Computer simulations suggest a compact space instrument could map methane and carbon dioxide at a much finer modeled ground scale than the CO2M reference used for comparison, while reaching similar simulated error levels. The study puts the modeled ground-resolution elements at approximately 250–300 metres, compared with 2 × 2 square kilometres for CO2M. At albedo 0.2, the comparison uses reference requirements of 10 parts per billion for methane and 0.7 parts per million for CO2. The result is a technology projection based on end-to-end simulations, not a demonstration by a satellite, flight or field instrument.
The manuscript is an arXiv preprint, version 1 dated 20 August 2026. It models an imaging spectrometer in the shortwave-infrared, or SWIR, range from 1590 to 1670 nanometres, alongside multiple detector configurations. The analysis generates filter transmission profiles, simulated detector signals and retrieval outputs rather than measurements from a built instrument. Here, retrieval means the calculation used to estimate gas amounts from those signals. Experimental validation of the filters and integrated detector assembly remains the next step.
The design behind the result
The design process began with approximately 6,500 physically realizable transmission profiles for photonic-crystal filters. Candidate sets were evaluated with Monte Carlo simulations before the instrument model settled on ten complementary filters. The architecture represents each ground element with ten integrated signals from task-optimized filters. The authors describe that compact signal set as enabling substantial data reduction before downlink.
Three detector arrangements were modeled: CHROMA-D, COBRA-L and SNAKE 3×3. The reported swaths were 150 kilometres, 100 kilometres and 115 kilometres, respectively. CHROMA-D and SNAKE 3×3 had the best overall simulated performance, with SNAKE showing a small advantage for CO2 retrieval. These were comparisons among modeled configurations, not tests of working instruments.
Performance comes with conditions
The strongest performance claim is conditional. For surface albedos of 0.2 or higher, CHROMA-D and SNAKE 3×3 achieved simulated methane errors below approximately 10 ppb and carbon-dioxide errors below approximately 0.7 ppm. Surface albedo is the model’s measure of how reflective the ground is; ppb and ppm mean parts per billion and parts per million. All simulations used a solar zenith angle of 50 degrees. The thresholds therefore describe the stated surface and viewing setup, not a universal result for every observation.
The paper’s CO2M comparison is about matching reference requirements, not benchmarking two operating instruments. The proposed design’s modeled ground-resolution elements are approximately 250–300 metres, versus 2 × 2 square kilometres in the CO2M reference, and its simulated errors are comparable to the reference requirements at albedo 0.2. That makes the spatial-resolution claim striking on paper, but the analysis itself treats the comparison as model-based rather than direct inter-mission validation.
The paper also tests a variable-projection retrieval, a fitting method that separates the linear albedo parameters from the nonlinear gas parameters and solves them independently. Across the investigated albedo range, residual bias—the leftover systematic offset—was approximately 5–10 times smaller than retrieval precision. The authors interpret that pattern as leaving measurement noise as the main contributor to total error. It is evidence about the modeled retrieval, not a test of whether fabricated filters and an integrated detector will behave in the same way.
What still has to be built
Beyond the gas estimates, the mission geometry is also projected rather than demonstrated. For a 150-kilometre swath, the authors project global coverage in four days using five satellites. That is an engineering projection from simulations, not an operational result. The ten-filter representation is likewise described as enabling substantial pre-downlink data reduction, but no quantitative downlink-volume reduction is reported in the supplied analysis.
The paper also says that full-physics retrieval would require auxiliary information, potentially including a multi-angle spectropolarimeter. This caveat puts the reported gas errors in the context of a specified modeled setup rather than a complete operational retrieval. Fabrication, detector integration and experimental tests will have to show whether the simulated filter responses and retrieval performance hold in hardware. Until then, the proposed mapping, coverage and data-volume benefits remain projections.
The broader message is therefore cautious but useful. The simulations indicate that task-optimized filters, alternative detector layouts and variable projection can be combined in a compact architecture with modeled gas-retrieval performance comparable to the stated global-mapping reference, while preserving much finer modeled ground sampling. They do not establish measured atmospheric performance, a flight-ready instrument or a quantified downlink saving. The preprint’s contribution is a testable design target for the hardware work that follows.
Paper data and sources
Original title: Developing a Compact SWIR Imaging Spectrometer for CO2 and CH4 Retrieval Using Photonic Crystal Filters
Authors: Marijn Siemons, Brecht Simon, Irina Malysheva, Ralf Kohlhaas
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text