---
title: AlphaEarth Foundations Dataset
url: https://www.emergentmind.com/topics/alphaearth-dataset
type: topic
---

# AlphaEarth Foundations Dataset

AlphaEArth Dataset, more commonly formalized as the annual embedding-field release of **AlphaEarth Foundations**, is a global geospatial representation in which each terrestrial 10 m pixel is encoded as a compact 64-dimensional annual vector summarizing multi-source Earth observation and contextual information over a chosen valid period [2507.22291]. In operational usage, the term refers both to the foundation model and to the analysis-ready embedding product distributed for downstream use—most prominently through Google Earth Engine as **“Satellite embedding v1 (annual)”**—and consumed as fixed features for mapping, forecasting, regression, segmentation, retrieval, and transfer-learning pipelines [2508.11739].

## 1. Naming, scope, and released product

The literature uses both **“AlphaEarth”** and **“AlphaEArth”** spellings, but papers that discuss the public release identify the formal dataset and Earth Engine asset with the **AlphaEarth** spelling [2509.07852]. A recurring distinction is between **AlphaEarth Foundations** as the upstream geospatial foundation model and the **public embedding dataset** as its practical output. Several downstream studies explicitly note that most users interact with the precomputed embeddings rather than the underlying model weights [2601.01558].

| Aspect | Reported specification |
|---|---|
| Formal release | AlphaEarth Foundations / “Satellite embedding v1 (annual)” |
| Public access path | Google Earth Engine, `GOOGLE_SATELLITE_EMBEDDING_V1_ANNUAL` |
| Spatial coverage | Global terrestrial coverage, including minor islands, approximately between $\pm 82^\circ$ latitude |
| Spatial resolution | 10 m |
| Representation size | 64 channels per pixel per year |
| Public cadence in release paper | Annual layers for 2017–2024 |

The release paper describes a near-global terrestrial product with annual layers from **2017 through 2024** [2507.22291]. Some downstream studies report adjacent distribution variants, including a **Source Cooperative** archive with annual data for **2018–2024** and application papers referring to availability through **2025** [2602.17250][2606.20034]. This suggests distribution-specific versioning or later catalog extensions, rather than a single immutable packaging.

At the dataset level, AlphaEarth is presented as an **analysis-ready** representation intended to reduce dependence on bespoke per-task preprocessing. Multiple application papers adopt the embeddings as off-the-shelf inputs and emphasize that the downstream workflow often consists primarily of spatial extraction, aggregation, and a lightweight predictor rather than end-to-end retraining of the foundation model [2605.21804].

## 2. Representation, temporal semantics, and source modalities

The core object is an embedding field $\mu(x,t_s,t_e)$ defined over space and a user-specified **valid period** $[t_s,t_e)$ [2507.22291]. Each embedding is a **64-channel** vector constrained to the unit sphere $S^{63}$, intended to summarize the temporal evolution of multiple observations and contextual signals over that valid period. Public annual layers instantiate this mechanism at yearly cadence.

AlphaEarth assimilates a broad set of modalities during training. The release paper lists the following sources: **Sentinel-2 L1C**, **Landsat-8/9 C2 Tier 1 TOA**, **Sentinel-1 GRD**, **ALOS-2/PALSAR-2 ScanSAR**, **GEDI LiDAR RH metrics**, **Copernicus DEM GLO-30**, **ERA5-Land monthly aggregates**, **GRACE mascons**, **NLCD land cover**, **geocoded Wikipedia articles**, and **GBIF taxa** [2507.22291]. Downstream papers typically summarize these as multi-source optical, radar, LiDAR/topography, climate, hydrology, and text-associated context [2601.00857].

The public dataset is annual, but the model itself is described as **continuous-time conditioned**. This matters because some applications treat the released layers as annual composites, whereas the underlying model formulation supports valid-period summaries beyond the annual public product. The release paper notes that some evaluations used **monthly valid periods** for evapotranspiration, even though the public dataset is annual [2507.22291]. By contrast, application papers repeatedly note that the publicly consumed embeddings are **annual summaries rather than near-real-time streams**, which constrains tasks requiring fine intra-annual timing [2605.21804].

The release also specifies a quantized public representation. Embeddings are stored as signed **int8** values with power-law companding and dequantized via
$$
s = \frac{y}{127.5}, \qquad x = \operatorname{sign}(s)\,|s|^2
$$
after retrieval [2507.22291]. One height-mapping study reports that a **Source Cooperative** distribution stores annual global embeddings as signed 8-bit integers in the range $[-128,127]$ and recommends a normalization to $[0,1]$ for local processing, while describing the Earth Engine catalog as accessible in double precision [2602.17250]. Taken together, these reports indicate that the logical embedding field is stable, but serving formats can differ across distribution channels.

## 3. Model architecture and training objective

AlphaEarth Foundations is built around a **Space–Time–Precision (STP) encoder** with **15 STP blocks**, each combining three parallel operators and learned Laplacian-pyramid exchanges [2507.22291]. The three paths are a **space path** with ViT-like spatial self-attention at $1/16$ input size, a **time path** with axial self-attention conditioned on sinusoidal time codes, and a **precision path** with $3\times3$ convolutions at $1/2$ input size. Final features are projected through a **von Mises–Fisher bottleneck** with fixed $\kappa = 8\times 10^3$, and per-source **implicit decoders** reconstruct sensor-specific observations from the embedding, metadata, and normalized time code.

Training uses a weighted combination of reconstruction, uniformity, teacher–student consistency, and text-contrastive terms. The release paper gives the objective as
$$
L_{\text{total}}=
\frac{a}{M}\sum_{i\in M} f_i(y_i,y'_i)w_i
+b\sum_{k=1}^{64}|u_k\cdot u'_k|
+c\left(\frac{1-\mu\cdot\mu_s}{2}\right)
+d\,f_{\text{CLIP}}(\mu,\mu_t),
$$
where $f_i$ is source-specific reconstruction loss, the second term enforces batch-wise uniformity on $S^{63}$, the third term is the teacher–student consistency penalty, and the fourth aligns embeddings with text embeddings through a CLIP-style loss [2507.22291].

The training corpus comprises **8,412,511 sequences** from **5,145,244 sites** and **3,047,520,515 frames** across sources, covering approximately **1.1% of Earth’s land surface** [2507.22291]. Two model variants were trained, approximately **1B** and **480M** parameters, with deployment favoring the **480M** model for efficiency. Training used **512 TPU v4**, **100k steps**, and **batch size 256 sequences**.

A central design goal is **generalization without re-training**. The release paper attributes this to continuous-time conditioning, source-specific decoders, robustness to missing or perturbed inputs via teacher–student training, batch-uniformity regularization, and text alignment, allowing the released embeddings to support simple downstream heads such as **kNN** and **linear probes** across heterogeneous tasks [2507.22291].

## 4. Distribution, access, and practical extraction workflows

The canonical public release is the Google Earth Engine collection **`GOOGLE_SATELLITE_EMBEDDING_V1_ANNUAL`** [2507.22291]. The release paper also provides **Zenodo** links for processed evaluation datasets and training-site coordinates, but not downloadable model weights for local re-inference [2507.22291]. A height-mapping study explicitly states that the **AlphaEarth Foundations model weights are not open-source**, even though the embeddings are accessible through public distribution channels [2602.17250].

Downstream usage is typically extraction rather than retraining. Studies aggregate embeddings to the task unit by **zonal means**, **polygon clipping**, **basin averaging**, or **buffer-based sampling**. Examples include averaging embeddings over field polygons for agricultural mapping [2605.21804], averaging across basin masks and across years 2017–2024 for hydrological descriptors [2601.01558], extracting 10 m pixels within **2–5 km** DHS buffers for health modeling [2605.10949], and consuming embeddings directly on a **25×25** quadrature grid for spatio-temporal point-process forecasting [2607.01082].

Preprocessing is notably heterogeneous and task-dependent. Some papers state that **no additional normalization** was required because AlphaEarth bands were already bounded or analysis-ready [2605.21804]; others apply **per-dimension standardization** before cosine similarity or predictive modeling [2601.01558]. Spatial harmonization is common: one landslide study resampled AlphaEarth rasters to **30 m × 30 m** to match conventional conditioning factors [2601.07268], while a subsurface study intentionally sampled the product at an effective **1 km** scale in Earth Engine to stabilize centroid-level features [2604.14756].

This use pattern has made AlphaEarth functionally similar to a large, annual geospatial feature store. Application papers repeatedly describe the practical pipeline as: query annual embeddings for the target year, aggregate or align to the study geometry, and fit a comparatively lightweight downstream model [2601.00857].

## 5. Downstream applications and empirical behavior

AlphaEarth embeddings have been applied across a wide spectrum of tasks, from classical land-cover mapping to point-process forecasting, subsurface inference, hydrology, public health, and urban analytics. Reported outcomes are highly task-specific, but several studies provide precise quantitative evaluations.

| Domain | Configuration | Reported outcome |
|---|---|---|
| Spatio-temporal EMS forecasting | LGCP with AlphaEarth as linear spatial context across eight held-out regions | Mean multiplicative density ratio **5.42×** at one week and **2.37×** at two weeks; gains remain positive at longer histories [2607.01082] |
| Tomato cropping systems | U-Net on annual 64-band chips from 9,484 California fields | **99.19%** pixel accuracy, **99.04%** F1, **98.11%** IoU, **99.02%** chip accuracy [2605.21804] |
| Prediction in ungauged basins | LSTM with AlphaEarth basin descriptors on CAMELS-US | OOS median **NSE = 0.612** vs **0.553** for CAMELS attributes; donor-basin selection peaks at **0.75** for selected basins at $k=200$ [2601.01558] |
| Urban indicator prediction | 64-d embeddings across six U.S. metropolitan areas, 2020–2023 | Global $R^2$ exemplars: drive alone **0.74**, transit **0.72**, obesity **0.69**, log violent crime **0.48** [2604.03456] |
| Landslide susceptibility | CNN1D/CNN2D/ViT against conventional conditioning factors | F1-score gains of approximately **4%–15%** and **AUC** increases of about **0.04–0.11** [2601.07268] |
| Agricultural benchmarking | Yield, tillage, and cover-crop tasks in the U.S. | Strong local performance, but weaker spatial transfer; soybean US$\rightarrow$Argentina AEF overall $R^2 \approx -2.13$ vs RS baselines $\approx 0.27$–$0.30$ [2601.00857] |

A clear empirical pattern is that AlphaEarth is particularly effective when downstream tasks benefit from **persistent landscape context** and when labels are sparse or costly. The emergency-medical-services forecasting study is explicit on this point: with only **1–2 weeks** of history, adding AlphaEarth to a fixed **log-Gaussian Cox process** produced approximately **2–6×** multiplicative predictive-density gains in held-out regions [2607.01082]. Hydrology and public-health studies report analogous benefits when embeddings are used as static or annually indexed descriptors of basin or cluster context [2601.01558][2605.10949].

At the same time, several benchmarks show that the dataset’s utility is not uniform. In agriculture, AlphaEarth-based models were competitive with purpose-built remote-sensing features for **local training**, but exhibited **limited spatial transferability**, especially across ecoregions and countries [2601.00857]. In fine-scale Local Climate Zone mapping, AlphaEarth simplified preprocessing and achieved strong performance, yet was consistently below **TESSERA** and showed larger sensitivity to cross-year transfer [2606.20034].

## 6. Interpretability, geometry, and recurring limitations

A substantial secondary literature analyzes the internal structure of AlphaEarth embeddings rather than using them only as predictors. One interpretability study over **12.1 million CONUS samples** reports that the full 64-dimensional space can reconstruct many environmental variables with high fidelity: **12 of 26 variables exceed $R^2>0.90$**, and temperature and elevation approach **$R^2=0.97$**; the same study finds a mean spatial block-cross-validation gap of **$\Delta R^2 = 0.017$** and mean inter-year correlation **$r=0.963$** for dimension–variable profiles [2602.10354]. This establishes that AlphaEarth dimensions are not arbitrary numerical codes, even if many applications treat them as latent and uninterpreted.

A separate geometric analysis characterizes the embedding manifold as strongly **non-Euclidean**. It reports global effective dimensionality **PR $\approx 13.3$** from 64 raw dimensions, local intrinsic dimensionality of approximately **10**, weak local–global alignment with mean **$|\cos\theta|=0.17$**, and substantial tangent-space rotation, with **84%** of locations exceeding **$60^\circ$** [2604.18715]. The same study finds that retrieval works substantially better than compositional vector arithmetic, and that local geometry predicts retrieval coherence with **$R^2=0.32$**. This implies that AlphaEarth is more naturally used as a **retrieval space** than as a globally linear concept space.

A third line of work argues that dimensions can be organized along a **functional hierarchy**. In global land-cover experiments, accurate classification at **98% of baseline performance** could be achieved with as few as **2 to 12** of the 64 dimensions, depending on the class, and dimensions were grouped into **specialist**, **low-generalist**, **mid-generalist**, and **high-generalist** categories based on their class associations [2603.16911]. This suggests substantial redundancy in the embedding space and motivates dimension selection for operational inference.

Recurring limitations are equally clear. Across applied papers, the most frequently noted caveats are **annual temporal granularity**, **limited time sensitivity for in-season tasks**, **weak direct semantics of individual channels**, and **variable spatial transferability** [2601.00857][2606.20034][2601.07268]. Some studies also report **tile gaps** or incomplete coverage in particular locations [2606.20034]. The embeddings encode **surface context** rather than direct subsurface or causal variables, so successful inference in domains such as site characterization or geothermal mapping remains indirect and dependent on robust validation design [2604.14756]. A plausible implication is that AlphaEarth is most reliable when used as a high-capacity contextual descriptor that complements, rather than universally replaces, domain-specific covariates and task-aware evaluation protocols.

Source: https://www.emergentmind.com/topics/alphaearth-dataset