Long-term patterns of wetland evapotranspiration across climates. The long-term patterns of ET in South American wetlands follow a climate gradient (Fig. 1). The combination of high precipitation and available energy (assessed here as net radiation; Rn) in equatorial wetlands (Fig. 2a and 2b) produces the highest annual ET rates (1296–1542 mm/year), and these areas also exhibit the greatest leaf area index (LAI) values (Fig. 2c). Meanwhile, the lowest annual ET rates occur in the temperate wetlands (743–1128 mm/year). Because of persistent cloud cover in the Amazon, ET was not estimated for the months of January to May; however, the available period is representative of the flood maximum and minimum stages, enabling us to understand the seasonal dynamics of ET in this region, while the small ET amplitude in Amazon enables us to estimate its annual rate. While annual Rn is relatively similar between equatorial and tropical wetlands, the higher water availability (precipitation) leads to higher ET in the former. In turn, the fraction of available energy that is converted into ET depends on surface water availability, i.e., the extent to which incoming waters accumulate on the terrain surface. Our regional-scale analysis indicates that the higher the wetland flood fraction, the higher the evaporative fraction, independent of climate type (Fig. 2d). Examining regional flood fraction values that exceed 0.3, however, reveals that the evaporative fraction reaches a plateau around 0.7–0.8. The fraction of precipitation that becomes evapotranspiration (ET/P) ranges from 0.5–0.7 in equatorial and most tropical wetlands to greater than 0.8 in temperate ones, with the highest values in the Pampas and Paraná floodplains (Fig. 2e). In the Pampas, relatively little runoff is routed out of the wetland through a consolidated river drainage network, and almost all precipitation turns into ET. An ET/P ratio greater than unity for the Pampas suggests a memory effect within the system that is associated with groundwater storage42. Meanwhile, the Paraná floodplain receives water from the upstream basin, and total water inflow exceeds precipitation.
Comparing ET in wetlands and adjacent uplands at the regional scale, we uncover greater differences among temperate wetlands, which are located within water-limited environments (E0/P > 1, where E0 stands for atmospheric evaporative demand, indicated here by the reference ET), with values reaching 29%, 23%, 13%, and 5% for the Paraná, Iberá, Pampas, and Chaco wetlands, respectively (Fig. 2e). For instance, for the case of Paraná river, this means that the floodplain inundation due to waters coming from upstream may increase the annual average latent heat flux by around 20 W/m², and reduce sensible heat flux by the same amount. This is equivalent to an annual evapotranspiration increase of around 250 mm/year in relation to uplands (see values for other wetlands in Supplementary Fig. S1). The increased surface water availability in these areas enables them to meet the evaporative demand, i.e., higher ET/E0 values in wetlands than uplands. In contrast, the ET/E0 rate is close to unity (i.e., points close to the 1:1 line in Fig. 2e) for both uplands and wetlands in the equatorial regions. The high flood fraction in large portions of the equatorial Magdalena Mompós depression and the tropical Pantanal wetlands (Fig. S2) produces significant wetland-upland differences (19% and 16%, respectively). Flooding in these areas stems from a combination of geomorphological processes (i.e., depressions associated with sedimentary basins) and large river inflows (the Magdalena and Paraguay rivers, respectively) within a mostly non-forest vegetation landscape. In contrast, the equatorial wetland forest in the Central Amazon exhibits only small differences (3–5%) because the large upland forest (“terra-firme”) maintains high ET rates throughout the year. Furthermore, trees in the Amazonian wetlands are less diverse and smaller and include a larger proportion of deciduous species than those in the adjacent uplands; therefore, wetland trees are expected to exhibit lower annual transpiration rates than upland ones16,43,44. The small wetland-upland ET difference (3–5%) is due to open water evaporation, and this effect may be further enhanced by reduced precipitation in the Amazonian floodplains compared to the uplands (-5%23), which increases energy availability in the floodplains because of decreased cloud cover. On the other hand, most tropical hyperseasonal savannas (Orinoco, Negro, Moxos, and parts of the Bananal and Pantanal) are surrounded by upland forests. In this case, while the flooded savannas exhibit higher ET rates than do the adjacent non-flooded savannas, the surrounding forests maintain high ET rates year-round and therefore exhibit annual ETs that exceed those of the flooded savannas (Fig. S3).
Seasonal patterns of wetland evapotranspiration and its environmental drivers. We investigate the role of environmental drivers on ET dynamics by correlating monthly ET estimates with six main drivers: flooding, precipitation, LAI, Rn, vapor pressure deficit (VPD), and wind speed (Ws; Fig. 3a; see detailed correlation matrices and scatterplots in Figs. S4 and S5). While precipitation is identified as the main driver of annual ET magnitude in all wetlands (Fig. 2), the available energy (Rn) is the main driver of ET seasonality for equatorial (high water availability throughout the year) and temperate wetlands (high Rn seasonality). In temperate climates, the wet season is in phase with Rn, producing the highest ET rates in the austral summer (Fig. 1). In turn, the tropical wetlands face a dry season water deficit; thus, water availability (measured via both flooding and precipitation variables) complements Rn as a major ET driver.
Wetland LST exhibits strong variation across the continent (Fig. 3b). While it is mainly driven by flooding in equatorial wetlands (a strong correlation between flood fraction and LST), a different pattern occurs in the temperate wetlands, where the strong Rn seasonality is responsible for the large LST amplitude (Fig. S4). Meanwhile, the tropical wetlands exhibit an intermediate pattern, with both Rn and flooding moderately correlated with LST. These differences have major implications for ET dynamics.
While, in average years, ET seasonality follows strong Rn variation, years with anomalous flooding lead to anomalously high ET in many South American wetlands (see the correlation for ET anomaly in Fig. 3). This is the case of the Argentine Pampas, which is characterized by an erratic interannual flooding pattern, with alternately flood-rich (2000–2004 and 2011–2015) and flood-poor periods (2005–2010). While the highest ET rates occurred in the flood years (Fig. 4a), high ET values persisted some years after the main flooding period (2000–2004), indicating memory effects on groundwater storage in the Pampas42.
A strong correlation between flood fraction and ET/E0 for many wetlands (Fig. 3) also corroborates the finding of our long-term scale analysis, i.e. that the flooding process generally enables the water supply to meet the evaporative demand. In the areas downstream of the Amazon and the Magdalena and the hyperseasonal savanna wetlands (Fig. S6), the highest ET/E0 values occur during the flood peak. In the temperate wetlands, however, ET/E0 is mainly driven by Rn. While the temperate wetlands exhibit the greatest annual amplitude and the highest monthly rates (e.g., 170 mm/month in December in Iberá; Fig. 1), confirming the hypothesis that grassland wetlands have ET rates as high as forested ones in some months45, the equatorial wetlands exhibit nearly constant ET rates (from 100 to 130 mm/month in the Central Amazon). Heavy cloud cover in the equatorial Amazon wetlands prevents available energy rates from increasing, especially during the wet season. Months with more significant flooding are associated with the smallest wetland-upland differences between hyperseasonal savanna wetlands and the surrounding upland forests (Negro, Orinoco, Moxos, and Bananal). However, the differences increase during the dry season, with the forests exhibiting higher ET (Figs. S1 and S7). During the wet season, the differences between the flooded savannas and non-flooded forests likewise decrease. Conversely, in the Central Amazon, greater flooding produces a greater difference.
The role of flood propagation on floodplain evapotranspiration. In terms of flooding mechanisms, inland wetlands can be classified into 1) interfluvial wetlands, which are associated with local runoff and vertical hydrological processes (endogenous processes), 2) river floodplains, where flooding is related to the overbank transfer of waters from upstream areas (exogenous processes), and 3) a combination of both7,46. The flood wave propagation along river floodplains produces a delay of many months between maximum precipitation and flooding at the farthest downstream reaches of the Paraná (two months), the Amazonian (three months), and the Pantanal (six months) wetlands. These delays derive from a combination of vertical soil wetting, which can produce a delay of up to two months, as seen in the interfluvial wetlands in Fig. 1, and flood wave translation, which can lead to longer delays depending on river hydrodynamics. We demonstrate that flood propagation largely affects the ET dynamics in the Pantanal, which experiences more than 100,000 km² of flooding annually6. While precipitation peaks in January across the entire region, the month of maximum flooding varies from March in the upstream reaches to July in the downstream ones (Fig. 4b). ET climatological behavior is driven by the complementary role of flooding and evaporative demand, which reaches its maximum between August and November in the entire region, according to VPD and Ws patterns (Fig. S8). Consequently, the upstream regions that experience their flood peak in March (i.e., Region 1 in Fig. 4b) exhibit ET peaks in both March and November. Conversely, the downstream regions, where evaporative demand and surface water availability due to floodplain inundation are in phase, reach their annual ET peak between August and November. The higher ET rates in the downstream regions are likely associated with open water evaporation. Meanwhile, the Pantanal’s unique geomorphology causes longer delays in its flood wave translation compared to the other two large floodplains addressed here (the Amazon and the Paraná). However, the effects of flood propagation on ET dynamics are also evident in these wetlands, where anomalous ET rates are strongly correlated with periods of anomalous flooding (Fig. 3).
Evapotranspiration of floodplain forests across biomes. The partition of wetland ET into vegetation transpiration and open water and soil evaporation is difficult to disentangle. Open water evaporation tends to increase with surface water availability and can potentially offset plant transpiration, which may, in turn, be reduced by flooding due to anoxic or hypoxic conditions, an increase of toxic compounds, or a decrease in the availability of nutrients47–49. These effects can induce stomatal closure, while flood adaptation measures, such as adventitious roots and aerenchyma, can, in contrast, increase stomatal conductance during flood peaks50,51. The total canopy conductance depends on stomata opening and total leaf area; therefore, the various adaptation strategies plants use to cope with alternating cycles of flooding and drying ultimately determine transpiration seasonality.
Here we compare ET processes in five floodplain forests across South America, using the MODIS Enhanced Vegetation Index (EVI; see Methods) as a proxy of stomatal activity and forest dormancy under flooding conditions52–54. Although the EVI signal may be affected by the inundation itself, its use here is reasonable given the high tree cover in the assessed floodplains. Both water excess and a water deficit can hinder wetland forest activity across South America, depending on plant adaptation and local-scale factors, including the soil’s water retention capacity. The highest ET rates occur during the wet season (or dry-wet transition in the Amazon, which corresponds to the floodplain leaf shedding period44) in all floodplains, while the flood peak leads to reduced vegetation activity or forest dormancy (low EVI) in all but the Paraná floodplain (Fig. 5). The phenological seasonality of the Paraná forests is predominantly driven by flooding, which occurs during the dry season (austral winter) due to the flood wave translation along the upstream river network, and its associated nutrient-rich sediments17. In the Paraná, the lowest EVI occurs during periods of receding waters, which correspond with periods of minimum energy availability, and EVI levels remain low until the onset of the wet season. In the Bananal forest, we observe that ET is not water-limited; in the adjacent floodable savannas, however, ET exhibits the opposite characteristic and decreases in the dry season (see the BAN flux tower in Fig. S9). A small—and, indeed, below the annual average—ET peak, which may be associated with soil evaporation,32,52 occurs in the month of maximum flooding in the Bananal and Pantanal floodplains. ET decreases in the Orinoco floodplains during flooding; in contrast, the large-scale flooded savanna experiences its maximum ET during the flood period (Fig. 1), which suggests the greater importance of direct surface evaporation in this region associated with limited vegetation activity in riparian forests during flooding. Finally, the assessed Pantanal floodplain is a Vochysia divergens monodominant forest, which is not water-limited during the dry season due to plant adaptation strategies and exhibits a relatively high soil moisture content throughout the entire year45,49. In this case, the reduced vegetation activity observed during the dry season may be related more strongly to a reduction in available energy.
Contrasting mechanisms in the Central Amazon. In the Central Amazon, the maximum wetland ET occurs at the transition between the dry and wet periods, which corresponds with maximum VPD, Ws and E0 values. This pattern is similar to the ET pattern that occurs in the Amazon uplands55, where ET is energy-limited and strongly correlated with Rn, the highest values of which occur in the dry-wet period transition when cloud cover is limited. In addition, as discussed in the previous section, forest transpiration in flooded areas tends to be limited by flooding, which peaks during the dry season because of the river flood propagation process. In contrast, the ET/E0 ratio peaks during maximum flooding in the most floodable, downstream reaches (Fig. S6), highlighting the role of open water evaporation. On a large scale, two compensating effects interact to determine the actual annual ET in the Central Amazonian wetlands (Fig. 6). First, more dense tree cover occurs in the upper reaches (roughly upstream from the city of Manaus; Fig. 6d), whereas the downstream reaches—with their lower tree height and a large proportion of native herbaceous plants—are associated with lower precipitation rates, a longer dry season, smaller flood depths, less nutrient availability56, and the conversion of forestlands to agricultural areas and pastures57. Second, the downstream reaches feature more lakes and experience flooding for a lengthier period of the year, and both of these characteristics increase ET. The combination of these two opposing effects produces the highest annual ET rates in the upstream reaches, which exhibit greater tree cover, while a decreasing ET trend is observed in the downstream direction, with an exception (i.e., an increase) only in the furthest downstream parts with the largest flood fraction. The downstream Amazon surpasses the upstream region only during the high flood period (June and July) because of open water evaporation (Fig. S1).