JAXA Earth API for Python: a simple way to work with satellite data

I found that JAXA has published a Python API for working with Earth observation satellite data, so I looked into it. Satellite data tends to sound like something that requires a lot of specialist knowledge, but with this API you can fetch and process data without having to care much about sensor specs or resolution details.

What Is the JAXA Earth API?

It is a service for using JAXA’s Earth observation data. It is distributed as a Python package and includes the core pieces you need, from fetching satellite imagery to masking, difference calculation, statistical processing, and visualization.

Official documentation: https://data.earth.jaxa.jp/api/python/v0.1.5/en/

Installation

You install it directly from JAXA’s package repository.

pip install --extra-index-url https://data.earth.jaxa.jp/api/python/repository/ jaxa-earth

If you want to pin the version:

pip install --extra-index-url https://data.earth.jaxa.jp/api/python/repository/ jaxa-earth==0.1.5

The dependencies matplotlib and requests are installed automatically.

Main Classes

The API is built around four main classes.

Class	Purpose
`ImageCollectionList`	Fetch and search the list of available satellite image collections
`ImageCollection`	Filter images by date, resolution, area, and band
`FeatureCollection`	Read and extract geographic data in GeoJSON format
`ImageProcess`	Perform masking, difference calculation, statistical analysis, and visualization

The basic workflow is a two-step approach: fetch the data with ImageCollection, then process it with ImageProcess.

Basic Usage

Fetching and displaying images

from jaxa.earth import je

# Get an image collection and filter it
ic = je.ImageCollection("GCOM-C.SGLI.L2.LAND.RSRF")
ic.filter_date("2021-12-01", "2021-12-31")
ic.filter_resolution(1000)
ic.filter_bounds([130, 30, 145, 45])  # around Japan
images = ic.get_images()

# Display the images
ip = je.ImageProcess(images)
ip.show_images()

The methods are chainable, so you can set parameters progressively with filter_date(), filter_resolution(), and filter_bounds().

Ways to specify a geographic area

Bounding box approach

Define a rectangular area with four longitude and latitude values.

ic.filter_bounds([110, 20, 160, 50])

GeoJSON approach

If you need to handle more complex shapes, you can load a GeoJSON file and use that instead.

fc = je.FeatureCollection()
fc.read("path/to/area.geojson")
ic.filter_bounds(fc)

Masking operations

There are three main patterns:

Range extraction: use elevation data to mask rainfall data
Exact-value extraction: extract only urban areas from land-cover classification data
Bit extraction: use a mask band to determine valid pixels

Difference images and time-series analysis

You can also compare NDVI, vegetation index, across different dates, merge data from multiple days, and display spatial statistics over time.

ip.diff_images()  # difference calculation
ip.calc_spatial_stats()  # spatial statistics
ip.calc_temporal_stats()  # time-series statistics

Claude Desktop Integration (MCP Server)

JAXA also provides MCP server support so that Claude Desktop can operate directly on JAXA satellite data.

Setup steps

Create a virtual environment and install the package
Place mcp_server.py
Add it to the Claude Desktop configuration

{
  "mcpServers": {
    "jaxa_api_tools": {
      "command": "C:\\path\\to\\venv\\Scripts\\python",
      "args": ["C:\\path\\to\\mcp_server.py"]
    }
  }
}

Available functions

search_collections_id - get collection information
show_images - display satellite images
calc_spatial_stats - calculate spatial statistics
show_spatial_stats - display statistical result images

Available Datasets

Land Cover Class (2019)
GCOM-C SGLI RGB imagery (December 2021)
Forest Non Forest (2017-2020)
JASMES Japan area (2020 to September 2022)

You can browse the full dataset list at https://data.earth.jaxa.jp/en/datasets/ .

About the Data Formats

COG (Cloud Optimized GeoTIFF): A format that lets you fetch only the needed portion with HTTP range requests
STAC (Spatio Temporal Asset Catalog): A hierarchical JSON catalog for accessing COG data

These formats make it possible to work efficiently with large satellite datasets.

Possible Ways to Use It

Use it directly from Python
Integrate with QGIS
Run it in Google Colab
Use it from Claude Desktop through the MCP server

Personally, the Claude Desktop integration is the part I found most interesting. It opens the door to natural-language instructions like “show me vegetation data for Japan.”