan December workshop demo session

- By end of December:
- Participate in stakeholder workshop with Met Office by leading the demo session
- Document feedback from workshop and prioritise improvements to pipeline and assessment notebook.
- Support handover of clim-recal management due to Greg leaving the project
- <s>Support comprehensive assessment of methods (with guidance from Met Office) using additional utility measures, for different variables/locations. Write a first version of the guidance document with Met Office/PDRA.</s>

- By end of March: (As handed over by Greg)
- Priority 1: With best bias correction algorithm, apply to whole of UK and all 13 independent runs to create dataset for Met Office Climate Data Portal, share with Met Office.
- Priority 2: Complete work on the clim-recal repo:
- Finalise documentation
- Add Python tests for main functions and classes, set up CI to run all tests (include R scripts if enough time)
- Tidy up the Python code, add any missing docstrings (include R scripts if enough time)
- A unified way of environment management (Docker or alternative)
- An assessment notebook that incorporates user feedback from December workshop
- At least 2 bias-correction algorithms
- Connect to UA technology platform: Make repo part of the UA organisation and write some high-level description for the platform web page.
- Work with PDRA to support investigation of impacts of climate change to health. The main task is to link SPC data to clim-recal data.
- Showcase at AI UK