Evaluation of the remote-sensing-based DIFFUSE model for estimating photosynthesis of vegetation

Abstract

Vegetation captures carbon from the atmosphere through photosynthesis, the rate of which varies across space, through time and is determined by both physical and biological factors. Methods for estimating photosynthesis (A) vary in their complexity and in which driving processes they capture. Whilst the effect of diffuse shortwave irradiance on A is well understood, few models have explicitly incorporated the diffuse effect into estimates of A. Here we present the DIFFUSE model, a simple, generic, diffuse-light-based method for estimating A at the monthly time scale. This model is based on the assumption that, at the monthly time scale, the majority of variability in A can be explained by the variability in total and diffuse irradiance and in the fraction of shortwave irradiance absorbed by foliage (f). Comparison of model estimates to eddy flux tower-derived monthly A showed that the majority (83%) of variability in observed A could be explained by the DIFFUSE model. The diffuse fraction contributed 5 to 10% of the model's accuracy across many of Australia's coastal regions, but contributed up to 50% in the monsoonal north. Various aspects of the DIFFUSE model were tested including its performance relative to an example of the more commonly used “stress-scalar” type of photosynthesis model. In all tests, the DIFFUSE model performed at least as well as more complex alternative models, and often outperformed them. The strengths of DIFFUSE are its physical basis, its simplicity and transparency, and its minimalist data requirements — all of which are expected to make it useful to a wide variety of contexts and applications.