A cross-biome comparison of daily light use efficiency for gross primary production

Vegetation light use efficiency is a key physiological parameter at the canopy scale, and at the daily time step is a component of remote sensing algorithms for scaling gross primary production (GPP) and net primary production (NPP) over regional to global domains. For the purposes of calibrating and validating the light use efficiency (epsilon(g)) algorithms, the components of epsilon(g) - absorbed photosynthetically active radiation (APAR) and ecosystem GPP - must be measured in a variety of environments. Micrometeorological and mass flux measurements at eddy covariance flux towers can be used to estimate APAR and GPP, and the emerging network of flux tower sites offers the opportunity to investigate spatial and temporal patterns in epsilon(g) at the daily time step. In this study, we examined the relationship of daily GPP to APAR, and relationships of epsilon(g) to climatic variables, at four micrometeorological flux tower sites - an agricultural field, a tallgrass prairie, a deciduous forest, and a boreal forest. The relationship of GPP to APAR was close to linear at the tallgrass prairie site but more nearly hyperbolic at the other sites. The sites differed in the mean and range of daily epsilon(g), with higher values associated with the agricultural field than the boreal forest. epsilon(g) decreased with increasing APAR at all sites, a function of mid-day saturation of GPP and higher epsilon(g) under overcast conditions. epsilon(g) was generally not well correlated with vapor pressure deficit or maximum daily temperature. At the agricultural site, a epsilon(g) decline towards the end of the growing season was associated with a decrease in foliar nitrogen concentration. At the tallgrass prairie site, a decline in epsilon(g) in August was associated with soil drought. These results support inclusion of parameters for cloudiness and the phenological status of the vegetation, as well as use of biome-specific parameterization, in operational epsilon(g) algorithms.