Expanding Copernicus’s Earth Observation Program With Hyperspectral Imaging

In June 2023, the European Commission and the European Space Agency awarded Kuva Space a €5M commercial contract to be the sole provider of hyperspectral data services for the Copernicus program. Amidst strong competition in the Earth Observation market, where lately a few commercial hyperspectral companies have entered, the choice of only one service provider to fill the gap until Copernicus’ CHIME mission launch is intriguing.

The Copernicus program is a European Union (EU) initiative for Earth observation and environmental monitoring. Its primary objective is to provide accurate, timely, and reliable information for environmental and security purposes. It uses Earth observation satellites, such as the Sentinel constellation, groundbased sensors, and in-situ data collection to support a wide range of applications, including climate monitoring, agriculture, forestry, urban planning, disaster management, and maritime surveillance. The collected data is made freely available to the public, policy-makers, and the scientific community to foster innovation, scientific research, and informed decision-making for addressing global environmental challenges.

AI-Enabled CubeSats Constellation

Kuva Space’s constellation of 100 CubeSats aims to provide hyperspectral images of any location on Earth at visible-to-near infrared (VIS-NIR, 450-950 nm) and visible-to-shortwave infrared (VIS-SWIR, 450-2,500 nm) wavelengths. Hyperfield, the first satellite of the polar sun-synchronous constellation will be launched in 2024. The Hyperfield’s payload is based on a novel Piezo-actuated Fabry-Perot interferometer (PFPI) where the FPI acts as a tunable bandpass filter. It consists of two semi-transparent reflectors separated by a gap that is controlled by three piezo-actuators in a closed capacitive feedback loop. The mirror substrates are at least 10 mm thick in order to achieve good flatness and adequate mirror quality. This design offers higher optical throughput and gives the company the flexibility to acquire data in different image acquisition modes through software.

It is anticipated that with the Kuva Space constellation and its automated AI-based processing chain, EO data will be available at affordable prices and in a timely and robust manner worldwide. Kuva Space’s image-processing architecture is designed to determine in less than 15 minutes, and deliver within an hour, the measured surface reflectance values within each image pixel from the acquired raw data after radiometric, geometric, and atmosphere calibrations. Hyperspectral image products of georeferenced Top-of-Atmosphere (TOA) radiance (L1B), orthorectified data (L1C), Bottom-of-Atmosphere reflectance (L2A), and geophysical parameters such as water and land biochemical indices (e.g. Chl-a, normalized difference vegetation index) will be delivered in a GeoTIFF format along with other products necessary for producing accurate analytical products such as clouds classification map.

Kuva Space processing tools are developed with deep learning and driven by artificial intelligence (AI) to achieve highly accurate data products and robust radiometric, geometric, and spectral consistency required by the European Space Agency. Artificial intelligence-driven machine learning methods, such as the geo-orthorectify tool, achieve subpixel accuracy in mountainous scenes (e.g. Himalayas, Alps) with low latency and computational costs that allow each image scene to be geolocated in less than four seconds.

Achieving Copernicus Program Goals

The Copernicus program seeks to achieve one of its main goals by making Sentinel data available for free to researchers and service providers so that tools can be developed for monitoring crop production and farming at national and field scales in order to predict seasonal yields and prevent hunger through careful harvest planning. Despite the massive success of Sentinel and its open-source data scheme, its seasonal yield predictions suffer from several shortcomings and inaccuracies originating from low spectral resolution and irregular and sparse time series due to cloud covers and weekly revisits. Consequently, the vegetation indices extracted from the multi-spectral resolution are insufficient to identify minor crops such as fruits, vegetables, tree nuts, nursery plants, or ornamental crops, or differentiate inner parcel variability.

The rich biochemical information contained within the hyperspectral data provides better measurements and estimates of vegetation properties required to evaluate crop growth stages and health conditions. It also enables the identification of inner-pixel variations, enforces the detection of under-representative crop species, and differentiates parcels based on agricultural practices. The highly accurate spectral data of Hyperfield would allow European stakeholders and Global service providers to automatically map land cover changes and crops, improve yield forecasting, and evaluate biomass volume and carbon content are necessary to address climate change and food security.

By supplying daily hyperspectral measurements to Copernicus, Hyperfield will enrich and extend the optical-based geophysical and biochemical parameters derived from Sentinel and other available Earth Observation instruments and support the development of global services, beyond crop monitoring and land use mapping. It will enhance the extraction of biogeochemical characteristics of the global oceans and regional seas, improve the monitoring of water quality, enable the identification of harmful algae species, and extend Blue Economy applications including aquaculture and blue carbon. Hyperspectral data can be used to produce accurate bathymetry maps (up to 30 m depth) and seafloor substrates (such as sand, rock, and coral) of coastal areas by analyzing the intrinsic optical properties of the water column in different spectral wavelengths.

It can improve the monitoring of sea and land ice surfaces in Arctic regions by providing accurate albedo (reflectivity) measurements and characterizing permafrost composition (e.g. ice, organic matter, and mineral content) based on spectral information. Additionally, Hyperfield data has the potential to advance the measurement of climate change indicators and the monitoring of atmospheric pollutants like methane and carbon dioxide. By delivering information pertinent to support progress in biomass measurement, biodiversity mapping, and accurate change detection in forest and urban areas, Hyperfield will improve the response of Copernicus services to health-related environmental phenomena.

Kuva Space also aims to support Copernicus safety and security services by providing near real-time observation of conflict areas and hazard events and monitoring their impacts and recovery processes, by providing spectral information pertinent to these activities. To achieve this goal, Kuva Space has also the intention to offer tasking capacity to Copernicus for ensuring the provision of data to disaster management, security and defense, and climate change services. Particularly, Hyperfield’s unique imaging technology will enable four acquisition modes: optimized scanning, collecting stereo images in off-nadir observation, producing enhanced-resolution images via micromovements, and “staring on” a specific spot for a longer period of time.

By extending the integration time during the staring mode, the signal-to-noise ratio can be improved significantly by a few hundred percent, and data quality can be enhanced to meet security requirements. Additionally, the company plans to integrate GPUs into its second-generation satellites to enable AI-based calibration and processing on board the satellite. This will allow Kuva Space to align all the spectral bands, create hyperspectral data cubes, detect scenes affected by clouds, and compress the data to reduce volume and increase the amount of data that can be transmitted to a network of ground stations.

Furthermore, the company seeks to develop its Sat-to-Sat and Sat-to-IoT mobile communications to accelerate the delivery of analytic products to end users in real-time through secured API to operational platforms. A good example of how important such a capacity is to security services is the detection and rediscovery of suspicious vessels. The marine surveillance services monitor and track vessels based on their automatic identification systems (AIS) and radio frequency information (RFI).

It is very well known that criminal, stealth or dark vessels tend to turn AIS off and do not transmit RF signals. In security operational systems, the absence of these signals produces an alert. With Hyperspectral, these vessels can be detected based on their spectral signatures, and then automatically rediscovered and tracked when they approach ports for parking and must turn on their AIS systems. In near real-time, the automatic processing on board the satellite provides the vessel’s position coordinates to mobile phones or operational platforms through secured channels.

During the next five years, the European Space Agency will evaluate the quality of Hyperfield data with several global use cases to ensure that the satellite data meets the technical, operational, and security requirements of Copernicus. Successful evaluation will grant Kuva Space a commercial contract to provide hyperspectral data to the Copernicus program on a permanent basis.

This article was written by Dr. Michal Shimoni, Director of Analytics and Applications, Kuva Space (Espoo, Finland). For more information, visit here  .

---