If the bottom chord in that arched truss is stretched, the top must contract under compression. It will not bow up, it will bow down.
Look at any king post, real or imaginary, in one of the trangular elements near the center. When the arch is stretched it will staighten out. The king post in the center will pull the center of the top chord down.
Now look at the king post at an end. Stretch the bottom chord and the king post will rise up, moving the vertex of the triangle up.
The thought experiment can be done by compressing the bottom chord. The top chord will bow upwards in that case.
Shapes can always be approximated by decomposing them in to triangles. In the arched truss, they just get smaller, then larger along the long axis of the truss. Finite element analysis often uses triangles, as "finite" elements, instead of points to model complex shapes and situations.
Holy cow! Check this out:
http://www.startribune.com/10204/story/1343624.html
“At the site, Hoeppner talked to construction workers who survived the fall. They had been doing repair work but expressed concern to him that the bridge had been wobbling several days before it collapsed. Every layer of concrete the workers removed, the bridge would wobble even more, they told Hoeppner.”