Scales of temporal and spatial variability of midlatitude soil moisture

Soil moisture observations from direct gravimetric measurements in Russia are used to study the relationship between soil moisture, runoff, and water table depth for catchments with different vegetation types, and to estimate the spatial and temporal correlation functions of soil moisture for different soil layers, For three catchments at Valdai, Russia, one with a grassland, one with an old forest, and one with a growing forest, the interannual soil moisture variations are virtually the same for the 31-year period, 1960-1990, The runoff is higher for the grassland than for the old forest, and the water table depth is not at deep. The runoff and water table for the growing forest vary from grassland-like during the first decade, when the trees are small, to old forest-like at the end of the period. The seasonal cycle of soil moisture is similar at all three catchments, but the snowmelt and summer drying begin a month earlier at the grassland than in the forests. A statistical model of both temporal and spatial variations in soil moisture is developed that partitions the variations into red noise and white noise components. For flat homogeneous plots, the white noise component is relatively small and represents solely random errors of measurement. For natural landscapes with variable vegetation and soil types, and complicated topography, this component is responsible for most of the temporal or spatial variance, The red noise component of temporal variability is in good agreement with theory. The timescale of this component is equal to the ratio of field capacity of soil to potential evapotranspiration, approximately 3 months, The red noise component of spatial variability reflects the statistical properties of the monthly averaged precipitation field, The scale of spatial correlation of this component is about 500 km. The estimates of scales of temporal and spatial correlation do not differ significantly for water content in the top 20-cm and 1-m layers of soil, These results have important implications for both remote sensing of soil moisture and soil moisture parameterization in climate models.