If you read a standard reference on the interaction of gamma rays with the Earth's atmosphere -- Weekes is one example, as is this paper by Diehl (

http://www2011.mpe.mpg.de/~rod/gamma-ray_processes.ps) -- you'll see the term "optical depth." Suppose that the optical depth of a medium is represented by the letter 't'. Then as radiation passes through that medium, the amount which penetrates the medium is given by

(amount penetrating to optical depth t) / (original amount) = exp(-t)

I've written "exp(-t)" to represent the negative exponential function; it is often written as "e" with a superscript to denote the power.

Suppose that a cloud of gas has an optical depth of t = 3 to some incoming radiation. That means that the fraction of the radiation which goes through the cloud is just exp(-3) = 0.0498, or about 5 percent.

If the cloud has optical depth t = 10, then the fraction passing through the cloud will be exp(-10) = 0.00004.

The optical depth of Earth's atmosphere to gamma rays depends on the energy of the gamma rays, but Diehl gives a rough estimate of about t ~ 100. That implies that the fraction of gamma rays from space which reach the ground is exp(-100) = 10^(-44). That's ... really small. That implies that one would need to release 10^(44) gamma rays on one side of the atmosphere for a single gamma ray to reach the other side.

(Digression: a megaton of gamma rays is only about 10^(28) photons, so this quick calculation seems at first to suggest that satellites in orbit should not be able to detect nuclear bomb blasts near the Earth's surface. Since satellites _do_ detect nuclear blasts, I must assume that the initial flash of gamma rays is able to modify the atmosphere in such a way as to render it somewhat transparent to gamma rays emitted a fraction of a second later. Interesting)

The bottom line is that only a teeny, tiny fraction of the gamma rays from a celestial object will pass through the atmosphere and reach the ground. Such a tiny fraction that no known and detected sources produce enough to be measured from the ground directly. Sure, if a supernova were to explode very close to the Sun, we might detect it ... but not GRBs from the far reaches of the universe.