[warning: this turned into a long one. Enjoy!]

I think the problem is you're thinking of two chunks of dirt with rocks sticking out. Instead, think of the surfaces of the two sides more as sandpaper. It's already broken up and pulverized. The two sides have forces pushing them together which locks the fault, a 'clamping force'. Some areas have more clamping force than others. Those that have less force don't require as much energy to get past the friction.

To continue the sandpaper analogy, imagine (or do it for real) having a block with sandpaper on it in each hand. If you only press the blocks together gently, it's not that difficult to overcome the friction of the sandpaper and slide them past each other. However, if you push the blocks together real hard, it's going to take a lot more pressure to slide the blocks across each other.

In a real fault, you might say that different areas are being pressed together (clamped) more than others. Also, the grain or grit # of the sandpaper isn't constant along the plane of the fault. Some areas are "grittier" than others.

The clamping force can be affected by other quakes in nearby areas. There were some 5'ers near and before Loma Prieta, right? Imagine it like this. The section of the San Andreas where Loma Prieta occured was clamped hard. So a lot of force was building up there but it still couldn't break the friction. Now comes along those two 5'ers that occured before hand. Those quakes shifted the stress field around them. Turns out those 5'ers may have reduced the clamping pressure that was holding the Loma Prieta section together. Now there's not as much clamping force, but there's still all that built up energy from years of wanting to slide. As the clamping becomes less and less, a point is reached where the built up energy is now sufficient to overcome the friction along the fault and WHAM!! We now have a large quake.

Much of the reason quakes are so hard to predict is we can't get down in the fault itself and measure how much clamping force there is in various areas nor determine the friction on the fault surface that keeps it from sliding. We can determine some of the forces trying to slide the blocks past each other (long term general crustal motion), and we can compute the change is stress fields after a quake has occured (coulomb stress modeling).

To use the sandpaper block analogy again, we know how hard the blocks are being pushed to slide past each other, but we don't know what grit is being used, nor how hard the blocks are being squeezed together. If we did, we could calculate what amount of sliding pressure would be sufficient to break the friction holding them together. We could 'predict' when the blocks would slide.

I feel earthquakes could be predicted, or at least forecast in the near term instead of 30 years if we knew more about the exact structure of the fault planes and the coulomb stresses in the ground. This is a hard problem to solve, though. We can't "see" into the earth that well. We need to see structures on the order of meters or less all along the fault plane from top to bottom and from end to end.

We also need to know the stress fields. Unlike the fault plane, which is two dimensional, the stress fields are in the 3d structure of the crust. We'd need to map that all out as well in detail, perhaps in 1km cubes or smaller. Then this all needs to be dumped into a computer and the simulation run. That's the next problem.

Some systems are so complex that a simulation of the system would take longer than the real thing occurs in real time. That is, the computations for a simulated event lasting 10 minutes might take hours to compute. Since the earth is a dynamic machine, all the variables are changing dynamically over time. By the time you get done running the simulation, everything has changed.

These types of problems came to the forefront during the hurricanes last year. The meterologists took the latest data and stuffed it into their supercomputers so they could see what the hurricane was going to do in the next hour. Problem was, the simulation took over two hours to complete. By then, it's too late.

The problem is much worse in seismology. Not only do we not have enough computing power to "think" fast enough to predict what's going to happen before it actually does, but we don't even have all the data necessary to make a reliable simulation.

What worse, it is said that some dynamic systems maybe so complex that the only way to simulate them is to do the real thing. Basically, that no computer could be built that coudl run fast enough to simulate the real thing quicker than letting mother nature do the real thing. This is where chaos theory comes in. If it turns out that fault systems are that complex, we may never be able to predict what they will do.

But, there may be clues and side effects caused by the stresses in the earth, and these may be useable as indicators of an impending quake. But not all clues occur at all faults; and not all faults behave the same way. What may be an indicator for one area may not apply to others. The system is too dynamic and variable. The Parkfield Experiment bore that out. The place was wired with nearly every type of sensor to look for those little clues, and none showed up.
If we were to wire dozens of other faults the same way, some of those sensors may have shown something. But each fault would proabably show different clues.
Each fault or fault system would have to be treated differently.

And this is where I think most quake predictors screw up. They look at one thing and one thing only. That one thing may work on one fault, but it may not work on all faults. Some predictors do look at multiple clues, but I feel they may not be looking at enough of them. Certainly, I feel most are not interpreting the clues correctly to begin with. There's also the old "garbage in, garbage out" problem.

And finally, I really think people are not giving "real" scientists enough credit. There is a saying, "this *IS* rocket science". Some stuff really is complex and it really does take a lot of knowledge to understand. To dismiss science simply because it's difficult to understand is poor thinking. Not everything in nature has a simple formula describing it yet.

I'm reminded of the "cargo cult" story. The poor natives didn't understand the amazing technology they were suddenly exposed to and could only understand it from their limited frame of mind. To us, it looked silly. A bunch of "dumb" natives mimicking what they saw us "smart" folks doing, and they couldn't understand why it wasn't working. They didn't understand the science, and therefore they got it wrong.

So in a sense, that's why scientists tend to laugh at many of the ideas put forth in prediction (or any science) by non-scientists. The scientists are like the smarter military folks in the cargo cult story, with their airplanes and radios and loud weapons, and the lowly "nutty" quake predictor is the native, who only sees these 'magical' machines, but doesn't understand what they are or how they work. Once you understand the science, you'll understand where your ideas went wrong and why. You'll see the forest for the trees, so to speak.

Science is simply a tool to make sure that when we look at a forest, we see a bunch of trees and not be deluded into seeing a bunch of cardboard cutouts painted like trees.

Wow....this sure turned into a little monologue, didn't it? Sorry if I went overboard, but I was on a role. I hope this is more helpful than confusing, though, for everyone on the forum.

Brian