Cold Season Surface Classification by Response to Snow Accumulation and Melt: An Active–Passive Microwave Perspective from GPM

Active and passive microwave remote sensing of snowpacks has a lengthy heritage, with substantial understanding and data records for snow water equivalent developed from passive microwave imagers such as SSM/I(S) and AMSR-E/AMSR2. Likewise, active sensors such as the Ku-band Seawinds scatterometer on Quikscat and the C-band ASCAT scatterometer have been used to identify snowpack accumulation, melt, and refreeze events. The Global Precipitation Measurement (GPM) mission core satellite hosts two sensors that provide unique microwave measurements in a 65° inclination orbit, both of which are sensitive to snowpack properties: The 13-channel, 10-183 GHz GPM Microwave Imager (GMI) and Ku+Ka band Dual-frequency Precipitation Radar (DPR).

Five years of GPM data have been collected since launch in February 2014 and processed with an emissivity retrieval algorithm. The algorithm uses the GMI 183+-3 and 183 +-7 GHz channels, to nudge thermodynamic profiles from reanalysis data providing better fidelity to the retrieved 89 and 166 GHz emissivities and optimal detection of cloud- and precipitation-contaminated retrievals. These emissivities were matched to the attenuation-corrected surface backscatter cross-section, measured between 0-18°. Snow water equivalent (SWE) was interpolated from the MERRA2 reanalysis at each DPR pixel to better understand the response of microwave characteristics to changes in SWE.

It was found that the emissivity and backscatter response varies depending the underlying topography and vegetation. Croplands and prairies show the strongest decrease in emissivity to low levels of SWE (< 10 mm) at 89 and 166 GHz. In this SWE range near-nadir backscatter tends to decrease over boreal forest. As SWE increases to the 10-100mm range, decreases in emissivity are evident at 18 GHz over prairies and dormant croplands; at 36-166 GHz this behavior is nearly ubiquitous albeit at varying magnitudes. Near-nadir there is a decrease in backscatter over the tundra regions of Siberia and Canada. Off-nadir, the volume scattering component dominates, and the backscatter response inversely mirrors the emissivity response, particularly at Ka-band. These responses are used in an unsupervised classification algorithm to identify self-similar surfaces by their response to snowpack accumulation, where surface-specific algorithm parameters might be derived to obtain SWE estimates from GPM or similar microwave measurements.