[Paper Review] Microwave emissivity of fresh water ice--Lake ice and Antarctic ice pack--Radiative transfer simulations versus satellite radiances

Peter Mills|arXiv (Cornell University)|Feb 18, 2012

Arctic and Antarctic ice dynamics15 references3 citations

TL;DR

This study validates a plane-parallel radiative transfer model for microwave emissivity using point-by-point comparisons between simulated and satellite-measured radiances over Lake Superior and the Antarctic ice pack. Despite neglecting scattering, the model shows strong correlation with observations, indicating its physical basis is sound, though biases arise primarily from unaccounted scattering in ice and atmosphere.

ABSTRACT

Microwave emissivity models of sea ice are poorly validated empirically. Typical validation studies involve using averaged or stereotyped profiles of ice parameters against averaged radiance measurements. Measurement sites are rarely matched and even less often point-by-point. Because of saline content, complex permittivity of sea ice is highly variable and difficult to predict. Therefore, to check the validity of a typical, plane-parallel, radiative-transfer-based ice emissivity model, we apply it to fresh water ice instead of salt-water ice. Radiance simulations for lake ice are compared with measurements over Lake Superior from the Advanced Microwave Scanning Radiometer on EOS (AMSR-E). AMSR-E measurements are also collected over Antarctic icepack. For each pixel, a thermodynamic model is driven by four years of European Center for Medium Range Weather Forecasts (ECMWF) reanalysis data and the resulting temperature profiles used to drive the emissivity model. The results suggest that the relatively simple emissivity model is a good fit to the data. Both cases, however, show large discrepencies whose most likely explanation is scattering both within the ice sheet as well as by cloudy atmospheres. Scattering is neglected by the model. Further work is needed to refine the scattering component of ice emissivity models and to generate accurate estimates of complex permittivities within sea ice.

Motivation & Objective

• To test the validity of a plane-parallel radiative transfer model for microwave emissivity using direct, point-by-point comparisons with satellite measurements.
• To reduce uncertainty in sea ice emissivity modeling by using fresh water ice—where complex permittivity is more predictable—instead of saline ice.
• To assess the impact of scattering in ice and atmosphere on emissivity model accuracy, especially in low-frequency microwave observations.
• To evaluate whether physical emissivity models can outperform purely statistical models in representing brightness temperature behavior over ice.
• To explore the feasibility of retrieving physical ice properties such as temperature, thickness, and complex permittivity from satellite microwave data.

Proposed method

• Simulating microwave brightness temperatures using a plane-parallel radiative transfer model driven by thermodynamic ice temperature profiles derived from ECMWF reanalysis data.
• Applying the model to two test cases: Lake Superior (freshwater lake ice) and the Antarctic ice pack, using AMSR-E satellite radiance measurements.
• Using temperature-dependent permittivity models: real part from Maetzler (2006), imaginary part from Hufford (1991), with effective permittivity for granular Antarctic ice via mixture models.
• Calculating transmission and reflection coefficients using Fresnel equations and exponential attenuation based on layer thickness, angle, and attenuation coefficient (α ∝ ν·imag(n)).
• Comparing simulated brightness temperatures with actual AMSR-E measurements across six frequency channels (6–89 GHz), analyzing bias, correlation, and weighting functions.
• Constructing a statistical model using 1000 temperature profiles and singular vector decomposition to compare with the physical RT model, assessing similarity in weighting functions and predictive accuracy.

Experimental results

Research questions

• RQ1 Does a physically based, plane-parallel radiative transfer model accurately simulate microwave emissivity over freshwater and glacial ice when compared to in-situ satellite radiance measurements at the pixel level?
• RQ2 To what extent do discrepancies between simulated and observed brightness temperatures arise from unmodeled scattering in ice and cloudy atmospheres?
• RQ3 How do the weighting functions of the radiative transfer model compare to those of a purely statistical model trained on the same data?
• RQ4 Can the model reliably retrieve physical ice properties such as temperature and complex permittivity, especially given the low variability of pure ice permittivity?
• RQ5 Why does the model overestimate brightness temperatures in the Antarctic ice pack despite accurate temperature profiles?

Key findings

• The radiative transfer model shows strong correlation with AMSR-E brightness temperature measurements over both Lake Superior and the Antarctic ice pack, despite neglecting scattering.
• Over Lake Superior, the model exhibits low bias and high correlation, indicating good performance for thin, fresh water ice with minimal scattering.
• Over the Antarctic ice pack, the model systematically overestimates brightness temperatures, with biases increasing at higher frequencies, particularly for horizontal polarization.
• The primary cause of bias is identified as scattering within the granular, high-density icepack (grain sizes up to 15 mm), which increases effective opacity beyond the model’s non-scattering assumption.
• Statistical models trained on the same data achieve similar accuracy to the physical RT model, but with different weighting functions, suggesting that scattering effectively alters the sensitivity of brightness temperature to subsurface temperature.
• Weighting functions from the RT model and statistical model are similar in shape but shifted upward, with the statistical model allowing negative weights—indicating that scattering decouples surface from deep ice temperature signals.