3.3 Budyko analysis for sub-basins
To analyze the phase lag patterns across the sub-basins, we applied the
Budyko framework (Figure 5). Since Budyko analysis is only applicable
over annual or longer time frames and for large-scale watersheds, we
analyzed the annual averaged data for each sub-basin. Almost all the
sub-basins are energy limited, but some sub-basins (#5, #9, #16, and
#33) in some years are water limited, which are located in the south of
Amazon (subplots on the bottom row of Figure 5). Plotting\(\phi_{ET-P}^{\text{subbas}}\) and\(\phi_{ET-TWSA}^{\text{subbas}}\) against PET / P shows
that the northern and southern basins are clearly separated (Figures S5
and S6). Except for sub-basins #2, #4, #6, and #13, where PET/ P is greater than ~0.65, the correlations
between \(\phi_{ET-P}^{\text{subbas}}\) and\(\phi_{ET-TWSA}^{\text{subbas}}\) are statistically significant
(Figure S7). When the linear correlation coefficients are greater than
~0.6, the correlations are significant (p
< 0.05 ) (Figure S8).
Water-limited sub-basins tend to have smaller\(\phi_{ET-P}^{\text{subbas}}\) (southern basins, Figures 4 and 5),
which implies a less delayed ET after P . For these
southern basins, the linear correlations between\(\phi_{ET-P}^{\text{subbas}}\) and\(\phi_{ET-TWSA}^{\text{subbas}}\) are more statistically significant,
indicating that if the immediate supply from P is insufficient to
maintain ET , groundwater plays an important role. However, if the
sub-basin is not water limited, the correlation between the phase lags
is not statistically significant, thus groundwater is less effectively
affecting ET than in those water limited sub-basins.