WP-2000 Calibration of EDA spread and adaptation of the observation error model

Title
WP-2000 Calibration of EDA spread and adaptation of the observation error model
Report
Date Published
02/2022
Series/Collection
ESA contract 4000130590/20/NL/IA
Author
Katie Lean, 
Sean Healy, 
Abstract

In this second phase of this project to investigate potential future constellations of small satellites carrying microwave (MW) sounding instruments, there are two subjects that have been explored. In the first part of this report, real MW data have been used to establish links between the observation impact using the Ensemble of Data Assimilations (EDA) method, described in the initial consolidation phase, and in Observing System Experiments (OSEs). Reductions in the EDA spread, which indicate a positive impact from the MW data, were compared with reductions in the standard deviation of forecast error. The forecast error has been evaluated by comparing the short-range forecast to a reference firstly provided by analyses from the operational system and secondly by radiosonde observations, taking into account the observation error. There is overall good qualitative similarity between EDA spread and both measures of the forecast error. Better agreement of the observation impact is found in the lower/mid-troposphere with possible under-estimation of the impact using the EDA spread in the stratosphere. While it is not useful to derive a quantitative relationship here, the results show that the EDA method provides a basis to investigate the relative impacts from the future simulated small satellite data and with the addition of reference points from the use of existing MW data.
The second part of the report presents the adaptation of the observation error model for the MW temperature sounding channels needed due to the unavailability of key low frequency channels on the small satellites. For channels with cloud sensitivity, the observation error model for the all-sky assimilation inflates the observation error in cloudy regions by using an indicator of the presence of cloud in the observations or model. The use of the 52.8GHz (channel 4 on AMSU-A) has been proposed as the basis to construct an alternative cloud indicator forMWtemperature sounding channels.
Evaluation using AMSU-A observations showed that a similar structure to the current observation error model can be achieved with the new indicator. Assimilation experiments showed some small but significant negative impacts in short-range forecasts as measured by low peaking temperature sounding channels of ATMS. There is also a small but not significant signal in some humidity sensitive observations, however, impacts were generally neutral. Neutral impacts at longer forecast lead times measured by verification against analyses showed that these short-range changes did not affect the medium-range forecasts. The results confirm that 52.8GHz provides a viable alternative to use for the small satellite temperature sounding observation error model. However, it is noted that the potential impact from the loss of the lower frequencies is not fully captured as the channels are not currently used in the operational assimilation of AMSU-A.

URLhttps://www.ecmwf.int/en/elibrary/81289-wp-2000-calibration-eda-spread-and-adaptation-observation-error-model
DOI
10.21957/1auh0nztg