The Mars Global Surveyor Mars Orbiter Camera (MGS MOC) has observed several atmospheric events repeat within a few degrees of Ls from one year to the next over 3 Mars years during a portion of the northern summer, southern winter seasons (Ls = 124°–140°). The repeatability suggests a previously unobserved lack of significant variation in the present Martian northern summer seasonal climate. Nonetheless, interannual variations in the seasonal south polar cap and in the residual north polar cap were observed. These appear to have had a negligible effect in perturbing Mars' weather during the northern summer season.
 Meteorological phenomena on Mars such as dust storms and cloud patterns are a direct result of atmospheric dynamics and thereby serve as proxies for monitoring seasonal and interannual atmospheric processes. Prior to the Mars Global Surveyor (MGS) mission, Earth-based observations of Martian weather phenomena were limited to periods of favorable viewing near opposition or areas accessible to spacecraft for limited times in elliptical, moderately inclined orbits around Mars. Earth-based observations, which document the recessional behavior of the polar caps and changes in large-scale cloud cover and dust storm activity, have revealed two important attributes of the Martian climate system: that the patterns of large-scale seasonal cap recession reoccur from year to year [Cantor et al., 1998; James and Cantor, 2001] and that large regional and global dust storms develop more or less randomly during a half year period referred to as the classical “dust storm season,” Ls = 150°–340° [Martin and Zurek, 1993], where Ls is the areocentric longitude of the Sun, measured from 0° at Mars' northern spring equinox. Ground-based millimeter measurements of dayside average atmospheric temperatures in Mars' low to middle latitudes during the 1990s [Clancy et al., 2000] suggested that perihelion regional or global dust storms occur in every Mars year, as suggested by Zurek .
 A chief goal of the MGS Mars Orbiter Camera (MOC) experiment was to obtain daily global images that would document the behavior of the Martian atmosphere over the course of one Mars year [Malin et al., 1992]. The great success of the MGS mission, including the mission extension beyond the first Mars year, has allowed for the first time the opportunity to examine aspects of the interannual variability in the present-day northern summer, southern winter seasonal weather. This paper examines some of those aspects of the interannual variability in the seasonal weather, which have been observed by MOC during a portion of the northern summer, southern winter seasons (Ls = 124°–140°) over parts of three consecutive Mars years (1997–2001). First, we will give a brief description of the MOC experiment, followed by a description of the annual repeatable atmospheric dust and cloud events observed by MOC between Ls = 124° and Ls = 140° and the interannual variable behavior of the polar caps. Section 5 will discuss the implications of the MOC observations in terms of interannual variability and forecasting of Martian weather.
2. MOC Observations
 MOC initially acquired a few global images during its approach to Mars in July and August 1997, before arriving at the planet in mid-September 1997. Owing to an extended period of orbital insertion and aerobraking [Malin and Edgett, 2001], during 1997 and 1998, the MOC acquired only sporadic views of the planet. When MGS reached its almost circular (1400 LT) polar orbit in March 1999, MOC began its mapping mission (1 Martian year of Mars observations), which ran through January 2001. An extended mission phase has followed, and MGS is currently in its second Martian year of daily, systematic observations.
 The MOC is a nadir-pointed instrument that consists of a broadband, visible wavelength (500–900 nm) narrow-angle camera (NA) with a maximum resolution of 1.4 m pixel−1 and two wide-angle cameras with red (580–620 nm) and blue (400–450 nm) band passes (WAR and WAB), both with a field of view of 140°, yielding limb-to-limb views with a maximum resolution of 230 m pixel−1. In their “daily global map” mode of operation the WA cameras have been continuously mapping the Martian surface at a constant resolution of 7.5 km pixel−1 (using a table-driven cross-track variable-summing algorithm) [Malin et al., 1992].
 Visual inspection of these daily global MOC images has resulted in the detection of a number of dust [Cantor et al., 2001] and condensate cloud events along with interannual variations in the seasonal behavior of the polar ice caps. Distinguishing between dust and (water ice) condensate clouds in the MOC images was accomplished by comparing corresponding WAR and WAB images. Condensate clouds (ice) have a higher single-scattering albedo and surface albedo than dust in the WAB band pass, which results in condensate clouds being more uniformly white [James, 1985] and therefore brighter at blue wavelengths than red wavelengths. The opposite is true for dust.
 Since its initial images in July 1997, MOC has observed Mars via global and narrow-angle imaging over portions of three different Martian years, providing overlapping spatial coverage over diurnal, seasonal, and now annual timescales. Such observations allow the opportunity to examine aspects of the interannual variability in the present-day Martian northern summer seasonal weather, using observations obtained of atmospheric dust and cloud events along with the monitoring of seasonal variations in the polar caps.
3. Atmospheric Observations
3.1. Annular Cloud
 MOC observations since 1997 of the northern summer, southern winter seasons have revealed that several distinct local-scale meteorological phenomena (clouds, dust storms, and dust devils) reoccur annually at nearly the same locations and about the same time of year (within ∼12 Mars days or 6° of Ls). The first meteorological phenomenon observed to repeat was a large annular cloud located in northern Vastitas Borealis (Figure 1). The annular cloud was initially seen by the Hubble Space Telescope (HST) Wide Field Planetary Camera 2 (WFPC2) on 27 April 1999 (Ls ∼ 130°), centered near 65°N, 85°W, and described as a large, slow, eastward moving cyclone (J. F. Bell III et al., STScI-PRC99-22, NASA, May 19, 1999). Owing to a problem with the MGS spacecraft [Malin and Edgett, 2001], MOC images of this location were not available until 30 April 1999. At that time, several large arc-shaped clouds were observed in the same area. Two days later, another annular cloud appeared centered at 65°N, 78°W and was observed moving northeast at ∼5 m s−1. MOC observations showed the annular cloud to be an early morning phenomenon, breaking up by midday into multiple clouds. Almost a full Martian year later, at Ls = 124° in March 2001, the early morning annular cloud phenomenon was observed again, this time centered near 67°N, 97°W, with the cloud-free core bordering on Alba Fossae and Tantalus Fossae to the south and Scandia Colles to the east. The phenomenon recurred irregularly over the next 2 weeks, forming in the early morning hours and then by midday either dissipating or breaking up into multiple clouds (Figure 2). The clouds typically moved eastward at ∼2 m s−1, downslope into the low-lying plains north of Acidalia Planitia. The 1999 and 2001 occurrences showed little evidence of large-scale cyclonic circulation. About 13 Martian days after the first annular clouds were observed in both 1999 and 2001, at Ls ∼ 136° and Ls ∼ 130° respectively, dust storms formed just south of the residual north polar cap and moved northward onto the cap between longitudes 0° and 30°W; each persisted longer than a single Martian day [Cantor et al., 2001]. The 2001 dust event was observed to be associated with the northern edge of the slow eastward moving remnants of the annular cloud (Figure 3).
3.2. Valles Marineris Dust Storm
 Interannual recurrence of other dust events has also been observed over the past 3 Martian years, particularly in the southern hemisphere just south of and within the Valles Marineris trough system (9°–15°S, 75°W) between Ls = 136° and Ls = 141° (Figure 4). The first of these dust storms was observed in 1997 by HST at Ls = 139° [Wolff et al., 1999] and by the MOC narrow-angle camera during approach to Mars at Ls = 141°. The HST and MOC images in 1997 showed a moderately large storm (large enough to be visible in images with a scale of 60 km pixel−1) already in Valles Marineris. At Ls = 139° the following Mars year (1999), MOC observed a dust storm [Cantor et al., 2001] south of and within Melas Chasma (one of the Valles Marineris) that moved farther into the canyons the following day and persisted as a dust haze for several more days. Because it had occurred twice before, during the extended MGS mission (2001), we specifically targeted the Valles Marineris region during the same time period. At Ls = 136.5° in 2001, MOC images captured the early stages of a dust storm extending northeastward from Sinai Planum and Solis Planum to just south of Melas Chasma. This storm entered Melas, Coprates, and Candor chasms one day later and persisted as a dust haze in these troughs for several days.
3.3. Dust Cells and Dust Devils
 MOC observations demonstrate that compact dust-raising events also regularly occur at specific times and locations. Two types of compact dust events are noted: “dust devils,” defined as columns of dust that are taller than they are wide and only a few square kilometers in areal extent at most, and “dust cells,” defined as dusty regions that are wider than they are high and typically of the order of a few hundred square kilometers. Dust devils, as seen, for example, in Amazonis Planitia (Figure 5), show a seasonal cycle that was observed in northern summer in May 1999 and repeated a Mars year later (March–April 2001). Dust cells often occur at the onset or along the leading edges of dust storms; this is the case for the Valles Marineris storms described above in May 1999 and March–April 2001. The seasonal coincidence of dust devils in Amazonis Planitia and dust cells in Syria and Solis Planum might suggest that dust cells are simply large dust devils, but two observations indicate otherwise. First, the largest dust devils observed anywhere on Mars (those in Amazonis) are a factor of 4–8 times smaller in surface area than the observed dust cells. Second, no dust devil has been observed to be associated with the formation of any dust storm, while dust cells are always associated with dust storms. This suggests that contrary to a popular hypothesis [Ryan, 1964], dust devils, although nearly ubiquitous [Edgett and Malin, 2000], are not directly responsible for initiating dust storms.
 In the north, MOC observations revealed that the summertime behavior of the residual north polar cap was very different between 1999 and 2001. During the 2000–2001 northern summer the residual north polar ice cap showed a more rapid recession than observed the previous Martian year in 1999 (Figure 7). The residual cap reached its minimal extent around Ls ∼ 97° [Cantor and James, 2001; James and Cantor, 2001] rather than at the previously observed Ls = 151°. However, in 2001, high albedo portions of the residual cap had by Ls = 119° grown back in some regions to the greater extent observed in 1999, although in other areas they were still smaller than in 1999. This early summer accelerated cap recession may have been caused by increased surface heating resulting from an increase in surface dust concentrations generated by polar dust storm activity during the middle of northern spring in late 2000 [Cantor and James, 2001]. It is interesting to note that while the residual cap behavior in early 2001 was very different than that observed in 1999 (and different from what was seen during the Viking missions of 1976–1980 [Bass et al., 2000]), the north polar residual cap observed by Mariner 9 in October 1972 was almost identical to the cap seen by MOC in early 2001 (years straddling or during the largest dust events ever recorded on Mars); see Figure 7. This suggests two things: One is that a small residual north polar cap is an indicator of recent or potential global dust activity, and the other is that the residual cap variations observed may be part of the cap's normal behavior over timescales >1 Martian year but perhaps less than a Martian decade or two.
 MGS MOC observations span portions of 3 Martian years, beginning with approach on 2 July 1997 and continuing through the most recent and ongoing extended mission global imaging. Thus far, interannual repeatability of dust storm, dust devil, and cloud events have only been recorded for the period Ls = 124°–140°. This period is known to be one of transition in seasonal climate: in the north it is summer, and atmospheric water content has already peaked and is quickly decreasing [Jakosky and Farmer, 1982; Davies, 1982; Bass and Paige, 2000] as temperatures begin to fall at the north pole owing to the decrease in polar daylight hours. Local dust storm activity persists around and over the residual north polar cap edge [Cantor et al., 2001] owing to strong daytime average temperature contrast between the cap (near 201 K) and the adjacent bare ground (near 236 K). (The average daytime surface temperature for the residual north polar cap and the surrounding bare ground were retrieved from the MGS Thermal Emission Spectrometer, thermal bolometer data obtained between Ls = 124° and 140°. Surfaces were assumed to radiate as blackbodies.) At about this same time, insolation is increasing in the southern hemisphere, where the seasonal CO2 south polar cap is starting to recede, releasing CO2 into the atmosphere and causing an increase in the atmospheric pressure globally [Hess et al., 1979]. Thus, after a long autumn and early winter absence of significant dust activity [Cantor et al., 2001], conditions during this period of transition are also conducive to the onset of dust storm activity in the southern hemisphere.
 The interannual repeatability of weather events observed by MOC during northern summer was not totally unexpected. It was during this same season that the meteorological conditions at the two Viking landing sites were observed to be fairly repeatable from one year to the next between 1976 and 1982 [Tillman, 1985, 1988; Zurek et al., 1992]. One example of this interannual repeatability was at the Viking 1 landing site where transient pressure variations were measured between Ls = 133°–144° and Ls = 144°–154° over several Mars years [Tillman, 1988]. These transient variations were assumed to be associated with passing dust storms [Tillman, 1988], but unfortunately, no other observations were available to confirm this. It is interesting to note that at least one of these transient events occurred within the same seasonal timeframe of the MOC observed event. This correlation possibly suggests that interannual repeatable weather (meteorological) conditions are not localized phenomena but may occur in a number of different locations across the planet during at least a short period of the northern summer season. The duration of this repeatability in the present Martian climate is not known, but it has been observed by the Viking 1 Lander [Tillman, 1988] and now by MOC to occur for a minimum of 3–4 consecutive Mars years. On the basis of these observations we predict that the MOC-observed events will repeat again for a fourth Mars year. They will occur in the same locations, during the same northern summer season (Ls = 120°–140°; December 2002 to February 2003) and will occur within several days of when they were observed in previous years. However, it should be noted that the repeatability of the clouds and dust events observed by MOC does not guarantee a high degree of repeatability and predictability over all seasons and regions on the planet. This was confirmed by the unexpected onset of a series of large dust storms (i.e., the 2001 “global dust storm”) at the beginning of northern autumn around Ls = 185°.
 As suggested above, it is possible that the repeatable dust and cloud activity occurs only during the Ls = 124°–155° period, when conditions are close to thresholds necessary for their occurrence. However, the repeatability of atmospheric events may also point to longer-term lack of variability in addition to seasonally restricted variations in the Martian climate. The interannual variations in the seasonal polar cap interior and residual polar cap configuration, both of which should reflect variations in the energy balance of the planet, appear in the longer term to have had a negligible impact on the repeatability of weather phenomena during the northern summer season. At this time the quantity of volatiles involved in the observed polar variations are highly uncertain; therefore it is premature to state the fractional change in the energy balance of the planet these polar variations could represent. However, it appears that variations, such as that represented by the retreat and resurgence of the residual north polar cap (noted above), are self-mitigating on fairly short timescales (e.g., less than seasonal). The self-mitigating north polar cap variation along with the interannual repeatable dust and cloud events observed by MOC suggests a state of limited variability in the present northern summer/southern winter seasonal climate.
 Since its initial images in July 1997, MOC has observed Mars via global and narrow-angle imaging over portions of 3 different Martian years, providing overlapping spatial coverage over diurnal, seasonal, and now annual timescales. Such observations allow the opportunity to examine global aspects of the interannual variability in Martian northern summer/southern winter seasonal weather, using observations obtained of atmospheric dust and cloud events along with the monitoring of seasonal variations in the polar caps.
MOC observations obtained over the past 3 Mars years, during part of the northern summer/southern winter seasons, have revealed that several local-scale meteorological phenomena (extratropical annular cloud, north polar and Valles Marineris dust storms, and Amazonis dust devils) reoccur annually at nearly the same locations and about the same time of year (within ∼12 Mars days or 6° of Ls). These reoccurrences suggest an interannual repeatability in select Martian weather events during the north summer season that has not been observed in any extent since the Viking 1 Lander from 1976 to 1982.
It now seems possible to predict, with some success, when and where a select group of large-scale atmospheric phenomena (dust storms, dust devils, and clouds) will form during the northern summer season.
The interannual variations in the south polar cap interior and in the recession of the residual north polar cap appear to have had a negligible effect in perturbing Mars' northern summer seasonal weather. The variations represented by the retreat and resurgence of the residual north polar cap appear to have been self-correcting on fairly short timescales, suggesting a state of limited variability in the present northern summer seasonal climate.
 We thank the MOC operations and support staff at MSSS: Michael Caplinger, Diana Michna, Elsa Jensen, Kim Supulver, Scott Davis, Jennifer Sandoval, Liliya Posiolova, Robert Zimdar, and Becky Williams, without whose efforts daily conduct of the MOC experiment would not be possible. We also thank the two referees for their helpful reviews of this paper. This research was supported by JPL contract 959060.