Search

A significant challenge in global change research is understanding how vegetation interacts with the environment to influence ecosystem gross primary productivity (GPP) through carbon assimilation. One emerging objective is to consistently predict GPP fluctuations worldwide by establishing a robust scaling relationship between GPP measured at flux towers and satellite spectral reflectance data. However, a major hurdle in achieving this goal is the discrepancy in spatial resolution between early satellite measurements and eddy flux measurements. By using a large set of growing season data covering 100 site-years in North and Central America, we explored the potential of transforming incident and reflected shortwave (Rg) and photosynthetically active radiation (PAR) measurements into a broadband normalized difference vegetation index (NDVI) and near-infrared (NIR) reflectance of vegetation (NIRv) which simultaneously explains the GPP variability. We found that the broadband NDVI and NIRv derived from Rg and PAR measurements at the daily time scale were highly correlated with Planet Fusion, Landsat-8/9, and Sentinel-2 narrowband NDVI and NIRv across a wide range of climate and ecological gradients. The differences between satellite and broadband NDVI and NIRv were found to be significantly associated with soil background variations, phenological stages, water stress and signal saturation of broadband NIR reflectance at high biomass. The seasonal variability of broadband NDVI and NIRv remarkably captured the seasonality of vegetation phenology, evaporative fraction, GPP and rainfall in different ecosystems. Although saturation of GPP at high NDVI was evident, a linear relationship between broadband NIRv times incident PAR versus GPP indicated the effectiveness of NIRv-based approach to capture the hidden light use efficiency impacts on GPP. Our study concludes that inexpensive measurement of Rg and PAR components can provide reliable information on NDVI, NIRv, and GPP uninterruptedly. This enhances the sensing capability of flux tower sites without requiring additional spectrometer measurements. The proposed in-situ vegetation indices make a compelling case on using radiation signals for handshaking between ecosystem-scale measurements and remote sensing observables relevant to carbon uptake.

Estimating gross primary productivity (GPP) over space and time is fundamental for understanding the response of the terrestrial biosphere to climate change. Eddy covariance flux towers provide in situ estimates of GPP at the ecosystem scale, but their sparse geographical distribution limits larger-scale inference. Machine learning (ML) techniques have been used to address this problem by extrapolating local GPP measurements over space using satellite remote sensing data. However, the accuracy of the regression model can be affected by uncertainties introduced by model selection, parameterization, and choice of explanatory features, among others. Recent advances in automated ML (AutoML) provide a novel automated way to select and synthesize different ML models. In this work, we explore the potential of AutoML by training three major AutoML frameworks on eddy covariance measurements of GPP at 243 globally distributed sites. We compared their ability to predict GPP and its spatial and temporal variability based on different sets of remote sensing explanatory variables. Explanatory variables from only Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance data and photosynthetically active radiation explained over 70 % of the monthly variability in GPP, while satellite-derived proxies for canopy structure, photosynthetic activity, environmental stressors, and meteorological variables from reanalysis (ERA5-Land) further improved the frameworks’ predictive ability. We found that the AutoML framework Auto-sklearn consistently outperformed other AutoML frameworks as well as a classical random forest regressor in predicting GPP but with small performance differences, reaching an r2 of up to 0.75. We deployed the best-performing framework to generate global wall-to-wall maps highlighting GPP patterns in good agreement with satellite-derived reference data. This research benchmarks the application of AutoML in GPP estimation and assesses its potential and limitations in quantifying global photosynthetic activity.