GeoDevelopment Observatory of Bolivia / Indicators
Satellite embeddings
What 333 municipalities look like from space, each compressed by a deep-learning model into 64 numbers: features for machine learning.
Sixty-four satellite embedding dimensions (A00–A63) for 333 of Bolivia’s 343 municipalities in 2017, averaged from the annual 10 m Google Satellite Embedding dataset (AlphaEarth Foundations). The model summarises Sentinel-2, Landsat and other imagery into abstract features that capture land cover, texture, roads and buildings. Two versions are published: emb, the area-weighted mean over each municipality, and embpw, weighted by population so that it describes where people live. The dimensions are not interpretable one by one; they are inputs to models, for example to predict poverty. The values were computed on DS4Bolivia’s polygons of the 2015 municipal map, so they are published for the 333 municipalities whose territory did not change; the six that lost territory to the four units created since (Colquechaca, Mizque, Porco, Tomave, San Ignacio and Santa Ana) and the four new units have no data.
Variables of emb
A00: Embedding A00 (area-weighted). Dimension 0 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A01: Embedding A01 (area-weighted). Dimension 1 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A02: Embedding A02 (area-weighted). Dimension 2 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A03: Embedding A03 (area-weighted). Dimension 3 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A04: Embedding A04 (area-weighted). Dimension 4 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A05: Embedding A05 (area-weighted). Dimension 5 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A06: Embedding A06 (area-weighted). Dimension 6 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A07: Embedding A07 (area-weighted). Dimension 7 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A08: Embedding A08 (area-weighted). Dimension 8 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A09: Embedding A09 (area-weighted). Dimension 9 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A10: Embedding A10 (area-weighted). Dimension 10 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A11: Embedding A11 (area-weighted). Dimension 11 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A12: Embedding A12 (area-weighted). Dimension 12 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A13: Embedding A13 (area-weighted). Dimension 13 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A14: Embedding A14 (area-weighted). Dimension 14 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A15: Embedding A15 (area-weighted). Dimension 15 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A16: Embedding A16 (area-weighted). Dimension 16 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A17: Embedding A17 (area-weighted). Dimension 17 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A18: Embedding A18 (area-weighted). Dimension 18 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A19: Embedding A19 (area-weighted). Dimension 19 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A20: Embedding A20 (area-weighted). Dimension 20 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A21: Embedding A21 (area-weighted). Dimension 21 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A22: Embedding A22 (area-weighted). Dimension 22 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A23: Embedding A23 (area-weighted). Dimension 23 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A24: Embedding A24 (area-weighted). Dimension 24 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A25: Embedding A25 (area-weighted). Dimension 25 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A26: Embedding A26 (area-weighted). Dimension 26 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A27: Embedding A27 (area-weighted). Dimension 27 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A28: Embedding A28 (area-weighted). Dimension 28 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A29: Embedding A29 (area-weighted). Dimension 29 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A30: Embedding A30 (area-weighted). Dimension 30 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A31: Embedding A31 (area-weighted). Dimension 31 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A32: Embedding A32 (area-weighted). Dimension 32 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A33: Embedding A33 (area-weighted). Dimension 33 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A34: Embedding A34 (area-weighted). Dimension 34 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A35: Embedding A35 (area-weighted). Dimension 35 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A36: Embedding A36 (area-weighted). Dimension 36 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A37: Embedding A37 (area-weighted). Dimension 37 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A38: Embedding A38 (area-weighted). Dimension 38 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A39: Embedding A39 (area-weighted). Dimension 39 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A40: Embedding A40 (area-weighted). Dimension 40 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A41: Embedding A41 (area-weighted). Dimension 41 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A42: Embedding A42 (area-weighted). Dimension 42 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A43: Embedding A43 (area-weighted). Dimension 43 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A44: Embedding A44 (area-weighted). Dimension 44 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A45: Embedding A45 (area-weighted). Dimension 45 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A46: Embedding A46 (area-weighted). Dimension 46 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A47: Embedding A47 (area-weighted). Dimension 47 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A48: Embedding A48 (area-weighted). Dimension 48 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A49: Embedding A49 (area-weighted). Dimension 49 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A50: Embedding A50 (area-weighted). Dimension 50 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A51: Embedding A51 (area-weighted). Dimension 51 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A52: Embedding A52 (area-weighted). Dimension 52 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A53: Embedding A53 (area-weighted). Dimension 53 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A54: Embedding A54 (area-weighted). Dimension 54 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A55: Embedding A55 (area-weighted). Dimension 55 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A56: Embedding A56 (area-weighted). Dimension 56 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A57: Embedding A57 (area-weighted). Dimension 57 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A58: Embedding A58 (area-weighted). Dimension 58 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A59: Embedding A59 (area-weighted). Dimension 59 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A60: Embedding A60 (area-weighted). Dimension 60 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A61: Embedding A61 (area-weighted). Dimension 61 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A62: Embedding A62 (area-weighted). Dimension 62 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).A63: Embedding A63 (area-weighted). Dimension 63 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with Earth Engine's mean reducer at 10 m, which weights each pixel by its area of overlap. An abstract feature without a physical unit: together the 64 dimensions summarise the year's multi-sensor appearance of the land surface (Sentinel-2, Landsat, radar and other inputs).
Sources
- Brown, C. F., Kazmierski, M. R., Pasquarella, V. J., Rucklidge, W. J., Samsikova, M., Zhang, C., et al. (2025). AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data. arXiv:2507.22291. https://arxiv.org/abs/2507.22291
- Runfola, D., Anderson, A., Baier, H., et al. (2020). geoBoundaries: A global database of political administrative boundaries. PLoS ONE, 15(4), e0231866. https://doi.org/10.1371/journal.pone.0231866
- Mendez, C., Gonzales, E., Leoni, P., Andersen, L., & Peralta, H. (2026). DS4Bolivia: A Data Science Repository to Study GeoSpatial Development in Bolivia [Data set]. GitHub. https://github.com/quarcs-lab/ds4bolivia
Variables of embpw
A00: Embedding A00 (population-weighted). Dimension 0 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A01: Embedding A01 (population-weighted). Dimension 1 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A02: Embedding A02 (population-weighted). Dimension 2 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A03: Embedding A03 (population-weighted). Dimension 3 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A04: Embedding A04 (population-weighted). Dimension 4 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A05: Embedding A05 (population-weighted). Dimension 5 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A06: Embedding A06 (population-weighted). Dimension 6 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A07: Embedding A07 (population-weighted). Dimension 7 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A08: Embedding A08 (population-weighted). Dimension 8 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A09: Embedding A09 (population-weighted). Dimension 9 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A10: Embedding A10 (population-weighted). Dimension 10 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A11: Embedding A11 (population-weighted). Dimension 11 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A12: Embedding A12 (population-weighted). Dimension 12 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A13: Embedding A13 (population-weighted). Dimension 13 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A14: Embedding A14 (population-weighted). Dimension 14 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A15: Embedding A15 (population-weighted). Dimension 15 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A16: Embedding A16 (population-weighted). Dimension 16 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A17: Embedding A17 (population-weighted). Dimension 17 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A18: Embedding A18 (population-weighted). Dimension 18 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A19: Embedding A19 (population-weighted). Dimension 19 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A20: Embedding A20 (population-weighted). Dimension 20 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A21: Embedding A21 (population-weighted). Dimension 21 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A22: Embedding A22 (population-weighted). Dimension 22 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A23: Embedding A23 (population-weighted). Dimension 23 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A24: Embedding A24 (population-weighted). Dimension 24 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A25: Embedding A25 (population-weighted). Dimension 25 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A26: Embedding A26 (population-weighted). Dimension 26 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A27: Embedding A27 (population-weighted). Dimension 27 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A28: Embedding A28 (population-weighted). Dimension 28 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A29: Embedding A29 (population-weighted). Dimension 29 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A30: Embedding A30 (population-weighted). Dimension 30 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A31: Embedding A31 (population-weighted). Dimension 31 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A32: Embedding A32 (population-weighted). Dimension 32 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A33: Embedding A33 (population-weighted). Dimension 33 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A34: Embedding A34 (population-weighted). Dimension 34 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A35: Embedding A35 (population-weighted). Dimension 35 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A36: Embedding A36 (population-weighted). Dimension 36 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A37: Embedding A37 (population-weighted). Dimension 37 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A38: Embedding A38 (population-weighted). Dimension 38 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A39: Embedding A39 (population-weighted). Dimension 39 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A40: Embedding A40 (population-weighted). Dimension 40 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A41: Embedding A41 (population-weighted). Dimension 41 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A42: Embedding A42 (population-weighted). Dimension 42 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A43: Embedding A43 (population-weighted). Dimension 43 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A44: Embedding A44 (population-weighted). Dimension 44 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A45: Embedding A45 (population-weighted). Dimension 45 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A46: Embedding A46 (population-weighted). Dimension 46 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A47: Embedding A47 (population-weighted). Dimension 47 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A48: Embedding A48 (population-weighted). Dimension 48 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A49: Embedding A49 (population-weighted). Dimension 49 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A50: Embedding A50 (population-weighted). Dimension 50 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A51: Embedding A51 (population-weighted). Dimension 51 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A52: Embedding A52 (population-weighted). Dimension 52 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A53: Embedding A53 (population-weighted). Dimension 53 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A54: Embedding A54 (population-weighted). Dimension 54 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A55: Embedding A55 (population-weighted). Dimension 55 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A56: Embedding A56 (population-weighted). Dimension 56 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A57: Embedding A57 (population-weighted). Dimension 57 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A58: Embedding A58 (population-weighted). Dimension 58 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A59: Embedding A59 (population-weighted). Dimension 59 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A60: Embedding A60 (population-weighted). Dimension 60 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A61: Embedding A61 (population-weighted). Dimension 61 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A62: Embedding A62 (population-weighted). Dimension 62 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.A63: Embedding A63 (population-weighted). Dimension 63 of the 64-dimensional Google Satellite Embedding V1 (AlphaEarth Foundations), annual 10 m embeddings for 2017, averaged over the municipality with each 10 m pixel weighted by its population (WorldPop 100 m, 2017, resampled to 10 m): Σ(embedding × population) / Σ population. It describes the places where people live. An abstract feature without a physical unit.
Sources
- Brown, C. F., Kazmierski, M. R., Pasquarella, V. J., Rucklidge, W. J., Samsikova, M., Zhang, C., et al. (2025). AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data. arXiv:2507.22291. https://arxiv.org/abs/2507.22291
- WorldPop & CIESIN (2018). Global High Resolution Population Denominators Project (100 m, 2000–2020). University of Southampton and Columbia University. Funded by the Bill and Melinda Gates Foundation (OPP1134076). https://doi.org/10.5258/SOTON/WP00645
- Runfola, D., Anderson, A., Baier, H., et al. (2020). geoBoundaries: A global database of political administrative boundaries. PLoS ONE, 15(4), e0231866. https://doi.org/10.1371/journal.pone.0231866
- Mendez, C., Gonzales, E., Leoni, P., Andersen, L., & Peralta, H. (2026). DS4Bolivia: A Data Science Repository to Study GeoSpatial Development in Bolivia [Data set]. GitHub. https://github.com/quarcs-lab/ds4bolivia