OEMC project use case: spatial enhancement of SIF for better GPP flux estimations
Description
Sun-induced chlorophyll fluorescence (SIF) is a signal that can be retrieved from space and that has been shown to correlate well with gross primary productivity (GPP), thereby providing an invaluable tool to monitor carbon uptake by terrestrial ecosystems. Such data thus has the potential to contribute to the quantification of the global carbon budget (GCB), the critical annual update revealing the latest trends in global carbon emissions. More specifically, the aim of this use case is to prepare tools that could be latter used by the 'REgional Carbon Cycle Assessment and Processes' (RECCAP) initiative coordinated by the Global Carbon Project, which seeks to gather and integrate regional data from 14 major global regions, ensuring enough harmonization to scale these budgets globally and facilitate regional comparisons. To do so, they rely on model simulations which would greatly benefit from detailed information on terrestrial carbon fluxes provided by satellite Earth Observation (EO), such as spatio-temporal proxies such as SIF. However, current and past satellite instruments from which this SIF signal can be derived have a very coarse spatial resolution (5 km at best) and are usually very noisy due to the subtle nature of the signal.
Stakeholders needs:
- The RECCAP efforts are done volutarily by many different and diverse participants, each with distinct approaches and looking at different regions, so they should have easy access to sub-sections of the data that they should be able to download on their computers
- High spatial resolution to discriminate better GPP by land cover type, possibly also to detect the effect of disturbances and of land cover change
- Coverage of coastal/island regions with finer spatial scale could be desirable as these are often misrepresented
Implementation:
In this use case, we tried to find solutions to these problems by proposed a open workflow to enhance the spatial resolution of SIF estimations (e.g. to 1 km) by leveraging on the synergistic use of other sources of remote sensing (e.g. LST, NIRv) with a semi-empirical approach in which a data-driven method is combined with known physical constraints governing the relationships between the input variables. We developed a first prototype workflow in Julia language based on a previous SIF downscaling approach proposed by Duveiller and Cescatti (2016) and Duveiller et al. (2020). We contributed to a new SIF retrieval based on Sentinel-5P TROPOMI data in which machine learning is used in the retrieval itself to enhance the retrieval quality and reduce noise. The result are ungridded instantanous SIF estimates that are then used for the spatial enhancement tool. This spatial enhancement tool was further ported to the Copernicus Data Space Ecosystem (CDSE) using the OpenEO framework, with the intention to make it available for users to do their spatial downscaling on-demand over their target areas based on their downscaling model of their choice. A dedicated notebook was made to illustrate how it can be done. To evaluate the use of increasing the spatial resolution of SIF for GPP estimation, efforts have also been dedicated to match the satellite grid cells to eddy covariance flux site ground measurements of GPP whilst quantifying the effect of spatial heterogeneity. Furthermore, we aim to deploy the downscaling approach on adjustable moving windows over a Discrete Global Grid System (DGGS) composed of hexagons, which better preserve the area and neighborhood of every cell, thus ensuring a finer representativity of the surface properties.
Recorded talks (ordered chronologically):
- Towards a better estimation of GPP for greenhouse gas accounting by Gregory Duveiller, Open-Earth-Monitor - Global Workshop 2023;
- Quantifying the (mis)match between in-situ and satellite time series by Daniel Pabon-Moreno, Open-Earth-Monitor - Global Workshop 2024;
- Rethinking the grid - Towards less distorted imagery and AI by Daniel Loos, Open-Earth-Monitor - Global Workshop 2024;
Technical info
Visit the use-case page on the OEMC website to learn more: https://earthmonitor.org/sif-based-high-spatial-resolution-gpp-flux-estimations/