Validating satellite phenology through intensive ground observation and landscape scaling in a mixed seasonal forest

Phenology is a key component of monitoring terrestrial ecosystem variations in response to global climate change. Satellite-measured land surface phenology (LSP) has been widely used to assess large scale phenological patterns and processes. However, the accuracy of LSP is rarely validated with spatially compatible field data due to the significant spatiotemporal scale mismatch. In this study, we employ intensive field observations specifically designed to address this deficiency. High density/frequency spring phenological observations were collected in a mixed seasonal forest during 2008 and 2009. A landscape up-scaling approach was used to derive landscape phenology (LP) indices from plot-level observations in order to validate Moderate-resolution Imaging Spectroradiometer (MODIS) based LSP. Results show that the MODIS Enhanced Vegetation Index (EVI) derived start of spring season (SOS) measure was able to predict LP full bud burst date with absolute errors less than two days. In addition, LSP derived SOS captured inter-annual variations and spatial differences that agreed with ground observations. Comparison of complete time series of LP and LSP revealed that fundamental differences exist between the two observation means, e.g., LP development had increased influence on LSP during the course of spring onset. Therefore, inferring continuous LP processes directly from LSP measures could be problematic. However, using LSP derived from techniques such as logistic curve modeling for extracting seasonal markers appears more promising. This study contributes to a more explicit understanding of the linkages between remotely sensed phenology and traditionally observed (ground-based) phenology.