Infrared radiance cloud detection diagnostic tool

This is an adaption for ACCORD DA tools of the IR cloud detection diagnosis tool done by (c) [email protected] (25th Novemeber 2022) available on ECMWF in ~fi9/f90/clddet/clddet.tar

Adaptation was made by Joan Campins ([email protected] on 9th May 2023)

1. Purpose

This diagnostic tool is intended for use in evaluation of the performance of the infrared radiance cloud detection scheme in the context of HARMONIE-AROME data assimilation systems. As of February 2022, the most relevant infrared sounders include IASI and CrIS. The cloud detection scheme follows broad principles that are outlined in McNally and Watts (QJRMS, 2003).

2. Content

The tool comprises two shell scripts:

• clddet_diagnosis.sh

• clddet_plotter.sh

and two other code files (in src):

• clddet_analyzer.f90
• clddet.py

3. Usage

The main production script is clddet_diagnosis. In the heading of this script, the user will need to specify:

1. the input directory (indir), which usually is the experiment identification name
2. the output directory (outdir), where large files are stored
3. date and time of the analysis that they are diagnosing (cycle)
4. name of the infrared sounder that they are working with (sensor)

It is assumed that the relevant experiment (input directory) exists.

Once these settings are correctly updated, the tool is run by command

./clddet_diagnosis -i indir -o outdir -c cycle -s sensor

Help can be obtained with ./clddet_diagnosis -h

The script needs some input files, called Fetch input files, taht are necessary to run clddet_analyzer.x:

• relevant source code as cloud_detect_setup.F90: stored in src, which is a copy from ~./src/arpifs/obs_preproc/cloud_detect_setup.F90

• namelist file: ${sensor}_CLDDET.NL: stored s stored in src, which is a copy from ~./nam/${sensor}_CLDDET.NL

• HM_Date log file: HM_Date_${cycle}.html: stored in ${indir}, which is a copy of ~./archive/log/HM_Date_${cycle}.html

• ASCII file derived from odb files (ECMA.*): clddet_ascii.dat_${cycle}: stored in ${indir}

  • clddet_ascii.dat_${cycle} can be obtained from ECMA by means:

  • tar -xf odb_stuff.tar odbvar.tar

  • tar -xf odbvar.tar odbvar/${odbname} ; where odbname="ECMA.${sensor}" (sensor= iasi, cris or airs)

  • cd ./odbvar/${odbname}

  • module load odb odb_api

  • odbsql -q "select satellite_identifier,degrees(lon),degrees(lat),vertco_reference_1,fg_depar,rank_cld,datum_event1.contam_cld_flag from body,sat,hdr,radiance_body" > clddet_ascii.dat_${cycle}

In usual circumstances, running the script will take several minutes. After the script run is completed, a symbolic link "clddet_sorted_smoothed.dat" should appear in the working directory. This link will point to a non-zero sized ASCII file placed in the output directory. The ASCII file contains data from all observations of the sounder of interest at the appropriate date and time, as far as these data are correctly stored in clddet_ascii.dat.

Graphical plotting of results can then be done with the command

./clddet_plotter -j ARG

where ARG is an integer number that represents a running index of an individual infrared satellite sounding. By default, ARG=0, which means that the output figure will by representative of average spectrum computed from the input data. The plotting script will save the produced figure in file clddet.png.

4. Interpretation

The cloud detection scheme makes use of the expectation that stratospheric- and upper-tropospheric-sensitive channels are more often clear of clouds than lower-tropospheric channels. The x-axis in the output figures consists of channels in vertically-ranked space such that the highest-peaking channels are at far left, and lowest-peaking channels at far right. If the scheme has identified some cloud contamination in the given sounding, the cloud-contaminated and cloud-free channels are distinguished from each other by the vertical dashed red line somewhere in the middle of the figure: to the right of this line, all channels are cloud-affected and will consequently not be used during the assimilation. Channels on the left-hand-side of the red vertical line are free from the cloud effect and can be assimilated.

Since the cloud detection relies heavily on O-B departure data, it can easily be interfered if bias corrections are not up to date for any reason. This will be most easily diagnosed on the stratospheric-peaking channels, that are most of the time not affected by cloud, and therefore should on average be very close to zero in (O-B). In a healthy system setup, the left-hand part of the smoothed O-B departure curve should be approximately horizontal. Considerable deviation from this pattern is likely indicative of a problem of some kind in the system setup.