Acting on long-held believes by Namias and others, the CPC, in the early 1990's, embarked on a program to quantify the impact of anomalous soil moisture conditions on subsequent monthly and seasonal atmospheric conditions. In the past this had been done only qualitatively as part of the subjective input to the monthly/seasonal forecast, or, at best, via linear correlations between antecedent precipitation and next month's temperature, thus only hinting at a role for the lower boundary. The task was further defined as one in which we had to produce a multi-decade integration of a soil moisture model - ‘climate' work requires long data sets. By necessity the soil model had to be simple, because the available fields of forcing required to run such models were very limited. The first such model, a ‘leaky' hydrologist's bucket, was described in Huang et al(1996), and its forcing consisted of monthly precipitation and evaporation, the latter evaluated from monthly temperature via an adjusted Thornswaite expression. Once tuned, it is not difficult to run such models on past data, say the Climate Divisions 1932-1993, and the resulting calculated soil moisture is, indeed, a much better local predictor than antecedent precipitation for subsequent temperature. Being in an operational center, however, required a daily update to bring soil moisture up to the latest possible date, before jumping off into the forecast. Thus emerged a hybrid monthly/daily integration of the leaky bucket. This activity put an additional focus on the daily precip analysis, which had been under debate for ages, a debate that continues to this very day. We will review the forecast tools that use soil moisture as an initial state, with emphasis on non-local versus local impacts. In a companion paper, by Fan et al, we will discuss the forthcoming transition from the leaky bucket model to the US Land Data Assimilation System