72 ANNUAL REPORTS OF DEPARTMENT OF A(jRlCULTURE. 



iications are madt' of a. iii. lorcciists based on o').sei'vatioiis taken at 

 8 a. m., seventy-lifth-nieridian time. These forecasts are made 36 

 hours in advance. It has lon«; been recofjnized that the complete 

 vori Heat ion of ^veather forecasts, by Avliich is meant the findin<j of the 

 exact relation between the conditions forecast and those which actu- 

 ally occur, involves insurmountable difficulties and that approxi- 

 mately accurate results only are possible according to a series of 

 relatively arbitrary rules. As verifications are based on the results 

 of two observations a da}' (8 a. m. and 8 p. m., seventy-fifth-ineridian 

 time), account is taken onlj' of precipitation or its absence during 12- 

 hour periods. For example, the forecast " Eain to-night, fair 

 AVednesday," would be counted a failure in the sei-ond period if some 

 rain fell after the morning observation, although the weather was 

 fair for the rest of the day; it would receiA^e a credit of only 50 per 

 cent by the rules of verification, but the. public would undoubtedl}- 

 regard the forecast as a complete success. In the same manner, a 

 rain forecast for both periods would fail if the rain stopped just 

 before the first observation of the second period. 



While general rules for the verification of weather forecasts will 

 alw'ays fail and succeed, more or less, in individual cases of the kind 

 cited, it is nevertheless generally admitted that in such cases the 

 failures and successes balance each other in the long run, so that after 

 all the general averages under empirical rules fairly represent the 

 true results. 



RIVER AND FLOOD WARNINGS. 



Increased cost of materials and services re(|uired in the u^^keep of 

 river gages, including demands for increased compensation on the 

 part of some of the gage readers, has made it difficult to maintain the 

 full program of activities in the safeguarding of life and property 

 from destruction by flood. 



The problem was met in part by a revision of the observational 

 work whereby the season Avas shortened, wherever it could be safeh' 

 done, and the small amount so saved was applied in granting small 

 increases in compensation at the most important points. At the 

 close of the year the revision had been completed and the maximum 

 possible service Avas being rendered for the minimum expenditure of 

 funds. 



The flood-Avarning service has functioned efficiently throughout the 

 year, notwithstanding the large number of unusual floods in various 

 parts of the country. In October and November, 1919. serious flood- 

 ing occurred in the region from northeast Texas to the upper Ohio 

 Valley, and in December of that year the most severe flood in 3") vp-'- 

 occurred in the streams of southeastern Mississippi, Alabama, and 

 western Oeorgia. Notwithstanding the fact that full publicity was 

 giA'en to the coming of theso floods, there was a small loss of life and 

 a great loss of property, mostly unaA'oidable. 



The floods of 1919 haA-e emphasized the fact that matured crops 

 groAvn on lands subject to OA^erfloAv are not safe until removed to 

 places not sul)ject to oA^erfloAv. This Avas particularly noticeable in 

 the case of the 1919 corn crop in the loAver Ohio Valley and elsewhere, 

 much of which AA^as lost by flood. Scarcity of farm labor may be the 



