A First-Order Characterization of Errors From Neglecting Stratigraphy in Forward and Inverse Passive Microwave Modeling of Snow

Large-scale snow hydrology has been studied via spaceborne passive microwave (PM) measurements for decades. Forward and inverse radiative transfer (RT) models of snow are utilized in this context but typically neglect snow stratigraphy. Our objective in this paper is to characterize the expected model error in PM brightness temperature (T_b) predictions due to neglecting stratigraphy over a range of snow cover conditions. For 191 snowpits ranging from prairie to alpine, we performed side-by-side RT model runs including and ignoring stratigraphy via mass-weighted averages across stratigraphic layers; error was estimated by comparing the two RT model runs. Neglecting stratigraphy at 37 GHz led to approximately 10-K root mean square error (RMSE) for moderately deep (alpine) snow cover and to approximately 5-K RMSE for shallower (prairie) snow. RMSE across all types of snow was 1.67 and 26.9 K at 18.7 and 89 GHz, respectively. At 37 GHz, there was a low bias for deep snowpacks and a high bias for moderate-to-shallow snowpacks. Bias magnitude bias was dependent on vertical grain size variability. Based on these results and estimates of sensitivity of T_b to snow depth, we estimated that snow depth RMSE due to neglecting stratigraphy approaches 50%.