How can we better benchmark global GPP data using flux tower measurements across timescales?

Gross Primary Productivity (GPP) is the largest flux in the terrestrial carbon cycle. Accurate GPP estimates across diverse spatio-temporal scales are essential for constraining the land carbon budget and understanding ecosystem feedbacks to climate change. While numerous global GPP products exist, they rely on models with differing complexities and assumptions. A persistent challenge remains: it is unclear how effectively these products capture photosynthesis across varying timescales—from rapid diel responses and seasonal dynamics to discrete extreme events and long-term inter-annual trends. This study moves beyond traditional comparisons that treat the entire time series as a whole to develop a robust methodological framework for benchmarking global GPP data across varying time domain. We leverage high-frequency measurements from eddy-covariance flux tower networks, which provide multi-decadal, half-hourly records across diverse plant functional types. Simultaneously, we integrate recent advances in Earth observation, specifically the hypertemporal sampling of geostationary sensors and the long-term consistency of cross-calibrated satellite Climate Data Records (CDRs). The core of our methodology involves developing scale-specific metrics designed to isolate uncertainties at different temporal resolutions. For high-frequency dynamics, we introduce metrics to evaluate the ability of satellite-derived GPP to resolve the diurnal cycle and its response to environmental stressors. For long-term dynamics, we propose diagnostic tools to assess whether products accurately capture the physiological "greening" or "browning" trends observed in multi-decadal tower records. By identifying whether discrepancies originate from structural model deficiencies, input errors, or scaling mismatches, this framework provides a deeper diagnostic understanding of GPP model performance.