Modelling solar-induced chlorophyll fluorescence (SIF) with terrestrial biosphere model QUINCY

Satellite observations provide unique opportunities for the development and evaluation of terrestrial biosphere models (TBMs). Solar-induced chlorophyll fluorescence (SIF) is one of the remotely sensed variables that is directly related to photosynthetic activity. Chlorophyll fluorescence (ChlF) is one of the pathways for the absorbed radiation within leaves, along with photochemistry and non-photochemical quenching (NPQ). However, the relationship between photosynthesis and ChlF depends on environmental conditions and on the level and saturation of NPQ. Therefore having a process-based representation of SIF within a TBM will help to fully exploit the data stream available from the remote sensing in carbon cycle studies.

 

We have implemented a leaf-level model of ChlF in the QUINCY ('QUantifying Interactions between terrestrial Nutrient CYcles and the climate system') TBM. Based on a previous study testing different alternatives for describing the radiative transfer of SIF, we used a radiative transfer model L2SM (developed by T. Quaife) for the SIF signal. Observed leaf level SIF spectra were used to convert the model output to observed units.

 

As data sources we used satellite observations of the SIF from TROPOMI as well as data products using previous satellite missions and machine learning that go further back in time, as the TROPOMI observations begin in 2018. We extracted satellite observations at the carbon dioxide (CO₂) flux tower sites in different ecosystems and used these satellite SIF observations, along with gross primary production (GPP) from the flux observations to evaluate and improve our model. Since the footprints of flux towers are different from TROPOMI observations, we focus on flux towers located in homogeneous landscapes.  We emphasize the use of data from research infrastructures,  such as ICOS, because they have up-to-date data coverage. In addition, we used some in situ tower-based SIF observations.

 

According to our results, the formulation of NPQ required different parameterizations for sustained NPQ in different temperature regimes and we also tested a new formulation for NPQ in drought conditions. The magnitude of the simulated SIF signal was too high at evergreen conifer sites when compared to the in situ -observations but it was at a similar level to the satellite observations. The seasonal cycle of grassland phenology was off in the model and satellite SIF provided another data stream to work on improving it. Satellite-based SIF proved to be a useful addition for model development and having a process-based representation for SIF enhances its usefulness.