[Illustration: Fig. 38.—Arch-shaped auroræ borealis, observed by Nordenskiöld during the wintering of the _Vega_ in Bering Strait, 1879] Masses of light at so low a level that the rocks behind them are obscured have frequently been observed to float in the air, especially in the arctic districts. Thus Lemström saw an aurora on the island of Spitzbergen in front of a wall of rock only 300 m. (1000 ft.) in height. In northern Finland he observed the auroral line in the light of the air in front of a black cloth only a few metres distant. Adam Paulsen counts these phenomena also as polar lights of the first class, and he regards them as phosphorescent clouds which have been carried down by convection currents to an unusually low level of our atmosphere. Polar lights of the second class are distinguished by the characteristic auroral rays or streamers. Sometimes these streamers are quite separated from one another (see Fig. 39); as a rule they melt into one another, especially below, so as to form draperies which are so easily moved and unsteady that they appear to flutter in the wind (Fig. 41.) The streamers run very approximately in the direction of the inclination (magnetic dip) needle, and when they are fully developed around the celestial dome their point of convergence is distinctly discernible in the so-called corona (Fig. 40). When the light is at its greatest intensity the aurora is traversed by numerous waves of light. [Illustration: Fig. 39.—Aurora borealis, with radial streamers] The draperies are very thin. Paulsen watched them sometimes drifting over his head in Greenland. The draperies then appeared foreshortened, in the shape of striæ or ribbons of light in convolutions. These polar lights influence the magnetic needle. When they pass the zenith their influence changes sign, so that the deviation of the magnetic needle changes from east to west when the ribbon is moving from north to south. Paulsen therefore concluded that negative electricity (kathode rays) was moving downward in these rays. These polar lights correspond to violent displacements of negative electricity, while polar lights of the first class appear to consist of a phosphorescent matter which is not in strong agitation. The streamers may penetrate down into rather low atmospheric strata, at least in districts which are near the maximum belt of the northern lights. Thus Parry observed at Port Bowen an auroral streamer in front of a cliff only 214 m. (700 ft.) in height. [Illustration: Fig. 40.—Aurora with corona, observed by Gyllenskiöld on Spitzbergen, 1883]