Performance of Singular Spectrum Analysis in Separating Seasonal and Fast Physiological Dynamics of Solar-Induced Chlorophyll Fluorescence and PRI Optical Signals

Abstract

Abstract High temporal resolution measurements of solar-induced chlorophyll fluorescence ( F ) and the Photochemical Reflectance Index ( PRI ) encode vegetation functioning. However, these signals are modulated by time-dependent processes. We tested the applicability of the Singular Spectrum Analysis (SSA) for disentangling fast components (physiology-driven) and slow components (controlled by structural and biochemical properties) from PRI , far-red F ( F 760 ), and far-red apparent fluorescence yield ( Fy * 760 ). The proof of concept was developed on spectral and flux time series simulated with the Soil Canopy Observation of Photochemistry and Energy fluxes (SCOPE) model. This allowed the evaluation of SSA decomposition against variables that are independent of physiology or are modified by it. Slow SSA-decomposed components of PRI and Fy * 760 showed high correlations with the reference variables ( R 2 = 0.97 and 0.96, respectively). Fast SSA-decomposed components of PRI and Fy * 760 were better related to the physiological reference variables than the original signals during periods when leaf area index ( LAI ) was above 1 m 2 m -2 . The method was also successfully applied to predict light-use efficiency ( LUE ) from the fast SSA-decomposed components of PRI ( R 2 = 0.70) and Fy * 760 ( R 2 = 0.68) when discarding data modeled with LAI $<$ 1 m 2 m -2 and short-wave radiation R in $<$ 250 W m -2 . The method was then tested on data acquired in a Mediterranean grassland. In this case, the fast SSA-decomposed component of apparent LUE * showed a stronger correlation with the fast SSA-decomposed component of Fy * 760 ( R 2 = 0.42) than with original Fy * 760 ( R 2 = 0.01). SSA-based approach is a promising tool for decoupling physiological information from measurements acquired with automated proximal sensing systems. , Plain Language Summary A fraction of the solar light, which is absorbed by leaves but is not used during photosynthesis, is released through heat or as chlorophyll fluorescence ( F ), a small emission of energy. Recently, it became possible to indirectly estimate the heat and F by measuring the solar light incoming and reflected from leaves using high-resolution optical instruments. Heat release can be monitored with the Photochemical Reflectance Index ( PRI ). While both PRI and F are theoretically linked to the processes associated with photosynthesis, there is a need to remove the disturbing effects from these signals. We tested whether the Singular Spectrum Analysis (SSA) method can identify at which temporal scale (e.g., seasonal, diurnal) physiological processes (i.e., photosynthesis) and vegetation biophysical changes (e.g., phenology) drive variability in PRI and F . We applied the method on artificial time series of PRI and F simulated with a model (Soil Canopy Observation of Photochemistry and Energy fluxes [SCOPE]) and found that SSA can successfully split these signals into several components recognized as slow (seasonally changing structure and pigments) and fast (physiological response to stress) processes. The method was also tested on time series collected in a Mediterranean grassland, yielding promising results in detecting physiologically driven changes in apparent fluorescence yield ( F normalized by photosynthetically active radiation). , Key Points Singular Spectrum Analysis (SSA) allowed to separate slow and fast temporal dynamics in time series of the Photochemical Reflectance Index and apparent fluorescence yield SSA successfully extracted the effect of de-epoxidation state of the xanthophyll cycle on leaf absorptance in time series simulated with Soil Canopy Observation of Photochemistry and Energy fluxes model SSA allows decoupling long-term biophysical and rapid physiological changes in high temporal resolution spectral measurements