Understanding droughts in a climate context remains a major challenge. Over the United States, different choices of observations and metrics have often produced diametrically opposite insights. This paper focuses on understanding and characterizing meteorological droughts from station measurements of precipitation. The Standardized Precipitation Index is computed and analyzed to obtain drought severity, duration and frequency. Average drought severity trends are found to be uncertain and data-dependent. Furthermore, the mean and spatial variance do not show any discernible non-stationary behavior. However, the spatial coverage of extreme meteorological droughts in the United States exhibits an increasing trend over nearly all of the last century. Furthermore, the coverage over the last half decade exceeds that of the dust bowl era. Previous literature suggests that climate extremes do not necessarily follow the trends or uncertainties exhibited by the averages. While this possibility has been suggested for droughts, this paper for the first time clearly delineates and differentiates the trends in the mean, variability and extremes of meteorological droughts in the United States, and uncovers the trends in the spatial coverage of extremes. Multiple data sets, as well as years exhibiting large, and possibly anomalous, droughts are carefully examined to characterize trends and uncertainties. Nonlinear dependence among meteorological drought attributes necessitates the use of copula-based tools from probability theory. Severity-duration-frequency curves are generated to demonstrate how these insights may be translated to design and policy.