2.2.4. Amazon Site

Measurements were conducted on a site located at Tapajós National Forest (TNF, 2◦51S, 54◦58W, Figure 1), near the Santarém-Cuiabá highway (BR-163). The TNF is limited by the Tapajós River in the west and by the BR-163 highway in the east, extending 150 km to the south of Santarém city, Pará state. At the eastern side of the BR-163, the landscape is dominated by agriculture. The tower was installed approximately 6 km west of the highway. The canopy has a significant number of large emergen<sup>t</sup> trees (to 55 m height), *Manilkara huberi* (Ducke) Chev., *Hymenaea courbaril* L., *Betholletia excelsa* Humb. and Bonpl., and *Tachigalia* spp., and a closed canopy at ~40 m [41]; this forest can be considered primary, or "oldgrowth" [42]. Analyzed data comprise CO2 and energy fluxes and meteorological data. Measurements comprise daily and monthly means of hourly data in the period from January 2009 to December 2011. CO2 fluxes were measured at 58 m in height through a closed-path analyzer (LICOR-6262) while a Campbell CSAT3 anemometer was used for tridimensional wind measurements. The 65 m micrometeorological tower is located at an area emerging from within the primary forest with a dense canopy of approximately 40 m in height, reaching up to 55 m for some emerging trees [43]. Rainfall was provided by a INMET station near of site (Belterra).

Figure 1 shows the location of the four studied sites and their respective biomes.

#### *2.3. Instrumentation and Data Processing*

Instrumentation in each of the sites is described in previous publications [3,27,44,45]. Gaps formed due to the exclusion of spurious data during the rigorous data screening process were filled using a marginal distribution sampling method (MDS) described by [46], which accounts for the covariance between fluxes and meteorological variables as well as the autocorrelation of fluxes. In this algorithm, three conditions are identified, and a procedure is adopted accordingly: (1) when flux data are missing, but meteorological data are available (Rg, Ta and VPD), missing data are replaced by the mean value in similar meteorological conditions over a seven-day window; (2) when only radiation data are available, the missing data are replaced by the mean value in similar meteorological conditions over a seven-day window; (3) when there are no meteorological data available, the missing data are replaced by the mean value in the last hour, thus accounting for the daily variability of each variable. If after these steps, data were still not filled, the procedure was repeated with larger window sizes until the gap could be filled. For the gap filling procedure, an automated online tool developed by the Max Planck Institute

for Biogeochemistry (http://www.bgc-jena.mpg.de/~MDIwork/eddyproc/, accessed on 2 February 2022) was used.
