384 Ferret — Measures of Intensity of Solar Radiation. 



with the correction given by the preceding formula the lack of 

 symmetry between the forenoon and afternoon branch of the 

 curve is greater than that in such curves generally from other 

 means of measuring the intensities, though these are generally, 

 though not always, less in the afternoon than in the forenoon. 



It is perhaps doubtful whether Crova's curves, rightly inter- 

 preted, would really give as much depression in the afternoon 

 as is inferred. From 10 o'clock on, through most of the after- 

 noon, the oscillations of the curve about the medium line are 

 very great, and such a line seems to have a great depression, 

 and even a second maximum ; but as the lowest points of the 

 curve represent only the exceptionable states of the atmosphere, 

 which are perhaps of comparatively short duration, the curve 

 representing the average intensities would have to be drawn 

 higher up, nearer to the upper extremes of the oscillations. 

 These oscillations, also indicated by the bolometer, are undoubt- 

 edly caused by what may be called invisible clouds, the atmo- 

 sphere during the warmer part of the day being in an unstable 

 state, in which local ascending currents are formed and the 

 vapor contained in them, in cooling, does not quite arrive at 

 the state of saturation and condensation ; or if so, the conden- 

 sation is not quite sufficient to cause a visible cloud at great 

 altitudes, as is frequently the case when numerous small patches 

 of cumulus cloud are formed. 



It is evident from an inspection of (11), that the observed 

 values of I in summer and in winter, or in low latitudes and 

 high latitudes, or low and high altitudes, or in any circum- 

 stances in which there is a difference of temperature, are not 

 comparable without a reduction to the same temperature. For 

 instance, using (13) instead of the first expression, though not 

 very accurate for large values z/ = r— r , and supposing A to be. 

 equal to 20°, we get I o = l'305 I, so that the observed value at 

 the temperature r, must be increased nearly one- third to make 

 it comparable with one observed at the temperature r , accord- 

 ing to the formula above. But it is not to be understood here 

 that this formula, with the numerical constants used, is claimed 

 to be a correct reduction, but simply one which gives roughly 

 the order of reduction which may be required, and it is given 

 rather as an illustration merely, than as a correct formula of 

 reduction, even in the case in which iron and copper are used 

 in the thermoelectric circuit. For other metals, of course, the 

 formula would be different. Some allowance, also, must be 

 made for the effects of neglected conduction and convection. 



In the bolometer the deflection of the galvanometer needle 

 is proportional to the differences of temperature of the bolome- 

 ter strips, and so the h'rst two factors into which I in (11) is 

 multiplied disappear. But there are the same relations between 



