[Paper Review] Reconstruction of a Long-term spatially Contiguous Solar-Induced Fluorescence (LCSIF) over 1982-2022

Jianing Fang, Xu Lian+4 more|arXiv (Cornell University)|Nov 25, 2023

Atmospheric and Environmental Gas Dynamics4 citations

TL;DR

This study reconstructs a long-term, spatially contiguous solar-induced fluorescence (LCSIF) dataset from 1982 to 2022 using bias-corrected AVHRR and MODIS reflectance data. By training a neural network on MODIS SIF and two reflectance bands (red and NIR), the authors extend SIF estimates back to 1982, revealing an accelerating global SIF trend of 0.0038 mW m⁻² nm⁻¹ sr⁻¹ per decade since 2001, with strong agreement to in-situ GPP and established vegetation indices.

ABSTRACT

Satellite-observed solar-induced chlorophyll fluorescence (SIF) is a powerful proxy for diagnosing the photosynthetic characteristics of terrestrial ecosystems. Despite the increasing spatial and temporal resolutions of these satellite retrievals, records of SIF are primarily limited to the recent decade, impeding their application in detecting long-term dynamics of ecosystem function and structure. In this study, we leverage the two surface reflectance bands (red and near-infrared) available both from Advanced Very High-Resolution Radiometer (AVHRR, 1982-2022) and MODerate-resolution Imaging Spectroradiometer (MODIS, 2001-2022). Importantly, we calibrate and orbit-correct the AVHRR bands against their MODIS counterparts during their overlapping period. Using the long-term bias-corrected reflectance data, a neural network is then built to reproduce the Orbiting Carbon Observatory-2 SIF using AVHRR and MODIS, and used to map SIF globally over the entire 1982-2022 period. Compared with the previous MODIS-based CSIF product relying on four reflectance bands, our two-band-based product has similar skill but can be advantageously extended to the bias-corrected AVHRR period. Further comparison with three widely used vegetation indices (NDVI, kNDVI, NIRv; all based empirically on red and near-infrared bands) shows a higher or comparable correlation of LCSIF with satellite SIF and site-level GPP estimates across vegetation types, ensuring a greater capacity of LCSIF for representing terrestrial photosynthesis. Globally, LCSIF-AVHRR shows an accelerating upward trend since 1982, with an average rate of 0.0025 mW m-2 nm-1 sr-1 per decade during 1982-2000 and 0.0038 mW m-2 nm-1 sr-1 per decade during 2001-2022. Our LCSIF data provide opportunities to better understand the long-term dynamics of ecosystem photosynthesis and their underlying driving processes.

Motivation & Objective

• To overcome the limited temporal coverage of satellite SIF observations, which are mostly restricted to the past decade.
• To extend high-resolution SIF estimates back to 1982 using long-term, bias-corrected surface reflectance data from AVHRR and MODIS.
• To develop a two-band neural network model that accurately reproduces SIF using only red and near-infrared reflectance, enabling long-term continuity.
• To validate the reconstructed LCSIF against satellite SIF and in-situ GPP, ensuring its reliability for ecosystem monitoring.
• To quantify long-term trends in terrestrial photosynthesis and assess the impact of climate and environmental change on global vegetation.

Proposed method

• Calibrate and orbit-correct AVHRR reflectance data against MODIS reflectance during the overlapping period (2001–2022) to reduce biases.
• Use a neural network model trained on OCO-2 SIF and corresponding red and near-infrared reflectance from MODIS to predict SIF.
• Apply the trained model to bias-corrected AVHRR reflectance data to generate SIF estimates from 1982 to 2000.
• Ensure spatial continuity by applying the same model across all global land surfaces using consistent input reflectance bands.
• Validate the reconstructed LCSIF against independent satellite SIF (OCO-2) and in-situ GPP measurements across diverse vegetation types.
• Compare LCSIF performance with widely used vegetation indices (NDVI, kNDVI, NIRv) to assess its representativeness of photosynthetic activity.

Experimental results

Research questions

• RQ1 Can a two-band reflectance model accurately reconstruct long-term SIF using only red and near-infrared reflectance?
• RQ2 How does the reconstructed LCSIF compare to satellite SIF observations and in-situ GPP measurements across different biomes?
• RQ3 What are the long-term trends in global photosynthetic activity from 1982 to 2022, and how have they changed over time?
• RQ4 Does the LCSIF product outperform or match the performance of traditional vegetation indices in capturing photosynthetic dynamics?
• RQ5 To what extent does the integration of AVHRR and MODIS data improve the temporal continuity and reliability of SIF records?

Key findings

• The LCSIF product shows a significant accelerating upward trend in global SIF, with a rate of 0.0025 mW m⁻² nm⁻¹ sr⁻¹ per decade from 1982 to 2000 and 0.0038 mW m⁻² nm⁻¹ sr⁻¹ per decade from 2001 to 2022.
• LCSIF demonstrates comparable or higher correlation with satellite SIF and in-situ GPP estimates than traditional vegetation indices (NDVI, kNDVI, NIRv) across diverse vegetation types.
• The bias-corrected AVHRR reflectance data enabled reliable SIF reconstruction from 1982 to 2000, extending the SIF record beyond the MODIS era.
• The two-band neural network model successfully reproduces OCO-2 SIF with high fidelity, confirming the robustness of the approach despite reduced input data.
• The reconstructed LCSIF dataset provides a continuous, globally consistent record of photosynthetic activity over 40 years, enabling new insights into long-term ecosystem dynamics.
• The study establishes a new benchmark for long-term SIF monitoring by integrating legacy AVHRR data with modern SIF observations through a physics-informed machine learning framework.