The Agricultural Understanding Platform provides agricultural data and insights
for stakeholders across the agricultural sector. This platform lets you identify
landscape features and monitor crop type, and sowing and harvest dates at the
individual field level for targeted decision-making and interventions.

> [!TIP]
> **Tip:** Ready to dive in? Skip straight to [Access modes](https://developers.google.com/agricultural-understanding#access-modes), or see [Basic retrieval](https://developers.google.com/agricultural-understanding/api/complete-examples/basic-retrieval) and run the [Example Colab](https://developers.google.com/static/agricultural-understanding/Example_Colab.ipynb) once you set up your [API key](https://developers.google.com/agricultural-understanding/api#setup).

## Key capabilities

The platform provides the following key capabilities:

- **Agricultural Landscape Understanding (ALU)**: ALU segments the landscape into different elements to identify multiple classes, including individual field boundaries, trees, woodlands, and minor irrigation structures like farm wells and dug ponds.
- **Agricultural Monitoring and Event Detection (AMED)**: For landscape features corresponding to fields, AMED provides crop monitoring history at the feature level, indicating sowing and harvest times as well as the likely cultivated crop.

## Data organization

To work with the data effectively, keep the following concepts in mind:

- **S2 cells:** The platform organizes spatial data into level 13 [S2 cells](https://developers.google.com/agricultural-understanding/understanding-data#data-indexing)
  (approximately 1x1 km grids). The system assigns features to a cell based on
  the feature's centroid location.

- **Data coverage:** Coverage varies by surface, with an initial focus on
  agricultural regions of the Global South. By default, the system filters out
  non-agricultural areas. For data access using the REST API, these areas can
  be [included in
  queries](https://developers.google.com/agricultural-understanding/api/request-format#non-agricultural-areas).

- **Confidence levels:** Confidence scores accompany feature geometries
  and crop predictions. Learn more in the data understanding guides for
  [ALU](https://developers.google.com/agricultural-understanding/understanding-data#alu-confidence) and
  [AMED](https://developers.google.com/agricultural-understanding/understanding-data#amed-confidence).

- **Temporal snapshots:** The system derives field boundaries and crop
  predictions from satellite imagery captured at specific times. You can query
  historical captures to observe year-over-year changes. See [Retrieve
  historical
  captures](https://developers.google.com/agricultural-understanding/api/complete-examples/retrieve-historical-captures)
  for more details.

## Choose your access mode

You can interact with Agricultural Understanding data through more than one
platform. Select the modality that best fits your workflow:

- **[REST API](https://developers.google.com/agricultural-understanding/api#setup)**: Best for low-latency, rapid point-lookups for live mobile or web applications.
- **[Google Earth](https://developers.google.com/agricultural-understanding/earth#setup)**: Best for immediate visual context, inspection, and spatial overlays without writing code.

## Try it now

For an interactive example of querying and visualizing data using the REST API,
see the [Example Colab](https://developers.google.com/static/agricultural-understanding/Example_Colab.ipynb).

> [!NOTE]
> **Note:** You must [set up an API key](https://developers.google.com/agricultural-understanding/api#setup) before running the Colab.

## Research and methodology

To learn more about the models and methodology powering the platform:

- For details on agricultural landscape understanding and panoptic segmentation, read the [ALU research
  paper](https://arxiv.org/abs/2411.05359).
- For details on crop monitoring and prediction approaches, read the [AMED
  research paper](https://arxiv.org/abs/2507.02972).