442 REPORTS ON THE STATE OP SCIENCE. — 1919. 



velocity, and thus sites which would, ceteris paribus, be tolerable in a sheltered 

 valley may be extremely unsuitable near the sea or in a high, windy situation. 

 At inland and not unduly windy stations a gauge may often be placed on 

 level ground with no artificial shelter from wind at all, but with increasing 

 wind velocity increasing shelter, especially on the side of the prevailing wind, 

 is imperative. The diminution in catch observed in gauges placed at more 

 than one foot above the ground becomes increasingly greater as the position is 

 more exposed. Should the gauge used be of non-standard pattern, with shallow 

 funnel, all defects due to wind-eddies are aggravated, and in extreme cases 

 the record becomes quite useless. 



Except in very wind-swept localities the loss of catch due to wind-eddies 

 is almost entirely confined to the winter months. In our paper on ' Isomeric 

 Rainfall Maps of the British Isles' [Q.J.B. Met. Soc, vol. xli., 1915, pp. 1-25), 

 Dr. H. R. Mill and I established the fact that on the average of a number 

 of years the percentage of the annual total rainfall occurring in any month 

 does not vary appreciably at adjacent stations, even though these are at 

 greatly different altitudes {e.g., Fort William and Ben Nevis Observatory), 

 depending upon geographical and not upon orographical factors. It follows 

 that the percentage of the summer rainfall to that of the winter in the same 

 neighbourhood, on the average of several years, is a constant. Frequent 

 examinations of the records from unsuitably exposed rain gauges have shown 

 that the percentage of the summer rainfall observed during the winter is always 

 smaller than that found at properly exposed stations. When computing 

 annual average values it is often advisable to ignore the records for the winter 

 months and to substitute for them values computed from the summer rainfall 

 by applying a ratio derived from neighbouring sound records. By this means 

 one is able to arrive at a reasonably accurate annual average for stations with 

 defective exposure.'!. This method has been utilized extensively in constructing 

 annual average rainfall maps, and has been found to yield satisfactory results. 



Over-exposure of rain gauges is probably the most fruitful source of error 

 in rainfall observing, and far too little attention has hitherto been paid to it 

 in selecting sites for rainfall stations. It is extremely difiicult to lay down 

 any simple instruction which will entirely meet the case. A system of inspection 

 by officials thoroughly conversant with the varying requirements of each 

 locality would do much to remedy the defect, but some time must elapse before 

 any such scheme could be put into effective operation. 



The opposite pole of danger in respect of faulty gauge exposure — viz., over- 

 shelter — is much easier to avoid. Whilst it is true that a degree of shelter 

 which would be harmful in one case would be much less so in another, 

 broadly speaking, the conditions are similar everywhere. It is usuaily safe to 

 suggest that the top of any object, such as a wall or other building, should 

 never subtend an angle greater than 45° with the gauge. In windy positions, 

 when the rain commonly falls at an acute angle, 30° is preferable to 45°. In 

 the case of growing plants, shrubs, or trees, the angle should in no case be 

 greater than 30°, that is to say, the distance of any such object should be 

 at least twice its height. This allows for growth, which is apt to be overlooked 

 as it takes place. It should be noted that whilst the error introduced by undue 

 shelter by a wall or building is always caused by the interception of part of 

 the rain, that caused by trees or shrubs may be either positive or negative, 

 mterception occuiTJng under certain conditions, whilst at other times water-drops 

 hanging on leaves may be blown into the gauge, or drip from overhanging 

 branches. I have come across at least one instance in which a gauge was 

 placed by a careless observer actually under trees, with the result that the 

 positive and negative errors practically balanced, and, until the gauge was 

 inspected, no fault in exposure was suspected. I do not recommend this method 

 of obtaining accurate records ! 



One of the greatest difficulties in securmg rainfall records free from the 

 defects which I have described — and this applies with most force to errors 

 due to over-exposure — lies in the fact that in any individual reading the amount 

 of the error is usually smaller than the difference from the reading at a neigh- 

 bouring station which may arise naturally. A systematic error becomes more 

 apparent when the totals for a considerable period are compared, but, even 



