3.5 Phase lag for three zones’ averaged variables
As can be noted from the spatial variations of phases,\(\phi_{ET-P}^{1km}\) and \(\phi_{ET-TWSA}^{1km}\), there are north-south and east-west patterns. These patterns could indicate that there are spatially varied differences in the interaction mechanism ofET , P , and TWSA . To further analyze the differences, we calculated the phases for zone-averaged variables based on kilometer scale phases (Figures 6 and 7). The detailed method for zone partitioning is described in the supporting information (SI). Zone 1 and Zone 2 are marked by the negative phase (blue) and positive phase (yellow), respectively, while Zone 3 is for the phase approaching zero (green) (Figure S9). Zone 1 covers the north half of the Amazon basin with some scattered regions in the mid-south area. The south half part (except the scattered area of Zone 1) and some northern-most regions constitute Zone 3 except for the southwest Andes mountain and the areas with low vegetation cover, which are marked by Zone 2. The results of the phase lag for the three zones’ averaged variables are shown in Figures 9, 10, and 11, respectively.
For Zone 1, the trends of \(\phi_{ET-P}^{Zone\ 1}\) and\(\phi_{ET-TWSA}^{Zone\ 1}\) are both relatively stable during 2002 to 2013 (Figure 9a). The monthly peak and trough of P and ETare out of phase (Figure 9b). The annual P in this zone is larger than those in other zones (Figures 9c, 10c, and 11c), which implies that larger P (and less radiation) suppresses ET . In addition, the phases slight increase in drought years, suggesting the suppression of P on ET decreases along with the reduction of Pand cloud-cover. The small variations of two phases relate to sufficientP , indicating the drought events may have weak impact on theET in Zone 1. The suppression of ET in Zone 1 is most likely that the cloudy conditions limit the energy available to driveET . Thus, this area is considered to be light limited.
As Zone 2 has mountain areas that are covered by low vegetation, snow and ice, and therefore ET is mainly driven by solar radiation. The drought event in 2005 promotes the impact of rainfall on ET(Figure 10a), indicating that Zone 2 was also water limited, which corresponds to the Budyko analysis. The time series of P andET are resonant before 2005. However, the monthly peak and trough of P and ET time series are offset and the amplitude ofTWSA variation slightly increases after 2005 (Figure 10b). These patterns could indicate that P has a positive effect on ETbefore 2005, and suppresses ET after 2005.
The variations of phases for Zone 3 averaged variables correspond well to the drought events (Figure 11). This is the regime with higher aridity index (the lowest annual averaged P among three zones), where P is not large enough to suppress ET and is therefore water limited, and this is corroborated in the Section 3.3. The variations of phases for Zone 3 are also relatively similar to those at the whole Amazon scale (Figure 3). The close response of\(\phi_{ET-P}^{Zone\ 3}\) to the variation of the annual Psupports the hypothesis that when P is small, ET relies on rapid evaporation of rainfall. The large variation and the similar response of \(\phi_{ET-TWSA}^{Zone\ 3}\) to the annualP indicate that groundwater supports ET during the dry periods (Figure 11a) via water supply mechanism (rooting depth and groundwater) and vegetation water requirement (Christoffersen et al., 2014).
ET presents different seasonality in different zones (Figure S9). In August and September, it peaks in Zone 1 while is lowest in Zone 2 with the largest variation. In Zone 3, the lowest ET occurs in June (Figure S10). The seasonality of precipitation in Zone 2 and 3 are similar, but it is slightly different in Zone 1 (Figure S11). The precipitation in Zone 3 is much more intense as there are many more days with daily rainfall larger than 20 mm, which is not the case for Zones 1 and 2. Therefore, controls of ET across the different area of the Amazon basin vary. Evaporation demand (especially net radiation) plays a more important role in wetter forests, and soil moisture (or P ) has larger affects in the relative drier area (Rocha et al., 2009). The soil water storage still remains relatively large after the start of the dry season (i.e. when rainfall is small). When the soils reach their lower water storage capacity, 3 months after the peak of the dry season, the rainy season has already started to provide enough water supply for plants. Therefore, the annual flux of ET remains relatively stable in dry years.