[1] The seasonal pattern of evapotranspiration (expressed as latent heat flux Q e ) for a 28-to 30-m-tall tropical transitional (ecotonal) forest was quantified over an annual cycle using eddy covariance measurement and micrometeorological estimation techniques. The study was conducted near the city of Sinop, in northern Mato Grosso, Brazil, which is located within the ecotone of tropical rain forest and savanna (cerrado). Although the majority of net radiation (Q*) was consumed by Q e (50-90%), seasonal variations in Q e were large and positively correlated with precipitation. Total daily Q e for the dry season (June-August) was on average 6.0 MJ m À2 d À1 , while daily Q e for the transition (October-November and April-May) and wet (December-March) season periods were 7.5 and 10.0 MJ m À2 d À1 , respectively. The seasonal variation in midday (0900-1500 LT) surface conductance (g s ) was also positively correlated with precipitation. Analysis of the ''decoupling factor'' ( ) indicated that the forest was strongly coupled to the atmosphere ( = 0.1-0.3) over the dry season and transition periods, suggesting that Q e was under relatively strong stomatal control. Although rainfall during the study period was above the long-term (30-year) average, our results indicate that the seasonal dynamics of Q e for the tropical transitional forest were more comparable to tropical savanna than to rain forest.

Summary 1.Tower-based eddy covariance and measurements of the vertical CO 2 concentration gradient within the canopy were used to quantify the seasonal variations in the net ecosystem CO 2 exchange ( NEE ) of a 28 -30 m tall transitional tropical forest (cerradão). The study was conducted near the city of Sinop, Mato Grosso, Brazil (11 ° 24·75 ′ S; 55 ° 19·50 ′ W), which is located in the ecotone of two major regional ecosystem types of South America (tropical rainforest and savanna). 2. The NEE during the dry season (August-September) was in balance, but during the transition period between the dry and wet seasons (October-November) the cerradão stand became a net source of 50-150 mmol m -2 day -1 CO 2 to the atmosphere. Measurements during the wet season (February, April) indicate that the forest was a net sink of between -55 and -102 mmol m -2 day -1 . 3. The NEE of the transitional tropical forest was more similar to that of tropical rainforest during the wet season, but during the dry season the NEE of the transitional forest was more similar to that reported for tropical savanna. The data suggest that seasonal variations in rainfall have important implications for the seasonal pattern of NEE in cerradão.

This research utilized tower-based eddy covariance to quantify the trends in net ecosystem mass (CO 2 and H 2 O vapor) and energy exchange of important land-cover types of NW Mato Grosso during the March-December 2002 seasonal transition. Measurements were made in a mature transitional (ecotonal) tropical forest near Sinop, Mato Grosso, and a cattle pasture near Cotriguaçú , Mato Grosso, located 500 km WNW of Sinop. Pasture net ecosystem CO 2 exchange (NEE) was considerably more variable than the forest NEE over the seasonal transition, and the pasture had significantly higher rates of maximum gross primary production in every season except the dry-wet season transition (September-October). The pasture also had significantly higher rates of whole-ecosystem dark respiration than the forest during the wetter times of the year. Average ( AE 95% CI) rates of total daily NEE during the March-December 2002 measurement period were 26 AE 15 mmol m À2 day À1 for the forest (positive values indicate net CO 2 loss by the ecosystem) and À38 AE 26 mmol m À2 day À1 for the pasture. While both ecosystems partitioned more net radiation (R n ) into latent heat flux (L e ), the forest had significantly higher rates of L e and lower rates of sensible heat flux (H) than the pasture; a trend that became more extreme during the onset of the dry season. Large differences in pasture and forest mass and energy exchange occurred even though seasonal variations in micrometeorology (air temperature, humidity, and radiation) were relatively similar for both ecosystems. While the short measurement period and lack of spatial replication limit the ability to generalize these results to pasture and forest regions of the Amazon Basin, these results suggest important differences in the magnitude and seasonal variation of NEE and energy partitioning for pasture and transitional tropical forest.

The net ecosystem CO2 exchange (NEE) of a 28–30 m tall transitional (ecotonal) tropical forest of the Brazilian Amazon was quantified using tower‐based eddy covariance. Measurements were made between August 1999 and July 2001 and were used to develop nonlinear statistical models to assess daily variations in ecophysiological parameters and provide annual estimates of NEE, gross ecosystem CO2 exchange (GEE), and respiration (Re). Diurnal trends in NEE were correlated with variations in photosynthetic photon flux density (Q), vapor pressure deficit (V), and temperature. Seasonal trends in the CO2‐flux components estimated from nonlinear regression (Amax and R0) were highly correlated with soil water availability and canopy structural properties (LAI and litter production). These results suggest that variations in soil water content can affect rates of canopy photosynthesis and whole forest respiration by altering both physiological processes and canopy structural properties. Estimates of the annual NEE suggest that the forest was in balance with respect to CO2 during the study period, which, in terms of rainfall, was a relatively typical period compared to the 30‐yr average rainfall regime. Our results also suggest that the warmer and dryer microclimate and decline in LAI that accompany land‐cover change will cause transitional forests to be sources of CO2 to the atmosphere.