Contiguous Chile KGPV at the native solar grid spacing#

The mainland Chile gallery provides regional maps, Atacama and central-valley zooms, a downloadable GeoTIFF, GeoJSON/Shapefile contours, sample input values and source provenance. It uses the repository’s existing KGPV classifier with fixed literature thresholds. 7,848,061 mainland pixels are classified; 2,621,175 remain unclassified.

Resolution and coverage#

The output preserves the 9-arcsecond (0.0025°, nominal 250 m) solar lattice. This is the finest verified, openly downloadable GHI grid adopted for this example, not a claim that no finer dataset can exist. No arbitrarily fine grid is created. Unlike the Canaries, Chile does not use CanaryClim: the regional Canary climate product does not cover Chile.

Component

Source and period

Native support and treatment

Monthly mean temperature

CHELSA v2.1, 1981–2010

30 arcseconds, about 1 km; nearest original cell

Monthly precipitation total

CHELSA v2.1, 1981–2010

30 arcseconds, about 1 km; nearest original cell

Annual horizontal GHI

Global Solar Atlas 2.0, 2019 Chile archive, 1999–2018

9 arcseconds, nominal 250 m; original pixel centres retained

Country contour

Natural Earth Admin 0, 1:10 million

Largest connected Chile polygon; vertices and holes retained

Display

Bounded categorical reads and overviews

Nearest-neighbour only; downloadable pixels are unchanged

In geographic coordinates, 9 arcseconds is about 278 m north–south and 164–265 m east–west across mainland Chile. The climate grid is roughly 927 m north–south. 250 m output spacing does not establish 250 m climate or KGPV accuracy. Natural Earth’s 10m denotes 1:10 million, not ten-metre accuracy.

Usable GSA Chile values stop at 45°S, although the file rectangle extends to 46°S (its southernmost valid pixel centre is 44.99875°S). The map includes the entire connected mainland outline to approximately 53.89°S, but land south of 45°S is unclassified. Coastal and other source-data gaps also remain gray. The example therefore does not provide complete classified mainland coverage. Missing land stays in the denominator of the area chart; it is never hidden as ocean. The source raster’s rectangle extends west to Easter Island, but all islands are excluded by the mainland mask.

Two potential replacements were investigated but not substituted: CHELSA rsds includes terrain-slope/aspect effects and is not interchangeable with horizontal GHI; the Explorador Solar SIG download’s published kWh/m² label could not be reconciled with the retrieved raster’s unlabelled values. Neither is silently blended into the southern gap. Higher-resolution or more complete results need verified horizontal irradiation and climate inputs, with units, periods, licensing and support recorded explicitly.

What “contiguous” means#

The analysis selects Chile (CHL) from the source Shapefile, then chooses its largest connected polygon by WGS84 geodesic area. This includes the connected mainland and Brunswick Peninsula. It excludes Tierra del Fuego, Chiloé, the coastal archipelagos, Easter Island, Juan Fernández and Antarctica. No polygon simplification or bridge across water is applied. Holes remain excluded.

Output cells are included when their centres lie inside that polygon. Coastal cells can therefore partly overlap water; the contour is not a surveyed coast. Area summaries sum WGS84 ellipsoidal areas of these selected whole cells, not fractional polygon intersections. This is more appropriate than comparing raw pixel counts across Chile’s large latitude range.

Scientific interpretation#

Climate covers 1981–2010 and GHI 1999–2018. These are different historical periods. This exploratory application is not a common-baseline climatology, future scenario or exact reproduction of the thesis map. It calls the existing cellwise classifier without weakening map_kgpv’s shared-baseline requirement.

Encoded CHELSA nodata (65535) is masked before scale/offset conversion. Temperature is converted from kelvin to Celsius; precipitation in kg/m²/month is equivalent to millimetres per month. The pinned GSA XML identifies annual GHI in kWh/m²/year; do not multiply it by 365. Southern Hemisphere seasons and the existing KGPV threshold/equality conventions are unchanged. There is no refitting of irradiation percentiles to Chile and no interpolation or filling of missing inputs.

Downscaled climatologies, mixed periods and cartographic masks can introduce errors, particularly along coastlines, steep terrain and classification thresholds. These maps support regional exploration, not site-level yield or loss predictions. Most repository development is AI assisted; the cited, developer-selected literature remains the basis for the method, while code and derived results still require scientific review.

Offline example and downloads#

uv sync --locked --all-packages --group docs
cd examples/analysis
uv run python plot_kgpv_chile.py

The gallery verifies checksums and independently recomputes twelve site samples from the supplied CSV without network access. Coyhaique, Cochrane and the inland Punta Arenas sample retain missing GHI and class zero. Download the three-band GeoTIFF and provenance from the gallery. The raster is EPSG:4326, 14,553 rows by 3,480 columns:

Band

Meaning

1

KGPV class 1–12, with labels in the zone_codes tag; 0 means unclassified/outside

2

Mainland mask: 1 inside, 0 outside; distinguish gray gaps from ocean using this band

3

All required inputs available: 1; otherwise 0

GeoJSON is a single-file contour download. For Shapefile use, retain all five companions (.shp, .shx, .dbf, .prj, .cpg) together. The acquisition command also generates these contours locally.

Reproduce from source data#

From the repository root:

uv sync --locked --all-packages --group docs --group dev
uv run python scripts/prepare_kgpv_chile.py --plan-only
uv run python scripts/prepare_kgpv_chile.py

--plan-only reads the bundled source manifest and makes no HTTP requests. The default cache is artifacts/chile-kgpv/raw; outputs are in artifacts/chile-kgpv/native. --cache, --output, --source-manifest, --tile-size (128–1024) and --workers (1–2) are configurable. If required by your environment, --use-environment-proxy explicitly uses HTTPS_PROXY; TLS certificate verification remains enabled.

CHELSA’s global source rasters are accessed through bounded HTTP ranges, not whole-world downloads. Every request uses the recorded strong ETag with If-Match; a changed object fails closed. The ETag is a version validator, not a SHA-256 checksum. Each regional crop also has a stable SHA-256 of its native little-endian uint16 pixel bytes. Its separate file SHA-256 detects local cache corruption. Original transforms, scales, units and full source sizes are recorded. Crop windows are integral; climate-grid agreement allows only 1e-10° floating-point tolerance, not meaningful spatial displacement.

The solar archive (145.4 MB) and boundary archive (4.9 MB) are pinned by full-file SHA-256. Only exact expected archive members are extracted. GSA’s temporary signed redirect remains in memory; provenance retains the canonical public URL. Raster processing uses bounded tiles and does not load the full 50-million cell output rectangle plus 24 climate bands into memory. Raw inputs stay outside Git; the losslessly compressed categorical TIFF is about 0.55 MB.

For other countries or custom contours, see country mapping. This study’s mainland selection and chosen source pins are Chile-specific.

Sources and attribution#

  • Method: Ascencio Vásquez (2021), doctoral thesis, and Ascencio-Vásquez et al. (2019), KGPV classification. See the method contract and central bibliography.

  • Climate: Karger et al. (2021), CHELSA v2.1 climatologies, data DOI, CC0 1.0; Scientific Data paper. Provider catalogue.

  • Solar: Global Solar Atlas Chile download. Data obtained from the Global Solar Atlas 2.0, a free, web-based application developed and operated by Solargis s.r.o. on behalf of the World Bank Group, utilizing Solargis data, with funding provided by the Energy Sector Management Assistance Program (ESMAP). CC BY 4.0 with additional mediation/arbitration terms, preserved in full in the provenance. Cropping, climate-cell sampling and KGPV derivation are modifications; no provider endorsement is implied.

  • Boundaries: Natural Earth countries, public domain.

  • Excluded radiation alternatives: CHELSA v2.1 radiation model, Section 2.2.3; Explorador Solar SIG.