Precipitation (amount and type) is arguably the surface sensible weather element of greatest
importance to society. Unfortunately, precipitation remains the most challenging weather element
to forecast skillfully. There are three primary reasons for this behavior, all of which can not be
ignored: 1) large model errors, 2) high levels of analysis-observational uncertainty and inherently
short predictability limits. The talk will touch upon all three aspects of the challenge. We argue
that inclusion of stochastic forcing in our NWP constructs may offer a viable, cost-effective
approach to improve certain aspects of the performance of ensemble forecast systems that avoids the
overhead associated with an operational center supporting numerous versions of models and
parameterizations. Incremental improvements in forecast skill and value will remain difficult to prove
unequivocally, however, in the face of the observational uncertainty and the short predictability limits
that characterize precipitation, especially heavy precipitation.