I suggest this aproach. Let fs(x,y) be your simplex noise function. Let's introduce a second function : f(x,y) = WIDTH * (1 - (X * X + Y * Y)) or any function that rassemble the following and that gives values from 0 to 1:

at this point take your simplex noise fs(x,y) :

and multiply per f(x,y) you get something like this :

finaly apply a colorgradient to obtain something like :

I suggest this aproach. Let fs(x,y) be your simplex noise function. Let's introduce a second function : f(x,y) = ((float)Math.Sin(((float)x/(float)WIDTH) * Math.PI) ) * ((float)Math.Sin(((float)y / (float)WIDTH) * Math.PI) ) or any function that rassemble the following and that gives values from 0 to 1:

at this point take your simplex noise fs(x,y) :

and multiply per f(x,y) you get something like this :

finaly apply a colorgradient to obtain something like :