PROTECTION AGAINST FROST 371 



records of the particular station. A thermometer reading at this median time, 

 subtracted from the afternoon maximum, gives, presumably, half the total 

 fall in temperature to be expected. Thus if the maximum were 70°F., the 

 median temperature 50°, the difference, 20°, taken from the median temperature, 

 would indicate the expected minimum to be 30°. Under conditions obtaining 

 at some stations this method seems the most reliable that has been tried. 

 In general it seems to give closer approximations to actual temperatures in 

 regions of very low humidity, not, perhaps, because the method works better 

 there than elsewhere, but possibly because the other methods do not work so 

 well. As indicated by Hallinbeck, with certain precautions in its application 

 it seems to work well at Roswell, New Mexico. Wherever compared with 

 the older method of assuming identity between evening dewpoint and morning 

 minimum it has proved superior. 



Still more accurate predictions were found possible in Ohio by Smith, using 

 the equation y = a -\- bR, where R is the evening relative humidity, y the varia- 

 tion of the morning minimum temperature from the evening dewpoint, while a 

 and b are constants derived from previous data accumulated at points with like 

 conditions. This linear equation, when plotted, fitted the Ohio data very satis- 

 factorily, but charts from certain other points were fitted more closely by a 

 parabola whose equation was modified by Smith to read v = x + by -\- cz in 

 which X, y and z are coefficients to be determined from previous data, b the eve- 

 ning relative humidity, c the square of the relative humidity and v the variation 

 of the minimum temperature of the following morning from the evening dew- 

 point. The value found for v is added to or subtracted from the evening dew- 

 point and the minimum temperature indicated. 



The method of obtaining the constants is explained in detail in Supplement 

 16 of the Monthly Weather Review. As has been suggested above, the constants 

 vary with the locality. As samples, the following may be cited: for the y = a 

 + bR equation, at Lansing, Mich., a = —11.2, b = 0.727, at Grand Junction, 

 Col., a = —7.01, b = 0.53; for the v = x -\- cz -\- by equation, Modena, Utah 

 (all nights, radiation or otherwise), x = 7.3, y = 0.18, z = 0.0057; for Montrose, 

 Col, X = -22.0, y = 0.383, z = 0.01167. 



The first equation has been found to give satisfactory results at some places, 

 the second has proved preferable at others; as stated above, the median tempera- 

 ture method seems best here and there, while in some cases still other me'thods 

 are used. Sometimes a mean between results secured by two methods has 

 proved more nearly accurate than either singly. One disadvantage of the 

 median temperature method as compared with the others outlined here lies in 

 the fact that the forecast cannot be made until several hours later than is possible 

 from the metheds based on hygrometric data. The fact that different methods 

 fit various places is probably an expression of the differences in topography and 

 in humidity, relation to centers of high pressure and other factors somewhat 

 pecuhar to particular locaUties but all combining in frost production. 



It should be borne in mind also that the methods outlined fit only 

 radiation nights and that occasionally fruit blossoms are damaged by 

 cold in other ways such as high cold winds or cold snow squalls. To 

 forecast these, reliance must be placed in the weather map. The problem 



