GIS Wind Energy & Sea-Level Rise (SLR)

This repository supports a geospatial workflow for UK onshore wind capacity under sea-level rise scenarios. It ingests public energy and boundary data, combines GIS-derived flood or inundation flags with site attributes, quantifies electrical capacity at risk, visualizes results, and applies a transparent prioritization rule for which farms merit protection attention first.

---

Highlights at a glance

Metric (from current Affected_Farms.csv)	1 m SLR	3 m SLR
Affected sites (binary flag)	41	47
Capacity at risk (MW)	681.6	784.7
Share of baseline capacity (%)	5.11	5.89

Prioritization (default rule): 10 farms flagged Priority_Save = 1 cover ~50.9% of the weighted risk score across all at-risk sites (47 farms with positive score). See Prioritization method below.

---

QGIS maps — SLR inundation & wind farm locations

Wind farm points overlaid on the DEM-derived inundation raster for the 1 m and 3 m sea-level rise scenarios (QGIS screenshots). The North Sea coast is the most ar risk zone out of the entire AOI, especially the coastal areas in what is knows as The Wash, the most prominent shallow bay on the east coast of England, described as the largest natural bay in the country.

1 m SLR	3 m SLR

---

Figures

Capacity under 1 m and 3 m scenarios

Percent of baseline capacity classified as affected:

Concentration of risk across farms

Do a few large farms dominate losses? Top-1 / top-3 / top-5 shares of affected capacity by scenario:

Lorenz-style curves (cumulative share of affected farms vs. cumulative share of affected capacity):

---

Repository layout

Path	Role
raw_data/	Original downloads (CSVs, shapefiles, GeoJSON, etc.)
processed_data/	Cleaned tables, QGIS exports, priority outputs
preprocessing/	Small ETL scripts (wind filter, UK offshore filter)
plots/	Summary CSVs and PNG figures from analysis scripts
Root *.py	DEM fetch, capacity tools, plotting, prioritization

---

Data

What we use

Dataset	Location (typical)	Role
UK renewable / wind sites	raw_data/renewable_power_plants_UK.csv	BEIS-style register; filtered to onshore wind → processed_data/land_wind_farms.csv
UK administrative / country boundaries	raw_data/UK_Boundries.geojson	AOI for DEM clipping (large file; may be gitignored locally, provided as zip)
EMODnet offshore wind (UK subset)	raw_data/EMODnet_.../ + preprocessing/filter_uk_only.py	Optional offshore UK wind farms → GeoPackage
SLR-affected onshore export	processed_data/Affected_Farms.csv	Wind farms + binary Affected_1m, Affected_2m, Affected_3m from QGIS
Priority-labeled export	processed_data/Affected_Farms_priority.csv	Full table + Priority_Score, Priority_Save
Priority-only table	processed_data/Priority_Farms_only.csv	Rows with Priority_Save == 1 only
DEM (example)	processed_data/UK_DEM_AWS_GLO30.tif (run get_uk_dem_aws.py script)	Copernicus GLO-30 mosaic clipped to UK AOI

Sources & attribution

• UK renewable installations (BEIS register)
  Public tabular data on renewable generating capacity (your copy in raw_data/renewable_power_plants_UK.csv). From Open Power System Data (OPSD).

• Copernicus Digital Elevation Model (GLO-30)
  Global 30 m DSM product. This project’s recommended scripted access uses the Earth Search STAC API (https://earth-search.aws.element84.com/v1, collection cop-dem-glo-30) and public https://copernicus-dem-30m.s3.amazonaws.com/... COGs.

• EMODnet Human Activities — wind farms

• UK boundaries
  The UK_Boundries.geojson file uses data from the UK Office for National Statistics (ONS) Open Geography Portal.
  Original source: ONS Open Geography Portal — Countries (December 2022) Generalised Clipped Boundaries UK BGC

---

Methodology (conceptual)

1. Build a consistent onshore wind table
   preprocessing/filter_wind.py reads the renewables CSV (with a robust CSV parser for occasional malformed rows), keeps Wind + Onshore, and writes processed_data/land_wind_farms.csv.

2. GIS overlay (QGIS)
   Combined the wind layer with SLR or inundation surfaces (different scenario layers). The export Affected_Farms.csv retains attributes such as electrical_capacity and adds binary affected columns per scenario.

3. Capacity accounting

   • Scenario totals: plot_capacity_by_slr.py sums electrical_capacity for affected vs. remaining rows for Affected_1m and Affected_3m (2 m omitted per researcher choice).
   • Multi-column tooling: calculate_available_capacity.py and compare_capacity_scenarios.py support single-flag or many-flag CSVs from QGIS.

4. Concentration / prioritization narrative
   analyze_priority_farms.py ranks affected farms by capacity, computes top-1/3/5 shares of affected MW, plots Lorenz curves and top-N bars, and applies a simple heuristic: if the top 3 farms account for ≥ 50% of affected capacity, site-specific protection is emphasized; otherwise broader portfolio measures are suggested.

---

Prioritization method

Script: prioritize_farms.py

Score

For each farm i, the script computes:

Priority_Score(i) = electrical_capacity(i) × ( w1 × Affected_1m(i) + w3 × Affected_3m(i) )

Defaults: w1 = 0.6, w3 = 0.4 (slightly higher weight on the 1 m flag).

• Farms with no exposure in both scenarios have score 0 and are never prioritized.

Binary label Priority_Save

1. Sort all farms with score > 0 by score descending.
2. Walk down the list and set Priority_Save = 1 until the cumulative sum of scores reaches a target fraction of the total score across all at-risk farms.
3. Default target: --cover-share 0.50 (50% of total weighted risk).

This is a greedy coverage rule: it finds a small set of high-score farms that explain about half of the weighted risk, which is useful for “protect first” discussions.

Coverage (example run on this repo’s Affected_Farms.csv)

Quantity	Value
Farms with Priority_Save = 1	10 / 737
Farms with any positive score	47
Share of total weighted score covered by priority set	~50.9%

Outputs:

• processed_data/Affected_Farms_priority.csv — full table + Priority_Score, Priority_Save
• processed_data/Priority_Farms_only.csv — only prioritized farms

Tune with:

python prioritize_farms.py --cover-share 0.40 --w1 0.6 --w3 0.4

---

Digital elevation model (DEM)

Recommended: get_uk_dem_aws.py — Copernicus GLO-30 via Earth Search STAC and public S3-hosted COGs (avoids Copernicus Data Space WAF rate limits for metadata-heavy STAC use).

.venv/bin/python get_uk_dem_aws.py --countries "England,Scotland,Wales,Northern Ireland" \
  --output processed_data/UK_DEM_AWS_GLO30.tif

Elevation values are in metres (vertical datum per Copernicus DEM documentation; commonly discussed relative to EGM2008). Horizontal CRS is typically geographic (e.g. WGS84 / EPSG:4326) depending on your clip pipeline.

---

Environment

Use a virtual environment (project convention):

python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install pandas geopandas matplotlib pystac-client rioxarray pyogrio shapely
# stackstac / xarray if using CDSE stackstac workflow

---

Common commands

Task	Command
Filter onshore wind	python preprocessing/filter_wind.py
UK offshore windfarms (EMODnet)	python preprocessing/filter_uk_only.py
Capacity for one flag column	python calculate_available_capacity.py processed_data/Affected_Farms.csv --affected-col Affected_1m
Many scenario columns	python compare_capacity_scenarios.py processed_data/your_export.csv --prefix Affected_
SLR capacity plots (1 m & 3 m)	python plot_capacity_by_slr.py
Concentration + Lorenz / top-N plots	python analyze_priority_farms.py --csv processed_data/Affected_Farms.csv --outdir plots
Priority labels	python prioritize_farms.py

---

Outputs cheat sheet (plots/ and processed_data/)

File	Description
plots/capacity_summary_by_slr.csv	Lost / remaining MW and % for 1 m & 3 m
plots/capacity_remaining_vs_lost_mw.png	Stacked bar chart
plots/capacity_loss_percent.png	Percent loss by scenario
plots/priority_analysis_summary.csv	Concentration metrics + heuristic recommendation
plots/priority_farms_Affected_*.csv	Ranked affected farms per scenario
plots/lorenz_capacity_*.png	Concentration curves
plots/topn_share_by_scenario.png	Top-1/3/5 shares

---

License

See LICENSE (Apache 2.0).