Spatio-temporal variability of soil water content on the local scale in a Mediterranean mountain area (Vallcebre, North Eastern Spain). How different spatio-temporal scales reflect mean soil water content

As a result of complex human-land interactions and topographic variability, many Mediterranean mountain catchments are covered by agricultural terraces that have locally modified the soil water content dynamic. Understanding these local-scale dynamics helps us grasp better how hydrology behaves on the catchment scale. Thus, this study examined soil water content variability in the upper 30cm of the soil on a Mediterranean abandoned terrace in north-east Spain. Using a dataset of high spatial (regular grid of 128 automatic TDR probes at 2.5m intervals) and temporal (20-min time step) resolution, gathered throughout a 84-day period, the spatio-temporal variability of soil water content at the local scale and the way that different spatio-temporal scales reflect the mean soil water content were investigated.Soil water content spatial variability and its relation to wetness conditions were examined, along with the spatial structuring of the soil water content within the terrace. Then, the ability of single probes and of different combinations of spatial measurements (transects and grids) to provide a good estimate of mean soil water content on the terrace scale was explored by means of temporal stability analyses. Finally, the effect of monitoring frequency on the magnitude of detectable daily soil water content variations was studied.Results showed that soil water content spatial variability followed a bimodal pattern of increasing absolute variability with increasing soil water content. In addition, a linear trend of decreasing soil water content as the distance from the inner part of the terrace increased was identified. Once this trend was subtracted, resulting semi-variograms suggested that the spatial resolution examined was too high to appreciate spatial structuring in the data. Thus, the spatial pattern should be considered as random. Of all the spatial designs tested, the 10×10m mesh grid (9 probes) was considered the most suitable option for a good, time-stable estimate of mean soil water content, as no improvement was obtained with the 5×5m mesh grid (30 probes). Finally, the results of temporal aggregation showed that decreasing the monitoring frequency down to 8h during wetting-up periods and to 1day during drying-down ones did not result in a loss of information on daily soil water content variations.