Process-Based Understanding of Ecosystem Carbon and Water Fluxes Using High-Resolution Remote Sensing

Abstract

Earth’s terrestrial ecosystems are central regulators of the global climate system, mediating essential exchanges of carbon, water, and energy between the land and atmosphere. At the core of these dynamics lie two fundamentally linked processes: photosynthesis, through which plants convert atmospheric carbon dioxide into biomass, and evapotranspiration (ET), the primary pathway by which water returns from terrestrial surfaces to the atmosphere. Monitoring these processes is critical for understanding the terrestrial carbon and water cycles, anticipating ecosystem responses to environmental stress, and supporting sustainable resource management under accelerating climate change. However, traditional remote sensing approaches, heavily reliant on greenness-based vegetation indices, remain insufficient to capture the true physiological functioning of vegetation. These indices primarily reflect canopy structure and often fail to detect subtle, early physiological responses to stress, thereby limiting our capacity for timely intervention and precise estimation of ecosystem fluxes. This thesis contributes to advancing the frontier of ecosystem monitoring by leveraging solar induced chlorophyll fluorescence (SIF), a radiative emission during photosynthesis that provides a direct physiological proxy for photosynthetic activity. At regional scales, the research investigates how the relationship between gross primary productivity (GPP) and SIF varies across India’s diverse agricultural ecosystems. It reveals significant spatial and seasonal heterogeneity: while strong correlations between GPP and SIF exist in homogeneous cropping areas, regions characterized by fragmented land use and lower crop area fractions, such as parts of southern and northeastern India, exhibit reduced seasonal variability in SIF signals. Further, this study show that environmental factors, including soil moisture, radiation, temperature, and canopy structure, critically influence the GPP–SIF relationship, with soil moisture emerging as a dominant driver during early growing periods and radiation and temperature gaining importance during peak and late growth phases. Such mechanistic insights deepen our understanding of how physiological signals integrate environmental variability to shape crop productivity, offering pathways for improving GPP estimation across India’s agricultural landscapes. Building on this mechanistic perspective, the study explores the innovative use of SIFyield the ratio of SIF to absorbed photosynthetically active radiation (APAR), as a early-warning indicator for detecting emerging drought stress. Unlike absolute SIF, which may remain elevated under drought conditions due to high canopy light absorption, SIFyield normalizes the fluorescence signal, exposing subtle declines in photosynthetic efficiency that precede visible symptoms. Across India’s diverse climate zones, empirical and causal analysis show that SIFyield consistently leads traditional stress indicators such as meteorological drought indices and soil moisture by one to two months, demonstrating its exceptional potential for providing early alerts of agricultural drought. These insights establish SIFyield as a transformative tool for proactive drought management, enabling interventions well before significant impacts on crop yields occur and contributing to greater resilience in agricultural systems facing climatic extremes. Recognizing that carbon and water fluxes in ecosystems are deeply intertwined, given that stomatal conductance simultaneously governs both photosynthesis and transpiration, the study further investigates the use of SIF for estimating water flux (ET). By comparing a semimechanistic approach, which integrates SIF, vapor pressure deficit (VPD), and the photochemical reflectance index (PRI) within the Penman–Monteith framework, to a machine learning (ML) model that incorporates SIF, VPD, soil water content (SWC), and PRI, the research demonstrates that ML consistently outperforms rigid mechanistic models. ML captures complex, nonlinear interactions between physiological signals and environmental drivers that semimechanistic models struggle to represent, especially under stress conditions. Explainable ML techniques, such as SHAP (Shapley Additive Explanations), further illuminate the dominant roles of SIF and VPD in determining ET under normal conditions, while highlighting the heightened influence of SWC and PRI during periods of environmental stress. These findings underscore the limitations of traditional modeling approaches and reveal the potential for hybrid systems that blend process understanding with data-driven flexibility, ultimately improving ET estimation across diverse ecosystems. Extending this physiological monitoring to the global scale, the research pioneers a novel approach for quantifying ecosystem functional resilience by applying critical slowing down theory to multi-decadal SIF time series. This innovative analysis translates temporal patterns in SIF variance and autocorrelation into recovery rates, offering the first global maps of photosynthetic resilience. The results reveal pronounced latitudinal gradients: boreal region exhibit strong recovery capacities, whereas tropical ecosystems show widespread declines in resilience, closely linked to rising temperatures, increasing vapor pressure deficits, and persistent soil moisture deficits. Furthermore, land-use and land-cover types display distinct resilience trajectories, reflecting both physiological traits and exposure to climatic stressors. These findings highlight the invaluable role of physiological indicators like SIF in detecting early signs of ecosystem instability, enabling more proactive management of global ecosystems under the pressures of climate change. Taken together, the work presented in this thesis establishes solar-induced chlorophyll fluorescence as far more than a novel remote sensing signal. It emerges as a powerful physiological fingerprint that bridges scales from individual leaves to entire global biomes, connecting the carbon and water cycles while revealing early signals of ecosystem stress and resilience. By transitioning from structural proxies to direct measurements of physiological processes, this research lays a transformative foundation for next-generation ecosystem monitoring, more accurate Earth system models, and timely interventions that can safeguard biodiversity, ensure agricultural sustainability, and support climate mitigation strategies in an era of unprecedented environmental change.