hack-the-climate-coextremes

Problem Description

When thinking about extreme weather conditions, it’s important to know whether or not variables can take their largest values simultaneously, as this information enables us to select appropriate statistical models. Changes in extremal dependence may also have important consequences for future risk and impact planning; a problem of particular interest is to understand how extremal dependence between weather variables at different locations may change in the future. In this project we examine the future development of the dependence between extreme events using the UKCP18 projections -- more concrete we look at the spatial extension of heatwaves (answering the question which regions will likely experience heatwaves together and whether or not heatwaves will become more localized or more widespread) and at the co-occurrence of high wind, conditional on low temperature: a problem for energy systems since wind-turbines may need to shut down, but energy demand could be high.

Work Description

Our work is split into two parts:

1. Analysing co-occurence of extreme event of high wind conditional to extreme low temperature, looking at two different time slices [2020,2030], [2050,2060]. The analysis looked at the Chi function (defined in Coles et al., 2000), considering 4 locations for the wind variables. The results consists of:

• Heat map of the Chi function
• The variability of the probability of joint extremes with respect to the base period [2010,2020]

2. Analysing co-occurence of extreme high temperature events over the next 3 decades, conditional to extreme at a given base point. Here we have down three things:

• Produce plots for Chi for an arbitrary basepoint in the UK and for the time-intervals [2020, 2030], [2030, 2040] and [2040, 2050].
• Heatwave buddy finder: find regions that are going to experience heatwaves parallely to you.
• Measuring size (statistically) of heatwaves via connected sets.

Results:

• The presentation of results in 1) can be found in the folder: coextreme_wind_temp/windfarm_coextremes.ipynb

• The presentation of the results in 2) in the file coextreme_heatwaves/heatwave_buddies.ipynb.

Preview:

Screenshot of the tool for 1) in coextreme_wind_temp/windfarm_coextremes.ipynb:

Screenshot of the tool for 2) in coextreme_heatwaves/heatwave_buddies.ipynb:

Environment

On UCL JupyterHub:

• required environment conda config --add envs_dirs /shared/groups/jrole001/geog0121/envs
• required data ln -s /shared/groups/jrole001/geog0121/UKCP18 $HOME/UKCP18

For our packages there is a hack.yml from which a conda-environment can be created. Additionally for the windfarm_coextremes.ipynb and for the heatwave_buddies.ipynb you will need to install gradio from pip: pip3 install gradio.

Additional Files - Calculations

• visualise.py: returns 3 plots (1 for every decade) showing the Chi values on the co-occurence of heatwaves conditional to extreme for a given base point. Usage demonstrated in visualise_test.ipynb
• VISUALISAION.ipynb: interactive generation of plots (as above) based on the chosen decade and UK postcode

Additional Files - Data

• chi_vals[...].txt : Chi values on the co-occurence of heatwaves conditional to extreme for all UK base points, for a given decade
• derotated_coordinates.csv: aggregated data on grid point coordinates + postcodes