access_access_DiagnosticToolsV1 - ACCESS-NRI/accessdev-Trac-archive GitHub Wiki
NESP2.5 milestone for June 2017: Diagnostic Toolkit version 1
This document describes the computer programs published as the NESP milestone# 2: Diagnostic Toolkit version 1. The programs were written and used for investigating the key systematic errors in a high resolution (60 km grid) version of the current ACCESS atmospheric model (with the GA7 climate configuration). The result of this investigation was documented in the NESP2.5 milestone# 1, which has been published as a NESP ESCC Hub technical report (Rashid et al. 2017; hereafter, RZS17). This report may be accessed from the NESP ESCC Hub web site: http://nespclimate.com.au/wp-content/uploads/2017/07/ESCC-TR001-ACCESS-1705.pdf
The computer programs are written mainly using the NCAR Command Language (NCL; http://www.ncl.ucar.edu), GraDS (http://cola.gmu.edu/grads/) and FORTRAN. The NCL programs use its standard library functions, the documentation of which may be found at: http://www.ncl.ucar.edu/Document/Functions/list_alpha.shtml. A brief documentation of the diagnostics and the programs used to calculate the diagnostics is given below. See the above NESP ESCC Hub tech report (RZS17) for a detailed scientific discussion of the diagnostics and the model systematic errors. The computer programs and some netCDF files containing the intermediate results of the computation are made available through a code repository (https://trac.nci.org.au/svn/access_tools/NESP_diagnostics/trunk). Note, you need to use your NCI account username and password to access this code repository. For further information, contact Harun Rashid (Harun.Rashid @ csiro.au) for the NCL programs and Hongyan Zhu (Hongyan.Zhu @ bom.gov.au) for the GrADS and FORTRAN programs.
Diagnostics for surface climate errors. The main diagnostics used for the surface climate are seasonal mean surface temperatures (tas), sea-level pressures (psl), and the shortwave (rss) and longwave (rsl) radiations (Figs. 1–4 in RZS17). The computer programs used to calculate these diagnostics are:
compTimeMean_CGCM-Obs_SurfFlds.ncl
compTimeMean_CGCM-Obs_SurfRadFlds.ncl
plot_SeasMeans_ObsMod_SurfFlds.ncl
plot_SeasMeans_ObsMod_SurfRadFlds.ncl
The first two programs calculate the time mean surface fields from three dimensional (longitude, latitude, time) climate variables from all months, as well as from months corresponding to four standard seasons (December-January-February (DJF), March-April-May (MAM), June-July-August (JJA), and September-October-November (SON)). The results of these calculations are output as two netCDF files:
ObsMods_SurfFlds_TimeMean.nc
ObsMods_SurfRadFlds_TimeMean.nc
These netCDF files are then used as inputs to two plotting programs (the third and fourth program files above) to produce geographical plots of the seasonal mean fields and the associated errors.
Diagnostics for upper-level circulations. The diagnostics used for examining errors in the upper-level circulations are: the eddy stream function (psi) and velocity potential (chi) at the 200-hPa level, and the storm tracks (v'^2^) at 500 hPa (Figs. 5–7 in RZS17). The programs that calculate and plot these diagnostics are:
plot_SeasMeans_ObsMod_Flds.ncl
comp_StormTracksSyn_ObsMods.ncl
The first program creates geographical plots of the seasonal mean fields and the associated errors. The inputs to this program are again the three dimensional (longitude, latitude, lev@200/500, time) circulation fields (psi, chi, and vsq). The psi and chi variables were calculated using the NCL function “uv2sfvpF” from the zonal and meridional winds. The second program calculates the storm track diagnostics (squares of the 2-8 day filtered eddy meridional wind anomalies at 500 hPa).
**Diagnostics for the Hadley and Walker circulations.**The Hadley and Walker circulations were represented by the meridional and (equatorial) zonal overturning stream functions, respectively (Figs. 8 and 9 in RZS17). The following programs calculate and plot these diagnostics:
compZonMeanHC_ObsMod_mon_LoRes.ncl
compEqMeanWC_ObsMod_mon.ncl
plot_ZonMeanHC_ModsObs.ncl
plot_EqMeanWC_ModsObs.ncl
The inputs to first two programs are the four-dimensional (longitude,latitude,level, time) meridional wind and the irrotational component of the zonal wind, respectively. The programs write out the computed meridional and equatorial zonal overturning stream functions, respectively, in two netCDF files. The latter are then used as inputs to the two plotting programs (the third and fourth above) to illustrate the seasonal mean Hadley and Walker circulations and the errors associated with them.
**Diagnostics for tropical rainfall errors.**The model errors in the geographical distribution of rainfall and the rainfall diurnal cycle (Figs. 10 and 12 in RZS17) are calculated in the following GrADS and FORTRAN programs:
shoa-color-pr-bias-Fig10a.gs
shoa-color-pr-diff-Fig10b.gs
write-diurnal-cycle-land-sea-MC.f90
DiurCycl-BORNIE-N216-N96-land.gp
DiurCycl-BORNIE-N216-N96-sea.gp
The first two (GrADS) programs calculate and plot the time-mean rainfall errors over the Maritime Continent region. The FORTRAN program (with “f90” extension) calculates the rainfall diurnal cycles separately over the land and sea points. The last two programs (with “.gp” extension) plot the computed diurnal cycles. All the input and output data are described in the program files.
The rainfall amount distributions (plotted separately for all, sea and land grid points) (Fig. 11 in RZS17) are calculated and plotted in:
plot_PrecipPDFs.ncl
The inputs to this program are three-dimensional (longitude,latitude,time) rainfall data from observations and model simulations. The output is postscript file containing the figure.
Single-column model (SCM). The single-column model (SCM) allows one to simulate the evolution of an atmospheric column without the complexities associated with horizontal heterogeneities and interactions between neighbouring grid points. Tendencies of temperature, humidity, and wind due to surface fluxes and the large-scale flow are supplied, thereby isolating the role of the model’s parameterization schemes. This makes the SCM an indispensable component of the diagnostic toolkit for model evaluation. Robert Warren (rob.warren @ monash.edu) and Christian Jakob (christian.jakob @ monash.edu) at the Monash Node have been using the SCM to investigate deficiencies in the ACCESS convection parameterisation.