Improving prediction accuracy of soil water storage through reducing sampling frequency

Abstract

Knowledge of spatiotemporal dynamics of soil water storage (SWS) is essential for hydrological modeling and vegetation restoration in semi-arid areas. However, characterizing the temporal stability of SWS at a regional-scale requires time-consuming and labor-intensive manual sampling. Moreover, the influence of soil depth on temporal stability of SWS is not systematic. This study aimed to investigate the influences of sampling frequency and soil depth on SWS and SWS temporal stability. We measured soil moisture at 20-cm intervals in the soil profiles to a depth of 3 m using a neutron probe at 135 locations along a 1340-m long transect on 14 dates from 2012 to 2013. Results showed that sampling frequency did not influence the mean SWS (P < 0.05), while sampling frequency significantly affected temporal stability characteristics including Spearman’s rank correlation coefficient (rs), standard deviation of mean relative difference (SDRD), the number of locations with SDRD < 5%, and the representative locations. Temporal stability of SWS increased with the increasing soil thickness and depth, which increases the possibility of the number of representative locations in deep soil. Although the mean SWSs of all soil depths can be predicted accurately at each sampling frequency, the prediction accuracy improved when sampling frequency was reducing. The values of R2 ranged from 0.769 to 0.978 at 15-day sampling frequency, and from 0.987 to 0.998 at 45-day sampling frequency. Soil moisture stability may be more important than the soil water regime during prediction of soil moisture. These findings can provide guidelines for optimizing soil moisture sampling strategies and benefit management of water resources in semi-arid watershed.