In case you already have a working startup.m file in your ~/matlab directory:

1. Add the line described above to your startup.m .

In case you can not mount ocp's directory (that's the case for everybody outside of LDEO) you can now download the whole stuff. Be warned, it's 160MB. Why that much ? It includes the full coastline database used also in GMT ! Where to find it ? Here.

In case you can or do not want to download the big archive, you can now find an html listing and the code for a number of the m-files here.

Various m-files:

Always try the help for any particular function before asking me ! And read it carefully !

Conventions in my help and in my functions (this differs from the standard look and feel of Matlab help because I find it too messed up) :

• matrix / array refer to at least 2-dimensional data fields.
• vector refers to a 2-dimensional array (all data fields in Matlab have at least 2 dimensions!) with one singular dimension.
• If there is one argument called optionsor op it stands for a Matlab-structure. That is a variable with sub-variables. You can access the sub-variables via op.blabla1 and op.blabla2 but then refer to the whole thing via op . This greatly simplifies the handling of functions with lots of arguments.
• Brackets like  [blabla]denote something optional or the default value.

You can find an html listing and the code for several of the following files here.

Graphics
loammap	Draw a nice map of scalars or vectors together with outlines of the continents.
contourpm	Draw a contour-plot differing between positive, zero and negative lines (color and b/w).
errorbar2	Draw a line plot with error bars. In difference to standard Matlab this function allows bars in 2 directions.
coledit	Interactive tool to create a new color map. (see also the coledit home page )
exportfig	Ever wondered how to get a figure on the paper with the same aspectratio as the one you had on the screen ? Try this function. It's supposed to handle that problem. (thanks Bruce!)
colorbar3	Thinner colorbar with label. Draws only used contourlevels when applied to a contourf graph.
stretch	Change the distribution of colors within a colormap. E.g. if you want to concentrate the resolution into one end of the color range or if you want to concentrate it around 0.
arrow	Draws nice arrows (read the help carefully!).
ceven	Center color range around 0
Data manipulation
nans	Replaces elements in an array (e.g. replace -999 by NaN).
remseason	Remove seasonal cycle from an array or vector.
Simple data analysis functions
nmean / nmin / nmax nmedian / nstd /ncumsum nsum	Applies the respective function to arrays despite NaN's.
rms	Calculates the root-mean-squared.
binav	Calculate bin-averages
lastval	Grab the last non-NaN number from an array (dimension can be specified)
Complex data analysis functions
objmap	Do an objective analysis of your data.
geteof	Calculate EOFs (principal components).
cancorr	Do a canonical correlation analysis.
getmssa	Do MSSA
lagcorr	Do a (lagged) correlation.
lagcov	Calculate the (lagged) covariance. (also useful for calculating regressions)
veccor	Calculate the correlation between two vector series/maps.
corrsig	Calculate the significance of a correlation coefficient
File handling
readnc	Load variables and attributes from a NetCDF file. Can load only parts of a variable !
ncload2	Like ncload (loads a variable from a NetCDF file), but replaces 'missing_value' by NaN
hdrload	Load an ASCII file which has a non numerical header.
textread / dlmread	Load mixed character/number ASCII files.
save2cdf	Save one variable with several attributes in a NetCDF file.
Geophysical functions
geowind	Calculate geostrophic winds from SLP

Tips and Tricks:

• Matlab is easiest to learn by doing. Reading the manual does, at least for me, not help a lot. If you, however, have no idea at all, read the first couple of pages of the getting started guide. It will introduce you to the basics of Matlab's scripting language. In case you have some experience programming everything except for the names of the functions should be straightforward from thereon.
• If you intend to do a lot of graphics, try to understand what figure, axes, axis, properties, and handles are.
• Don't type your stuff directly in!  Use little scripts!  Using CPM (cut-paste-modify) programming you will be a lot faster than by retyping everything.
• Comment EVERY m-file. You won't be able to understand what you did last year without comments.
• Write functions for reusable pieces of code.
• Ask somebody who has been using Matlab for a long time for existing functions. Virtually everything has already been written, you only don't know that it already exists.
• Use the built-in help a lot. A substantial part of the questions I have been asked are answered in the help. Remember to write a good help when you write functions which are to be used by other people. This will also help you when you have to reuse a function after 5 years.
• Organize your growing collection of m-files in different sub-directories. Include these directories into your search path (have a look at gerdstartup5 to learn how). This will prevent you from creating a mess in your working directories, from duplicating files, and from creating several different versions of m-files. (or it might actually lead to just that ;-)
• If you have written a nice and useful function, bring it to me and we will include it in my directories. This will save work for others.
• Where possible use matrix calculations instead of for loops !  Matlab is extremely slow with loops.
• If you really have to use loops, preallocate the result matrix by e.g. res=zeros(10,10) for a result matrix which is 10 x 10 in size. That's much faster than without preallocation.
• Avoid pcolor , if you want to print the picture afterwards (it is fine for quick displaying on the screen). When you display a large matrix it is extremely slow in postscript files. Use contourf instead.
• There is a matlab newsgroup comp.sys-soft.matlab . They sometimes have interesting discussions.
• 2000/11/17

If you are writing an m-file/function and have an error but would like to have a look at the variables right before the error occurs, use keyboard. It stops the execution at that point and waits for your input through the keyboard. You can look at single variables or even change them. With return you resume the execution of the m-file/function.

Links:

•  mexcdf toolbox allows to read and write NetCDF files from Matlab. Used by readnc and save2cdf .
•  m_map toolbox  allows to draw maps in all kinds of projections. Used by loammap .
•  EPSONDE collection includes a couple of useful m-files.
•  SEA-MAT collection has links to various oceanography related m-files and toolboxes.
•  Mathtools a former competitor (now swallowed by Mathworks) has made a comprehensive list of Matlab-toolboxes on the web. In case you are looking for something you're likely to find it here. This is a lot better than the contributed files provided by Mathworks itself.

Questions and Answers:

• How do I use two (or more) different colormaps in one figure ?

For the Matlab-purist, have a look at the Using MATLAB Graphics manual on the last pages. They give an example. Another, easier, way is to simply make multiple figures, print them piece by piece, and paste them together. This can be done by programs like Xfig or, within LaTex, with multiple \psfig commands within one \centerline command.
• How do I change the thickness of the lines in a contour-plot ?

contour(x,y,z,...,'linewidth',0.1)   default is 0.5 or 1.0 or so.
• How do I get grid lines at other locations than the labels (e.g. more grid lines than labels) ?

Method 1: Use set(gca,'xtick',[tickvector]);  labelmatrix = get(gca,'xticklabel'); labelmatrix(unwantedlabels,:) = ' '; set(gca,'xticklabel',labelmatrix);
Method 2: Draw dotted lines with the plot command at the locations you want, don't use grid.
• Why is plotting with loammap very slow ?

This has two possible reasons. First, loammap has a hold on internally. If you do not clear the figure before rerunning a script it piles up drawing above drawing, the last of which you only see. This makes it slower and slower. Solution is to add a clf at the beginning of the script. Second, you are drawing too many contourlevels. A number of 50 or so gives reasonable results in speed and appearance.
• How do I get a low height, wide width figure ?

Use axes('position',[0.1,0.1,0.8,0.1]). Modify the 4 numbers to change the size of the figure. They are [x coordinate and, y coordinate of the lower left corner, and x, and y width/height]. All are given in screen values (i.e. lower left corner of the figure is [0,0], upper right [1,1]). Leave some space for the labels (i.e. [0,0,1,1] will fill the whole figure without leaving space for labels). This a more versatile command than subplot.
• How do I plot two curves with two different ranges on top of each other ?

Use plotyy. In addition to plotting the two curves it gives you as output the handles to two different axes which are plotted on top of each other. To control the axis range of these two plots use set(axis1,'ylim',[min_y,max_y]). Replace axis1 with the handle of the plot you're changing and min_y and max_y with the respective ranges.
• I have several profiles of different length with data at different depths. How do I merge them into on array for easy plotting ?

Use interp1 to interpolate the data onto the same depth levels and store every profile as one column in the array. pcolor or contourf no give nice graphs of the data.  interp1 gives NaN for values outside the 'known' range. This might look strange with contourf but shows when/where you have data and when/where not.
• I want only a some tickmarks labeled.

With get(gca,'xtick') you can get the positions of the x-tickmarks (use set to change them). With get(gca,'xticklabels') you will get the respective labels. By default Matlab labels all tickmarks. The output of the second command is a character array. Replace the labels you don't want with blanks and feed it back with set. Watch out, 'xticklabels' wants the same number of rows as 'xtick' has elements. Replace x by y or z to change the other axes.
• When I do a complicated plot with filled contours everything appears white on the screen.

This appears to be a bug in Matlab. The postscript files are usually ok. To see it correctly on the screen you can try set(findobj('tag','m_grid_color'),'face_color','none').

mailing-list:

• I created a mailing list through which I can update you on changes in the libraries and warn you when I have to take the server down. If you want to be added, send me an e-mail.

last modified June 17th, 2004.
send comments and question to  Gerd