We measured net carbon flux (F(CO2)) and net H2O flux (F(H2O)) by the eddy-covariance method at three Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco)-western hemlock (Tsuga heterophylla (Raf.) Sarg.) sites located in the Wind River Valley of southern Washington State, USA. Stands were approximately 20, 40 and 450 years old and measurements were made between June 15 and October 15 of 1998 in the 40- and 450-year-old stands, and of 1999 in the 20- and 450-year-old stands. Our objectives were to determine if there were differences among the stands in: (1) patterns of daytime F(CO2) during summer and early autumn; (2) empirically modeled relationships between local climatic factors (e.g., light, vapor pressure deficit (VPD), soil water content, temperature and net radiation) and daytime F(CO2); and (3) water-use efficiency (WUE). We used the Landsberg equation, a logarithmic power function and linear regression to model relationships between F(CO2) and physical variables. Overall, given the same irradiance, F(CO2) was 1.0-3.9 mol m-2 s-1 higher (P < 0.0001 for both seasons) at the two young stands than at the old-growth stand. During summer and early autumn, F(CO2) averaged 4.2 and 6.1 mol m-2 s-1 at the 20- and 40-year-old stand, respectively. In contrast, the 450-year-old forest averaged 2.2 and 3.2 mol m-2 s-1 in 1998 and 1999, respectively. Increases in VPD were associated with reduced F(CO2) at all three stands, with the greatest apparent constraints occurring at the old-growth stand. Correlations between F(CO2) and all other environmental variables differed among ecosystems, with soil temperature showing a negative correlation and net radiation showing a positive correlation. In the old-growth stand, WUE was significantly greater (P < 0.0001) in the drier summer of 1998 (2.7 mg g-1) than in 1999 (1.0 mg g-1). Although we did not use replicates in our study, the results indicate that there are large differences in F(CO2) among Douglas-fir stands of different ages growing in the same general area, and that variations in age structure and site conditions need to be considered when scaling flux measurements from individual points to the landscape level.