This study summarizes the results from a large number of forecast experiments with a global model with the objective of tagging errors that arise from different components of model dynamics and physics. To sort out such errors in non-linear systems is generally very difficult. In this study, we show examples of model forecasts that illustrate deficiencies (i.e., errors) that arise in the parameterization of cumulus convection and in the planetary boundary layer (PBL) physics. Our proposed Factor Separation method enables us to tag the systematic errors in our PBL formulation of moisture convergence. We note that the PBL scheme overestimates by roughly a factor of 2 the convergence near the tropical cloud base. These errors were the largest over meso-convective regions of deep convection. With such efforts at error tagging further improvements in modeling and forecasts can be achieved. We also address further work in this area of research.

Introduction

Factor Separation (FS), introduced by Stein and Alpert (1993) and Alpert et al. (2006), permits an explicit separation of atmospheric synergies among several salient features of a forecast model. They have exploited the FS method covering many scales of modeling to answer questions on issues such as the role of water and carbon dioxide for the understanding of global change. Their proposed method has many wide-ranging applications, as seen from the contents of this book. Our proposed study follows the same rationale in asking how one can sort out and tag errors that arise from different components of a non-linear model during the evolution of weather or climate.