C&GS machine was geared for 37 frequencies between one cycle 

 per year to eight cycles per day. Although most periods 

 were incommensurable, the gearing was done so precisely 

 that no constituent was off more than 2 at the end of a 

 year's predictions. The left side of the machine summed 

 the cosine curves. The right side summed the first deriva- 

 tives of the cosine curves; therefore the amplitude scales 

 on the right were weighted for the constituent speeds. When 

 the derivative side showed a zero sum, tb(- operator recorded 

 the times and heights of the predicted high or low tide for 

 use in the tide tables. When the C&GS bought its first 

 electronic computer, an IBM 65O, an attempt to program tide 

 predictions on the 65O disclosed that the mechanical, hand- 

 cranked machine, almost a half century old, could turn out 

 the predictions faster. The comparability vanished quickly 

 as faster electronic computers were developed and tide pre- 

 dictions are now prepared routinely on the latter. The 

 printed output is reproduced directly for publication. 



Other countries use more constituents, particularly 

 for shallow water areas where important non-linear combina- 

 tions introduce the need for additional tidal constituents. 

 An extreme case of the latter came up in trying to improve 

 tide predictions for Anchorage, Alaska. The range of tide 

 is very large, about 25 feet, and the discovery of oil in 

 Cook Inlet brought in deep-draft oil tankers. These re- 

 quired more accurate tide predictions; Figure 4 shows the 



26-9 



