This project analyzes changes in nighttime light intensity across Chile using NASA’s VIIRS Black Marble monthly product (VNP46A3).
The workflow builds annual mean composites for 2014, 2019, and 2024, standardizes them for comparability, computes pixel-wise changes (deltas) between periods, and summarizes results at the regional (Admin-1) level.
- Jupyter Notebook:
notebooks/01_chile_night_lights_change.ipynb - Web report (HTML): https://crgalleguillos.github.io/chile-night-lights-change/
The HTML version provides comfortable reading and full navigation, and is recommended for a general overview of the project and results.
- Produce annual nighttime-lights composites for 2014, 2019, 2024
- Detect spatial change using:
- Δ(2019–2014) and Δ(2024–2019) (continuous deltas)
- categorical change maps (increase / decrease / stable)
- Rank Chilean regions by:
- mean delta
- % pixels increasing / decreasing / stable
- Data download
- Download monthly VNP46A3 granules for each study year using a Chile bounding box.
- Annual compositing
- Aggregate monthly granules into annual mean rasters (pixel-wise mean; nodata handled; optional quality mask).
- Georeferencing strategy
- Apply a BBox-based fallback georeferencing (WGS84 / EPSG:4326) when needed and validate visually.
- Clipping
- Clip rasters to Chile’s national boundary to remove ocean/irrelevant areas.
- Normalization
- Apply Min–Max normalization (0–1) per year to support visual comparability.
- Change detection
- Compute deltas:
- Δ(2019–2014)
- Δ(2024–2019)
- Use robust visualization scaling (e.g., 99th percentile of |Δ|).
- Compute deltas:
- Categorical change maps
- Convert deltas into:
- Increase (Δ > +0.03)
- Decrease (Δ < −0.03)
- Stable (|Δ| ≤ 0.03)
- Convert deltas into:
- Regional aggregation
- Compute per-region metrics (mean/median/std of Δ and % increase/decrease/stable) and export CSV tables.
- Net change is small at the national scale, with distributions centered near zero, indicating that most pixels exhibit minor variations.
- Localized hotspots of increase/decrease exist and drive the tails of the delta distributions.
- Regional metrics reveal heterogeneous behavior across Chile, with northern regions often ranking among the strongest mean increases, while central macrozones show important signals in the later period.
Note: Because normalization is performed per year, delta values represent change in normalized units rather than absolute radiance units.
- Nighttime lights: NASA VIIRS Black Marble VNP46A3 (monthly)
- Administrative boundaries (regions): GADM v4.1 (Chile Admin-1)
- Language: Python
- Geospatial: GeoPandas, Rasterio, RioXarray, Fiona
- Scientific computing: NumPy, Pandas, Xarray
- Visualization: Matplotlib
- Environment: Conda, Jupyter Notebook
- Version control: Git & GitHub
Cristián Andrés Galleguillos Vega
Data Scientist | Biologist | MSc in Data Science & Big Data | MSc in Natural Resources Engineering
This project demonstrates an end-to-end, reproducible geospatial pipeline to quantify and communicate nighttime-light change across Chile.
By combining pixel-level deltas, categorical interpretation, and region-level aggregation, the notebook provides an interpretable baseline for future extensions such as city-level analysis, cross-year scaling strategies, and integration with socioeconomic or infrastructure datasets.