Skip to main content

Plots and analyses atmospheric profile data from UWyo database

Project description

SkewT provides a few useful tools to help with the plotting and analysis of upper atmosphere data. In particular it provides some useful classes to handle the awkward skew-x projection.


18 June 2015. I have been delaying creating a new release for a long time, but here it is! What’s new in SkewT version 1.1.0?

  • Fixed some bugs in CAPE calculations: we were mixing ambient temperature-derived significant levels with virtual temperature formulation for CAPE, which was producing some weird results in special cases.

  • Fixed plotting and CAPE calculations for soundings that had a tropopause above 100hPa

  • A few minor plotting issues, in particular things look better now if you want to plot a different temperature or pressure range (use tmax/tmin pmax/pmin kwargs to make_skewt_axes()).

  • Responses to bugs raised by community (Thanks dudes).

SkewT has undergone a major overhaul in order to implement the new features available in Matplotlib 1.4. It also has a bunch of new features that we have been meaning to implement since inception. We hope you like it more than ever!

Important Notice

  • Since version 1.0, SkewT has explicitly included the new SkewX classes that are showcased on the matplotlib website: This stuff is completely fundamental to and we are greatful to Ryan May from Unidata for providing this to the Python community.

Sounding Data

The easiest way to get some sounding data is to visit the University of Wyoming’s website:

To get some sounding data, follow the link and find the date, and location you are interested in, view the data as a text file and just save the file to your system. If you want to get loads of data please be considerate about the way you go about doing this! (Lots of wget requests makes the server admins unhappy).

You can also pass your own data to SkewT by writing a text file in identical format to the University of Wyoming files, which are constant-width columns. Here’s a sample of the first few lines of one of the bundled examples:

94975 YMHB Hobart Airport Observations at 00Z 02 Jul 2013

    hPa     m      C      C      %    g/kg    deg   knot     K      K      K
 1004.0     27   12.0   10.2     89   7.84    330     14  284.8  306.7  286.2
 1000.0     56   12.4   10.3     87   7.92    325     16  285.6  307.8  286.9
  993.0    115   12.8    9.7     81   7.66    311     22  286.5  308.1  287.9

Alternatively you can create a dictionary with the column headers as keys and the data as 1D python arrays (preferably use ma.masked_array). There’s more about this under the “Running SkewT” section below.

Installing SkewT

We recommend that you download the tarball (big green button on this page) and run:

python install

If you want to put it somewhere different than your system files, you can do:

python install --prefix=/path/to/local/dir

Just remember, if you use a non-standard location you’ll have to tell python about where you install it. An easy way to do this is to add the environment variable PYTHONPATH to your bashrc (you can read about this elsewhere).

Because SkewT is written purely in python, you don’t even have to install it try it out! Just download the tarball and extract it somewhere convenient, and navigate to SkewT/skewt, and everything you need is right there.

You can also install using the package manager, but I have had some complaints about dependency issues (All you should need is matplotlib and numpy).

Running SkewT

There are three basic ways to run SkewT. You can execute it from the command line with a text file name as an argument, or you can import it as a module and pass it a text file name, or you can pass it data directly.

Running from command line

From the command line (navigate to SkewT/skewt) you can type:

python /path/to/sounding_filename.txt

What you’ll get is all of the default settings. If you do this with the bundled example in SkewT/skewt/examples/bna_day1.txt, you’ll get this graphical output.

If that’s what you want, well and good, but if you want to tweak things like the colours, read on…

Import SkewT as a module

Assuming you have installed the package on your system and your sounding file is in your working directory, typical usage of SkewT could look like this (I use IPython):

In [1]: from skewt import SkewT
In [2]: S=SkewT.Sounding("bna_day1.txt")
In [3]: S.plot_skewt(color='r')

The function plot_skewt() is a wrapper for a bunch of other functions. This will give you exactly the same plot as running SkewT from the command line, but you have immediate access to all of the matplotlib plot options for the profile traces and the barbs, but you don’t get any control over anything else.

The full sequence of commands to get what plot_skewt wraps is this:

In [1]: S.make_skewt_axes(tmin=-40.,tmax=30.,pmin=100.,pmax=1050.)
In [2]: S.add_profile(color='r',bloc=0)
In [3]: parcel=S.get_parcel()
In [4]: S.lift_parcel(*parcel)

You don’t have to put the tmin and other keyword arguments in to make_skewt_axes() unless you want to plot against different values from the defaults shown here. The keyword argument bloc stands for ‘’barb location’’ and allows you to shift the wind barbs to the left or right. This is handy if you want to plot multiple profiles on the one Skew-T diagram, for example, to compare today’s and yesterday’s soundings:

In [1]: S=SkewT.Sounding("./skewt/examples/bna_day1.txt")
In [2]: T=SkewT.Sounding("./skewt/examples/bna_day2.txt")
In [3]: S.make_skewt_axes()
In [4]: S.add_profile(color='r',bloc=0)
In [5]: S.soundingdata=T.soundingdata      # replace the sounding data in S with that from T
In [6]: S.add_profile(color='b',bloc=1)

Import as a module and run with your own data

Got sounding data from another source? Want to make Skew-T diagrams of model output? Look no further. All you need to do is define a python dictionary like so:

In [1]: mydata=dict(zip(('hght','pres','temp','dwpt'),(height_m,presssure_pa,temperature_c,dewpoint_c)))
In [2]: S=SkewT.Sounding(soundingdata=mydata)

At a minimum we require pres, temp and dwpt to generate the profile traces, and hght is required for parcel calculations (although a future implementation will use a hydrostatic atmosphere assumption). The other keys accepted are those listed in the University of Wyoming sounding data header above.

Parcel Ascent

As of version 1.0, SkewT has a full parcel ascent routing including automatic parcel definitions and CAPE/CIN and significant level calculations.

Automatic Parcel Definition

There are three standard parcel definitions used in predicting severe weather (see

  • Surface Based ('sb'): The surface conditions. Found by taking the lowest level where all data is available. This may not represent the convective potential of the sounding very well but is commonly used.

  • Mixed Layer ('ml'): A parcel representing the mean potential energy in the lowest 100-mb of the atmosphere. Found by averaging potential temperature and the water vapour mixing ratio.

  • Most Unstable ('mu'): The most unstable parcel of air found within the lowest 300-mb of the atmosphere. Found by calculating CAPE for conditions at all levels in the sounding data, and determining the equivalent surface parcel by adiabatic descent. (Note: if CAPE is 0 for all levels this routine defaults to the surface based parcel)

To calculate one of these parcels for your sounding, use the get_parcel() routine, which is a wrapper for surface_based_parcel(), mixed_layer_parcel() and most_unstable_parcel(). Optionally pass it the parcel type you want (default is 'mu'):

In [1]: S=SkewT.Sounding("./skewt/examples/bna_day1.txt")
In [2]: parcel=S.get_parcel('mu',depth=300)
In [3]: parcel
Out[3]: (1000.0, 23.037, 13.626, 'mu')
In [4]: S.lift_parcel(*parcel_2)

Or, you can define your own parcel (the fourth item is just some text which appears on the Skew-T diagram):

In [5]: parcel_2=(1000.0, 25.0, 18, 'user')
In [6]: S.make_skewt_axes(); S.add_profile();
In [7]: S.lift_parcel(*parcel_2)

CAPE/CIN calculation

Definitions in this section are based on Markowsi and Richardson (2010).

The lift_parcel() routine above is a wrapper for the get_cape() routine, but it also handles the graphics. The get_cape() routine, by itself, will calculate significant levels and CAPE/CIN:

In [8]: P_lcl,P_lfc,P_el,CAPE,CIN=S.get_cape(*parcel)
In [9]: print P_lcl,P_lfc,P_el,CAPE,CIN
870.560154927 859.695806371 382.117602258 427.793216382 -8.64938413185

In [10]: P_lcl,P_lfc,P_el,CAPE,CIN=S.get_cape(*parcel_2)
In [11]: print P_lcl,P_lfc,P_el,CAPE,CIN
902.773891386 902.773891386 178.058628014 2540.55724083 0.0

get_cape() complains a bit if there are any dew point temperatures missing in the profile, but its default behaviour is to fill these with the minimum dewpoint in the column, and this will have a minimal effect on the CAPE calculation.

The lifted condensation level (LCL) is found by solving for the intersection of the temperature for dry adiabatic ascent for the parcel, and a line of constant water vapour mixing ratio.

To find the level of free convection (LFC), the parcel is lifted along a moist adiabat from the LCL. For details, please see the moist_ascent() routine in All intersections of the parcel temperature and the environmental temperature are identified. Strictly speaking, all such levels are equilibrium levels. There are basically three possible scenarios:

  • Parcel cooler than environment at LCL and no equilibrium levels: There are no unstable levels in the profile above the LCL, so the LFC does not exist.

  • Parcel warmer than environment at LCL: This means that LFC=LCL, and there must be at least one stable equilibrium level, which could be as high as the tropopause.

  • Parcel cooler than environment at LCL and at least two equilibrium levels: This means that the parcel is initially stable at the LCL, but further lifting will bring it to a condition where it becomes unstable. The LFC is defined as the first point at which this occurs.

The term Equilibrium Level (EL) is often used to describe the first stable equilibrium level above the LFC, if this exists. Once the LCL, LFC and EL have been defined, we can calculate the Convective Available Potential Energy (CAPE) and Convective Inhibition:


This expression only applies to the region where T_parcel>T_environment between the LFC and the EL. trapz is a basic trapezoidal integration routine from numpy.` Similarly for CIN:


Which applies to the region where tparcel<=tempenv between the surface and the EL.

The example above (bna_day1.txt) is a perfect demonstration of why this behaviour might not be desirable. Using the textbook definition (i.e. totalcape=False) of the EL, you get practically no CAPE, but it’s clear that there is a large layer of instability aloft. However, if you define the highest equilibrium level as the EL (i.e. totalcape=True), you get an answer that is more representative of the conditions of the day.

The keyword argument totalcape lets you override the default definition of the so-called ‘Equilibrium Level,’ (EL) which I took from Markowsi and Richardson (2010, p. 33): “The equilibrium level is defined to be the height at which a buoyant lifted parcel becomes neutrally buoyant, that is, the height above the LFC at which the parcel temperature is equal to the environmental temperature.”

Working Examples

We have bundled in a set of example soundings in the SkewT/skewt/examples directoy. You can run them like this:

$ python example1

Substitute digits 1-4 to get the different examples. The code for these is right down the end of the file so you can have a look and play around with them if you want without affecting how SkewT works on import.

  • Example 1: Two soundings from Hobart that I used to develop al ot of the initial code base

  • Example 2: Total CAPE vs. Textbook CAPE

  • Example 3: Some severe weather events in Australia, with automatic parcel definitions.

  • Example 4: Use of custom parcels

  • Example 5 (new in v1.1.0): High tropopause sounding

The sounding files and output graphics for the examples are all hosted here.

To-Do List

  • More column diagnostics.

  • Hodographs? Anyone?


  • Ross Bunn from Monash University is actively developing and finding all my warty bugs.

  • Gokhan Sever (North Carolina) is a keen user and has been encouraging me to add more stuff. It’s thanks to him that I have finally implemented the CAPE routines.

  • Simon Caine.

  • Hamish Ramsay (Monash) has promised to at least think about adding some extra diagnostics.

  • Holger Wolff as tester

Thanks for your interest in this package and I’d love to hear your feedback: thomas.chubb AT

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

SkewT-1.1.0.tar.gz (52.9 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page