On the basis of an analysis of the microphysical processes involved in pressureless sintering of particulate materials and an examination of the physical conditions likely to prevail on Triton, it is argued that a well-annealed transport nitrogen layer can form on Triton at 37 K on a seasonal time scale (~100 Earth years), provided the initial grain diameter is less than 1 &mgr;m. Slightly larger grains (in the 1--10 &mgr;m range) are allowed if grain growth does not hinder densification, but grains larger than 100 &mgr;m will certainly not achieve seasonal densification. If densification is to occur simultaneously with deposition, an initial grain diameter of <0.3 &mgr;m is required. The proposed layer is polycrystalline, the final grain diameters being 10--30 &mgr;m. In the north this layer is currently forming, whereas in the south it is the remnant of the layer formed during the last winter. The following observations can be explained by the presence of this layer: (1) the 2.16-&mgr;m absorption feature, provided that in analogy to the visible absorptivity of water ice, grain boundaries do not reduce the absorptivity of flawless nitrogen ice; (2) the smoothness of the surface at the <100 &mgr;m scale, indicated by the absence of the opposition effect and a low thermal emissivity inferred from the analysis of Voyager ultraviolet, infrared, and imaging data; (3) the dark and bright appearance of the northern and southern hemisphere, respectively, despite the fact that frost should be currently condensing in the north and subliming in the south; if nitrogen forms a transport layer upon condensation, Voyager may have seen a dark substrate in the north and a permanent nitrogen cap in the south; and (4) the plumes; a well-annealed nitrogen layer would provide the ''seal'' required by the solid-state greenhouse model.

In addition to explaining these observations, two conclusions can be reached from the analysis presented in this paper. First, if the equatorial collar is the site of a recent snowfall, its fresh, bright and blue apperance can be understood if the characteristic size of the snowflakes is ~100 &mgr;m, as calculated in a recent model. For such large grains the annealing time scale exceeds 104 Earth years. A recent appearance of this collar on a dark frost-free surface would help to explain the decrease in the degree of reddening observed between 1977 and 1989. Second, the dust devils are unlikely to work on Triton if micron-sized nitrogen grains are intermixed with the dust: the rapid sintering of such grains would provide the cohesive forces whose absence is essential in the model. An observational test of the ''annealed-layer'' hypothesis to be performed during the current decade is proposed.