Comparing global passive microwave freeze/thaw records: Investigating differences between Ka- and L-band products

Informing Improvements in Freeze/Thaw State Classification Using Subpixel Temperature

“…While STAs are effective for FT classification, they do come with limitations, which predominantly exists over regions with less well-defined frozen and thawed reference states. This includes areas with drier climates, abundant surface water, high subgrid variability, and regions with transitional climates [22], [30], [48], [49]. Several additional algorithms have been applied to capture the complexity of FT classification and improve upon STAs, such as decision trees [43], multiband discriminant and probabilistic algorithms [50], [51], moving window/change detection approaches [52], [53], estimation of frozen soil fraction [28], the use of multiple MW band reflectance values [54], and the combination of surface skin temperature with satellite Tb [35], [36].…”

“…This is especially true in regions experiencing frequent FT transitions, such as the continental United States (CONUS). Still, band and algorithm differences between SMAP and FT-ESDR records have been shown to be the dominant driver of classification divergence [22].…”

“…FT observation across large areas is constrained by limited in situ surface skin and soil temperature observations, especially at high latitudes, making the use of satellite products crucial to spatially distributed FT cycle monitoring. Though, due to coarse resolution satellite observations, the heterogeneous nature of the land surface, microwave (MW) observing bands with fundamentally different emitting layers, and varied classification algorithms, binary FT classifications come with considerable uncertainty [20]- [22]. The grid spacing of global passive-MW FT records from MEaSUREs Freeze Thaw Earth System Data Record (FT-ESDR) [23] and the Soil Moisture Active Passive (SMAP-FT) [24] are limited to 25 and 36 km, respectively.…”

“…Recent studies have highlighted frozen classification variations between lower frequency (SMAP, L-band 1.41 GHz) and higher frequency (FT-ESDR, Ka-band 36.5 GHz) radiometer observations. Compared to the FT-ESDR, SMAP-FT has been shown to observe FT transitions deeper into the soil, to be less impacted by vegetation and surface snowmelt, though also being increasingly affected by arid conditions and having limitations due to a larger sensor observational footprint [21], [22], [29], [30]. In addition, regions with ill-defined FT seasonality (i.e., mid and lower latitudes), persistent snow cover, abundant surface water, varied topography, and complex vegetation/soil layers have been identified as challenging regions to accurately assess surface FT state generally due to the presence of mixed FT conditions [12], [22], [26], [31]- [34].…”

“…Compared to the FT-ESDR, SMAP-FT has been shown to observe FT transitions deeper into the soil, to be less impacted by vegetation and surface snowmelt, though also being increasingly affected by arid conditions and having limitations due to a larger sensor observational footprint [21], [22], [29], [30]. In addition, regions with ill-defined FT seasonality (i.e., mid and lower latitudes), persistent snow cover, abundant surface water, varied topography, and complex vegetation/soil layers have been identified as challenging regions to accurately assess surface FT state generally due to the presence of mixed FT conditions [12], [22], [26], [31]- [34]. Current approaches provide a general landscape FT state, which is defined by a general classification of the entire radiometer footprint area and a mixture of all components therewithin.…”