136 E.R. C. REYNOLDS AND L. LEYTON 



Standard errors were very high (Table 4) but, since the amount collected 

 formed only a small part of the total throughfall (Table 2, column 4), its 

 contribution to the variability of the latter was negligible. If stemflow is 

 investigated as a separate entity, because of its local significance, the 

 regression techniques adopted by Wicht (1941), and by Wilm and 

 Niederhof (1941), offer means of studying the causes of variation, which 

 Voigt (i960) has examined experimentally. 



INCIDENT PRECIPITATION 



Commonly, throughfall estimates are deducted from the precipitation 

 incident upon the canopy to determine interception, that is, the amount or 

 proportionof the precipitation which fails to reach the ground. Measure- 

 ment of incident precipitation, however, is beset with many complica- 

 tions. Usually, it has been measured in adjacent open ground, but this is 

 not always available and the occasional attempts which have been made, 

 using gauges mounted above the trees, have proved generally unsatisfac- 

 tory. In neither case is it sufficient to append statistics of variabihty to the 

 mean catch of such gauges; evidence must be presented to support the 

 contention that the measurement is not subject to errors due to difference in 

 location or, in the case of 'tree-top' gauges, to instrumental errors. Instru- 

 mental errors associated with rain gauges above trees are considered to be 

 particularly great in areas where wind speeds are appreciable. Under his 

 relatively exposed conditions. Law (1959, impubhshed) suggests the addi- 

 tion of 15% to the catch of gauges elevated 10 ft above the tree canopy; 

 this correction was based on the relationship of catches of gauges at different 

 heights over short grass. 



For critical investigations into the water balance, it is essential to have 

 estimates of incident precipitation and of its spatial variabihty over the 

 plantation. 



In the present work, therefore, sixteen gauges of various types were 

 installed above the tree crowns (cf Plate I) ; these gauges were at random 

 locations. The water was conducted down P.V.C. tubing to ground 

 level and collected in enclosed bottles. Six unshielded 5-in. gauges were 

 placed in the canopy itself, as low as possible to obtain shelter from the 

 wind, yet not so low as to be affected by interception or splash. All other 

 gauges were 38 ft above the ground, 10 ft above the mean tree height at the 

 beginning of i960. Eight of these gauges were 5-in. gauges, two being 

 surrounded by 60° Nipher sliields with a horizontal wire mesh border, and 

 six by Alter shields such as are used with the International Standard rain 



