Microwave remote sensing of snow and the lower atmosphere in polar regions

Abstract

Europe faces an increasing vulnerability to frequent extreme weather events, requiring enhanced predictive capabilities within the high latitudes through improved Numerical Weather Prediction (NWP) models. Currently, limited data availability due to sparse weather station networks prevents high-latitude predictive accuracy. Microwave sounding radiances, notably at 50 and 183 GHz, offer vital atmospheric temperature and humidity profile data, vital for correcting forecast initial conditions. However, challenges arise from variable snow and sea ice emissions, inhibiting data assimilation into forecast models. The refinement of snow emissivity modelling enables atmospheric observation assimilation, thereby improving NWP models.

This thesis presents a coupling of two models to illustrate the assimilation of microwave radiances into NWP models: the Factorial Snow Model (FSM) simulates snow microphysical properties influencing microwave emissions and the Snow Microwave Radiative Transfer (SMRT) model accurately simulates microwave emissivity across frequencies.

Initial research integrates and validates these models using data from Trail Valley Creek (TVC), Northwest Territories, Canada. Comparison of in-situ snow pit data with FSM-generated snowpack profiles reveals a precise simulation of density and snow grain size profiles characteristic of Arctic tundra snowpacks. However, FSM fails to capture variability in snow pit profiles. Coupling FSM with SMRT, simulated brightness temperatures (T₈) at 89 GHz are compared to ground-based radiometer observations. SMRT's superior performance using FSM-simulated snow inputs (mean error: -4.4 K) contrasts with snow pit data (mean error: 11.9 K). Sensitivity tests indicate snow grain size as the most influential factor in snow surface emissivity.

In the thesis's concluding phase, SMRT-driven simulations expand to the NWP grid scale across frequencies from 10.65 to 234 GHz. T₈ generally increases with frequency, except at atmospheric window channels (157 and 243 GHz). Spatial variability, frequency-dependent, is minimal at the lowest (≤ 18.7 GHz) and highest frequencies (≥ 118 GHz) due to reduced scattering. Intermediate frequencies (37 and 89 GHz) exhibit greater spatial variability due to increased snowpack scattering. In summary, large spatial and temporal variations in Arctic tundra snow emissivity highlight the necessity for precise emissivity simulations grounded in accurately modelled or observed microphysical snow properties. This approach contrasts with the static emissivity values prevalent in many NWP systems.