An L-shaped antenna array with a large aperture is used for characterizing the spatial distribution and temporal variability of the noise environment for several frequencies in the high-frequency (HF) band. The process of whitening is used to compare spatial covariance matrices constructed from data collected on the array. The eigenvalue spread and the maximum likelihood method (MLM) spectrum are measures applied to these spatial covariance matrices to characterize the spatial distribution of the sources of HF noise. MLM spectra of the HF noise environment show a fairly flat distribution, but there is higher noise power in the directions of the continental U.S. In the temporal variability analysis it is determined that the HF noise environment at lower HF frequencies (8 and 10 MHz) has an underlying structure that is fairly stationary; the noise environment at 16 MHz does not show this stationarity. Models for HF noise fields are compared to the data. The model including both a symmetrical sky noise field and receiver thermal noise fits the data the best of the models examined.