Significance Statement

Gross primary production (GPP), the amount of carbon fixed during photosynthesis in a given length of time, is the fuel of life and one of the key components of the global carbon cycle. Although numerous efforts have been made to estimate GPP at the global scale, its uncertainty has remained high for decades. It is necessary to evaluate the performance of relevant models and supporting datasets against high-quality field measurements to find out where the loose parts are, and then tighten the loose ends. Our study reveals that large biases exist in some of the models and driving spatial data products, and addressing these biases is a high priority in modeling GPP from site to global scales.