Forecasts during persistent valley cold pools in the Bonneville Basin by the North American Mesoscale Model

Forecasts from the 12-km North American Mesoscale (NAM) model for three winter seasons are considered to assess how well this model forecasts low-level temperature and humidity during persistent cold pool events in the Bonneville Basin of Utah. Statistical comparisons of forecast to observed 2-m temperature and dew point show forecast accuracy does not significantly decrease as lead time increases. Additionally, for stations in the basin, errors in temperature and dew point are not statistically different from 0 K at 0000 UTC, but forecasts valid at 1200 UTC have a strong warm bias. There is a strong moist bias in dew point at either time in the western part of the basin. Along the basin rim, there is a cold bias that is especially pronounced during the day. Sensitivity experiments are conducted with the NAM model to determine the causes of the errors. Within the basin, improved forecasts are observed when the land use and soil-water freezing threshold are adjusted to better approximate typical conditions during winter over an alkali flat. The strong daytime cold bias along the rim is the result of inaccurate parameterization of low-level temperatures over snow in the land-surface model.