Cumulative temperature metrics such as heating degree days and cooling degree days have utility to many applications. Cumulative temperature (CT) forecasts are derived directly from daily mean temperature forecasts – however, the error characteristics differ because the forecast errors for each day are not independent. A bias-corrected ensemble temperature forecast (eCast) is used to study these error characteristics and compare different methods of deriving the cumulative temperature forecast. For a deterministic forecast, the derivation of a cumulative temperature forecast is the sum of of the individual days' forecasts. Comparison of the individual daily mean forecast errors of each forecast length to the forecast errors of the cumulative temperature forecast over an entire length confirm errors correlated in large part to the timing of weather events. With an ensemble forecast, a cumulative temperature forecast can be computed by accumulating the ensemble mean forecasts for each individual day or by taking the ensemble mean of each members' accumulated temperature forecasts. These two methods are compared and help characterize the error of cumulative temperature forecasts. Comparison of the ensemble calibration implementations on the daily mean temperature forecasts and on the CT forecasts can also be used to study the properties of the forecast error. Using the accumulated temperature forecast for each ensemble member, forecast errors due to the timing of weather events are explicitly accounted for and provide a sharper, more reliable probabalistic forecast. In this way, an ensemble forecast can a priori provide an explicit estimate of the cumulative temperature forecast errors. Other cumulative temperature metrics such as HDDs and CDDs are also commonly used, but have different error characteristics due to nonlinear behavior at the degree day reference threshold. These metrics are compared to the cumulative temperature.