From Sen2Like to ChimeLike: Transferring Sentinel-2 Harmonisation Principles to Establish CHIME as a Hyperspectral Reference Standard

Abstract

Challenge The Copernicus Hyperspectral Imaging Mission for the Environment (CHIME) is designed to provide systematic hyperspectral observations of terrestrial surfaces and coastal zones, with a primary objective to support the development of new and improved operational applications and services that are related to food security, sustainable agriculture, and raw materials. As a secondary objective, CHIME will support operational applications and services related to inland and coastal waters quality, biodiversity, snow and ice properties, environmental degradation, and hazards. CHIME together with other forthcoming hyperspectral missions, will complement an expanding constellation of spaceborne imaging spectrometers, including EMIT, PRISMA, DESIS, and EnMAP. Their combined operation would enable increased temporal sampling, supporting the stringent observational requirements of precision agriculture, ecosystem monitoring, and snow characterization. However, exploiting this increased temporal density in a consistent manner requires the harmonisation of observations acquired by different hyperspectral sensors with different observation conditions. This process of harmonisation would ensure interoperability and the generation of seamless, analysis-ready hyperspectral data products. To address this challenge, experience gained from the harmonisation of Sentinel-2 data is directly relevant. The Sen2like framework was developed to generate radiometrically and geometrically consistent Sentinel-2 time series by integrating observations from compatible missions, in particular Landsat-8 and Landsat-9. Methodology The proposed Chimelike concept builds on the modular Sen2like architecture, which includes geometry refinement, spectral harmonisation, radiometry, geometry fusion, atmospheric correction, cloud screening and anisotropy correction. The idea is to reuse that concept with the necessary adaptation to the CHIME Hyperspectral mission as depicted in figure1. Harmonisation shall be understood as keeping the native geometry of the input mission, while fusion is understood as a process including the geometric transformation of input mission to CHIME one. The target mission for the harmonisation/fusion is CHIME, the same way it is Sentinel-2 for Sen2Like. A summary of the required information from any candidate hyperspectral mission is established concerning the auxiliary data, mission orbital parameters, instrument characteristics, product level and pixel level metadata, atmosphere metadata (atmospheric and meteorological), uncertainties, together with technical documentation. All this information serves as inputs for the differents harmonisation processing blocks mentionned above and a set of recommendations is given to optimize the quality of the harmonisation process, e.g. the Digital Elevation Model, the Orthorectification Basemap, the Grid Tiling System, the Solar Irradiance Model and Masking strategy. Study area and data Sen2like demonstrated its flexibility and capability to generate dense time series over country scale area like Belgium for year 2023 with input data from 4 differents instruments: Sentinel-2A, Sentinel-2B, Landsat-8, Landsat-9. It can also be used to generate Level-2 harmonised data on smaller zones defined by a region of interest or covering an entire Sentinel-2 MGRS tile. Within CHIME, mission, the objective is to develop a CHIME Level 2H/F Demonstration Processor that will output harmonised CHIME-like surface reflectance products using as input CHIME L1C data or other similar Hyperspectral mission L1C data. The Chimelike demonstration processor integrates CHIME Level-2 components for atmospheric correction and cloud screening, together with an anisotropy correction module producing nadir-adjusted reflectance (NBAR) based on the High-resolution Adjusted BRDF Algorithm (HABA). The other modules will be adpated from existing Sen2like modules. Discussion and Conclusions This introduction to Chimelike would be the occasion to discuss the potential systematic biases and uncertainties that can propagate through the remote sensing data processing pipeline, when combining data sets from different sources. The discussion focuses then on the identification of relevant quality metrics for assessing harmonised surface reflectance products and on their suitability for terrestrial monitoring applications relying on multi-mission hyperspectral time series.