Zenith wet delay (ZWD) is a key parameter in analyses of microwave propagation delays caused by atmospheric water vapor. This paper explores the possible use of numerical weather analyses and numerical weather forecasts to estimate the ZWD in nearly real time or, more likely, after the fact. The predicted precipitable water (PW) field is transformed into a predicted ZWD field using the relationship ZWD=PW/&Pgr;, where &Pgr; is a function of various physical constants and a weighted mean temperature of the atmosphere, Tm, that is also predicted using the numerical weather analysis or forecast. We use the +12-hour forecasts of the Nested Grid Model to predict ZWD and test these predictions against values derived from independent radiosonde observations. By performing these comparisons at more than 100 North American radiosonde stations over a 6-month period we find that forecast errors tend to increase as the magnitude of ZWD increases. This provides a simple framework for characterizing the maximum level of error associated with a ZWD estimates obtained by interpolating in time between analytical ''nowcasts'' and numerical weather predictions. We develop a preliminary approach to estimating the ZWD forecast errors at any time between zero and 12 hours from model initialization. ¿ American Geophysical Union 1996