Mars Orbiter Laser Altimeter (MOLA) data provide compelling evidence that the martian lowlands, below the smooth and sparsely cratered northern plains, are extremely old, far older than the plains which cover them. The evidence is in the form of a very large population of “Quasi-Circular Depressions” (QCDs), many of which are very evident in the MOLA elevation data but generally not visible in available imagery. We interpret these “invisible” QCDs to be buried impact basins. Cumulative number versus diameter curves for lowland QCDs suggests the buried lowland surface is older than the visible highland surface and that the lowland plains are a relatively thin (1–2 km) veneer overlying this much older surface. We conclude that the martian lowlands have been low and stable for nearly all of martian history.
 The fundamental unsolved problem in martian crustal evolution is also the oldest: origin of the basic two-fold division of ancient, high-standing, mostly southern cratered terrain and the contrasting, smooth and sparsely cratered younger lowland terrain in the north. The northern lowlands of Mars are covered with plains units that are mostly Hesperian age and younger [Scott and Tanaka, 1986; Greeley and Guest, 1987; Tanaka and Scott, 1987]. Little is known about what underlies these plains, and explanations for the origin of the lowlands allow both an older surface as well as the possibility that the entire crust is similar in age to the younger surface units. Here we show the northern plains are a thin veneer overlying a surface that is much older, perhaps as old as the highland surface to the south.
 Mars Orbiter Laser Altimeter (MOLA) data [Smith et al., 1999] from the Mars Global Surveyor mission [Albee et al., 1998] have confirmed the suggestion by McGill  that a very large and ancient impact basin lies in the Utopia region of eastern Mars. This basin is largely responsible for the low elevation of this part of Mars. Large impacts have been suggested as one way to form the northern lowlands early in martian history [Frey and Schultz, 1988, 1990]. Internal causes, including plate tectonics, have also been proposed [Wise et al., 1979; McGill and Dimitriou, 1990; Sleep, 1994, 2000; Zhong and Zuber, 2001]. Many internal mechanisms have in common a relatively long duration (several hundred million years) and in general suggest the lowland crust is relatively young throughout its thickness like the plains themselves.
2. Quasi-Circular Depressions
 High precision gridded topographic data from MOLA have revealed the presence a very large number of roughly circular basins in both the martian highlands and lowlands [Frey et al., 1999, 2000, 2001]. We call these Quasi-Circular Depressions (QCDs). For many of these there is little or no evidence of a corresponding structural feature seen in image data. Profile data from the early portion of the Mars global Surveyor mission revealed the first of these, a 450 km wide, 2 km deep basin in the Arabia portion of Mars with no observable rim structure [Frey et al., 1999]. We suggested this was a buried impact basin, and suggested that similar features must be present elsewhere on Mars that might be detectable as the mission continued to collect data. We developed interactive computer graphics tools [Roark et al., 2000; Roark and Frey, 2001] that shift and stretch the color representation of the MOLA gridded data to systematically search for such candidate buried impact basins and found that QCDs are common over the highland portions of Mars. Furthermore, the QCDs not visible on images generally outnumber visible impact basins of the same size [Frey et al., 2000].
 We believe the MOLA-found QCDs without visible structural expression to be buried impact basins. This is supported by: the widespread distribution of the features; their close association with known impact basins, mostly in the heavily cratered highlands of Mars; lack of association with obvious tectonic and volcanic centers; generally circular, bowl-like shape, softened profiles and subdued relief; and similarity of shape of their cumulative number versus diameter curves (see Figure 3) to that for known impact basins. The large number of buried basins found suggests that early impact bombardment in the >200 km size range was much more intense and early resurfacing which buried early impact basins was more active than had previously been believed. This has important implications for the stratigraphy and chronologies that are used to describe martian evolution [Tanaka, 1986; Tanaka et al., 1992].
3. Lowland QCDs and Age of the Lowlands
Figure 1 shows one example of a large number of QCDs in the northern lowlands which are easily detectable in MOLA elevation data but which are not visible in Viking imagery. Many of these are isolated features, but, like craters and basins in the highlands, they sometimes overlap. Contours help resolve complicated structures. We searched the entire northern lowlands for QCDs >50 km using 32 x 64 pixels/degree gridded data. We avoided both the polar caps and the area immediately along the crustal dichotomy boundary zone where it is sometimes difficult to decide where lowland plains begin and the transition zone ends.
 Our survey revealed 644 QCDs larger than 50 km diameter in the northern lowlands (Figure 2). Only 90 of these are visible on Viking MDIMs, including many nearly buried but partially protruding craters in the knobby plains of Amazonis. The largest of the visible QCDs is a 365 km wide basin marked by an incomplete ring of mountainous material called Erebus Montes which Schultz and Frey  considered the inner ring of a much larger basin. 554 (85%) of the QCDs shown in Figure 2 are not readily visible in Viking MDIMs. These likely buried basins range up to 1075 km in diameter. Floor-to-rim relief in the buried features is typically several hundred meters, compared with expected relief ≥1.6 km for fresh impact craters ≥50 km in diameter [Garvin et al., 2000].
 QCDs are well distributed throughout the lowland plains. This supports the hypothesis that the QCDs (both hidden as well as visible) are impact basins. However, we found no QCDs of any size north of the large central volcanic structure Alba Patera and west and southwest of Olympus Mons. These are areas of relatively young lava flows whose thickness may be great enough to completely bury all relief associated with possible pre-existing impact basins.
 Where basins smaller than 500 km wide are still detectable by their relic relief, the overlying cover cannot exceed about 5–6 km or no relief would be preserved (based on [Garvin et al., 2000]). Where 50 km wide QCDs are abundant, the plains are likely less than 1.5 km thick. Based on the large number and distribution of basins of this size and even smaller, the northern lowland plains are generally 1–2 km thick, but in some places <1 and elsewhere >5 km thick, if the absence of large basins is due to burial by the overlying plains.
Figure 3 shows cumulative frequency curves for the northern lowland QCDs compared to QCDs in the Arabia highlands. Over the diameter range 200 to 500 km, buried highland basins are 3–4 times more numerous (in a cumulative sense) than the visible basins of the same size. Buried highland basins follow the −2 power law trend of visible basins over large diameters, but fall low at smaller diameters (as expected for a buried population). Lowland visible basins plot very low in Figure 3 compared to highland visible basins, consistent with their overall Hesperian and younger age. Buried basins in the lowlands plot much higher: at 100 km diameter they are 15 times more abundant in a cumulative sense than the visible lowland basins. More significantly, buried lowland basins plot above visible highland basins for diameters >100 km. At smaller diameters the lowland buried basins fall off from the −2 slope, as expected for a buried population.
 The nominal interpretation of the curves shown in Figure 3 is that the buried surface represented by the hidden lowland basins (under the lowland plains) is older than the visible highland surface. The highland surface on average is Middle Noachian age [Scott and Tanaka, 1986; Greeley and Guest, 1987; Tanaka and Scott, 1987]. Therefore, the lowlands beneath the plains are Early Noachian, dating from the earliest period of martian history, and pre-dating most of what is seen on the martian surface. This is a new and fundamental constraint on the age of the lowlands basement beneath the plains deposits.
 These results imply not that only are the lowlands extremely old, but apparently they have been stable (as a surface to preserve craters) for nearly all of martian history. The implications of this are profound. The highland-lowland crustal dichotomy is a primordial feature on Mars, dating from the Early Noachian, not later as has often been thought (e.g., [McGill and Dimitriou, 1990]). This is consistent with the age of the Utopia impact basin [McGill, 1989], which accounts for the dichotomy elevation difference in eastern Mars. QCDs older than Middle Noachian are superimposed on this basin (Figure 2); therefore the Utopia basin and the dichotomy predate the QCDs and must be Early Noachian in age. The Isidis impact basin, which is Early Noachian and superimposed on both the Utopia Basin and the dichotomy boundary in eastern Mars (Figure 2), also implies the crustal dichotomy dates from the Early Noachian.
 These results constrain not only when, but perhaps also how, the crustal dichotomy was established. The existence of buried basins does not directly indicate the mechanism by which the lowlands were formed, but does favor processes that operate quickly and early. Large size impact cratering [Frey and Schultz, 1988, 1990] associated with the end of accretion is consistent with this requirement. Internal mechanisms, if they occurred, had only a short time to operate before the formation of the preserved buried impact basins in Utopia.
 It appears there were northern lowlands throughout essentially all of martian history. At whatever early time liquid water may have been available on Mars, there was a basin in the north into which it could drain. Mars may have had a primordial shallow ocean in the Noachian, as Clifford and Parker  suggest, in which eroded ancient highland material may have been deposited [Hynek and Phillips, 2001].
 MOLA data provide compelling evidence that the present young lowland plains are a relatively thin veneer overlying a much older surface. The lowlands appear to have been low and stable for most of martian history. This means the fundamental character of the martian crust, its highland-lowland crustal dichotomy, is a primordial feature established soon after accretion and preserved throughout all subsequent martian geologic evolution.
 We wish to express our admiration and appreciation for the dedicated and exemplary efforts of the MOLA Instrument Design and Data Analysis Teams who made possible this superb data set. We thank several members of the MOLA Science Team for discussions and encouragement of this work, and also acknowledge the constructive suggestions of two reviewers.