it from the spectroscope, and proves a disappointment to the observer who depends entirely upon the rainband. The small amount of the dry-air band at a must also be considered, and an increase of the a with less distinctness of the B. All these features make up the special spectrum noticeable more or less in advance of thunderstorms.

I well recollect one instance in which all these features were so very strongly marked at the time of observing, about 10 a.m., that I confidently prophesied a storm would take place, although none of the other instruments, nor any particular aspect of the sky, nor high temperature with oppressiveness, indicated that such an aërial disturbance was about to occur; yet certain enough, in the evening, about seven o'clock, the storm did come, much to the astonishment of a great many, including the friends with whom I was staying.

The prismatic colours, more especially the

blue and yellow, are useful guides as to the presence or absence of ozone in the air. If the air is very pure, the blue will be clear, and the F line will stand out distinctly; the yellow will show itself a true yellow, or have an orange tinge. On the other hand, when the reverse is the case, the yellow will have a greenish hue, and the blue will be deep and dark, with an indistinctness of the hydrogen line F. In most cases this peculiarity in the colours will be found to precede the atmospheric conditions.

The

I should recommend the observer to make it his first care, when using the spectroscope, to take note of any alteration in the true colours, as they will be found to change after looking for a short time through the instrument. next thing is the amount of rainband (on sky free of cloud, if possible), at about 13 degrees all around, and then at the zenith. After this, begin at the blue end and work towards the red, carefully scanning the appear

ance of the constant lines, moist and dry air bands; notice if they are clear, indistinct, hazy, well-defined, or prominent, and finish up by noting the depth of shade at the red endfor a change of weather in certain conditions of the atmosphere is frequently heralded by its becoming clearer.

For the satisfactory prediction of weather it is highly necessary to keep a record of observations on a chart and in a note-book. On the former ought to be dotted down the percentage of rainband for eight points of the compass and zenith; also that in the direction of the wind, or towards the quarter it is most likely to blow from in the course of the next twelve hours. The reason of the last is obvious, for we necessarily must expect the weather to come from that direction. In the note-book enter the peculiarities of the spectrum already noticed, and finally calculate, from what you have observed, the probabilities of rain or fine.

After a few months' practice, it is astonishing with what rapidity one can conduct a systematic observation. I use four spectroscopes of different dispersion in my atmospheric studies. A large two-prism one, with photo

graphed micrometer (vide Fig. 5), arranged on a wooden alt-azimuth stand, with mirror for viewing the sun at different altitudes. It has three powers: the highest enlarging the rainband enormously, and enabling you to just discern two or three lines between the

two D lines. A micro-spectroscope of small dispersion (Fig. 6). The first form of rainband spectroscope (Fig. 1), which, being accustomed to, I feel reluctant to disqualify

it for Grace's new and better form of hygrospectroscope (Fig. 3). To the last I have adapted a stand (originally intended to hold a condenser), to steady it and meet the require