Climate data induced uncertainty in model-based estimations of terrestrial primary productivity. (13th June 2017)

• Record Type: Journal Article
• Authors: Wu, Zhendong; Ahlström, Anders; Smith, Benjamin; Ardö, Jonas; Eklundh, Lars; Fensholt, Rasmus; Lehsten, Veiko
• Is Part Of: Environmental research letters. Volume 12:Number 6(2017:Jun.)
• Publication Date: 2017-06-13
• Subjects: climate datasets -- GPP -- uncertainty -- LPJ-GUESS -- apparent model sensitivity -- climate data range -- global carbon cycle; Environmental sciences -- Periodicals; Human ecology -- Research -- Periodicals; Environmental health -- Periodicals; 333.7
• DOI: 10.1088/1748-9326/aa6fd8
• Languages: English
• ISSNs: 1748-9326

Abstract

Abstract: Model-based estimations of historical fluxes and pools of the terrestrial biosphere differ substantially. These differences arise not only from differences between models but also from differences in the environmental and climatic data used as input to the models. Here we investigate the role of uncertainties in historical climate data by performing simulations of terrestrial gross primary productivity (GPP) using a process-based dynamic vegetation model (LPJ-GUESS) forced by six different climate datasets. We find that the climate induced uncertainty, defined as the range among historical simulations in GPP when forcing the model with the different climate datasets, can be as high as 11 Pg C yr −1 globally (9% of mean GPP). We also assessed a hypothetical maximum climate data induced uncertainty by combining climate variables from different datasets, which resulted in significantly larger uncertainties of 41 Pg C yr −1 globally or 32% of mean GPP. The uncertainty is partitioned into components associated to the three main climatic drivers, temperature, precipitation, and shortwave radiation. Additionally, we illustrate how the uncertainty due to a given climate driver depends both on the magnitude of the forcing data uncertainty (climate data range) and the apparent sensitivity of the modeled GPP to the driver (apparent model sensitivity). We find that LPJ-GUESS overestimates GPP compared to empirically based GPP data product in all land cover classes except for tropical forests. Tropical forests emerge as a disproportionate source of uncertainty in GPP estimation both in the simulations and empirical data products. The tropical forest uncertainty is most strongly associated with shortwave radiation and precipitation forcing, of which climate data range contributes higher to overall uncertainty than apparent model sensitivity to forcing. Globally, precipitation dominates the climate induced uncertainty over nearly half of the vegetated land area, which is mainly due to climate data range and less so due to the apparent model sensitivity. Overall, climate data ranges are found to contribute more to the climate induced uncertainty than apparent model sensitivity to forcing. Our study highlights the need to better constrain tropical climate, and demonstrates that uncertainty caused by climatic forcing data must be considered when comparing and evaluating carbon cycle model results and empirical datasets.