The statistical analysis shows that the spectral amplitude (defined as the maximum spectral energy density) varies from 0.1 × 10 4 m 2 s −2 (cycle/m) −1 to 3.2 × 10 4 m 2 s −2 (cycle/m) −1 , generally smaller than the theoretical saturated spectral amplitude, with remarkable temporal and spatial variations. Due to the latitudinal and seasonal variations of the background winds and wave sources, the amplitude exhibits obvious quasi-biennial, annual and semi-annual cycles at low latitudes, and the annual cycle at middle and high latitudes.The spectral slope, ranging from −1.6 to −3.0, is systematically less negative than the canonical value of −3, and exhibits a quasi-biennial cycle at low latitudes, annual cycle at middle and high latitudes, and semi-annual cycle at boreal high latitudes. Different from the amplitude, the slope shows weaker temporal and latitudinal variations, strongly indicating a more universal feature of the slope, which mainly results from the wave propagation process (such as the wave-wave or wave-background flow interactions) rather than the wave source characteristics. Furthermore, we statistically study the longitudinal variation of the spectral parameters. The spectral amplitude at different latitudes shows an obvious and distinctive longitudinal distribution, which is attributed to the latitude-dependent jet and front mechanisms and/or orography. However, the slope has no evident longitudinal distribution except over around 120 • W, 67 • W and 85 • E. This may be accounted for by the fact that orography-excited gravity waves are easily influenced by the background winds due to their low frequencies. Interestingly, the amplitude and slope have negative correlation at most latitudes. Larger spectral amplitudes could mean that the gravity wave spectrum is more strongly saturated, resulting in a spectral slope closer to the theoretical value of −3.
KEYWORDS