Drought is one of the most costly natural disasters. The impact of a given drought event can be characterized by its severity, area and duration. We define drought severity based on percentiles of runoff and soil moisture relative to their climatologies. Observations of soil moisture, in particular, are quite sparse; however, land surface hydrologic models can estimate the spatial distribution of soil moisture associated with observed precipitation, temperature, and other land surface variables. The University of Washington's national Surface Water Monitor provides daily nowcasts and weekly forecasts of soil moisture and runoff percentiles as estimated by the Variable Infiltration Capacity (VIC) model over the continental U.S. at ½ degree spatial resolution. Although these estimates compare well with observations in the limited cases where such comparisons are feasible, in some cases drought persistence appears to be underestimated. Because groundwater responds more slowly than the unsaturated zone to surface conditions and the lower soil layers respond to groundwater conditions, a possible explanation for this discrepancy in drought duration is the absence of a representation of groundwater in VIC. In order to improve our estimates of hydrologic states, we incorporated the Simple Groundwater Model, SIMGM, of Niu and colleagues at the University of Texas, into VIC. We tested this revised model against water table depth observations over Illinois. Resulting estimates of evaporation, soil moisture, and runoff, as well as drought persistence, are compared to those of the standard VIC model.