It is very, very unlikely any curvature could be distinguished in that picture.
For the sake of discussion, let's say that the horizon was 20 miles wide in that picture. I've seen a chart on this site showing the drop in feet at varying distances (as listed in ENaG) and at 20 miles there would be 266 feet of drop. Also consider, that the center of the horizon would be a high spot, so the sides would only drop half that distance from the peak in the center (ten miles in each direction from the center). Based on the number of feet in a mile the drop from the curvature would be approximately 400 times less than the width. Since the picture is 800 pixels wide, and the drop would only be half in each direction, it would mean that the horizon at the sides of the picture would be 1 pixel lower than the center.
Considering the resolving power of the cameral lens, imperfections in focus, artifacts from jpeg processing, the fact that the building blocks a large part of the horizon, small differences in elevation of the terrain, foliage, buildings, etc., etc. it is virtually impossible that any curvature would be visibly detected in that picture. Add to this the issue of any curvature introduced by lens distortion and any detected curvature is basically irrelevant in that picture
That 20 miles is just a random number for discussion. If the horizon was 50 miles wide, the drop would be a larger portion of width, but it would still be something like 2-2.5 pixels of drop, still difficult to easily detect.