Search

Measurements of radar backscatter are sensitive to snow water equivalent (SWE) across a wide range of frequencies, motivating proposals for satellite missions to measure global distributions of SWE. However, radar backscatter measurements are also sensitive to snow stratigraphy, to microstructure, and to ground surface roughness, complicating SWE retrieval. A number of recent advances have created new tools and datasets with which to address the retrieval problem, including a parameterized relationship between SWE, microstructure, and radar backscatter, and methods to characterize ground surface scattering. Although many algorithms also introduce external (prior) information on SWE or snow microstructure, the precision of the prior datasets used must be high in some cases in order to achieve accurate SWE retrieval. We hypothesize that a time series of radar measurements can be used to solve this problem and demonstrate that SWE retrieval with acceptable error characteristics is achievable by using previous retrievals as priors for subsequent retrievals. We demonstrate the accuracy of three configurations of prior information: using a global SWE model, using the previously retrieved SWE, and using a weighted average of the model and the previous retrieval. We assess the robustness of the approach by quantifying the sensitivity of the SWE retrieval accuracy to SWE biases artificially introduced in the prior. We find that the retrieval with the weighted averaged prior demonstrates SWE accuracy better than 20 % and an error increase of only 3 % relative RMSE per 10 % change in prior bias; the algorithm is thus both accurate and robust. This finding strengthens the case for future radar-based satellite missions to map SWE globally. [ABSTRACT FROM AUTHOR]

Measurements of radar backscatter are sensitive to snow water equivalent (SWE) across a wide range of frequencies, motivating proposals for satellite missions to measure global distributions of SWE. However, radar backscatter measurements are also sensitive to snow stratigraphy, microstructure and to surface roughness, complicating SWE retrieval. A number of recent advances have created new tools and datasets with which to address the retrieval problem, including a parameterized relationship between SWE, microstructure, and radar backscatter, and methods to characterize surface scattering. Although many algorithms also introduce external (prior) information on SWE or snow microstructure, the precision of the prior datasets used must be high in some cases in order to achieve accurate SWE retrieval. We hypothesize that a time series of radar measurements can be used to solve this problem, and demonstrate that SWE retrieval with acceptable error characteristics is achievable by using previous retrievals as priors for subsequent retrievals. We demonstrate the accuracy of three configurations of the prior information: using a global SWE model, using the previously retrieved SWE, and using a weighted average of the model and the previous retrieval. We assess the robustness of the approach by quantifying the sensitivity of the SWE retrieval accuracy to SWE biases artificially introduced in the prior. We find that the retrieval with the weighted averaged prior demonstrates SWE accuracy better than than 20 %, and an error increase of only 3 % relative RMSE per 10 % change in prior bias; the algorithm is thus both accurate and robust. This finding strengthens the case for future radar-based satellite missions to map SWE globally. [ABSTRACT FROM AUTHOR]

In studying the mass balance of polar ice sheets, fluctuations in firn density near the surface is a major uncertainty. In this paper, we explore these variations at locations on the Greenland Ice Sheet and at the Dome C location in Antarctica. Borehole in situ measurements, snow radar echoes, microwave brightness temperatures, and modeling results from the Community Firn Model (CFM) are used. It is shown that firn density profiles can be represented using three processes: "long-scale" and "short-scale" density variations and "refrozen layers". Consistency with this description is observed in the dynamic range of airborne 0.5–2 GHz brightness temperatures and snow radar echo peaks in measurements performed in Greenland in 2017. Based on these insights, a new analytical partially coherent model is implemented to explain the microwave brightness temperatures using the three-scale description of the firn. Short- and long-scale firn processes are modeled as a 3D continuous random medium with finite vertical and horizontal correlation lengths as opposed to past 1D randomly layered medium descriptions. Refrozen layers are described as deterministic sheets with planar interfaces, with the number of refrozen-layer interfaces determined by radar observations. Firn density and correlation length parameters used in forward modeling to match measured 0.5–2 GHz brightness temperatures in Greenland show consistency with similar parameters in CFM predictions. Model predictions also are in good agreement with multi-angle 1.4 GHz vertically and horizontally polarized brightness temperature measured by the Soil Moisture and Ocean Salinity (SMOS) satellite at Dome C, Antarctica. This work shows that co-located active and passive microwave measurements can be used to infer polar firn properties that can be compared with predictions of the CFM. In particular, 0.5–2 GHz brightness temperature measurements are shown to be sensitive to long-scale firn density fluctuations with density standard deviations in the range of 0.01–0.06 gcm-3 and vertical correlation lengths of 6–20 cm. [ABSTRACT FROM AUTHOR]