Temporal resolution of vegetation indices and solar-induced chlorophyll fluorescence data affects the accuracy of vegetation phenology estimation : a study using in-situ measurements

Abstract

Vegetation phenology plays a critical role in inter-annual changes of the terrestrial carbon cycle. Land surface phenology (LSP) has been widely used to monitor vegetation phenology from remotely-sensed data (RSD) across multiple spatial scales. However, it remains unclear how the temporal resolution of RSD influences the accuracy of LSP estimation. This study systematically analyzed the influences of temporal resolution, including the observation temporal resolution (OTR) and composite temporal resolution (CTR), of RSD on LSP estimation from continuous ground-based remote sensing observations. Specifically, this study quantitatively assessed the sensitivity of LSP estimation to temporal resolution from both structural indicators including normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and near-infrared reflectance of vegetation (NIRV), and physiological indicators including photochemical reflectance index (PRI), gross primary productivity (GPP), and solar-induced chlorophyll fluorescence (SIF). The results showed that the effects of temporal resolution of RSD on LSP estimation can be divided into systematic error and random error. The systematic error of CTR was caused by the methods of LSP estimation and affected by data compositing methods, while the random error of both OTR and CTR was caused by data noise. The sensitivity of SIF and GPP to temporal resolution (both OTR and CTR) in LSP estimation was higher than that of vegetation indices (NDVI, EVI, NIRV, PRI) due to higher data noise. Furthermore, in LSP estimation, the selection of the required temporal resolution of RSD was directly related to the data quality. The results highlight the importance of the temporal resolution in LSP estimation from RSD and provide two possible insights to reduce the errors of LSP estimation. First, in the case of the appropriate data compositing strategies and very little data noise, temporal resolution (both OTR and CTR) can be considered to have little influence on LSP estimation. Second, most of the difference in sensitivity of different indicators to temporal resolution comes from external factors, including observation and model or algorithm errors, rather than the properties of indicators.