Radar Equivalent Snowpack: reducing the number of snow layers while retaining its microwave properties and bulk snow mass

Abstract. Snow water equivalent (SWE) retrieval from Ku-band radar measurements is possible with complex retrieval algorithms involving prior information on the snowpack microstructure and a microwave radiative transfer model to link backscatter measurements to snow properties. A key variable in a retrieval is the number of snow layers, with more complex layering yielding richer information but at increased computational cost. Here, we show the capabilities of a new method to simplify a complex multilayered snowpack to less than or equal to 3 layers, while preserving the microwave scattering behavior of the snowpack and conserving the bulk snow water equivalent. The method is based on a K-means clustering algorithm to group the snow layers based on the extinction coefficient and the height of the layer. Then, a weighted average using the extinction coefficient and the thickness was applied to the snow properties. We evaluated our method using snow properties from simulations of the SVS-2/Crocus physical snow model at 11 sites spanning a large variety of climates across the world and the Snow Microwave Radiative Transfer model to calculate backscatter at 17.25 GHz. Grouping and averaging snow stratigraphy into 3 layers effectively reproduced the total snowpack backscatter of multi-layered snowpacks with overall root mean squared error = 0.5 dB and R² = 0.98. Using this methodology, SWE retrievals can be applied to simplified snowpacks, while maintaining similar scattering behavior, without compromising the modeled snowpack properties. Reduction in the mathematical complexity of SWE retrieval cost functions and reduction in computation of up to 80 % can be gained by using fewer layers in the SWE retrieval.