Geostationary satellite-based solar radiation analysis for meteorology & climate, renewable energy, agriculture, social science, and disaster resilience.
AMATERASS transforms geostationary meteorological satellite observations into spatially continuous solar radiation information. Developed, implemented, and long-term operated by since 2007.
AMATERASS is a physical analysis system for geostationary meteorological satellite observations, rather than an ordinary image-processing or empirical-model approach. Solar radiation reaches the Earth's surface through absorption by gases such as water vapor and ozone, scattering by aerosols and cloud particles, and Rayleigh scattering by molecules. AMATERASS analyzes satellite observations by accounting for these effects, including multiple reflection between the surface and atmosphere.
The system grew from radiative-transfer calculations and neural-network acceleration, then expanded into sensor calibration for stable physical use of geostationary satellite data, geolocation correction algorithms enabling latitude-longitude gridded format conversion, and applications spanning photovoltaics, solar thermal, land-surface and hydrology, agriculture and vegetation, social systems, and disaster resilience — forming a foundation for cross-disciplinary use of solar radiation data.
Quasi-real-time processing of geostationary meteorological satellite observations into spatially continuous surface and top-of-atmosphere radiation products. Data is available approximately 10 minutes after satellite observation.
Wide-area solar radiation products from geostationary satellite observations, updated every 10 min. These maps provide spatial context for cloud fields, terrain effects, and regional radiation variability, forming a basic input layer for renewable-energy, climate, and environmental applications.
Solar radiation and photovoltaic power estimates at high spatial and temporal resolution, including every 2.5 min analysis.
Challenge toward multi-geostationary satellite analysis and global solar-radiation products. Extending the analysis from Himawari to multiple geostationary satellite domains supports wide-area, high-frequency monitoring across the tropics and mid-latitudes.
AMATERASS is a physical analysis system for geostationary meteorological satellite observations, rather than an ordinary image-processing or empirical-model approach. It interprets observations through absorption, scattering, cloud effects, and surface–atmosphere radiative exchange.
The goal is to convert observed radiances into physically meaningful solar-radiation quantities—such as global, direct, and diffuse components—by combining radiative-transfer knowledge, in-situ observations, site-scale validation, and high-frequency satellite data.
Radiative-transfer knowledge, in-situ observations, site-scale validation, and high-frequency satellite data are combined to estimate global, direct, and diffuse solar-radiation components.
Satellite radiances are interpreted through atmospheric absorption and scattering by gases, aerosols, clouds, and molecules.
The research was built step by step from ground radiation measurements and site-scale estimation above observation points toward wide-area satellite analysis.
Satellite-based surface solar radiation estimates are evaluated against ground observations across multiple sites to verify quantitative reliability.
AMATERASS originates from the idea of learning radiative-transfer calculations with neural networks, rather than relying only on large lookup tables. The learning algorithm was developed to estimate solar radiation from radiative parameters while retaining a physics-based interpretation.
This foundation connects radiation physics, neural-network acceleration, and quasi-real-time satellite analysis.
Neural networks were developed to learn radiative-transfer calculations from radiative parameters and accelerate physically grounded solar-radiation estimation. This learning approach became the computational bridge between detailed atmospheric radiation physics and operational processing of high-frequency satellite observations.
The Distortion-BP method was developed to learn radiative-transfer calculations and estimate solar radiation from radiative parameters. This algorithmic foundation reduced dependence on large lookup tables and connected physical radiation theory to fast quasi-real-time satellite analysis.
A neural-network solver trained on radiative-transfer calculations accelerated physically based radiation estimation by about a factor of 1,000. This speed-up made it practical to analyze wide-area, high-temporal-resolution geostationary satellite observations and produce operational atmospheric radiation-budget products.
The AMATERASS story is not a single dataset story. Starting from atmospheric physics research aimed at understanding the radiation budget, it progressed through fast radiation calculation, satellite sensor calibration, quasi-real-time operation, and geolocation correction for high-accuracy gridded satellite products. The 2011 Great East Japan Earthquake triggered a decisive turn toward photovoltaic applications and cross-disciplinary collaboration — and the story continues with international research with NASA.
Neural-network acceleration of radiative-transfer calculations enabled fast, physically grounded solar radiation estimation.
Calibration research with the Japan Meteorological Agency contributed to stable physical use of geostationary satellite data and WMO/GSICS-related activities (JMA/MSC calibration guide).
Quasi-real-time solar radiation analysis using Himawari observations started on July 7, 2007, and continued through successive satellite generations.
The Geo-information Correction method enabled high-accuracy gridded geostationary satellite products for quantitative analysis.
After the 2011 Great East Japan Earthquake, AMATERASS expanded toward satellite-based photovoltaic estimation, actively sharing data across disciplinary boundaries with energy, power-system, climate, and social-system researchers.
The system supports research and applications across meteorology & climate, renewable energy, agriculture, social science, and disaster resilience.
Geostationary satellites appear nearly stationary from Earth, enabling high-frequency observation of the same region — a major advantage for analyzing rapidly evolving atmospheric phenomena such as clouds. However, scanning mirror motion, satellite attitude variations, and navigation errors cause frame-by-frame positional shifts. Geo-information Correction method addresses this problem by correcting geolocation and enabling gridded-format data suitable for downstream analysis.
The geolocation-correction and disk-to-grid conversion capabilities developed through AMATERASS-related work formed part of the technical lineage leading to subsequent gridded geostationary satellite products, including the NASA GeoNEX L1G product workflow. The requirement for physically comparable, stable satellite data that drove AMATERASS development connects directly to the broader need for wide-area gridded products.
Geostationary satellites observe the Earth as disk-shaped data in satellite projection. AMATERASS-related programs convert these observations into latitude–longitude gridded data, conceptually transforming Level-1A-equivalent disk data into Level-1B-equivalent gridded products.
Phase-only correlation and gridded-format processing correct positional shifts in Himawari and GOES-class geostationary observations. The correction stabilizes the satellite image sequence and supports quantitative mapping, compositing, and time-series analysis on latitude–longitude grids.
AMATERASS-origin geolocation correction and gridded satellite product generation programs provided to NASA NEX were used in the NASA GeoNEX L1G product workflow, together with further development, integration, and execution by the GeoNEX team.
Radiation budget, clouds, aerosols, satellite calibration, and climate-model evaluation.
Photovoltaic output estimation, solar resource assessment, and distributed power systems.
Crop and fruit environments, radiation-driven productivity, and climate impact analysis.
Regional energy planning, demand-side analysis, urban systems, and data-driven decision support.
Distributed generation, blackout resilience, water resources, and emergency energy planning.
These examples are arranged broadly in chronological order.
Distributed microgrids and photovoltaic variability reduction supported by AMATERASS/EXAM-related solar-radiation information.
doi:10.9746/jcmsi.6.281 →
Satellite-derived solar radiation improves radiation, heat, and water-budget components in Japanese land-surface analysis.
doi:10.3178/hrl.9.14 →
High-spatial-resolution solar radiation data support voltage control and service restoration in distribution networks with photovoltaic systems.
doi:10.1061/(ASCE)EY.1943-7897.0000352 →
Ground observations validate Himawari-8 surface solar-radiation estimates and support confidence in satellite-derived products.
doi:10.5194/amt-11-2501-2018 →
An activity-based residential energy model is applied nationally across Japan at postcode-district scale, using AMATERASS-derived meteorological input.
doi:10.26868/25222708.2019.211024 →
Multi-geostationary gridded products demonstrate the broader product ecosystem linked to AMATERASS-related geolocation and gridding workflows.
doi:10.3390/rs12081267 →
Neural-network-based day-ahead scheduling links renewable-energy uncertainty with electricity-market decision support.
doi:10.9746/sicetr.56.57 →
Geostationary satellite observations are fused with numerical weather prediction output to improve surface solar irradiance forecast skill.
doi:10.1016/j.solener.2021.05.055 →
High-frequency geostationary observations reveal seasonality in Amazon evergreen forest greenness under persistent cloud conditions.
doi:10.1038/s41467-021-20994-y →
Regional AMATERASS-based solar-irradiance analysis across Asia-Pacific supports active and strategic use of solar energy.
doi:10.1016/j.solener.2024.112678 →
Himawari-8/9 data are used to model diurnal gross primary production across East Asia, linking flux-tower observations and carbon-cycle assessment.
doi:10.1016/j.rse.2025.114866 →AMATERASS analysis products have been widely used as fundamental solar-radiation and renewable-energy datasets across multiple research fields and projects.
AMATERASS and related satellite-based solar-radiation research have been introduced through newspapers, institutional news, and public outreach articles.
Hideaki Takenaka specializes in satellite remote sensing, atmospheric radiation, and solar-radiation energy applications. He studied atmospheric radiation at the Center for Environmental Remote Sensing (CEReS), Chiba University, and received his Ph.D. in Science in 2009.
He developed AMATERASS, a quasi-real-time solar radiation analysis system using geostationary satellite data. Since July 7, 2007, he has continuously operated quasi-real-time solar radiation analysis based on Himawari geostationary meteorological satellite observations.
He has also contributed to the physical reliability of geostationary satellite analysis through joint research with JMA/MSC, where the developed vicarious calibration method was proposed to WMO/GSICS by JMA as Japan's calibration technique. After the 2011 earthquake, he expanded into renewable-energy and power-system studies. He also developed a phase-only-correlation-based geolocation correction method, later connected to collaboration with NASA Ames and the NASA GeoNEX L1G product workflow. The analysis results of AMATERASS have been widely utilized as fundamental data across various research fields.