Statistical Analyses of Land–Atmosphere Feedbacks and Their Possible Pitfalls

Abstract

In some regions of the world, soil moisture has a typical memory for atmospheric processes and can also feed back to the latter. Thus, a better understanding of feedbacks between soil moisture and the atmosphere could provide promising perspectives for increased seasonal predictability. Besides numerical simulations, statistical analysis of existing GCM simulations or observational data has been used to study such feedbacks. By referring to a recent statistical analysis of soil moisture?precipitation feedbacks in GCM simulations, the authors illustrate potential pitfalls of statistical approaches in this context: (i) most importantly, apparent soil moisture?precipitation feedbacks can often as well or even better be attributed to the influence of sea surface temperatures (SSTs) on precipitation and (ii) the discrepancy between different GCMs is large, which makes the aggregation of individual model results difficult. These aspects need to be carefully evaluated in statistical analyses of land?atmosphere coupling. Results for soil moisture?temperature feedbacks complement the precipitation analysis.