The sensitivity of models of gross primary productivity to meteorological and leaf area forcing: A comparison between a Penman–Monteith ecophysiological approach and the MODIS Light-Use Efficiency algorithm

• Published in: Agricultural and forest meteorology 2016-03, Vol.218-219, p.11-24
• Main Author: Alton, Paul B.
• Format: Article
• Language: English
• Subjects: Carbon cycle; FLUXNET; Leaf Area Index (LAI); Light-Use Efficiency (LUE); Moderate Resolution Imaging Spectroradiometer (MODIS); Process-based models
• ISSN: 0168-1923; 1873-2240
• DOI: 10.1016/j.agrformet.2015.11.010

•Impact of forcing uncertainties approaches that of model formulation.•However, model accuracy is most limited by inconsistent calibration datasets.•Ecophysiological models are more sensitive than LUE algorithms to meteorology.•LUE algorithms may allow acceptable simplification of complex ecophysiology.•Latest MODIS product underestimates field LAI especially for grass and crops. The current trend in land-surface and carbon modelling development is largely dichotomous: simple algorithms which minimise the number of biophysical parameters and meteorological drivers versus complex ecophysiologically based models which do not. Understanding the sensitivity of both types of approach to current uncertainties in Leaf Area Index (LAI) and meteorological forcing is an important step in producing accurate model predictions of land–atmosphere carbon exchange. We force two quite disparate models (the Moderate Resolution Imaging Spectroradiometer (MODIS) Light-Use Efficiency (LUE) algorithm and the ecophysiological model JULES-SF) with two LAI forcings (satellite and site-normalised) and two meteorologies (tower-based and reanalysis). Simulations are conducted for 67 sites and 10 vegetation classes. The sensitivity of modelled Gross Primary Productivity (GPP) to both LAI and meteorological forcing, thus derived, is compared with model bias against observed carbon fluxes. Our most novel findings are as follows: uncertainty in model formulation (LUE versus ecophysiological) is at least as important (20% change in simulated GPP) as that pertaining to LAI and meteorological forcing (10–20% change). However, all these uncertainties are modest compared to both model bias (≤30%) and inconsistencies between observational datasets used for model calibration (45%). The ecophysiological model is more sensitive to meteorology (20% change in simulated GPP) than the LUE algorithm (10%) owing to the former's reliance on precipitation and shortwave radiation to calculate, respectively, the internal balances of water and energy.