Historical photographs are windows into the past. By comparing historical and modern photographs we can measure change; but how do we find the original camera location to repeat a photograph? Historical, geographic, topographic, and other clues may be used, but we lack a numerical method suitable for any user to find the general location of a historical photograph. We derive a geometric method that can be applied in the field or prior. The method uses measurements of at least three points of reference (POR) in the historical photograph and corresponding geographic locations measured via a compass or on a map. rePhoto, an open-source R package that applies the method and outputs spatial KML files for use with Google Earth, is provided. The geometric method was tested on 20 photographs with known locations and validated by independent users. The effectiveness of the method varied among users, but overall predicted a search area containing the original camera location 70% of the time (the prediction accuracy) and typically predicted search areas 99.5% smaller than the total region evaluated by the method. The method was robust regardless of whether three or four POR were used, and worked well even when POR were more than 30 km away. The proposed method only works for photographs with at least three identifiable geographic POR, thus other methods are illustrated for use more generally. California and the western USA have numerous iconic historical photographs, for which many locations could be found and re-photographed using methods described here.