Exploring spatiotemporal patterns and physical controls of soil moisture at various spatial scales

Soil moisture variability of various spatial scales is analyzed based on empirical orthogonal function (EOF) method using soil moisture datasets with various spatial resolutions: 1 km eco-hydrological model simulation, 0.25° passive microwave (Advanced Microwave Scanning Radiometer for the Earth Observing System, AMSR-E) dataset, and 0.5° land surface model simulation from Climate Predictor Center (CPC). All three datasets generate EOFs that explain similar variances with those generated from in situ observations from agro-meteorological network. Using AMSR-E product and eco-hydrological model simulation, it is found that the primary spatial pattern of soil moisture obtained from watershed scale has a strong connection to topographic attributes, followed by soil texture and rainfall variability, as suggested by the correlation between the primary EOF mode (EOF1) of soil moisture and landscape attributes. However, the EOF analysis of both AMSR-E and CPC datasets at regional scale reaches the conclusion that soil texture indices, such as sand and clay content, is of higher importance to soil moisture EOF1 spatial pattern (explaining 61 % variance) than topography is. Furthermore, correlation between soil moisture EOF1 and soil property is higher in spring than in autumn, which indicates that soil water-holding and drainage capabilities are more important under dry conditions. At national scale, the combined effects of topographic feature and soil property are clearly exhibited in EOF1. The study results reveal that different emphases should be placed on accurate acquisition of landscape attributes for soil moisture estimation according to various spatial scales.