Understanding Uncertainty of Snow Radiative Transfer Modeling Within a Mixed Deciduous and Evergreen Forest

Theodore Letcher,Alexandre Roy,Alexandre Langlois,Carrie Vuyovich

doi:10.1109/jstars.2021.3099944

Abstract

Satellite-based passive microwave observations provide the best available continuous observational estimates of global snow water storage due to their broad geographic footprint and low sensitivity to clouds and precipitation. However, these observations are subject to substantial uncertainty due to the complex radiative properties of snow and from interference in forested areas. Physical radiative transfer models can be leveraged to improve the fidelity of these observations and as a data-assimilation tool. In this article, the Dense Media Radiative Transfer model with Multiple Layers (DMRT-ML) is used to simulate snow brightness temperatures from data collected from snow pits excavated during a two-day-long field study performed a temperate forest in the Northeast United States. The simulations are evaluated against surface-based radiometer observations collected at the snow pits. The DMRT-ML is configured with varying complexity to determine the snowpack characteristics most essential toward simulating brightness temperature within a temperate forest with complicated snow stratigraphy. In general, the single-layer configurations were not sufficiently complex to accurately simulate snow brightness temperature without significant tuning. The most accurate simulation was a two-layer configuration with a prescribed ice layer separating the snow layers. This simulation had a root-mean-square error $<; $15 K for the 37-GHz frequency. More complicated snowpack stratigraphy configurations did not substantively improve the results over the two-layer model configuration. The DMRT-ML was also used to examine differences between redundant datasets of density and grain size. It was determined that similar snow data collection and radiative transfer model configuration techniques are critical to ensure cross-study comparability.

Highlights

Numerous estimates of global snow water equivalent (SWE) based on these observations are used operationally (e.g., GlobSnow [11]–[13];S EASONAL snowpacks cover on average 31% of the globe AMSR2 [7])
We used snow pit and surface-based radiometer (SBR) data collected during the CrustEX field experiment in the Hubbard Brook Experimental Forest (HBEF) to shed light on how snow-input complexity and model stratigraphy configuration affects simulated snow brightness temperature from the DMRTML in a mixed forest environment
It was shown that the inclusion of an ice lens in the model substantially improved the simulated 19-GHz horizontally polarized brightness temperature, reducing the root-mean-square error (RMSE) by approximately 5 K

Summary

Introduction

Numerous estimates of global SWE based on these observations are used operationally S EASONAL snowpacks cover on average 31% of the globe AMSR2 [7]) These products can vary by as much as a and are a critical part of the global water cycle [1]; factor of 2 due primarily to uncertainty in the PM measurements (see [14]–[16]). Recent studies indicate that physical snow radiative transfer models (RTMs) can be used to improve existing empirical PM SWE algorithms (see [17]–[22]) through a better understanding program and in part by the U.S Army Program Element 0603734A—Military Engineering Advanced Technology. Several large and small field projects have generated comprehensive datasets of in situ snow observations alongside of radio-

Results

Discussion

Conclusion