How to find a particular value from 3-d matrix?

Hey MATLAB universe!

I have a netcdf file of sea surface temperatures (360x40x444) and found the mean of it. I have obtained a 2-d matrix and need help with the following:

I want to find out how many cells have a particular value (say, 1.75) from the entire 360x40 matrix.

My main aim is to find how many grid points (and what area) is covered by a temperature of 1.75 degree celsius. I have attached a .mat file.

Thanks!!

Answer 1

[Walter Roberson]

None of the values are 1.75. The closest is 1.74979725818628

tolerence = 0.01;
target = 1.75;
nnz(abs(mean(:)-target)<tolerence)

%or

tolerence = 0.01;
target = 1.75;
nnz(ismembertol(mean, target, tolerence, 'datascale', 1))

The advantage of the second of these is that target could be a vector of values, each of which would be searched for:

tolerence = 0.01;
target = [1.75, 1.1];
[isinrange, idx] = ismembertol(mean, target, tolerence, 'datascale', 1);
nnz(idx==1)
nnz(idx==2)

Answer 2

[dpb]

If you're certain the value 1.75 exactly exists in the file and that's all you want exactly (to machine precision) then simply

V=1.75;
[ir,ic]=find(sst==V);

does the trick. As floating point differences can cause 1.75+/-2.2E-16 to not match because is >eps(1.75) away from 1.75, you may find

tol=1E-5;      % say, how close is "close enough"???
V=1.75;
[ir,ic]=find(ismembertol(sst,V,tol));    % find those within tolerance, tol

ismembertol is just "syntactic sugar" for writing the explicit test for abs(A-V)>tol

[Keegan Carvalho]

Thank you dpb. I will try the code and see how it works.

Thank you both for your help as I have got a better understanding of this

[Walter Roberson]

Without the 'datascale' option, ismembertol() works with relative tolerance rather than absolute tolerance.

[Rik]

That is something that still confuses me about the ismembertol function. Why would you not want to use absolute tolerances? It makes more sense to me to allow the relative tolerance as an optional parameter, instead of how the function currently works.

[Walter Roberson]

If you want to express then you do not generally want to have to know their range ahead of time in order to construct an absolute tolerance, and you do not want to have to keep track of which vector elements are which range in order to know what the appropriate absolute tolerance is for each one. You want to know if A is the same as B to "within round-off" where "within round-off" is related to a small multiple of eps of the one with the larger absolute value -- or to within the relative fraction if that is specified.

ismembertol(A,B,0.001) -- are they the same to within 1 part in 1/1000?

ismembertol(A,B,0.001,'datascale', 1) -- are they the same to within 0.001

Asking if two numbers are the same to within some absolute tolerance is common, it it does seem to be a bit ackward to have to spell out 'datascale', 1 for that common case. Perhaps an ismemberabstol() would have been a useful interface.