Space Seed Holdings Releases Open-Source Toolkit for Satellite Data Analysis
Space Seed Holdings Inc. (“SSHD”) released satellite-landuse-toolkit on GitHub under the MIT License on September 13, 2026, and obtained a DOI for it on Zenodo. The toolkit analyzes public satellite datasets to summarize land cover, annual tree-cover loss and changes in surface water within a selected area. It records the inputs and processing settings so that the analysis can be repeated.
SSHD developed the toolkit as part of its research and development related to SpaceAgent, a platform for conducting research autonomously in space and on the ground. It plans to use the toolkit in regional projects and is exploring educational and research applications with Spacenome Lab., Inc. (“Spacenome”), in which SSHD has invested.
Background
Land-cover maps at 10-meter resolution, more than 20 years of annual tree-cover loss data, surface-water records from 1984 to 2024, and optical satellite images with a nominal five-day revisit interval are all free to use. In practice, however, working with them often requires cloud accounts, API keys and specialist GIS knowledge, and results can be hard for others to reproduce.
Interpreting the numbers is harder still. Tree-cover loss data do not account for regrowth, so they cannot show how much forest remains. The stated accuracy of a global land-cover product does not guarantee its accuracy for a particular area. Comparisons of images from two dates are affected by clouds, seasons and alignment.
How the tool works
The user supplies an area, as a GeoJSON file or a rectangle defined by latitude and longitude. The toolkit provides five commands to find and download the public data for that area and analyze it. It runs on a local computer, without Earth Engine or other cloud platforms and without API keys.
– fetch: finds and downloads public tiles covering the area and searches Sentinel-2 scenes
– landcover (ESA WorldCover, 10 m, 2021): area by land-cover class
– forest (Hansen Global Forest Change, 30 m, 2001¨C2025): annual tree-cover loss, tile coverage, threshold sensitivity and hotspot cells of about 1 km
– water (JRC Global Surface Water, 30 m, 1984¨C2024): area by water transition class and the connected water pixels of a chosen lake
– change (Sentinel-2 L2A, 10 m): bare-ground change between two dates, using only pixels valid on both dates
Safeguards against misreading the data
– Reflectance checks: if the reflectance correction of Sentinel-2 data cannot be determined from metadata, the tool stops rather than guessing.
– Consistent pixels: two dates are placed on one grid, and only pixels free of cloud and water on both dates are counted. The share of valid pixels is reported.
– Coverage reporting: the tool shows how much of the area the tiles or images actually covered.
– Run records: input files, data versions, decision criteria and thresholds are written to a run record, so others can repeat the analysis.
– Interpretation checklist: a checklist in English and Japanese covers points such as the difference between mapped area and a statistical area estimate.
Use in education
The toolkit and its data require no fees or registration, and it runs locally on a computer with the required software installed. Students can enter a rectangle around their own town and check its land cover, where tree-cover loss has been detected since 2001, and how lakes and rivers have changed. Because the inputs and processing settings are recorded, teachers can use the process as well as the results as teaching material. SSHD has prepared a Japanese step-by-step guide based on a wetland area in Japan and is considering adapting it for education.
Plans with Spacenome
SSHD is considering two uses with Spacenome, with details and timing still to be decided:
– Education: an inquiry-based program in which junior and senior high school students choose a region, study its land use with satellite data, and compare the results with statistics and local information.
– Research with companies and local governments: a shared method for discussing regional data, for example when adding value to farm, forest and fishery resources or tracking land-use change.
How it was developed
An AI agent wrote and tested the code to a specification, and a separate AI agent reviewed it independently. This cycle was repeated six times before release. The reviews found problems that can be missed when testing only one region, including tile merging, tile naming rules, image alignment, reflectance correction and clipping of the area. Each was fixed, and the tool now has 26 regression tests. For the published example run, the hashes of the six input files matched in an independent environment, and the six output files were identical byte for byte.
SSHD and Spacenome are developing SpaceAgent to automate research, with one AI agent carrying out the work, another checking it, and people making the decisions. This project applied that approach to satellite data analysis.
Release overview
– Name: satellite-landuse-toolkit (Python package name: slandu)
– Version: v1.0.0 (September 13, 2026)
– Repository:
– DOI: 10.5281/zenodo.22730573 (all versions); 10.5281/zenodo.22730574 (v1.0.0)
– License: MIT (code). Each downloaded dataset is subject to its provider’s license.
– Requirements: Python 3.10 or later; rasterio, numpy, scipy and Pillow; the curl command
Comment
Kengo Suzuki, Representative Director and CEO, Space Seed Holdings Inc., and Director of SPACE LAB., Spacenome Lab., Inc.
“Anyone can now obtain satellite data. The hard part has been analyzing and interpreting it correctly. Looking at your own town from space, analyzing it, and thinking about what you can and cannot claim is useful experience for students and for people working in companies alike. I want to explore that possibility with Spacenome, which has long worked on space education.”
Next steps
SSHD plans to use the tool in regional projects in Japan and abroad, and to consider joint presentations on land-use analysis with research institutions in those regions. It will also formalize the development cycle used here, with AI implementation, independent AI review and human decisions, as a verification procedure for its experiment automation platform.
About Space Seed Holdings Inc.
Space Seed Holdings is a space deep-tech venture builder whose mission is “Turning science fiction into nonfiction.” Its activities span investment, research and venture creation, and include the Fermentation and Longevity Fund program, which supports the practical use of fermentation and longevity technologies. Its long-term goal is to help develop the technologies needed for humans to live in space by 2040, working with partners across different fields.
Website:
