We have to guess what you want to achieve, because all we see is the failing code. If you mention the intention (preferrably in a comment, such that you can read it also, when you try to maintain the code in 2 years), suggesting an improvement is more reliable.
Look at this expression:
(Fh(:,i) / 8) * (Reh(:,i) - 1000) * Prh(i))
This is the multiplication of two [15 x 1] vectors, but this is not allowed (in Matlab and mathematically). What is the wanted output? Either use the elementwise multiplication:
(Fh(:,i) / 8) .* (Reh(:,i) - 1000) * Prh(i))
or transpose the 1st vector:
(Fh(:,i).' / 8) * (Reh(:,i) - 1000) * Prh(i))
The division by / will suffer from the same problem. Again either adjust the array dimensions or use the elementwise operation ./ .
You can solve such problems by your own: Split the complicated command into small blocks and process them one by one:
a = Fh(:,i) / 8;
b = (Reh(:,i) - 1000) * Prh(i);
c = sqrt(Fh(:,i) / 8); % SQRT is much faster than ^0.5
... etc
d = a * b;
Nuh(:,i) = (d / (1+12.7*c.*(Prh(i).^(2/3))-1);
Now you can check the dimension of the smaller blocks much easier and find problems efficiently.
Note: Inserting spaces around operators increase the readability.
