In this paper we present a systematic study of the Mars topography focusing on the statistical distributions and maps of the 5°Â 5°cell-averaged mean elevations, standard deviations, skewness and kurtosis coefficients, and power spectra. Altogether, the obtained data suggest that at a 5°Â 5°cell scale a large portion of the Martian surface may be reasonably considered as a Gaussian random field with a three-range spectrum consisting: (1) a high-energy low-wavenumber range (% 0:003 < k < % 0:03 km À1 ) where the spectrum may deviate from a power law and attain a maximum;
(2) scaling range 1 (% 0:03 < k < % ð0:2-0:3Þ km À1 ) where the spectrum may be well approximated as SðkÞ / k Àb 1 ; and
(3) scaling range 2 (% ð0:2-0:3Þ < k < % 1:67 km À1 ) where the spectrum may be also approximated as a power function but with a different exponent, i.e., SðkÞ / k Àb 2 . The most probable values for the exponents are b 1 ¼ ð2:2-2:4Þ and b 2 ¼ 3:8. The data show that the separation of these two scaling ranges most frequently occurs at L c % 3:3 km. At a scale larger than the 5°Â 5°cell scale the topography is highly intermittent with patchy spatial distributions of the key statistical moments. This patchiness is superimposed with systematic north-to-south trends in statistical properties, reflecting the crustal dichotomy of the planet and large-scale differences in the surface-forming processes.