The knowledge of soil moisture spatio-temporal variability is highly relevant for water resources management. This is especially significant in highly demanding areas of arid and semiarid regions where sensitive ecosystems, agriculture and recreational areas are competing for water resources. This study reports an analysis of the spatial-temporal variability of soil moisture data for a small to medium-scale network in a coastal wetland of southwestern Spain. Measurements were taken from 5 sites located in the Doñana National Park, with an extension of approx. 25 × 3 km over the time-period of one hydrological year from September 2017 to September 2018. Absolute soil moisture was separated in two components: the temporal mean (time-invariant term) and the deviations from the mean or anomalies (time-variant term). The time-invariant component was generally the main contributor to the total spatial variance of absolute soil moisture data and it was mostly controlled by the groundwater levels in the area. Nevertheless, the time variant terms have a huge effect on soil moisture variability in very dry states at all 5 sites. Time-dependent patterns were found between average soil moisture state and the variance and covariance of both the time-variant and time-invariant components of soil moisture, which could be used for the up and downscaling of soil moisture from satellite data. Those patterns were only affected by heavy rainfalls giving rise to hysteretic behavior. This study shows that even though groundwater level is a time-variant variable, it significantly affects soil moisture’s time-variant and time-invariant terms.