| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 4459267 | Remote Sensing of Environment | 2012 | 13 Pages |
Plot-scale brightness temperature (TB) measurements at 6.9, 19, 37, and 89 GHz were acquired in forest, open, and lake environments near Churchill, Manitoba, Canada with mobile sled-based microwave radiometers during the 2009–2010 winter season. Detailed physical snow measurements within the radiometer footprints were made to relate the microwave signatures to the seasonal evolution of the snowpack, and provide inputs for model simulations with the Helsinki University of Technology (HUT) snow emission model. Large differences in depth, density, and grain size were observed between the three land cover types. Plot-scale simulations with the HUT model showed a wide range in simulation accuracy between sites and frequencies. In general, model performance degraded when the effective grain size exceeded 2 mm and/or there was an ice lens present in the pack. HUT model performance improved when simulations were run regionally at the satellite scale (using three proportional land cover tiles: open, forest, and lake) and compared to Advanced Microwave Scanning Radiometer (AMSR-E) measurements. Root mean square error (RMSE) values ranged from approximately 4 to 16 K depending on the frequency, polarization, and land cover composition of the grid cell. Snow water equivalent (SWE) retrievals produced using forward TB simulations with the HUT model in combination with AMSR-E measurements produced RMSE values below 25 mm for the intensive study area. Retrieval errors exceeded 50 mm when the scheme was applied regionally.
► We describe unique seasonal passive microwave and physical snow measurements. ► Snow emission model simulations were assessed through a full snow season. ► Model performance degraded when grain size > 2 mm and/or an ice lens was present. ► SWE retrievals with a forward TB model show promise for operational implementation. ► Coarse AMSR-E data mean individual retrievals contain a high level of heterogeneity.
