*3.1. Meteorological Conditions*

Precipitation and air temperature data in all study sites (Figure 2), as well as daily observations of solar radiation, air temperature and relative humidity (Figure 3), presented a seasonality that was consistent with the local climatology. Highest radiation incidence was registered in the Caatinga and Pantanal sites, which makes them the warmest sites since air temperature is strongly correlated with solar radiation. The highest monthly temperature was found in the Pantanal site (Figure 2A), with 31.6 ◦C being registered in October, while the Amazon site presented the lowest temperatures (25.4 ◦C in April), which is also probably associated with the fact that radiation is lower (Figure 3A). The Caatinga and Pantanal sites are in phase regarding solar radiation (Figure 3A) and air temperature (Figure 3B) patterns, despite VPD values being three-fold higher in the Caatinga (Figure 3C), while VPD in the Pantanal is similar to that observed in the Amazon. VPD in the Amazon practically doubles until September, along with lower observed accumulated rainfall and higher air temperatures.

**Figure 2.** Monthly variation of mean air temperature (lines) and accumulated rainfall (bars) for each site and period: (**A**)—Amazonia: 2009–2011 (rainfall from INMET station); (**B**) Caatinga: January 2014−July 2015 (rainfall from INMET station); (**C**) Cerrado: 2004−2006; (**D**) Pantanal: 2015−2016).

**Figure 3.** Seasonal variation of the daily mean of meteorological variables: (**A**) global incident radiation (W m<sup>−</sup>2), (**B**) air temperature (◦C) and (**C**) vapor pressure deficit (kPa). The shaded colored areas indicate data standard deviation.

Mean annual precipitation in the Amazon site was approximately 2221 mm, 38% higher than in the Pantanal site (1381 mm) and 25% higher than in the Cerrado site (1668 mm). Caatinga presented an annual accumulated rainfall value of 698.9 mm in 2014 and 376.3 mm until July 2015. The number of months with monthly precipitation <10% of total annual precipitation in each site (Figure 2) varied locally. In the Amazon, seven months met this criterion, while in the Cerrado and Pantanal, six months did as well. In the Caatinga, a total of nine months presented less than 10% of total annual rainfall. Besides the highest precipitation totals, the Amazon also presents the greatest monthly variability of precipitation, particularly in March (Figure 4). September is the month where precipitation in all four sites is the most similar, with the Caatinga site registering null precipitation and the other sites registering precipitation below 50 mm. In March, the difference between the wettest and the driest sites accounts for over 300 mm. Only in the months of June and July did the Caatinga site feature monthly accumulated rainfall higher than the Cerrado and Pantanal sites. Only the Amazon site surpassed 300 mm of precipitation in the months from February to April.

**Figure 4.** Monthly accumulated rainfall boxplots for each site and studied period (Amazonia: 2009–2011; Caatinga: January 2014−July 2015; Cerrado: 2004−2006; Pantanal: 2015−2016).

#### *3.2. Water and Energy Fluxes*

The annual cycle of observed daily ET (Figure 5) showed different patterns in the biomes: (1) Cerrado and Pantanal with maximum water vapor flux in October, decreasing toward March and April as the wet season ends; (2) a well-defined ET cycle in the Amazonia site, peaking in September during the dry season. ET cycles in the Cerrado and Pantanal are similar, with maximum daily ET of approximately 7.0 mm day−1. In July, daily ET values are similar in the Cerrado, Pantanal and Amazon sites, with a similar pattern to what was previously observed for solar radiation and air temperature (Figure 3A,B). The Caatinga site stood out, with values down to three-fold lower if compared to the other sites, despite the similar seasonal pattern (higher daily ET between October and April, lower daily ET between May and September). Maximum ET coincides with higher energy availability (radiation) and higher temperatures in the sites, between October and November.

**Figure 5.** Monthly daily evapotranspiration (mm day−1) boxplots for each site and studied period (Amazonia: 2009−2011; Caatinga: January 2014−July 2015; Cerrado: 2004−2006; Pantanal: 2015−2016).

The monthly seasonal patterns of net radiation and sensible and latent heat fluxes differ strongly from site to site. In the Pantanal, Caatinga and Cerrado, net radiation (Rn) presents values higher than 120 W m<sup>−</sup><sup>2</sup> (Figure 6A) in January and May, with a decreasing pattern the following months. The Amazon site, located further north than the other sites

and therefore more influenced by the Intertropical Convergence Zone during these months, presented lower values, of approximately 100 W m<sup>−</sup>2, with increasing values from May until the peak of the dry season in September. Hourly Rn patterns are similar, with the Amazon site featuring approximately 50 W m<sup>−</sup><sup>2</sup> less than the other sites. However, latent heat flux (LE) (Figure 6B) and sensible heat flux (H) (Figure 6C) have distinct hourly and monthly patterns. Maximum LE in the Pantanal is three-fold higher than in the Caatinga, while the H pattern is the inverse. Rn, H and LE patterns are in phase regarding Pantanal and Amazonia sites, particularly because LE and H sharply increase in the dry season, following the increase in available energy. An opposite relationship is found between Rn and LE in the Cerrado and Caatinga sites, which increase in September and October while LE reduces in the same period.

**Figure 6.** Hourly and monthly variation of energy fluxes: (**A**) net radiation (Rn, W m<sup>−</sup>2), (**B**) latent heat flux (LE, W m<sup>−</sup>2) and (**C**) sensible heat flux (H, W m<sup>−</sup>2).

#### *3.3. Carbon Fluxes and WUE*

According to the daily GPP analysis shown in Figure 7, seasonal changes in GPP are more intense in the Caatinga and Pantanal sites if compared to the Amazonia and Cerrado sites. The coefficient of determination (R2) between observed GPP and evapotranspiration (Figure 8) was also higher in the Caatinga and Pantanal. Maximum values reach up to 9.0 gC m −2 d−<sup>1</sup> in the Cerrado and Amazonia, while the lowest GPP values reach approximately 0.5 gC m<sup>−</sup><sup>2</sup> d−<sup>1</sup> in the Caatinga. The seasonality and intensity of GPP values in the Amazon and Cerrado are similar, while in the Caatinga and Pantanal, they are similar only between June and July. Regarding maximum values, GPP peaks in April in the Cerrado, coinciding

with the period of least radiation availability after the site inundation, while in the Amazon, GPP peaks in October, characterized by high radiation, air temperature and an increase in precipitation after an annual minimum in September.

**Figure 7.** Monthly GPP (gC m<sup>−</sup><sup>2</sup> day−1) boxplot for each site and study period (Amazonia: 2009−2011; Caatinga: January 2014−July 2015; Cerrado: 2004−2006; Pantanal: 2015−2016).

**Figure 8.** Correlation between daily GPP averages and evapotranspiration for each site. The size of the circles indicates the intensity of water-use efficiency (WUE) estimated at each day (gC kg H2O day−1).

We also assessed observed GPP data by comparing them with MODIS satellite data on a monthly scale (Figure 9). Results show a high overestimation of satellite data in drier periods, especially in the Amazonia site (Figure 9A). In the Caatinga site (Figure 9B), measurements are more similar from September to November if we consider the median values and monthly variability. In the Cerrado site (Figure 9C), MODIS GPP underestimates observed data in the dry season (August and September), while in the Pantanal (Figure 9D), daily variability of MODIS data is much more prominent than observed data, despite the coherence in representing the seasonal cycle. Given these patterns, the best correlations were found in the Pantanal (R<sup>2</sup> = 0.31) and Caatinga (R<sup>2</sup> = 0.27) sites (Figure 10), with values similar to what is found in the general literature regarding MODIS GPP assessment (Table 1). Due to the importance of investigating specific local water cycles and the effect of drought on the water balance and carbon uptake, we calculated WUE (Figure 11), showing interesting variability aspects depending on the studied biome. WUE is lower in the site

with the highest evapotranspiration (Pantanal), while a remarkable variability and the highest WUE values are found in the Caatinga site. WUE in the other sites did not present grea<sup>t</sup> seasonal differences, with values reaching approximately 4.0 gC kg H2O day−<sup>1</sup> in the driest months (between September and October). WUE surpassed 15 gC kg H2O day−<sup>1</sup> in the driest month (September) of the driest site (Caatinga), sharply decreasing in October with the occurrence of rainfall. Mean annual WUE values estimated in this study delineate new thresholds if compared to previously reported values in the literature (Table 2), with the Pantanal site featuring WUE lower than 1.0 gC kg H2O year<sup>−</sup><sup>1</sup> and the Caatinga site featuring values near 5.8 gC kg H2O year<sup>−</sup>1, which is much higher than in other dry forests studied in the literature.

**Figure 9.** Monthly GPP boxplot (gC m<sup>−</sup><sup>2</sup> day−1) for eddy covariance observed data (Tower) and MODIS—derived data (Satellite) for the: (**A**) Amazonia; (**B**) Caatinga; (**C**) Cerrado; (**D**) Pantanal sites.

**Figure 10.** Correlation between daily tower GPP × MODIS GPP for each site. The size of the circles indicates the intensity of daily accumulated rainfall (mm).



**Figure 11.** Monthly WUE (gC kg H2O day−1) boxplot for each site and studied period: (Amazonia: 2009 −2011; Caatinga: 01/2014 −07/2015; Cerrado: 2004 −2006; Pantanal: 2015 −2016).


**Table 2.** Comparison between mean annual WUE retrieved in the present study with values for different types of forests found in the literature.

**Table 2.** *Cont.*

