Autocorrelation within surface soil moisture (SSM) data may be used to produce high-resolution spatial maps of SSM from point samples. The objective of this study was to characterize the temporal and spatial properties of SSM (0-5 cm) in a Beltsville, MD cornfield using capacitance probes. The range of spatial autocorrelation was approximately 10 m and the highest sill values were found at water contents (theta) between 20-27%. Nugget values represented a significant portion of the total variance (up to 50% for theta > 20% and 73% for theta < 12%). The patterns of SSM under wet conditions exhibited large, continuous polygons while drier conditions resulted in smaller, discreet regions. Early season (< 60 days) Auto-Regressive Moving-Average (ARMA) forecasts of SSM plotted against observed data resulted in R2 values from 0.15-0.26, while late season (>80 days) forecasts improved to 0.46-0.65. Forecasts were improved by autoregressive coefficients and additional SSM datasets.