All-Sky Assimilation of Hyperspectral Infrared Sounder Observations in GEOS

Abstract

Hyperspectral infrared (IR) radiance observations have been one of the major
data sources assimilated into data assimilation systems over the past 20 years.
However, observations peaking in the lower and mid-troposphere are
underutilized in clear-sky radiance data assimilation, as the quality control
procedure removes a significant portion of cloud-affected observations.
To enhance the impact of these hyperspectral IR observations, particularly during
high-impact weather events, one approach is to assimilate them directly under all-sky
conditions.

The framework for all-sky assimilation of hyperspectral IR radiance observations
has been developed within GEOS. The symmetric cloud effect has
been used as a cloud proxy in observation error modeling, providing a
balanced representation that mitigates discrepancies between observations
and model simulations for IR all-sky assimilation. Observation errors vary with
different cloud conditions and have been modeled as both cloud-amount dependent
and inter-channel correlated.


The evaluation of simulated cloud-affected infrared (IR) brightness temperature (BT) has been
conducted, along with an analysis of its sensitivities to all hydrometeors. The simulated cloud-affected
BT tends to exhibit broader structures and lacks small-scale details, with excessive cloud generation
sometimes observed in the model. To address this issue, the optical properties of simulated clouds
have been revisited, including cloud effective radius, the cloud-overlap scheme, and the cloud look-up
tables (LUTs) in the CRTM. The fit between the cloud-affected background and observed BT has been improved,
 particularly in terms of bias. As a result, the overall impact of all-sky assimilation has been positive,
not only improving the simulation and forecasting of hurricanes in terms of their paths and intensity but
also enhancing the accuracy of numerical weather prediction (NWP) forecasts within GEOS.