Photosynthetic light use efficiency of three biomes across an east-west continental-scale transect in Canada

Light use efficiency (LUE) is used widely in scaling and modeling contexts. However, the variation and biophysical controls on LUE remain poorly documented. Networks of eddy covariance (EC) towers offer an opportunity to quantify epsilon(g) the ratio of P, gross primary productivity, to Q(a), absorbed photosynthetically active radiation (PAR), across climate zones and vegetation types. Using data from the Fluxnet Canada Research Network (n = 24 sites) in 2004, we examined the relationship between daily and yearly epsilon(g) driving variables, and site characteristics on a site-specific and plant functional type (PFT) basis using tree regression and linear regression. Data were available for three biomes: grassland, forest, and wetland. Yearly epsilon(g) values ranged from 0.1 to 3.6 g C MJ(-1) Q(a,) overall. Daily eg was highest in the grassland (daily median +/- interquartile range: 3.68 +/- 1.98 g C MJ(-1) Q(a)) intermediate in the forested biome (0.84 +/- 0.82 g C MJ(-1) Q(a)), and lowest for the wetlands (0.65 +/- 0.54 g C MJ(-1) Q(a)). The most important biophysical controls were light and temperature, to the exclusion of water-related variables: a homogeneity of slopes model explained C. 75% of the variation in daily epsilon(g). For a subset of sites with diffuse PAR data, the ratio of diffuse to total PAR, a proxy for cloudiness, was a key predictor. On the yearly time scale, epsilon(g). was related to leaf area index and mean annual temperature. Aggregating to PFTs did not show functional convergence within any PFT except for the three wetland sites and the Picea mariana toposequence at the daily time step, and when using the Koppen climate classification on a yearly time step. The general lack of conservative daily epsilon(g) behavior within PFTs suggests that PFT-based parameterizations are inappropriate, especially when applied on shorter temporal scales. (c) 2006 Elsevier B.V. All rights reserved.