On the Comparison of Multiple Types of Soil Moisture Memory

The temporal persistence of soil moisture (SM) anomalies, known as SM memory (SMM), is a potential source of predictability in climate systems and weather forecasting. The concept of SMM was originally applied to monthly basis data and quantified by lagged autocorrelation (LAC-SMM). More recently, with daily SM data against the backdrop of the enhancement of computer resources and the broadening implementation of in situ and satellite sensors, quantification based on drying rates during rainfall-free drydown periods has been investigated (DDP-SMM). This study examines the differences between these two forms of SMM with a theoretical framework and analysis of numerical model outputs from the Coupled Model Intercomparison Project phase 6 (CMIP6). It is found that the spatiotemporal variability of the DDP-SMM can be explained mainly by land surface properties, including SM, evaporation, and runoff. In contrast, LAC-SMM is additionally influenced by the temporal variability of precipitation, incorporating hydrologic cycle feedbacks. The interseasonal and interannual variabilities of the DDP-SMM are minor compared to those of LAC-SMM, and the former is quite consistent between the offline and coupled simulations, as land model properties strongly determine DDP-SMM. LAC-SMM reflects a coupled aspect of memory, encompassing both land and atmosphere. As a verifiable metric, DDP-SMM supports the paradigm of hierarchical system development, where the land model is developed offline before a fully coupled experiment, while the LAC-SMM reflects not only land surface properties but also the temporal variability of model precipitation and land–atmosphere interactions.