资源环境科技发展态势分析平台

Considerable variability exists in simulations of the future climate. This variability is caused by differences in the parameterisations across general circulation models (GCMs), the initial conditions used and the different assumptions made as to how emissions will evolve in the future. As a result, there is considerable disagreement between available projections of climate variables, which can be used to quantify the uncertainty each variable exhibits. This leads to the question-which variables (or set of variables) are more reliable for use in climate change impact assessments. This research presents a framework to quantify the relative reliability amongst a range of upper air atmospheric variables. This is made possible by pooling simulations across multiple models, trajectories (scenarios) and initial conditions in a rank-transformed space. A metric named the variable reliability score (VRS) assesses the relative reliabilities across different atmospheric variables on a common scale. The VRS has been applied to calculate the total reliability as well as reliability from each source of uncertainty, namely model, scenarios and initial conditions. This comparison helps to decide if more models, scenarios or ensembles are required for uncertainty analysis of climate change impact assessment.

The variables compared include geopotential height and its north-south difference, specific humidity, eastward wind and northward wind, all at the 500 and 850 hPa pressure levels. These variables were chosen based on availability of data and their documented use in previous climate change impact assessment studies worldwide. A regional assessment of VRS over 21 regions around the world shows that though the magnitude of VRS varies spatially, the ranked reliability of the variable rank remains relatively similar. On average, the lowest reliability is associated with geopotential height, whilst wind speeds and the north-south difference of geopotential height have higher reliability.