Beyond the solution to this particular diagram, can you explain if there is a general set of steps that you would follow to obtain this code?
My guesses: Would conversion to vector graphics give some guidelines?
Are there parts of the process that can be automated? Etc.
I tried: Using Inkscape to convert to vector graphics and somehow get a bunch of coordinates that I could use as a guideline for the tikz code.
I got the following so far (I took some code from this answer):
\documentclass{article}
\usepackage{pgfplots}
\usetikzlibrary{decorations.text}
\pgfmathdeclarefunction{gauss}{3}{%
\pgfmathparse{1/(#3*sqrt(2*pi))*exp(-((#1-#2)^2)/(2*#3^2))}%
}
\begin{document}
\begin{tikzpicture}
\draw[draw=none](0,4)--node[anchor=center, above] (A) {Marginal
distribution of Brownian motion through time} (10,4) ;
\draw[->,very thick](0,0)--(0,4) node[anchor=south, above left,
rotate=90] {Spacial distribution};
\draw[->,very thick](0,0)-- node[anchor=north, below] {Time} (10,0);
\draw[dotted, domain=0:9,samples=35,thick] plot ({\x+0.5},
{sqrt(\x)/2+2});
\draw[dotted, domain=0:9,samples=35, thick, smooth, postaction=
{decorate, decoration={raise=-10pt, text along path,text
align=center,text={Spread of Willow Tree}}}] plot ({\x+0.5},
{-sqrt(\x)/2+2});
\draw[very thin, domain=-0.5:2.5,samples=35] plot({gauss(\x, 1, 1)+1},
{\x+1}) node (B) {};
\draw[very thin, domain=-0.5:2.5,samples=35] plot({gauss(\x, 1, 0.5)+3},
{\x+1}) node (C) {};
\draw[very thin, domain=-0.5:2.5,samples=35] plot({gauss(\x, 1, 0.25)+5},
{\x+1}) node (D) {};
\draw[very thin, domain=-0.5:2.5,samples=35] plot({gauss(\x, 1, 0.2)+7},
{\x+1}) node (E) {};
\draw[->] (A) to (B);
\draw[->] (A) to (C);
\draw[->] (A) to (D);
\draw[->] (A) to (E);
\end{tikzpicture}
\end{document}
which produces:




pgfplotsmanual and cover image conversion, automation and unspecified further issues. Apart from anything else, that is much too broad. It cannot possibly be answered reasonably in this format. Also, what are you converting to vector and how? And why would that produce code? We don't know anything about the starting point so automation is essentially out of the question. – cfr Jul 29 '15 at 15:58MWE. I had assumed, given that you'd obviously been asking/answering questions here for a while, that you would know what I meant. However, it seems nobody bothered to mention it before. This is unfortunate, because it makes it much easier to ask questions, solve problems, and get answers! – cfr Jul 29 '15 at 20:52