Comparison of multiple remote sensing proxies for gross primary production across North America

Accurate estimation of gross primary productivity (GPP) is critical for understanding terrestrial carbon cycling and its responses to environmental change. Although numerous remote sensing–based proxies for GPP exist, their comparative performance across ecosystems remains uncertain. We assessed six proxies over North America, including three MODISbased vegetation indices (NDVI, NIRv, NIRvP) and three reconstructed solar-induced chlorophyll fluorescence (SIF) datasets (CSIF, GOSIF, RTSIF). Overall, CSIF and GOSIF achieved the highest correlations with GPP. In tree-covered biomes, GOSIF generally outperformed CSIF, with maximum R² in DBF (0.89), ENF (0.74), and WSA (0.69), and secondhighest in MF (0.74). In non-forest biomes (CRO, GRA, OSH, SAV, WET), CSIF consistently showed the strongest correlations (R² = 0.52–0.82). NIRv and NIRvP were similarly correlated with GPP, lower than SIF but higher than NDVI. Compared with NIRv, NIRvP improved performance in ENF (0.73 vs. 0.66) and WSA (0.54 vs. 0.40), likely due to incorporating the photosynthetically active radiation (PAR). Drought effects did not differ markedly among indices, but ENF and GRA exhibited pronounced R² declines, suggesting greater physiological sensitivity to water stress. These results improve understanding of proxy performance for GPP estimation across biomes and underscore the value of integrating physiological (SIF) and structural (VIs) signals. Future work may integrate SIF with NIRvP to overcome SIF’s spatial resolution limits and enhance large-scale ecosystem productivity monitoring under changing climates.