GRSG 34th Conference 2023

Title: Marigold on the DL Platform: An Online Cloud Compute Solution for the Management and Processing of High Volume Hyperspectral Data for Mineral Exploration Mapping in the Green Energy Transition

Author: Lori Wickert

The needs of the global population are evolving in response to climate change. The mining industry must adapt by seeking more raw materials to facilitate the transition to green energy, while embracing sustainable practices that minimize environmental impact. Fortunately, we find ourselves in an era marked by the rapid deployment of advanced spaceborne technologies poised to help provide new data sources to support this green transition. However, alongside these innovations, there is an imperative need for technological advancements to enhance data processing speed and efficiency. In the last couple of years, after decades of planning, several space agencies have launched new spaceborne capabilities for hyperspectral imaging (HSI), featuring hundreds of spectral bands optimized for regional-scale mapping.

Notable examples include the Italian Space Agency (ASI) with PRISMA, the German Space Agency (DLR) with EnMap, and NASA (USA) with EMIT. Yet, this is only the inception, as numerous additional sensors are in the pipeline, such as NASA’s SBG and ESA’s CHIME, accompanied by various commercial missions like PIXXEL, Orbital Sidekick, and HySpecIQ, scheduled for the near future. These advancements build upon the technology developed to address the signal-to-noise challenges of the first spaceborne sensor Hyperion (NASA, 1999), with many commercial missions promising mapping capabilities at higher spatial resolutions. For the mineral exploration community, this complements the existing technology of HSI at the local scale, attainable via airborne hyperspectral sensors from entities such as SpecTIR, HyVista, ISDAS, and others, which have been in operation for many years. Altogether the volume of data generated by these missions is set to make a substantial contribution to the global data landscape, with exponential growth projected.

In fact, it is anticipated that the global data generated over the next decade will surpass the cumulative data generated up to the present. To harness the full potential of these data sources, we must establish robust data management and processing solutions. Online cloud computing, offering the speed of supercomputer-like infrastructure and innovative data solutions, holds promise for managing the high data volumes associated with hyperspectral data. At Descartes Labs (DL), we have leveraged this approach to develop solutions capable of scaling to meet the demands of processing high density datasets, including complex operations like unmixing and spectral mapping, involving HSI dataset sizes ranging from gigabytes to terabytes.

This approach facilitates the simultaneous processing of entire or mosaiced datasets, significantly accelerating project-level data interpretation and problem-solving capabilities. In the realm of mineral exploration, hyperspectral imaging technology, with its broad spectral range and narrow spectral resolution, offers systematic species-level compositional mapping solutions. Sensors capturing reflectance (VNIR-SWIR) or emissivity (TIR) data can be combined to meet the mapping needs of different deposit types. HSI data yields intricate insights into mineral systems, encompassing factors such as abundance, position, and shape of pathfinder alteration halos. This information aids in deriving compositional features like vector-to-ore relationships and offers valuable insights into orebody fluid-flow pathways. Extracting this information from vast hyperspectral datasets presents formidable challenges, which DL has tackled through its innovative hyperspectral compute capabilities within our online remote sensing processing platform for mineral exploration, known as Marigold. We have redefined how geologists access remote sensing data by developing an online processing capability customized for mineral exploration, and have effectively addressed data volume and speed challenges for hyperspectral data.

The interface encompasses a suite of standard/multispectral and hyperspectral-specific processing tools, with the ability to load and compare other project datasets, such as geophysics data, within the same interface. DL’s hyperspectral solutions can also extend beyond mineral exploration, reshaping how geospatial data is accessed and utilized across the mining chain. This presentation will showcase several examples illustrating how our Marigold analytics interface meets the hyperspectral processing demands of the future.

The needs of the global population are evolving in response to climate change. The mining industry must adapt by seeking more raw materials to facilitate the transition to green energy, while embracing sustainable practices that minimize environmental impact. Fortunately, we find ourselves in an era marked by the rapid deployment of advanced spaceborne technologies poised to help provide new data sources to support this green transition. However, alongside these innovations, there is an imperative need for technological advancements to enhance data processing speed and efficiency. In the last couple of years, after decades of planning, several space agencies have launched new spaceborne capabilities for hyperspectral imaging (HSI), featuring hundreds of spectral bands optimized for regional-scale mapping. Notable examples include the Italian Space Agency (ASI) with PRISMA, the German Space Agency (DLR) with EnMap, and NASA (USA) with EMIT.

Yet, this is only the inception, as numerous additional sensors are in the pipeline, such as NASA’s SBG and ESA’s CHIME, accompanied by various commercial missions like PIXXEL, Orbital Sidekick, and HySpecIQ, scheduled for the near future. These advancements build upon the technology developed to address the signal-to-noise challenges of the first spaceborne sensor Hyperion (NASA, 1999), with many commercial missions promising mapping capabilities at higher spatial resolutions. For the mineral exploration community, this complements the existing technology of HSI at the local scale, attainable via airborne hyperspectral sensors from entities such as SpecTIR, HyVista, ISDAS, and others, which have been in operation for many years. Altogether the volume of data generated by these missions is set to make a substantial contribution to the global data landscape, with exponential growth projected. In fact, it is anticipated that the global data generated over the next decade will surpass the cumulative data generated up to the present. To harness the full potential of these data sources, we must establish robust data management and processing solutions.

Online cloud computing, offering the speed of supercomputer-like infrastructure and innovative data solutions, holds promise for managing the high data volumes associated with hyperspectral data. At Descartes Labs (DL), we have leveraged this approach to develop solutions capable of scaling to meet the demands of processing high density datasets, including complex operations like unmixing and spectral mapping, involving HSI dataset sizes ranging from gigabytes to terabytes. This approach facilitates the simultaneous processing of entire or mosaiced datasets, significantly accelerating project-level data interpretation and problem-solving capabilities. In the realm of mineral exploration, hyperspectral imaging technology, with its broad spectral range and narrow spectral resolution, offers systematic species-level compositional mapping solutions. Sensors capturing reflectance (VNIR-SWIR) or emissivity (TIR) data can be combined to meet the mapping needs of different deposit types. HSI data yields intricate insights into mineral systems, encompassing factors such as abundance, position, and shape of pathfinder alteration halos.

This information aids in deriving compositional features like vector-to-ore relationships and offers valuable insights into orebody fluid-flow pathways. Extracting this information from vast hyperspectral datasets presents formidable challenges, which DL has tackled through its innovative hyperspectral compute capabilities within our online remote sensing processing platform for mineral exploration, known as Marigold. We have redefined how geologists access remote sensing data by developing an online processing capability customized for mineral exploration, and have effectively addressed data volume and speed challenges for hyperspectral data. The interface encompasses a suite of standard/multispectral and hyperspectral-specific processing tools, with the ability to load and compare other project datasets, such as geophysics data, within the same interface. DL’s hyperspectral solutions can also extend beyond mineral exploration, reshaping how geospatial data is accessed and utilized across the mining chain. This presentation will showcase several examples illustrating how our Marigold analytics interface meets the hyperspectral processing demands of the future.