We have conducted an experiment to assess the real time skill in monthly and seasonal predictions based solely on patterns of antecedent hydrological information over the United States. The hydrological information is contained in a proxy for soil moisture at 102 locations over the lower 48 states. This soil moisture is calculated over the years 1931 to present from a local hydrological equation taking monthly precipitation (P) and temperature (T) as input, and producing soil moisture (w), evaporation (E), runoff (R), and loss to groundwater (G) as output. The initial condition (IC) for the forecast procedure is soil moisture over the United States at the end of the month (w30). We constructed an analogue to the w30 fields, i.e., made linear combinations of soil moisture fields at the same time of year in years past to reproduce the IC to within a small tolerance. The coefficients assigned to the years past are then made to persist, and the subsequent development in the historical years is linearly combined to form a forecast. This method has been running at CPC in real time since 1998, and we added 1981--1997 in retroactive real time mode to form a large enough sample. In total, we considered both seasonal and monthly forecasts at leads of -1 to +6 months for 1981--2001, for the elements w30, E, T, and P. From the outset, we wanted to investigate nonlocal forecast methods, considering local effects, on evaporation and temperature mainly, as being established already and well documented <Huang et al., 1996>. In a nonlocal method we entertain the possibility of precipitation (the response) falling downstream of a soil moisture anomaly (forcing). We found that we have about a 0.6 correlation in forecasting monthly soil moisture with a lead of one month (i.e., July at the end of May). This figure is higher in spring and somewhat lower in the early fall. The capability to forecast evaporation anomalies is very seasonal. During the cold half of the year, when E anomalies resemble T anomalies, the correlation is only 0.2--0.3, but in summer, when E anomalies resemble w anomalies, the skill of forecasts goes up to 0.6. We thus have some insight into patterns of anomalous water vapor input from the land surface into the lower atmosphere on a continental scale. Skill of forecasting T is modest, reaching 0.2--0.3 in many months and seasons, but there is no clear seasonal dependence that relates to the presumed physics of land atmosphere interactions. Skill in forecasting P is quite low, barely 0.1 in correlation, but +ve in all months and seasons. We did alternative experiments where the constructed analogue was built on E, T, or P instead of w and verified the forecast of all elements likewise. We found initial w to be the best for forecasting w itself and indeed for forecasting the other fields as well! This is important testimony that soil moisture is indeed the key, as has been suspected by many for ages.