Updated 28 Jan 2019
This toolbox contains a collection of Matlab functions for calculating the skill of model predictions against observations. Its primary value is in producing target and Taylor diagrams.
Evaluation of the predictive skill of models generally rely on analyzing the values provided by a variety of statistical metrics. While this analysis may be relatively straightforward when using a few metrics, it can become complicated when considering multiple model predictions, either for different model parameterizations, with respect to multiple references, or many models. To aid in this analysis, mathematical diagrams have been designed to graphically indicate which models provide the best predictive skill relative to a chosen reference. Two particular types of diagram that are simple to interpret and are widely used are the target and Taylor diagrams. These diagrams provide a means to compile statistical measures of the predictive skill of multiple models into a single graph that facilitates comparison and analysis.
Peter Rochford (2019). PeterRochford/SkillMetricsToolbox (https://www.github.com/PeterRochford/SkillMetricsToolbox), GitHub. Retrieved .
For further information regarding the status of the transparency of markers option (alpha), refer to my MarkerTransparency toolbox at https://www.mathworks.com/matlabcentral/fileexchange/65194-peterrochford-markertransparency.
The transparency of markers option (alpha) that allows the blending of symbol face color (0.0 transparent through 1.0 opaque) works up to Matlab R2017b but fails with version R2018b. This has been reported as a bug to MathWorks.
This is great for plotting error metric of coastal numerical models - hydrodynamics, morphology etc.
Refer to the Skill Metrics Toolbox FAQ at http://github.com/PeterRochford/SkillMetricsToolbox/wiki/FAQ. Please let me know if there is a case that I have not covered. Be sure to download the latest version 1.5 and refer to the example M-files that I have included in the toolbox.
How can I change the fontsize of the titles??
The version 1.4 update is motivated by user feature requests implemented in my Python version. If you use this toolbox please post feedback on what features you use (e.g. produce Taylor diagrams, GitHub Wiki, etc.) along with a rating. This will help guide my future development of the toolbox. At 1,188 all time downloads this toolbox must be getting used for something :-)
From an Internet search and some testing I found you can change the font size for all text in a single figure to 18pt by adding the following statement after the call to taylor_diagram:
This hopefully meets your need.
I am trying to figure out how to modify the font size for the correllation coefficient and standard deviation tick labels. For instance in taylor1_example (in the packaged Examples folder), there are correlation coefficients 0:0.1:0.9, 0.95, 0.99 and1. Currently they are set to some default font size but I would like to make them bigger, say size 18. I presume maybe this can be done with a call to one of the exported file handles, but I can't quite seem to figure out what it would be. I see these font sizes don't vary in any of the examples. Any suggestions?
I tried to test the code with target data however, no success. The target data has loading problem. tried different version of Matlab and various other loading options (ascii etc) however nothing worked. Wondering if anyone else has the same issue? Here is the error message that i get
Error using load
Unable to read MAT-file
mydirectory\target_data.mat. Not a binary
MAT-file. Try load -ASCII to read as text.
Thanks for this code, really useful but when i tried it the model are overlapping and i can't separate them.
A Python 2.7 equivalent of this toolbox is now available. A wiki describing the package can be found at
To install the Python package simply use the pip command:
pip install SkillMetrics
Here is the basic resource for Taylor diagram, the paper has an example to explain it.
Amazing.Since I'm definitely a newbie, could you please send me a copy of Taylor_diagram_examples.pdf? It's not included in the contracted files.Thank you sincerely.
Brilliant. Completely caning this for all sorts of analysis of model-data errors. One of the most useful downloads I have come across on Matlab.
My first downloading file !
Added ‘CMapZData’ option that allows markers to be color scaled according to other measures such as bias. Added two Taylor diagram examples: 1) use of legend with multiple columns, and 2) color scaling of markers using bias value.
Changed toolbox to use xlabel/ylabel, etc., rather than custom use of text function. This allows easier control of font size for labels using returned handles as documented on the FAQ: http://github.com/PeterRochford/SkillMetricsToolbox/wiki/FAQ.
Added options to adjust marker symbol face color (transparent through opaque) and marker size. Axes font size and line widths are now adjustable via default figure properties. Implemented better default angle for placement of RMSD contour values.
Added options for displaying observation standard deviation on the axis, a label for the point, and a circle.
A Wiki of the toolbox tutorial can be found at https://github.com/PeterRochford/SkillMetricsToolbox/wiki.
Corrected toolbox name.
Version 1.2 is compliant with GNU Octave, version 4.2.0. All of the latest example plots were generated with Octave and may not be as good as those using Matlab.
Newly added statistical metrics are bias, Brier score, Brier skill score, and RMSD.
Corrected typographic errors.
Create scripts with code, output, and formatted text in a single executable document.