Variability of soil moisture memory for wet and dry basins

Soil moisture memory (SMM) is not only important for atmospheric weather/climate forecasting, but may also be useful in flood and drought prediction. Despite their importance, SMM studies are restricted in certain regions due to the scarcity of soil moisture data. To overcome this limitation, this study explains the variability of SMM in wet and dry basins, and shows an alternative way to predict the basin scale SMM using observed precipitation and potential evapotranspiration information only. This study presents the basin average SMM in the form of a timescale that indicates the duration of significant auto-correlations at 95% confidence intervals. The soil moisture autocorrelations were calculated using observed precipitation, potential evapotranspiration, streamflow and soil moisture data sets simulated using the XinAnJiang (XAJ) model, for 26 river basins across the USA. The XAJ model's capability to simulate seasonal cycles (temporal anomalies) of soil moisture was validated against cycles from the observed data set of the Spoon River basin of Illinois State, USA. Based on the validation experience, the XAJ model was thereafter used to simulate soil moisture data for the analysed basins. Basin scale SMM timescale ranges were computed from 11 to 133 days. The SMM timescale is highly influenced by precipitation variability and exhibits strong seasonality. Dry basins tend to show the highest memory during the winter months (December to February) and lowest in late spring (May). In contrast, wet basins have the lowest memory during winter and early spring (December to April) and highest in the late summer and early autumn (July to September). The SMM timescale displayed an exponential relationship with the basin aridity index, with an r(2) value of 0.9. This relationship could be a cheap source of basin scale SMM prediction from widely available observed data sets (actual precipitation and potential evapotranspiration), and thus, could afford some knowledge of SMM under no knowledge conditions. (C) 2015 Elsevier B.V. All rights reserved.