Introduction

Today Mars is cold and dry with a thin CO₂ atmosphere. It is a dusty, desert planet with no stable liquid water on its surface. Yet spacecraft data suggest a different climate prevailed in the past. Images of fluvial features on its oldest surfaces, for example, suggest that very early in its history water flowed freely. Warmer and wetter conditions under a thicker greenhouse atmosphere are implied. Isotopic data and surface mineralogy support this view. Lakes and even oceans may have existed, powered perhaps by a sustained global hydrological cycle involving evaporation, rainfall, and runoff. This, however, has been theoretically difficult to demonstrate given the faint young Sun and limitations on the sources and sinks of plausible greenhouse gases. Consequently, alternative ideas for warming early Mars involving impacts, volcanism, clouds, and orbital changes have been suggested. Yet a widely agreed upon solution remains elusive. In more recent geological times, thick polar layered terrains, debris-covered glaciers, and latitude-dependent mantle deposits indicate that ice ages have come and gone similar to those on Earth. Although rainfall is not implicated, cyclic mobilization and redistribution of large surface ice reservoirs, a process similar to what occurred on Earth, is implicated. And for the vast majority of the billions of years of Mars’ existence, a gradually brightening Sun, chaotic orbital variations, and episodic volcanism strongly implicate an ever-changing climate system. Clearly, Mars today is not representative of what it was in the past. This article reviews the evidence for climate change on Mars and constructs a picture of how its climate might have evolved over time.

Historical Background

Before the spacecraft era, telescopic observers of Mars worked out its size, basic orbital properties, and noted its variable features including clouds and polar ice caps. Most famously, Percival Lowell claimed to see canals crisscrossing the surface that in his mind were constructed by intelligent beings trying to survive on a drying planet (Figure 1; Lowell, 1906). Lowell’s ideas were immensely popular and generated much interest in the red planet, but they were not scientifically sound. He recognized that the atmosphere was thin, for example, but argued for a warm climate, an inconsistency pointed out a year later by Alfred Wallace (1907).

Open in new tab

Figure 1. Color drawing of Mars by Percival Lowell in 1905. Lowell believed that an intelligent civilization had built canals on the surface of Mars to bring water from the poles to the rest of the planet.

Source: Lowell Observatory Archives.

Nevertheless, the possibility of life on Mars remained foremost in public discourse until the Mariner flybys of the 1960s. These missions revealed a much harsher Martian environment than Lowell envisioned. The seasonal polar caps were made of CO₂ ice, not water ice; its atmosphere was primarily CO₂, not nitrogen or oxygen; and the surface pressure was estimated to be very low (~7 hPa), much lower than previously thought. Most important, craters—not canals—dominated its surface. There was no evidence for life of any kind much less an advanced civilization engaged in a global-scale canal project. The landscape instead resembled the Moon, and like the Moon, it suggested that Mars was a dead planet whose story ended soon after it formed.

This perception changed with Mariner 9, the first spacecraft to orbit Mars giving a much greater view of its surface. When it arrived in November 1971, Mars was in the middle of one of its famous global dust storms. But when the storm cleared, Mariner 9’s cameras revealed volcanoes, canyons, flood features, dried-up riverbeds, layered sediments, and many landforms shaped by ground ice. The surface, it turned out, was much more geologically diverse than indicated by the earlier flybys. Mars had been active for much of its history, and water and climate change were implicated.

With each successive spacecraft mission, evidence for climate change accumulated. Many of the fluvial features discovered by Mariner 9 and mapped at higher resolution by subsequent missions are best explained by flowing liquid water in a warm climate. Furthermore, they are frequently found on surfaces that cratering statistics suggest date to the Noachian period between 4.1 and 3.7 billion years ago (Gya). Thus, billions of years ago, a warm, wet climate apparently existed on Mars. The hypothesis that emerged was that Mars had a thicker CO₂ atmosphere then—one that was capable of providing a strong enough greenhouse effect to warm the surface to the melting point despite the faint young Sun (Pollack et al., 1987). Evidence also emerged for relatively recent climate change. Both polar regions on Mars are geologically young (<100 My) and consist of thick deposits of layered sedimentary material composed predominantly of water ice. These were interpreted as evidence for periodic climate change associated with variations in Mars’ orbit parameters (e.g., Pollack, 1979; Toon et al., 1980). Thus, there is evidence for two different types of climate change on Mars: one related to a change in the mass and composition of the atmosphere early in its history, and one related to changing orbital properties relatively late in its history. The connection between these two epochs implies that the atmosphere thinned over time and that the climate system was dominated by changes in radiative forcing. This picture remains the general consensus, although the details are still under debate.

Climate Controls

The most important controls on a planet’s climate system are the mass¹ and composition of its atmosphere, its orbital properties, and the luminosity of the star it orbits. These are important because mass and composition determine the strength of the greenhouse effect and global circulation patterns; orbital properties determine the diurnal, seasonal, and latitudinal distribution of sunlight at the top of the atmosphere; and solar luminosity sets the total amount of energy available to drive the system. Each of these climate forcings changes with time: mass and composition because of outgassing, escape to space, and atmosphere–surface interactions; orbital variations because they depend on precession and gravitational interactions with solar system bodies that are in constant motion; and solar luminosity because nuclear reactions in the Sun’s interior consume the material that fuels them. A planet’s size is also important because it affects heat loss and mantle convection and, hence, the ability to outgas material from the planet’s interior to its atmosphere. Escape is also easier on small planets because gravity is weaker, and less energy is needed to expel atmospheric gases into space. Thus, size plays a key role in determining the mass and composition of the atmosphere at any given time (for a review, see Ehlmann et al., 2016).

For a given mass and composition, a convenient measure of an atmosphere’s greenhouse effect is how much it raises the mean surface temperature above the effective temperature. The latter is defined as

where $S_{\mathrm{o}}$ is solar insolation, $A$ is the planetary albedo, and σ‎ the Stephan–Boltzman constant (5.67 × 10⁻⁸ W m⁻² T⁻⁴). $T_{\mathrm{e}}$ is the temperature a perfect emitter would radiate to space the energy it absorbs from the Sun. The difference between the surface temperature and the effective temperature of Mars today is only ~5 K. By comparison, it is ~33 K for the Earth. The small greenhouse warming on Mars today is due to the fact that there is only one major infrared absorption band (the CO₂ 15-μ‎m band), and although it is nearly saturated, it is not wide enough to absorb a significant fraction of the energy radiating up from the surface. Hence, much of the surface infrared emission radiates directly to space.

Mass and composition also affect circulation patterns because they determine the radiative response time of the atmosphere, which is roughly given by

where $p_{\mathrm{s}}$ is surface pressure, $c_{\mathrm{p}}$ is specific heat at constant pressure, $\overline{T}$ is the mass-weighted average atmospheric temperature, and g is the acceleration of gravity. For Mars today, ${\tau }_r\sim 3\times {10}^5$s, which is approximately an order of magnitude smaller than it is for Earth. Thus, the thin Martian atmosphere responds quickly to thermal perturbations, and as a consequence, the thermal structure of the atmosphere is dominated by radiative processes. In a thicker atmosphere, such as might have existed early in Mars’s history, the radiative time constant would be longer and dynamical processes associated with horizontal motions would have a greater influence on the atmospheric thermal structure.

Gravitational interactions of Mars with other planets in the solar system lead to ~10⁵ year periodic variations in the planet’s “Milankovitch” parameters: obliquity, eccentricity, and longitude of perihelion. These parameters, illustrated in Figure 2, control the latitudinal and seasonal distribution of sunlight at the top of the atmosphere and hence exert a controlling influence on the thermal structure of the atmosphere, its global circulation, and the stability of surface volatile reservoirs. Without the stabilizing effect of a large moon, resonant interactions between the spin axis and orbital precession cause very large variations in Mars’s obliquity and eccentricity. These variations must cause dramatic changes in the planet’s climate system. Table 1 summarizes their key values. Orbital calculations indicate that during the past 20 My, the obliquity has varied between ~15° and 45°, and the eccentricity has varied between 0 and 0.13 (Figure 3). Beyond 20 My, the parameters may be chaotic and unpredictable (Laskar et al., 2004). However, Laskar et al.’s ensemble statistical calculations suggest that over the lifetime of the planet, the obliquity could have varied between 0° and 82° and the eccentricity between 0 and 0.20, with the most probable value being 41.8° for obliquity and 0.068 for eccentricity, although crater orientations suggest a lower mean obliquity (Holo et al., 2016). These numbers alone indicate that the climate of Mars today cannot be representative of its past.

Open in new tab

Figure 2. Depiction of orbital parameters for Mars. Obliquity, eccentricity, and longitude of perihelion are present-day values.

Open in new tab

Figure 3. Martian obliquity and eccentricity for the past 20 Myr and projected forward for 20 Myr. Shaded regions represent the range of values for Earth over the same time period.

Source: Data from Laskar et al. (2004); adapted from Forget et al. (2017).

Since it formed ~4.5 Gya, the Sun’s luminosity has steadily increased by ~30%. Thus, the total flux of solar energy at Mars average distance from the Sun has been steadily increasing from ~440 W m⁻² in the beginning to ~590 W m⁻² today. The reason for the increase is that temperatures in the Sun’s core must rise to maintain hydrostatic balance as fusion of hydrogen to helium increases the molecular weight of the interior. Like the orbit parameters, this slow but steady increase in the Sun’s output must have had a major impact on the climate system.

Mars Today

From the Viking mission it is now known that the mean annual surface pressure on Mars is ~6.1 hPa. Carbon dioxide makes up 95% of its mass, with Ar and N comprising roughly equal parts of the remainder. Water vapor is present in seasonally varying trace amounts (15–1,500 ppm), as are micron-sized silicate dust particles with typical near-surface concentrations ranging from ~0 to 5 particles cm⁻³. Dust particles strongly absorb solar and infrared radiation and therefore significantly affect the planet’s energy budget, atmospheric heating rates, circulation intensities, and overall thermal structure.

Mars is colder than Earth because of its thin atmosphere and greater distance from the Sun (~1.52 AU). Mean annual surface temperatures are ~200 K compared to ~288 K for Earth. However, its highly eccentric orbit, Earth-like obliquity, rapid rotation rate (88775 s), and bone-dry surface lead to large temporal and spatial variations in surface temperature. During winter at the poles, for example, surface temperatures plunge to below 150 K, freezing out CO₂ in the atmosphere to form the familiar polar caps. At the other extreme, daytime surface temperatures at the subsolar point at perihelion can exceed 300 K. Although this is well above the freezing point of water, evaporative cooling rates are so high that sunlight cannot provide enough energy to melt pure ice even if it existed in these locations (Ingersoll, 1970). In the tropics, typical daily temperatures range from 190 K at night to 270 K during the day. These large temperature swings drive powerful atmospheric thermal tides, which are a major component of the global circulation.

Seasonal variations in surface temperatures give rise to the dust, water, and CO₂ cycles, which characterize the present climate system (Figure 4). These cycles are linked through the global circulation, which itself is affected by the seasonal cycles and the radiative effects of the changing mass and aerosol composition of the atmosphere (dust and clouds). Annually, the polar caps wax and wane, with the seasons cycling almost 25% of the atmosphere into and out of the polar regions each year. These large seasonal changes in atmospheric mass are reflected in the surface pressure data (Figure 4, top), which show a pronounced semiannual variation due to the CO₂ cycle. During the perihelion season (northern fall and winter), the vigor of global circulation systems increases as the planet moves closer to the Sun and absorbs more solar radiation. The intensified surface winds subsequently lift more and more dust into the atmosphere, raising global visible opacities to near unity (Figure 4, middle). At aphelion, the circulation is less vigorous, and global opacities decline by more than a factor of 2. During this season (late northern spring and summer), the CO₂ cap in the northern hemisphere completely disappears, exposing an underlying “residual” water ice cap, which sublimates water into the atmosphere. Much of this water is transported into the southern hemisphere (Figure 4, bottom). Some of it condenses in the equatorial atmosphere, forming the “aphelion cloud belt.”

Open in new tab

Figure 4. Seasonal cycles of CO₂ (top; Pa), dust (middle; visible dust opacity), and column water vapor (bottom; precipitable micrometers) on Mars. From Martinez et al. (2017), Kahre et al. (2017), and Montmessin et al. (2017), respectively.

During southern summer, however, the south CO₂ cap does not completely disappear. Thus, there is no exposed surface water ice in significant quantities that can serve as a major source for atmospheric water. The remnant CO₂ ice cap is somewhat offset from the pole and exists for the duration of the summer season. This asymmetry in the behavior of the polar “residual” caps, where one is composed of water ice and the other is made of CO₂ ice, is not well understood. Adding to the mystery is the presence of a deeply buried deposit of CO₂ ice in the south polar region that is roughly equivalent to the mass of the present atmosphere (Phillips et al., 2011). However, the year-round presence of CO₂ ice at the south pole in contact with the atmosphere has important climate implications not only for present-day Mars but also for Mars in the past. As first discussed by Leighton and Murray (1966), a perennial CO₂ ice cap in equilibrium with the atmosphere can control the average surface pressure through its heat balance if the cap is massive enough. Although subsequent measurements have shown that the present-day residual cap is not massive enough to significantly buffer surface pressures as Leighton and Murray envisioned, this concept factors heavily into the orbitally forced climate changes that have dominated much of the planet’s history.

Evidence for Climate Change

The Noachian

This is the earliest epoch and provides the most compelling evidence that the climate then was at least episodically warmer and wetter than it is today. Noachian terrains are much more eroded than younger surfaces (Carr, 1992). There are fewer small craters (<10 km), and the larger ones have degraded rims and are partially filled with sediment (Craddock & Maxwell, 1993; Matsubara et al., 2018). Figure 5 shows an example. Fluvial modification is most likely the dominant eroding agent, although there are few directly observable traces of such activity. Consequently, other processes, such as impact debris, aeolian modification, and mass wasting, could account for some of the degradation and sediment production. The sediments often exhibit a rhythmic layering (Malin & Edgett, 2000b), implying a quasi-periodic mode of deposition.

Open in new tab

Figure 5. Themis daytime IR mosaic of a heavily eroded crater at 24.3°S, 24.7°E. Note the lack of ejecta relief, the degraded rim, and the flat floor filled in by sediment.

Source: Adapted from Matsubara et al. (2018). Image courtesy of Alan Howard.

The contrast in crater preservation between Noachian and Hesperian surfaces implies a rapid decline in erosion rates sometime around the end of the Noachian. Estimates put the change from ~10²–10⁴ nm/year during the Noachian (an amount comparable to the driest regions on Earth) to ~10⁰–10¹ nm/year in the Hesperian (Golombek et al., 2006). This dramatic change in erosion rates suggests the climate transitioned from relatively warm and wet to cold and dry within a few hundred million years sometime near the Noachian/Hesperian boundary. The implication is that the atmosphere thinned during this time by escape to space, incorporation into the crust, or some combination of each.

Incised into the Noachian degraded terrains are the valley networks (Figure 6). This implies they postdate the degradation and are not the cause of it. Howard et al. (2005) argue that degradation occurred throughout the Noachian but that a late-stage burst of fluvial activity cut the networks toward the end. Crater-counting ages support a late Noachian–early Hesperian time frame for the end of valley network formation (Fassett & Head, 2008). The morphology of the valleys resembles terrestrial river systems, suggesting that precipitation (rain or snow) and runoff were involved, although a groundwater source for some networks is possible and seasonal melting of surface ice may have also played a role (Wordsworth et al., 2013). Formation times are estimated to be 10⁵–10⁷ years (Hoke et al., 2011). The valleys are typically 1–4 km wide, 50–200 m deep, and several hundred kilometers long. Their estimated peak discharge rates of 10²–10⁴ m³ s⁻¹ are comparable to those of terrestrial rivers, and they are common throughout Noachian terrains. However, their drainage systems did not mature enough to allow regional integration into the Amazon-like basin systems on Earth. Instead, they most often drain into local topographic lows forming closed lakes, suggesting that precipitation rates were modest rather than intense during this time.

Open in new tab

Figure 6. THEMIS image mosaic of Warrego Valles (42°S, 267°E) showing a dense drainage network possibly formed by rainfall and surface runoff during the Noachian. The scene is ~100 km across.

Source: NASA/JPL-Caltech/Arizona State University.

Lakes are also common fluvial features of ancient Martian surfaces (Cabrol & Grin, 2001, 2010). Most date to the late Noachian and early Hesperian, but some can be found on even younger terrains. They form in local depressions and impact craters. Some are open basin lakes, as in Figure 7, in which an outflow channel(s) can be seen (e.g., Fassett & Head, 2008); others are closed basin lakes in which no drainage occurs, and outflow channels are not observed. The open basin lakes, particularly those fed by valley networks, likely formed during the late Noachian period, whereas the closed basin lakes formed later in Mars’s history (Goudge et al., 2016). Often associated with these lakes are deltas and alluvial fans. The sizeable deltas in some of the larger lakes indicate significant amounts of water and sediment flowed into the basin. The delta in Jezero Crater, shown in Figure 8, is an example and is now being studied by the 2020 Perseverance rover.

Open in new tab

Figure 7. Valley network fed open basin lake on Mars at 20.5°S, 87.0°E. Image shows MOLA gridded topography (color shading) overlaid on a CTX mosaic and a HiRise Nadir image. North is to the right.

Source: Adapted from Goudge et al. (2016). Image courtesy of Giulia Alemanno.

Open in new tab

Figure 8. A delta deposit in Jezero Crater (18.4°N, 77.6°E), the landing site of the Mars 2020 “Perseverance” rover. The site was chosen because it was once flooded with water, creating this ancient river delta that may have harbored life. Yellow and blue colors indicate basaltic minerals, green colors indicate clays, and purple–brown colors indicate no distinctive spectral features. The scene is ~20 km across.

Source: Credit NASA/JPL-Caltech/MSSS/JHU-APL.

The implication of lakes for climate is ambiguous. Some are consistent with rainfall and runoff, whereas others could be sourced from groundwater or seasonal melting of surface ice. However, if the valley networks, lakes, deltas, and fans formed in a sustained warm and wet climate, then a global hydrological cycle is implied, which requires large bodies of water or oceans to power the system (Haberle et al., 2017). The existence of a Noachian ocean has been proposed (Clifford & Parker, 2001) but is controversial. The observational evidence is based on the tentative identification of shorelines, which are difficult to date, not well defined, and in some cases deviate substantially from an equipotential surface (Head et al., 1999; Sholes et al., 2019). Most of the discussion about possible oceans on Mars has focused on the Hesperian period because of the many large outflow channels that date to this period. These outflow channels could have been the source of water for an ocean.

Another line of evidence for the action of liquid water during the Noachian is the presence of hydrated minerals on the surface. As shown in Figure 9, hydrated minerals, such as clays, carbonates, chlorides, sulfates, and silica, have been detected at many locations in the ancient terrains (Ehlmann, 2010; Ehlmann & Edwards, 2014). They are alteration products of basalt and require the action of liquid water to form. Clays and carbonates are found mostly on Noachian surfaces, whereas sulfates, chlorides, and silica are found on both Noachian and Hesperian surfaces. The carbonates found in the 4.1 Ga Allan Hills meteorite (ALH84001) also date to this period (Borg et al., 1999). The implication for climate depends on whether the alteration took place on the surface through runoff or in the subsurface through impacts and circulating hydrothermal systems. Available data suggest that most clay minerals formed in subsurface hydrothermal systems (Ehlmann et al., 2011), whereas rover/lander data imply substantial surface aqueous alteration (Arvidson, 2016). In some cases, the distinction can be difficult to determine. The formation environment of ALH84001, for example, favors above freezing temperatures (~18°C) either in short-lived surface waters or in an ephemeral subsurface aquifer heated by impacts (Halevy et al., 2011). Nevertheless, even if all the hydrated minerals were formed in groundwater systems, it is reasonable to assume that recharge by percolating liquid water, either from rainfall or snowmelt, was needed (e.g., Clifford & Parker, 2001). Thus, the climate must have at least occasionally been warm enough to permit rainfall or melt surface ice.

Open in new tab

Figure 9. Global distribution of aqueous minerals on Mars. Dust cover contributes to the apparent patchiness of the observed distribution. The pink hatched box in the upper left denotes the location of dunes enriched in gypsum.

Source: From Ehlmann and Edwards (2014).

It is important to note that Noachian surfaces are not completely weathered—physically or chemically. Although this is the period in Mars’s history when erosion rates were comparatively high, they are still far below those typical on Earth (Golombek et al., 2006), which is consistent with the preservation of large ancient features on the planet’s surface (Carr & Head, 2010). And although hydrated minerals are found on Noachian surfaces, so are unaltered basalts such as olivine and pyroxene. Because the timescale to completely weather surface olivine is less than several million years (Olsen & Rimstidt, 2007), this implies that on average the Noachian global hydrological cycle on Mars was less intense than that on Earth.

The global mean surface pressure on Mars today is ~6.1 hPa, which is very close to the triple point pressure of water. If liquid water flowed during the Noachian, then the surface pressure must have been higher than the triple point to prevent rapid boil off. How much higher is difficult to ascertain, but a plausible estimate is ~100 hPa (McKay, 2004). If greenhouse warming played a role, as many researchers believe, then a substantially thicker CO₂ atmosphere is implied. Support for a much thicker early atmosphere comes from isotopic data (Table 2). H, N, Ar, and Xe are all isotopically heavy compared to those on Earth, which is best explained by the preferential escape of the lighter isotopes (Haberle et al., 2017). Escape can occur thermally (e.g., Jeans escape) or nonthermally (e.g., sputtering or dissociative recombination). Present measured escape rates by the MAVEN spacecraft could account for the loss of 800 hPa of CO₂ over the history of the planet (Jakosky et al., 2018). Thus, it is highly likely that the surface pressure during the Noachian was much higher than it is today. How much higher is not known.

Table 2. Key Isotope Ratios Indicating Loss of an Early Atmosphere on Mars

Open in new tab

Isotope Ratio	Mars	Earth	Possible Interpretation
36Ar/38Ar	4	5.3	Mass fractionating nonthermal escape
40Ar/36Ar	1,800	300	Early loss of 36Ar (impact erosion?)
129Xe/132Xe	2.5	0.97	Very early loss of 132Xe (hydrodynamic escape?)
14N/15N	170	272	Mass fractionating nonthermal escape
D/H	~4–8 × VSMOWa	SMOW	Mass fractionating thermal escape

Open in new tab

^a Vienna Standard Mean Ocean Water D/H (~1.56 × 10^–4).

The Hesperian

Two significant changes appear to have occurred during this epoch. The first is that the climate cooled and dried, as indicated by the rapid decrease in erosion rates and valley network formation at the end of the Noachian and the cessation of most fluvial/lacustrine activity by the end of the Hesperian. This is most easily explained by a thinning atmosphere and/or a loss of greenhouse gases. The second is that there was a change in global aqueous chemistry from a clay-forming environment to a sulfate-forming environment. This change probably began in the late Noachian. Clays favor neutral or alkaline conditions, whereas sulfates favor more acidic environments. The transition to a more acidic environment in the late Noachian/early Hesperian may be the consequence of a peak in volcanic outgassing of sulfur gases associated with the construction of the Tharsis Plateau (Bibring et al., 2006).

In addition to sulfates, there are other water-related features that date to the Hesperian. Some valley networks, for example, are clearly Hesperian aged (Magnold et al., 2004). Some lakes also date this period. Gale Crater, for example, formed at the Noachian/Hesperian boundary. Therefore, the sediments that subsequently accumulated in it, shown in Figure 10, must be younger. The Curiosity rover team studied these sediments and found mudstones, sandstones, deltas, alluvial fans, and fluvial conglomerates that they interpret as evidence for the existence of multiple episodes of long-lived (>1,000 years) lakes fed by surface runoff and groundwater discharge (Grotzinger et al., 2015). The Gale Crater data indicate that intermittent lake-forming environments may have existed throughout the Hesperian period (Milliken et al., 2014; Palucis et al., 2016).

Open in new tab

Figure 10. Ancient delta sediments in Gale Crater, Mars. Outcrop in the foreground is ~6 m wide.

Source: From Grotzinger et al. (2015). Credit: NASA/JPL-Caltech/MSSS.

Deltas are other evidence of water activity in the Hesperian. Deltas generally form by the deposition of sediment at the mouth of a river when the flow slows and stagnates. Thus, as with the lakes, a warm climate with rainfall or snowmelt is implied. The time needed to form a delta is not well constrained but is estimated to be thousands to millions of years (Moore et al., 2003). Many deltas are Hesperian in age and are sometimes found in closed basins. The delta in Eberswalde Crater is a striking example. It has many of the same characteristics of terrestrial deltas and is late Hesperian in age (Magnold et al., 2012). The delta in Jezero Crater (see Figure 8) is also believed to be Hesperian in age. Approximately 50 deltas have been observed along the upland boundary of the hemispheric dichotomy at roughly the same elevation and have been attributed to sediments deposited by rivers emptying into a northern ocean (Di Achille & Hynek, 2010). Others, however, favor a lacustrine origin (Rivera-Hernández & Palucis, 2019). In either case, warm conditions and rainfall/snowmelt are implied.

Whether an ocean existed during the Hesperian is uncertain (Carr & Head, 2019). The possibility is linked, in addition to the dichotomy boundary deltas, to the presence of many Hesperian-aged outflow channels that drain into the northern lowlands. An apparent shoreline of a possible ocean created by such drainage is shown in Figure 11. These channels, which start abruptly, may have originated from the catastrophic release of groundwater trapped below an icy cryosphere that formed as internal heat flow during the Noachian declined (e.g., Carr, 1979). Thus, they do not necessarily indicate a warm, wet climate. An example is shown in Figure 12. Cumulatively, they could fill a northern ocean with 110 m global equivalent layer (GEL) of water. Some of the previously mentioned shorelines are young enough to indicate a Hesperian ocean, and geomorphic features near the shorelines have been interpreted to be the result of impact-generated tsunamis (Rodriguez et al., 2016). However, the possibility of a Hesperian ocean remains controversial. The principal skepticism is related to climate expectations and shoreline data (e.g., Sholes et al., 2019). The abrupt decreases in erosion rates and valley network formation at the end of the Noachian suggest a cooling and drying climate system associated with a decline in surface pressure and/or greenhouse gases. In such an environment, the eruption of water from subsurface aquifers would rapidly freeze and migrate to cold traps (Kreslavsky & Head, 2002; Turbet et al., 2017). However, the geologic record strongly suggests that conditions remained favorable for at least occasional long-lived surface water throughout the Hesperian. This issue remains unsolved.

Open in new tab

Figure 11. THEMIS daytime mosaic of a possible shoreline along the outer contact of the Vastitas Borealis Formation at 20°N, 120°E. The sharp boundary between darker and lighter material can be traced at a constant elevation for more than 2,000 km. However, the origin of the boundary is uncertain. The prominent crater in the lower left is ~17 km across.

Source: From Carr and Head (2019).

Open in new tab

Figure 12. THEMIS mosaic of the Ravi Vallis outflow channel on Mars (1°S, 318°E). This is one of several dozen outflow channels that sourced large volumes of water into the northern plains during the Hesperian. The abrupt beginning of the channel and lack of tributaries suggest the water was released from below the surface under great pressure. The prominent crater in lower right is ~50 km across.

Source: NASA/JPL-Caltech/Arizona State University.

The Amazonian

By the beginning of the Amazonian (~3 Gya), liquid water rarely existed on the surface of Mars, indicating that the climate had cooled substantially, despite the higher solar luminosity. Large outflow channels and valley networks almost completely ceased forming. Erosion and weathering rates declined to near present values (Golombek et al., 2014), and surface changes due to impact cratering and volcanism were modest compared to earlier periods (Carr & Head, 2010). Fluvial action was limited to gullies (Malin & Edgett, 2000a) and (if they do require water to form) recurring slope lineae (McEwen et al., 2014). These are small, localized features that have no implications for past climates. Instead, most of the evidence for Amazonian climate change comes from the action of ice on the surface (Forget et al., 2017).

Amazonian ice-related features, such as lobate debris aprons, are observed at many nonpolar latitudes in both hemispheres (see Butcher “Water Ice at Mid-Latitudes on Mars”). The presence of lobate debris aprons, lineated valley fill, concentric crater fill, and pedestal craters is indicative of the former presence of glaciers. Figure 13 shows some examples. Because ice is stable only in the polar regions at the present time, this implies a different climate in the past—one that favored the mobilization and redistribution of surface ice between the poles and lower latitudes. Orbital variations are the most likely mechanism responsible for this mobilization and redistribution (e.g., Toon et al., 1980).

Open in new tab

Figure 13. Examples of Amazonian nonpolar icy flow features in the lowland plains of Mars’ northern midlatitudes (45°N, 37°E). The lobate debris aprons near the bottom extend from a steep headwall on the side of a flat-topped mesa. The small valleys that cut the headwall contain glacier forms that flow into the aprons. The lobate debris apron at the top of the image is fed by lineated valley fill.

Source: NASA/JPL-Caltech/Malin Space Science Systems. Image annotated and provided by Frances Butcher.

Both polar regions are characterized by extensive, kilometer-thick, layered deposits (Figure 14). The continuity and extent of these layers suggest they were formed by the sedimentation of dust and ice from the atmosphere modulated by periodic climate change associated with orbital variations (Byrne, 2009; Cutts, 1973; Kieffer & Zent, 1992; Laskar et al., 2002; Pollack & Toon, 1982; Toon et al., 1980). Radar data indicate that water ice dominates the composition of both deposits, with the south polar deposits having a greater total volume and higher concentration of dust (~1.6 × 10⁶ km³ and 15%, respectively; Zuber et al., 2007) compared to the north (~0.8 × 10⁶ km³ and <5%, respectively; Grima et al., 2009). Although the polar deposits appear to be geologically young, their accumulation ages are very different. The north polar layered deposits likely accumulated within the past ~5 My (Levrard et al., 2007), whereas the southern deposits have taken much longer, possibly as long as ~100 My (Koutnik et al., 2002). These ages are much longer than the timescales associated with orbital variations, indicating that many cycles have likely contributed to overall growth of the deposits.

Open in new tab

Figure 14. HIRISE image of exposed layers in the north polar region during summer suggesting orbitally driven climate change similar to the ice ages on Earth. Image centered at 82.5°N, 58.6°E.

Source: NASA/JPL/University of Arizona.

Radar data further reveal that sequestered within the south polar layered deposits near the pole is a vast buried deposit of CO₂ ice (Bierson et al., 2016; Philips et al., 2011; Putzig et al., 2018). If released to the atmosphere, it would roughly double the present surface pressure. The origin and burial of this deposit are unclear but are most likely associated with low obliquity when polar insolation is low and conditions are favorable for the formation of permanent CO₂ ice caps (e.g., Buhler & Piqueux, 2021; Toon et al., 1980).

Images also reveal a latitude-dependent, meters-thick smooth mantling deposit that further implies orbitally forced periodic climate change (Kreslavsky & Head, 2000; Mustard et al., 2001). This mantling deposit extends down to ~30° in each hemisphere and consists of multiple layers of ice-rich material believed to have been laid down at times of higher obliquity in the recent past but that are now eroding back. The presence of near-surface ice at mid to high latitudes is corroborated by the Odyssey orbiter data (Boynton et al., 2002; Feldman et al., 2002; Mitrofanov et al., 2002), the Phoenix lander (Mellon et al., 2009; Smith et al., 2009), and images of fresh impact craters (Byrne et al., 2009). The inferred ice abundances from these data are generally in excess of that expected from the filling of pore space alone, suggesting that the ice is very pure. Although the tropics are generally ice-free, fan-shaped deposits on the northwest flanks of the Tharsis volcanoes appear to be remnants of cold-based tropical glaciers that existed in the past (Head & Marchant, 2003). Global circulation modeling simulations show that such glaciers could form at times of high obliquity (Forget et al., 2006). These simulations clearly demonstrate that obliquity variations can mobilize and redistribute surface ice deposits from the polar regions to the tropics. Thus, there is ample evidence for Amazonian climate change, and a plausible forcing mechanism has been identified.

Theories of Climate Evolution

The origin and evolution of the Martian atmosphere and climate system are not yet well understood, but there are plausible scenarios (e.g., Lammer et al., 2013). The following is intended to provide an overview of such scenarios, recognizing that uncertainties remain and alternatives may emerge. The narrative is chronological and includes geophysical topics relevant to the discussion. A timeline is given in Figure 15.

Open in new tab

Figure 15. Timeline of major events affecting the origin and evolution of the Martian atmosphere and climate.

From isotope systematics of NWA 7034, a 4.43-Gy-old brecca meteorite from Mars known as “Black Beauty,” accretion, core formation, and magma ocean crystallization were completed less than 20 My after the formation of the solar system (Bouvier et al., 2018). Thus, a stable primordial crust existed on Mars very early in its history. During accretion, impact devolatilization probably created a temporary steam atmosphere that could have contained many kilometers (GEL) of water (Elkins-Tanton et al., 2005). Under these circumstances, the high extreme ultraviolet (EUV) output of the young Sun would have powered an intense phase of hydrodynamic escape, leading to removal of most of the steam atmosphere. Xenon fractionation data support this possibility. Thus, an early impact-generated atmosphere that accumulated during accretion was likely rapidly lost to space (Scherf & Lammer, 2021).

A secondary atmosphere then developed from outgassing of the planet’s interior. As the planet cooled, convection within the core generated a dynamo and magnetic field that would have offered some protection to the growing atmosphere from solar wind stripping, but not EUV. The composition of this secondary atmosphere would have depended on the mantle redox state. As is the case for Mars, rapid core formation would quickly separate iron from volatiles favoring a more oxidized mantle like that of Earth where outgassed volatiles would have been dominated by H₂O, CO₂, and SO₂. However, some of the SNC meteorites² indicate a more reduced mantle than that of Earth, in which case H₂, CO, CH₄, and H₂S would have been favored. Complicating matters further is the fact that the early outgassing history is not known and EUV-driven thermal escape (Tian et al., 2009) could have significantly depleted any pre-Noachian atmosphere. Thus, the mass, composition, and fate of a secondary pre-Noachian atmosphere between 4.1 and 4.5 Gy are highly uncertain, and this atmosphere may even have been largely absent (see Scherf & Lammer, 2021).

By the beginning of the Noachian (4.1 Gy), however, rapid thermal escape would no longer be operative because of the decline in the EUV flux. However, impact erosion during a possible Late Heavy Bombardment period³ could be a factor (Melosh & Vickery, 1989). And the loss of the magnetic field would have exposed the atmosphere to the solar wind. Nevertheless, the buildup of a thick atmosphere would have been possible if outgassing rates exceeded loss rates. Because the Tharsis volcanic province is believed to have been constructed during the Noachian, enhanced outgassing rates were likely at this time. Phillips et al. (2001) estimate that Tharsis could have outgassed 1.5 bars of CO₂. Thus, it is plausible that a thick atmosphere developed during the Noachian. How thick is uncertain, but estimates based on the observed crater size distribution (Kite et al., 2014), isotopic modeling (Kurokawa et al., 2018), the ⁴⁰Ar/³⁶Ar ratio in ALH84001 (Cassata et al., 2012), lava flow volumes (Craddock & Greeley, 2009; Phillips et al., 2001), atmospheric escape (Hu et al., 2015), and magma composition (Lammer et al., 2013) suggest that the Noachian atmosphere was likely composed primarily of CO₂, having surface pressures between ~0.5 and 2 bars. It is worth noting that above ~2 bars, CO₂ begins to condense in a pure CO₂ atmosphere (e.g., Forget et al., 2013; Kasting, 1991), which limits its abundance in the atmosphere; above ~3 bars, permanent CO₂ ice caps form and surface pressures are buffered by their heat balance.

The main challenge for this period of Martian history, however, is finding an explanation for the comparatively high erosion rates and fluvio-lacustrine features. Rainfall and runoff in a warm early climate are an obvious explanation, but the circumstances producing such conditions are in dispute. The main problem is overcoming the faint young Sun, which was ~30% less luminous during the Noachian. An atmosphere producing a strong greenhouse effect can overcome the faint young Sun problem, but it must have enough greenhouse power to raise the surface temperature 77 K above the effective temperature to reach the melting point of water (Haberle, 1998). Considering that Earth’s atmosphere, whose greenhouse is powered by CO₂ and water, provides 33 K of greenhouse warming, reaching the 77 K mark is clearly a challenge. Complicating matters further, Mars climate models cannot produce such a strong greenhouse effect for atmospheres consisting of CO₂ and water alone (e.g., Forget et al., 2013; Wordsworth et al., 2013).

One way to overcome this problem is with supplemental greenhouse gases. Gases such as SO₂, NH₃, and CH₄ are good greenhouse gases, but raising and sustaining their concentrations to the required amounts is problematic given plausible estimates of their sources and sinks. Although each of these gases could be supplied by volcanic eruptions, SO₂ and NH₃ have strong removal mechanisms: SO₂ is soluble and rapidly converts to sulfate aerosols (Kerber et al., 2015), whereas photolysis destroys NH₃ on timescales of years (Kuhn & Atreya, 1979). CH₄ too is susceptible to photolysis, but it could last for up to hundreds of thousands of years if the surface pressure is high enough and the source strength strong enough (Kite et al., 2017). This timescale is the minimum formation time for the valley networks as estimated by Hoke et al. (2011). Thus, the lifetimes of SO₂ and NH₃ are too short to explain the valley networks, whereas CH₄ could last long enough under the right circumstances.

Collision-induced absorptions between CO₂–H₂ and CO₂–CH₄, which absorb broadly in the infrared, are promising candidates for this line of thinking if the atmosphere is thick enough (~0.5 bars or greater) (Ramirez et al., 2014; Turbet, Boulet, et al., 2020; Wordsworth et al., 2017). They can provide significant greenhouse warming with modest concentrations of H₂ and CH₄, and H₂ has a long lifetime because its major sink is escape to space, which at the diffusion limit—the fastest possible rate—is at least ~10⁵ years. If the mantle was reducing, as indicated by some SNC meteorites, then H₂ and CH₄ could have been supplied by volcanic outgassing. If the outgassing rates were Earth-like, climatically significant abundances could have accumulated in the atmosphere (e.g., Batalha et al., 2016; Ramirez et al., 2014). Other sources for these reduced gases include crustal serpentinization and subsequent seepage into the atmosphere (Chassefière et al., 2013), the impact of iron-rich bolides into water-rich surfaces (Haberle et al., 2019), and radiolysis (Tarnas et al, 2018). Serpentine has been detected, although is not widespread (Ehlmann, 2010), and impact degassing of reduced gases is possible if the meteorites contain sufficient iron or other reductants that are heated to high temperatures in the presence of water (Haberle et al., 2019). Thus, there are multiple sources for these reduced gases. However, a thick CO₂ atmosphere of at least 0.5 bars is still required.

Another source of greenhouse power is water ice clouds (Haberle et al., 2012). If the clouds are optically thick, have particles in the right size range (~10 μ‎m), form at high altitudes where temperatures are cold (<200 K), and are global in extent, they can provide significant greenhouse warming. Urata and Toon (2013) and Kite et al. (2021) demonstrated the potential for a cloud greenhouse on early Mars with general circulation models and found they could raise surface temperatures to near the melting point under the right conditions. Clouds, however, are notoriously difficult to model, so more studies are needed to assess this possibility.

Much of the debate about early Mars concerns the duration of the warming periods. Some have argued that the characteristics of valley networks are best explained by the existence of a long-lived warm semi-arid climate system with an active hydrological cycle involving rainfall and runoff (e.g., Craddock & Howard, 2002; Ramirez & Craddock, 2018; Ramirez et al., 2020). A major constraint for a long-lived continuously warm and wet climate is that it cannot produce too much rainfall because the valley networks are not well-integrated and many late Noachian terrains contain unaltered mafic minerals (Ehlmann et al., 2011). Partly for this reason, others have argued that this period of Martian history was predominantly cold with water in the form of surface ice and that the fluvio-lacustrine features were carved during transient periods of warming triggered by a variety of external forcings, such as volcanic eruptions or impacts (e.g., Halevy & Head, 2014; Segura et al., 2008; Wordsworth, 2016). Indeed, the cumulative effect of impact-generated climate change predicted by one-dimensional models was thought to provide enough rainfall and runoff to carve the valley networks (Segura et al., 2008), but more sophisticated three-dimensional models suggest this mechanism may fall short of the required rainfall (Steakley et al., 2019; Turbet, Gillmann, et al., 2020). Other more speculative transient forcings include methane bursts triggered by the destabilization of a subterranean methane clathrate reservoir during obliquity transitions (Kite et al., 2017) or limit cycles that result from a feedback between CO₂ outgassing and weathering rates (Batalha et al., 2016). The challenge for the episodic hypotheses is matching the timing of fluvial activity and producing enough water to form the observed fluvial features.

Whatever the mechanism, liquid water did flow on the surface during the late Noachian and into the early Hesperian. Valley networks were cut, lakes filled and drained, deltas formed, and an ocean possibly existed. Volcanic activity was believed to be at its peak, with sulfur emissions changing the nature of surface mineralogy from clays to sulfates (e.g., Bibring et al., 2006). Gradually, the valley networks, lakes, and deltas ceased forming as the atmosphere thinned. As interior heat flow declined, the cryosphere thickened, and subsurface liquid water occasionally burst forth onto the surface, forming the widely observed catastrophic outflow channels and possibly a temporary ocean or sea. By the late Hesperian and early Amazonian, the climate could no longer sustain liquid water on the surface, and water from the outflow channels quickly froze and eventually migrated to the poles.

From then on, orbital variations and a gradually increasing solar flux determined the climate of Mars. During the Amazonian’s roughly 3-billion-year history, escape and slow carbonate formation would outpace degassing and continue to draw down the atmosphere until a balance was achieved. When, or if, this balance was achieved is uncertain. As surface pressures continued declining, they became increasingly controlled by the heat balance of the polar regions (Leighton & Murray, 1966) and the adsorption capability of the regolith (Buhler & Piqueux, 2021; Fanale & Cannon, 1978).

At low obliquity, polar insolation decreases and surface temperatures fall until they reach the frost point of CO₂. Permanent polar caps form, and surface pressures decline until vapor pressure equilibrium with the caps is achieved. Estimates put this lower limit at less than 0.5 hPa for an obliquity of 5° (e.g., Newman et al., 2005). In these regimes, the atmosphere would be clear and dry. Low surface pressures would reduce near-surface atmospheric densities to levels insufficient to lift dust off the surface, and polar water ice deposits now covered by permanent CO₂ caps would not be exposed to the atmosphere, thereby cutting off a source of atmospheric water.

A much different regime would prevail at high obliquity. At these times, permanent CO₂ caps would disappear, and annual mean surface pressures would be comparatively high. How high would depend on the uncertain regolith capacity. If, for example, the CO₂ stored beneath the south polar cap was released to the atmosphere at high obliquity, as seems possible, and regolith capacity was minimal, surface pressures could double from present-day values (Phillips et al., 2011). If, on the other hand, regolith capacity was sizeable, peak surface pressures would be diminished but still higher than those of present day (Buhler & Piqueux, 2021). In these high obliquity regimes, dust storms would be frequent and the atmosphere much wetter and cloudier than at low obliquity (Haberle et al., 2003; Madeleine et al., 2009, 2014). Dust lifting would be easier with higher surface pressures, and the polar water ice caps would no longer be protected by a permanent year-round covering of CO₂ ice. Furthermore, they would be heated to comparatively high temperatures, which would drive up sublimation rates and greatly moisten the atmosphere. If the obliquity is high enough (approximately >35°), polar water ice might no longer be stable and could migrate to tropical latitudes (e.g., Forget et al., 2006). The main uncertainty in these high obliquity regimes is the extent to which a sublimation lag might develop, thereby choking off the water supply (Mischna & Richardson, 2005); the capacity of the regolith to buffer surface pressures (Buhler & Piqueux, 2021); the nature of the coupling between the dust and water cycles (Kahre et al., 2015); and whether the clouds would warm or cool the surface (Haberle et al., 2012).

These obliquity-driven alternating epochs of clear and dry at low obliquity, and dusty and wet at high obliquity, would modulate the transport and deposition of dust and ice in the polar regions and are the most plausible explanation for the ubiquitous layered terrains so readily visible in orbital images (e.g., Kieffer & Zent, 1992; Smith et al., 2016; Toon et al., 1980; see Figure 14). Although many details need further study to more precisely understand how orbital changes lead to the observable surface features attributed to them, the orbital-forcing theory for Amazonian climate change is widely accepted.

Conclusion

Mars is a planet whose climate has changed throughout its history. For the first 400 million years of its geological record, Mars was at least occasionally warm and wet, with a thick CO₂ atmosphere. Whether by rainfall, snowmelt, groundwater discharge, or impacts, valley networks were carved and lakes were filled. Erosion rates were high, and oceans may even have existed. How the atmosphere produced such conditions in the presence of a faint young Sun is not well understood, but like the Earth, warming by greenhouse gases or clouds is the prevailing explanation. Then sometime around 3.7 Gya, erosion rates began to decline, perhaps indicating that the atmosphere bad begun to thin. Escape to space and incorporation into the surface would continue drawing down the atmosphere. However, lakes and deltas continued forming until ~3.0 Gya, when most liquid water activity finally ceased altogether. By then, ice became the dominant form of surface water, and for the remaining 3 billion years—the majority of the planet’s existence—the large variations in Mars’s orbital parameters would control the planet’s climate system. They created the quasi-periodic Martian ice ages by mobilizing surface ice and moving it around the planet; built up the polar layered terrains by modulating the dust, water, and CO₂ seasonal cycles; and dramatically changed the surface pressure in response to the changing polar heat balance. Ironically, the overall temperature trend on Mars throughout its history is opposite the trend in solar luminosity: It was warm early on when the Sun was faint, but it was cold and dry at the end when the Sun was bright.