GAKDNER.] GEOGRAPHY— ELEVATION OF DATUM-POINTS. 631 



If the lines were all railroad-levels, the following points were consid- 

 ered as favorable to the character of any line or connection of lines : 



First. That the official reports should be recent and detailed. 



Second. That they should be reports pf the final computation of the 

 construction-levels, or a releveling of the completed line, or, best of all, 

 that we should have both of these reports agreeing closely. 



Third. That there should be few connections of lines formerly inde- 

 pendent to make up the present through-line. 



Fourth. That, where the line was made up by joining several parts 

 formerly independent, the connection between them should not be open 

 to any doubt. 



If, of several lines between two points, one disagreed largely from the 

 others of apparent equal weight, it was considered as probably in error. 



If several railroads, running from a common point, cross an important 

 river, the fall of the stream was determined by the very best lines, and 

 those were rejected which made it run up hill or gave an improbable 

 fall. 



If several parallel railroads were cut by a cross-line, well connected, 

 their agreement upon this common line was considered as an important 

 test. 



If several lines of levels between two points start from a common 

 datum or directrix, and end also at a common datum, the connection for 

 com]Darison is far more reliable than when the ends of the lines merely 

 came into the same city, and then have to be joined by connecting the 

 depots by city -levels. 



The results of the application of these standards of accuracy showed 

 that recent official reports of the final computation of the construction- 

 levels were generally reliable so far as any one line is concerned. The 

 Pa. E. E. may be given as one of the best examples. It has been 

 recently releveled, and though there was a discrepancy between the 

 new and old elevations of Pittsburgh of 11 feet, it was found on a third 

 leveling of a part of the line that this was due to erroneous connection 

 of two leveling-parties, and all occurred at one point. And now, when 

 a final computation of the old construction-levels, and of the new and 

 corrected line, is made, the elevation of Pittsburgh by the new line is 

 within a foot of the old. Great credit is due to Mr. Wilson, the consult- 

 ing engineer of this road, for the interest he has taken in investigating 

 the discrepancies of the Pa. E. E. profiles, and through his exertions 

 we have at last a correct report of the profile of that important railroad 

 so many years after the levels were run. The profiles that seemed from 

 their dates to be first calculations of constructed lines were often found 

 unreliable, and do not generally agree with the final calculation of the 

 levels when we have reports of both. The profiles of preliminary lines 

 of survey were, of course, found very unreliable. The elevations of Cai- 

 ro, 111., and Columbus, Ky., havehitherto rested on a preliminary survey 

 of the M. & O. E. E. from Mobile Bay, and seem to be 10 to 15 feet 

 below the better determinations. 



The worst of all the profiles, and the most jierplexing to the geographer, 

 are those made up in the offices of some of the railroads by putting 

 together data from old and new printed reports and from all the manu- 

 script profiles in the office, and treating them as if the same datum 

 of levels was referred to in all these sources of information. The mix- 

 ture thus produced generally defies the most ingenious power of anal- 

 ysis in the searcher after truth. 



In general, I am satisfied that the important errors in our railroad and 



