4. Conclusions
Using wavelet coherence analysis, we found that Amazon-scale averaged evapotranspiration (ET ) has strong correlations with precipitation (P ) and the total water storage anomaly (TWSA ) at the annual and multi-year (~2 to ~ 4 year) periodicities. But the strong annual periodicity was no longer significant due to irregular short-term pattern of ET after 2010 (drought event). Moreover, at multi-year scale the correlation from 2006 to 2012, possibly indicating that external climate drivers (drought and El Nino events) have enhanced the impact of groundwater on ET . The spatial distributions of\(\phi_{ET-P}^{1km}\) and \(\phi_{ET-TWSA}^{1km}\) across the Amazon basin have clear large south-north and small east-to-west patterns, and our Budyko framework analysis demonstrates that the water and energy limitation conditions vary significantly between the northern and southern sub-basins of the Amazon. Although the Amazon basin is generally energy limited, some southern sub-basins are water limited in some years.\(\ \phi_{ET-P}^{\text{subbas}}\) and\(\phi_{ET-TWSA}^{\text{subbas}}\) are also well correlated in the southern Amazon. The strength and significance of their correlations are affected by the aridity index (PET / P ) of each sub-basin. The spatial heterogeneity of \(\phi_{ET-P}^{1\ km}\) is negatively correlated with the spatial variation of annual ET , which implies that the spatial variation of ET is the primary cause for the changing pattern of \(\phi_{ET-P}^{1\ km}\) inter-annually.\(\phi_{ET-P}^{1\ km}\) and \(\phi_{ET-TWSA}^{1km}\) show the dynamic changes of the spatiotemporal correlation among ET ,P , and TWSA . The water limited area gradually decreases due to the frequent deforestation in the southern Amazon basin.
The results of the three zones analysis also confirm that the effect of drought on ET has the south-north patterns. The drought events have weak impact on ET with sufficient annual P in the northern Amazon basin, and in the southwest Amazon basin as well, where the main source of ET in the mountain areas is snow and ice cover. But the variation of phase in the southeast Amazon basin is closely related to the drought events. During the drought year (2010), the decreased phases (\(\phi_{ET-P}^{\text{Amazon}}\) and\(\phi_{ET-TWSA}^{\text{Amazon}}\)) likely indicate that ET was supported by both rainfall and groundwater to maintain the same yield compared to the years with sufficient P . After the drought year, when the watershed was no longer water limited,\(\phi_{ET-TWSA}^{\text{Amazon}}\) increased rapidly, possibly implying the groundwater system had recovered and ET was not immediately supported by groundwater but by P , since\(\phi_{ET-P}^{\text{Amazon}}\) is much smaller than\(\phi_{ET-TWSA}^{\text{Amazon}}\) during this recovery period. Thus, Amazon-wide annual ET is possibly not limited by rainfall availability since groundwater plays an important role during dry years.
The energy and water limitations are switched in some regions over time. Whether ET is driven by light-limited or water-limited needs to be studied separately in specific areas. This work expounds a deeper understanding of the control of ET in different regions in the Amazon basin by studying the phase lag between two variables at different scales. However, in most cases, the seasonality of ETis driven by the balance between radiation, rainfall, and vegetation regulations, rather than being completely limited by any one of these factors. Vegetation phenology further increases the complexity of studying the relative importance of controlling ET factors in the Amazon basin. Future, we expect to add datasets about radiation and vegetation to further analyze the factors of controlling ET in the Amazon basin.