What is driving the phenology across the Amazon? A benchmark for land-surface models

We present a study which has used a Fourier-based time-series analysis method applied to 8 years of EO-derived observations of phenology (vegetation indices) and their potential drivers (downward shortwave radiation and precipitation). We use these data sets to test the correlation (coherency) of the phenology to the driving variables and to determine the relative timing of their seasonal cycles. This has led to a better understanding of the linkages between phenology and their driving factors. Typically Amazonian canopy vegetation has varying timing in phenology on small (sub-pixel) scales. However, at the spatial resolution of 1 km2 spectral analysis shows that the sub-pixel phenology is well synchronised for a large proportion of the Terra Firme forest. Across the whole of northern South America about 60% of the land shows a significant seasonal cycle in grassland- and forest-biomes. Within this area having seasonal cycles, 43% has phenology in-phase with radiation of which 75% is Terra Firme-type forest and 25% is mainly grassland. 37% of the area with seasonal cycles is in-phase with precipitation, of which 40% is grassland and the remainder is forest. These results are in agreement with recent research that suggests much of the Amazon over humid tropical forests may be radiation-driven. However, we also support the hypothesis that in places both radiation and precipitation are influential, but may not be completely in-phase with the phenology. We identify core areas where the phenology-radiation and phenology-precipitation relationships are most apparent. The information regarding spatial controls of phenology provides a benchmark for land-surface modellers.