NIRv and SIF better estimate phenology than NDVI and EVI: Effects of spring and autumn phenology on ecosystem production of planted forests

Phenology plays an important role in affecting carbon sequestration in terrestrial ecosystems in the context of climate change. Remote sensing techniques have been widely used to investigate land surface phenology and the effects of phenology on ecosystem production at regional and global scales. Recently, the near-infrared reflectance of vegetation (NIRv) and solar-induced chlorophyll fluorescence (SIF) have been shown to be more promising metrics of gross primary production (GPP) than the normalized difference vegetation index (NDVI) and the enhanced vegetation index (EVI). However, there is a lack of comparison in the performance of these techniques for deriving phenological metrics. In this study, we explored the consistency in phenological metrics derived from both remote sensing approaches (NDVI, EVI, NIRv, and SIF) and flux tower GPP at six plantations (two broadleaf forests (BF) and four coniferous forests (CF)) in eastern China over the period 2006–2020. The vegetation indices (NDVI, EVI, NIRv) were derived from MODIS data, and SIF was based on the global, OCO-2 based SIF product (GOSIF). We further evaluated the effects of spring and autumn phenology on GPP. The results showed that the flux tower GPP was effectively tracked by NDVI, EVI, NIRv, and SIF (P < 0.001). Meanwhile, the phenological metrics derived from EVI, NIRv, and SIF, including the start of growing season (SOS), the end of growing season (EOS), and the length of growing season (GSL), had significant relationships with those derived from GPP at the six plantations (P < 0.05). NIRv and SIF were more effective at estimating phenological information than NDVI and EVI. In addition, the root mean squared deviation (RMSD) values between the GPP- and NIRv-retrieved phenological dates were less than those derived from NDVI, EVI, and SIF at the BF sites. However, the differences among RMSD values of NDVI, EVI, NIRv, and SIF were not significant at the CF sites. The linear regression analysis showed that the advance of SOSGPP (i.e., SOS derived from GPP) significantly increased GPP (R2=0.29, P < 0.05) over the period from March to April, and the delay of EOSGPP (i.e., EOS derived from GPP) remarkably enhanced GPP (R2=0.61, P < 0.001) over the period from September to October at the BF sites. The relationship of EOSGPP with GPP (R2=0.90, P < 0.05) over the period September-October was strong at the CF sites. In addition, the variations of annual GPP could be captured by GSLGPP × GPPmax, GSLNIRv × NIRvmax, and GSLSIF × SIFmax effectively across the BF and CF sites. These findings can help us understand the potential ability of NIRv and SIF in estimating phenological metrics and in revealing the effects of vegetation phenology on the carbon cycle.