Uncertainty in predicting the spatial pattern of soil water temporal stability at the hillslope scale

摘要

Soil water temporal stability is subject to spatial variation, which influences the prediction of mean soil water status on a hillslope. However, estimation of the spatial pattern of soil water temporal stability and quantification of the uncertainties associated with the predictions are often ignored. In this study, volumetric soil water contents at 10 and 30 cm depths on tea garden and forest hillslopes were monitored across 17 dates from January 2013 to April 2014. Soil moisture maps on these 17 dates were interpolated using ordinary kriging and then the spatial distribution of the relative difference of soil moisture was assessed. Based on these maps, standard deviations of relative difference (SDRD) of soil moisture were calculated to represent the spatial variation of soil water temporal stability. Uncertainties in predicted patterns of SDRD due to the limited number of sampling days used for calculating SDRD (U1) and for spatial interpolation of soil moisture by ordinary kriging (U2) were investigated using bootstrap and sequential Gaussian simulation techniques respectively. Results showed that soil water content on the forest hillslope generally exhibited stronger spatial variability than that on the tea garden hillslope. The SDRD substantially varied in space at the hillslope scale. Temporal stabilities of soil water content at 30 cm depth were significantly (P < 0.05) stronger than those at 10 cm soil depth, regardless of the land use type. However, differences in soil water temporal stabilities on these two land use hillslopes were not significant. In addition, U2 was generally more important than U1 on both hillslopes. This suggests that additional sampling sites and more robust interpolation methods rather than additional sampling days should be developed to reduce SDRD prediction uncertainty on the study hillslopes.