Efficient screening of climate model sensitivity to a large number of perturbed input parameters

摘要: [1] Modern climate models contain numerous input parameters, each with a range of possible values. Since the volume parameter space increases exponentially number parameters N, it is generally impossible to directly evaluate model throughout this even if just 2–3 values are chosen for parameter. Sensitivity screening algorithms, however, can identify having relatively little effect on variety output fields, either individually or in nonlinear combination. This aid both development and uncertainty quantification (UQ) process. Here we report results from sensitivity algorithm hitherto untested modeling, Morris one-at-a-time (MOAT) method. drastically reduces computational cost estimating sensitivities high dimensional because sample size grows linearly rather than N. It nevertheless samples over much N-dimensional allows assessment interactions, unlike traditional elementary (EOAT) variation. We applied EOAT MOAT Community Atmosphere Model (CAM), assessing CAM's behavior as function 27 uncertain related boundary layer, clouds, other subgrid scale processes. For radiation balance at top atmosphere, rank most similarly, but identifies that underplays two convection operate nonlinearly model. MOAT's ranking robust modest algorithmic variations, qualitatively consistent experience.