Snow Depth Retrieval on Arctic Sea Ice From Passive Microwave Radiometers—Improvements and Extensions to Multiyear Ice Using Lower Frequencies

Abstract

Snow on sea ice influences the Arctic energy and heat budgets and is therefore important for Arctic climate studies. Methods to derive snow depth based on satellite-borne microwave radiometer observations have existed since the 1990s. However, in the Arctic the most widely used algorithm can only be applied over first-year ice (FYI) and does not make use of the lower frequencies, which are available since 2002.Here we present three improvements to the current passive microwave snow depth retrieval: (a) We derive new coefficients based on a regression analysis using 5 years of Operation IceBridge airborne snow depth measurements. (b) We extend the algorithm to take advantage of the lower frequency channel at 7 GHz. (c) We consider an extension of the snow depth retrieval to multiyear ice (MYI) during spring. Our results show that the gradient ratio, GR(19/7) is most suited for deriving snow over both Arctic FYI (R =°À0.73) and MYI (R = À0.57). An evaluation of the new retrieval with Operation IceBridge snow depth measurements from March and April 2015 shows a good agreement over FYI (difference = À2.1 cm; 93% of the differences below 5 cm). Over MYI the difference is À4.0 cm and 56% of the differences are below 5 cm, that is, the root mean square distance (RMSD) is larger over MYI than over FYI. We demonstrate for the first time that spring snow depth measurements can be derived from passive microwave observations over both FYI and MYI.Plain Language Summary Snow on Arctic sea ice plays an important role in the Arctic climate system. It reflects the majority of the incoming solar radiation and isolates the sea ice from warm air in summer. However, the vast area and the extreme weather conditions make it difficult to monitor changes in snow depth during the Arctic winter. In this study, we develop a retrieval for snow depth on Arctic sea ice based on satellite observations. The advantages of satellite observations are that they provide daily coverage of the whole Arctic. We compare satellite observations in the microwave regime to airborne snow depth measurements, obtained in March and April from 2009 to 2015. We find a good agreement between changes in the signal observed by the satellite and changes in the measured snow depth when the ice is smooth. Over multiyear ice, ice that has survived at least one summer melt and that is often rough, the agreement is not as good. We demonstrate for the first time that spring snow depth estimations for the whole sea ice covered Arctic can be derived from satellite observations. While a clear signal can be found for changes in sea ice area, changes in snow depth on sea ice or snowfall are hard to quantify. In situ measurements and reanalysis data suggest a decline in summer snowfall (June to August) over the sea ice covered Arctic (Screen & Simmonds, 2012). Webster et al. (2014) analyzed spring ROSTOSKY ET AL. 7120

Modeling the Microwave Emission of Snow on Arctic Sea Ice for Estimating the Uncertainty of Satellite Retrievals

“…For GR(37/19), the influence of the MYI properties is one magnitude larger than for GR(19/7) and the potential error due to the variability of MYI properties is 29%. This is consistent with earlier findings that GR(37/19) cannot be used to retrieve snow depth for MYI, since the signal of the MYI at GR(37/19) is of similar strength than the signal of snow (e.g., Brucker & Markus, 2013;Rostosky et al, 2018). Rostosky et al (2018) found a reasonable correlation between GR(19/7) and snow depth on MYI, indicating that it is possible to derive a retrieval for snow depth on MYI using GR(19/7), even though the correlation between GR(19/7) and snow depth was lower for MYI than for FYI and high errors were found for MYI.…”

“…Rostosky et al () found a reasonable correlation between GR(19/7) and snow depth on MYI, indicating that it is possible to derive a retrieval for snow depth on MYI using GR(19/7), even though the correlation between GR(19/7) and snow depth was lower for MYI than for FYI and high errors were found for MYI. Also, the results found in this study indicate that MYI properties have only little influence on GR(19/7).…”

Estimating snow depth on Arctic sea ice using satellite microwave radiometry and a neural network

“…In contrast to other machine learning techniques, they are designed to extract relevant features and their weighting in the model themselves. Deep neural networks allow us to learn higher-order representations, tackle more complex problems and outperform other means of machine learning in terms of accuracy (Schmidhuber, 2015). Neural networks can be viewed as a universal system to represent any function.…”

Space‐Based Observations for Understanding Changes in the Arctic‐Boreal Zone

“…Rostosky et al () tackled the problem of expanding snow depth retrievals from satellite passive microwave data to include snow depths over multiyear ice, taking advantage of the 6.9 GHz measurements from the AMSR‐E and AMSR2 sensors. Comparisons of the derived snow depths with Operation IceBridge springtime measurements yielded good agreement, although better with first‐year ice than multiyear ice (Rostosky et al, ).…”

“…Rostosky et al () tackled the problem of expanding snow depth retrievals from satellite passive microwave data to include snow depths over multiyear ice, taking advantage of the 6.9 GHz measurements from the AMSR‐E and AMSR2 sensors. Comparisons of the derived snow depths with Operation IceBridge springtime measurements yielded good agreement, although better with first‐year ice than multiyear ice (Rostosky et al, ). Maaß et al () examined the use of lower frequency passive microwave data, at 1.4 GHz (L‐band), from the ESA's Soil Moisture and Ocean Salinity (SMOS) satellite to retrieve snow thickness estimates over thick Arctic sea ice, detailing both the complications and the sense that this could be an approach worth pursuing further.…”