It is a bit of a didactic distinction without foundation, maybe tied to history.
Typically, intro calc is divided into differential calc then integral calc. Differential involves only derivatives, built up on the ##\lim \frac{f(x+h)-f(x)}{h}## as h goes to zero. this is slope identification. Integral calc is finding the area under a curve using the limiting sum of rectangles. It is then argued that, by some miracle, the integral is functionally the same as the anti-derivative from differential calc. This gives us a quick way to avoid laying out rectangles and summing, since we can, by rules and insight (much like factoring in algebra) figure out how to quickly get the anti-derivative of a function we wish to integrate. This is all calculus 1.
Multivariate calc is a lot more complex and subtle. Usually in calc 1 there are only two dimensions, X and Y, and y = f(x). Mathematicians would more accurately say this one dimensional calculus, where functions map from ##R^1 \rightarrow R^1## - the real line to the real line. In multivariate you now have an n-dimensional domain mapping to another space, typically (but not always!) on dimensional.
The level of abstraction is quite a bit higher. In addition to all this space mapping, one ends up dealing with new 'things', like vectors and partial derivatives. On top of that there are things like changes of coordinates. For example, a lot of work goes into mapping from Euclidean (rectangular) to polar and cylindrical coordinates. One finds that these mappings themselves are not volume-preserving, so one needs to account for a whole bunch of structure and subtlety that never happens in single varialbe (search 'Jacobian'). In calc 1, the only coordinate change one might run into is a trivial one like ##x' = x + a##, which really means only changing the limits of integration. But, move from rectangular to polar and a whole lot more changes...
If I had to describe the pieces, I would do it as such:
1) Calc 1 - diff and integral single variable
2) multivariate
3) complex calculus (this gets really, really weird)
4) measure-theoretic (Riemann-Stieltjes) where one now no longer deals with unit differentials (so you can integrate w.r.t. things like dx^2)