Upgrades to the land-surface model (a.k.a. NOAH LSM) used in the NCEP Eta Analysis and Forecast System are tested in order to address shortcomings related to known low-level (e.g. 2-meter) air temperature and relative humidity biases in Eta model forecasts. These biases are due in part to limitations in certain aspects of the parameterization of the NOAH LSM physics. Previous off-line (e.g. atmospheric-forced, one-dimensional, site-specific) NOAH LSM testing suggests that parameterization improvements can significantly reduce these biases, with improvements related to: (1) soil heat flux, (2) canopy conductance, (3) 'bare' soil evaporation, and (4) cold season processes (including snowpack). These previous limitations in NOAH LSM physics can yield different biases. For example, over wet soils with sparse green vegetation common during early spring, the choice of a non-optimal formulation for soil thermal conductivity leads to excess soil heat flux which results in a dampened diurnal temperature cycle. During summer with dense green vegetation, an underprediction of canopy conductance and thus transpiration leads to a high bias of air temperatures throughout the day. Including frozen soil processes during winter ameliorates the typical underestimation (when frozen soil processes are ignored) of soil temperature (and thus surface and air temperatures) during soil freezing periods, and overestimation of temperatures during thawing periods. Over a melting/retreating snowpack, accounting for subgrid patchiness of a shallow snowpack allows for greater surface temperatures, upward heat flux and air temperatures greater than freezing. We will present results of testing in the coupled setting of the Eta Analysis and Forecast System using winter, early spring, and summer case studies under conditions of minimal large-scale forcing in order to demonstrate reductions of these model biases.