A comparative study of a multilayer and a productivity (light-use) efficiency land-surface model over different temporal scales

Several recent studies suggest that a simple productivity efficiency model (PEM), based on daily or weekly light-use efficiency, is sufficient to represent the exchange of carbon, water and energy at the land-surface. At the same time, many global land-surface models are becoming more process-based, simulating at high temporal resolution the interception of direct and diffuse sunlight at different depths within the canopy. Quantifying the accuracy and limitations of both types of model has become of great importance. The current study compares a PEM with a more complex (and computationally expensive) multilayer model operating at timesteps as short as 30 min (JULES-SF). Each model is optimised against observed fluxes (net carbon exchange, latent heat, sensible heat and net radiation) within the FLUXNET archive for an unprecedented number of sites (30) and site-years (71). Our main finding is that, after optimisation, the process-based multilayer model performs significantly better than the PEM on all timescales (daily and seasonal). However, the difference in model performance appears to diminish with an increase in measurement timescale. Thus, on average, the modelling efficiency increases from 0.32 (daily) to 0.46 (seasonal) using the PEM approach (r2=0.53→r2=0.53→0.71), whilst it remains close to 0.6 for JULES-SF on both timescales (r2=0.69→r2=0.69→0.75). We find that the maximum number of biophysical parameters that can be tuned against site fluxes (4 observables) is quite limited (typically 3–4). This inference applies to both models despite their considerable difference in complexity.