Canary Islands KGPV at native climate pixel spacing#

The Canary Islands gallery maps all eight main islands and nearby islets, with individual island panels, Tenerife close-ups, sample monthly inputs and seven downloadable native GeoTIFFs. It applies the existing energy_methods.kgpv.classify_kgpv thresholds to public climate and solar rasters. 304,733,809 native land pixels are classified.

What the resolution means#

The example preserves the finest grid present in the acquired CanaryClim files; it does not create an arbitrarily finer grid. Pixel spacing and source accuracy are different quantities:

Component

Source grid or evidence

Treatment

Monthly temperature and rainfall

CHELSACanaryClim v1.0, 2025 COG release: 0.00004684877106756975°, about 5.2 m north-south by 4.6 m east-west

Every native pixel centre is retained

Published climate-model comparison

Patiño et al. (2023) evaluated 100 m resolution

The finer COG spacing does not establish 5 m climate accuracy

Annual GHI

Global Solar Atlas 2.0, 9 arcseconds / 0.0025°, nominally 250 m

Nearest source-cell value; no invented finer solar information

Web figures

Bounded nearest-neighbour reads and categorical overviews

Display reduction only; downloadable TIFFs retain native results

Outlines

Native climate-input footprint

Raster coverage contour, not a surveyed or administrative coastline

The full results occupy approximately 2.7 MB with lossless compression. Their small file size reflects large contiguous categorical regions, not a reduction of the native grid. A fine climate boundary within a solar pixel can change KGPV class, but does not imply an independent solar observation there.

Periods, units and missing data#

Climate covers 1979–2013; the downloaded 2019 Spain solar archive covers 1994–2018. This is an exploratory mixed-period application, not a common-baseline climate map or reproduction of the published worldwide map. The application calls the cellwise classifier and records both periods. It does not bypass or weaken map_kgpv’s requirement for a shared climatological baseline.

Temperature and precipitation are stored as encoded unsigned integers. The reader masks raw nodata first, applies each band’s scale and offset, converts temperature from K to °C, and reads precipitation in kg/m²/month as mm/month. The solar file already contains annual GHI in kWh/m²/year; multiplying by 365 again would be incorrect. Source months, variables, units and aligned grids are validated before processing. No thresholds are refitted to the Canaries.

Gray indicates land without all required inputs. The source temperature footprint contains 310,567,549 pixels, including 5,833,740 unclassified pixels. Checks found missing summer rainfall values among these gaps; no gap filling or coastal extrapolation is applied. Pixels outside the source footprint are transparent over the ocean background. The example is a climate classification, not a site-specific yield, shading, soiling or snow-loss prediction.

Recreate the native maps#

From the repository root:

# Inspect the download size without fetching data.
uv run python scripts/prepare_kgpv_canary.py --plan-only

# Fetch 24 monthly climate COGs and one solar archive, then process all islands.
uv run python scripts/prepare_kgpv_canary.py

# Reuse the cache and process only Tenerife, with smaller working tiles.
uv run python scripts/prepare_kgpv_canary.py --islands tenerife --tile-size 256

Initial downloads total approximately 1.92 GB and need no account. The raw cache is artifacts/canary-kgpv/raw; outputs default to artifacts/canary-kgpv/native. --cache and --output select other locations. All 24 climate files are still required for an island subset. The processor reads small tiles rather than materializing the roughly four-billion-cell climate source grid. Two downloads can run concurrently; --workers 1 selects sequential acquisition. In a proxy-dependent environment, explicitly pass --use-environment-proxy to use HTTPS_PROXY.

Allowed --islands values are el_hierro, la_palma, la_gomera, tenerife, gran_canaria, fuerteventura (including Lobos), and lanzarote (including La Graciosa and the Chinijo islets). Multiple selections can follow the flag. These windows use the source raster mask rather than clipping to a coarser country polygon. The separate country mapping guide covers administrative shapefile contours and selected-country grids.

The supplied provenance pins source SHA-256 hashes, sizes and canonical public URLs. Cache corruption and changed provider bytes fail validation. The solar download follows only the provider’s exact Spain object redirect; temporary signed URLs are not saved. Provider outages or changed archives may require a reviewed source update. Acquisition and analysis tests use offline fixtures.

Reuse the GeoTIFFs#

Open kgpv_canary_<island>.tif in QGIS or Rasterio. Each file is EPSG:4326:

Band

Values

1

KGPV class IDs 1–12 in the existing ZONE_CODES order; 0 unclassified

2

January temperature’s native valid footprint: 1 within coverage, 0 outside

Both bands have nodata 0. Band 2 distinguishes missing land inputs from ocean; it is a source coverage mask, not an independent land survey. Embedded tags record zone codes, both periods, resolution limits and source attribution. Overviews use nearest-neighbour classes. Keep the accompanying kgpv_canary_provenance.json when sharing or adapting the derived rasters.

References and attribution#

The primary scientific reference remains Ascencio Vásquez (2021), doctoral thesis, with the KGPV classification of Ascencio-Vásquez, Brecl and Topič (2019). See the bibliography for the method and datasets.

  • Karger et al. (2025), CHELSACanaryClim-climatologies, v1.0, DOI 10.16904/envidat.692, CC0 1.0.

  • Patiño et al. (2023), Spatial resolution impacts projected plant responses to climate change on topographically complex islands, DOI 10.1111/ddi.13757.

  • Global Solar Atlas 2.0: data obtained from the free web application developed and operated by Solargis s.r.o. on behalf of the World Bank Group, utilizing Solargis data, with funding from ESMAP. Source: globalsolaratlas.info. Solar data use CC BY 4.0 with the provider’s additional mediation/arbitration terms; the complete downloaded license is retained in the provenance.

Changes comprise regional extraction, nearest-cell sampling and derived KGPV classification. Provider endorsement is not implied. This example also uses Rasterio for windowed raster I/O, nearest-neighbour reprojection and categorical overviews.