CHRONOLOGirAL VARIATIONS OF RAIN AND WIND. 



185 



the 80-«;allecl probable error or departure for auy one year, or the index of variability 

 of annual precipitation. The results are given in the accompanying Table VIII, and 

 are interpreted in the following paragraph : 



Table VIII. 



S(;iUoiis. 



Num- 

 ber of 

 years. 



Le.iveiiwortli 



Spicelaiid 



AYiisliirigtou . 



46 

 22 

 42 



Average 

 total an- 

 nual precip 

 itation. 



Ti)c]ies. 

 32. 48 

 39.40 

 39.48 



Probable error 

 of annual precipi- 

 tation. 



Inches. 

 6.02 

 5.47 

 5.20 



Percent. 

 18 

 14 

 13 



Probable error of 

 mean of 49 years. 



Inches. 

 0.86 

 0.78 

 0.74 



Per cent. 

 2.6 

 2.0 

 1.9 



The mean annual catch at the Fort Leavenworth gauge is 32.48 inches, as given by 

 46 years of observations, which, however, differ among themselves from year to year 

 so much that it is an even chance that the catch of any one year will differ from this 

 mean by more or less than 18 per cent of its value or by 6.02 inches ; this 18 per cent 

 is in part due to actual irregularities in rainfall and in part to the variable effect of 

 the wind and the irregular proportions of snow and rain ; the actual rainfall is larger 

 than this catch by an unknown amount depending on the character of the precipi- 

 tation and the strength of the wind at mouth of the gauge. 



It is therefore evident that any conclusion as to a change of climate during these 

 years involving quantities less than the probable errors of the mean rainfall must be 

 entirely illusory. 



EECOM.MENDATIONS. 



Our study of the rain gauge .(section 6) and its errors would have a melancholy 

 conclusion did it not afford us some suggestion as to the proper methods of determin- 

 ing and allowing for these errors. In view of our present knowledge we now see 

 that in establishing new stations better methods of exposure should be adopted and 

 such as are in fact very different from those that have hitherto been considered 

 allowable. We must closely Imitate the conditions prevailing at the average surface 

 of the ground, that is to say, in the order of preference the exposure would be: (1) 

 the pit gauge ; (2) the protected or the shielded gauge near the ground ; (3) several 

 protected or shielded gauges distributed over a flat roof; (4) the shielded gauge on 

 posts considerably elevated above slanting roofs. Moreover, in no case should a 

 single gauge be relied upon, but in all cases at least two similar gauges at very dif- 

 ferent heights should be observed. From the records of these two gauges we can 

 compute the catch of the normal pit gauge by the formula previously given. 



As this formula is also applicable to the ordinary and in fact any form of gauge, we 

 furthermore see that an approximate "correction, needed to reduce valuable past rec- 

 ords to the normal gauge, may now be determined, if these old gauges are still being 

 recorded, by at once estaVdishing near them two or more similar gauges at consid- 

 eraldy different heights; from the records of all these gauges for the next few years 

 wc may determine, at least approximately, a correction applicable to the past years 

 of historical records. Finally, we are warned against attempting to draw from past 

 records conclusions that are finer than the data will justify. 



APPLICATION TO FORESTRY. 



To this presentation of the rain-gauge question I will add that although the ideas 

 here given may not be altogether new to those especially interested in rainfall, yet 

 their application to the special precaution of establishing rain gauges in pairs at two 

 different altitudes has only been carried out by Ebermayer in Bavaria, and more 

 recently by Brandis and Blauford in the forests of the central provinces of India, but 



