For a function of two variables (surface) that can be written as $z=f(x, y;\boldsymbol{\theta})$ where x, y & z are scalars and
$\boldsymbol{\theta})$ is a vector of parameters (i.e. $x,y,z \in \mathbb{R}$ and $\boldsymbol{\theta} \in \mathbb{R}^n$)
this returns the function $g : \mathbb{R} \to \mathbb{R}^n; x,y \mapsto g(x,y)$, which is the function's (curves') sensitivity
to its parameters, i.e. $g(x,y) = \frac{\partial f(x,y;\boldsymbol{\theta})}{\partial \boldsymbol{\theta}}$
The default calculation is performed using finite difference (via
ScalarFieldFirstOrderDifferentiator) but
it is expected that this will be overridden by concrete subclasses.