Understanding the history of water on Mars is of a key importance to determine the evolution of its surface and atmosphere and its potential habitability. Thanks to the increasing quantity of data sent to Earth from the spacecraft (orbiters or landers/rovers) that reached Mars, a considerable variety of evidence of past water flow on the surface of the planet has been observed.
Owing to the actual Martian environmental conditions, liquid water cannot be present on the current planet surface for extended periods of time. However, several observations suggest that probably Mars was not always like it is today. Among these observations is the widespread presence of fluvial and lacustrine features on its surface (Ansan & Mangold, ; Baker, ; Cabrol & Grin, ; Carr, , ; Craddock & Howard, ; Irwin et al., ; Masursky, ). These features are classified on the basis of their morphology in valley networks, longitudinal valleys, outflow channels, and valleys on volcanoes (Carr, ). It is also possible to observe the presence of open and/or closed basin lakes associated with these valleys (Fassett & Head, ). This suggests that at the time of their formation, superficial temperature, and atmospheric pressure were different from the current conditions (Hynek et al., ). In particular, many authors believe that, during the Noachian Era, Mars had a warmer and wetter climate due to an intense volcanic activity, which made the atmosphere thicker (Carr, ; Kargel, ; Pollack et al., ; Sagan et al., ). This led to a greenhouse effect strong enough to produce a warming of the planet. Later, the progressive reduction of the volcanic activity (Carr, ) and the loss of atmospheric gases, mainly due to sputtering by the solar wind (Jakosky et al., ), caused a thinning of the atmosphere and, by consequence, a drastic reduction of the greenhouse effect.
A warm and wet Noachian Era is supported by many observations. While Hesperian and Amazonian terrains have well preserved impact craters, Noachian terrains are heavily eroded. Golombek and Bridges () estimated a decrease in the erosion rates from 3 to 6 orders of magnitude towards the end of the Noachian period. These authors also observed that in the first billion of Martian history not only the erosion velocity was much greater but also the sedimentation and deposition velocities. It is not clear where all the eroded sediments went, but there are no doubts that wind action alone cannot have produced such a large volume of eroded material: A likely explanation is the water flow in warm climatic conditions, very different from the present ones (Hynek et al., ).
This climatic change is also supported by the observation of phyllosilicates in Noachian areas and not in younger regions (Bibring et al., ). Phyllosilicates are hydrated silicates, showing the presence of the hydroxyl ion OH−, which are produced, on the surface of a planet, only in presence of a significant quantity of liquid water (Pollack et al., ; Schaefer, ). So the presence of these minerals is a strong indication that in the past, liquid water flowed on the surface of Mars. As a result, the presence of the valley networks and of these mineral deposits in the Noachian‐aged terrains have frequently been associated with relatively warm and wet climatic conditions during the Noachian Era (Barnhart et al., ; Craddock & Howard, ), requiring a vertically integrated hydrologic system (Head & Marchant, ).
However, the hypothesis of a warm and wet Noachian has also been questioned by a series of observations. Sometimes the valleys observed on the Martian surface maintain their width downstream, opposed to what happens in the terrestrial case where valleys tend to become wider and less deep downstream. These valleys also have shallow tributaries of small amplitude with abrupt terminations (Carr, ). These features, along with U‐shaped cross sections, are more characteristic of structures originated from basal sapping (Baker, ; Laity & Malin, ; Pieri, ), for example, the collapse of the terrain produced by the underground ice or water, a process which, according to Goldspiel and Squyres (), can virtually operate also under climatic conditions similar to the present ones.
Most valley networks show a poor integration (Carr, ; Stepinski & Coradetti, ) and even though those valleys are in some cases connected with open basin lakes, this immaturity could suggest short‐term fluvial activity (Aharonson et al., ). In addition, erosion and degradation rates typical of the Noachian are very low if compared with terrestrial standards (Golombek et al., ).
It is likely that the morphology of the valleys as they appear today is not original. Following their incision in the terrain, such structures may in fact have been changed by mass wasting processes, or by the failure of the side walls of the valley, giving rise to a morphology typical of groundwater sapping processes (Gulick & Baker, ). Erosion, faulting, and dust covering could then have further modified these valley systems. So, in most cases, we can see only the remnants of the original drainage path.
Clearly, the hydrological regimes in Noachian areas and in those post‐Noachian were very different. As previously mentioned, the Noachian surfaces are largely cut by valleys and show the signs of a considerable erosion while the post‐Noachian have not undergone significant erosion and exhibit local runoff channels. So we know that among the Noachian era and the following period, there was a noticeable change, but we do not know the cause. In addition, the valley networks seem to have formed by rainfall (Hynek & Phillips, ; Irwin et al., ; Irwin & Howard, ), but it is unclear under what conditions their formation took place.
Recent climate modeling studies (Forget et al., ; Wordsworth et al., , ) suggest an alternative scenario according to which Noachian Mars was characterized by mean annual surface temperatures below 0 °C. The proposed cold and icy climate contrasts with the previous climate hypothesis and is generally not compatible with precipitation‐driven valley network formation. For this reason, several mechanisms have been proposed to explain how valley networks formed under these harsh climatic conditions. Scanlon et al. () suggest an origin of the valley networks due to snowfall rather than rainfall. Other authors explain the formation of these structures considering episodic melting of icy deposits in the highland terrains owing to temporary warm conditions produced by the volcanic activity (Cassanelli & Head, ; Halevy & Head, ).
Moreover, Ehlmann et al. () showed that a significant component of the phyllosilicates in the Noachian crust (Fe, Mg clays) seems to be related to hydrothermal subsurface alteration. In this view, groundwater, in fact, could have had a key role in the formation and diagenesis of clay minerals, and its upwelling produced large deposits of sulfates, hematite, and chlorides. In addition, carbonate‐bearing rocks associated with olivine have been detected but carbonate is present in very small quantity on the Martian soil (Ehlmann & Edwards, ).
To explore the climate history of Mars through the study of the valley networks, an approach that uses the mapping of these structures on a global planetary scale is necessary. Up to now some global maps have been produced (Carr & Chuang, ; Hynek et al., ; Luo & Stepinski, ; Scott et al., ). Carr and Chuang () produced a global manual map (assessing the drainage networks by hand) based on Viking photographic data. Later, Luo and Stepinski () created a map of Martian valleys using automated routines based on topographic MOLA (Mars Orbiter Laser Altimeter on board of Mars Global Surveyor) data. The algorithm collects and maps all the troughs detected in the MOLA data. The produced map was then manually edited to remove false detections (Luo & Stepinski, ).
An update of this map was manually produced by Hynek et al. () using not only MOLA data but also a photographic THEMIS (Thermal Emission Imaging System on board of Mars Odyssey) mosaic with a resolution of 200 m per pixel. Here we present a global map of Martian valleys obtained using data at higher resolution. The approach, while more time consuming than computerize procedures, has successfully identified more drainage networks and smaller tributaries than computerized approaches were able to recognize, while correcting false positive identification of certain drainage networks. To improve our data set, we also used a recent published geologic map of Mars to obtain a rough estimation of the spatial distribution of valleys by their age.
We mapped all the Martian valleys with a total length greater than 20 km using photographic and topographic data obtained from different NASA missions: Mars Global Surveyor, Mars Odyssey, and Mars Reconnaissance Orbiter.
Among the various instruments onboard of the Mars Global Surveyor (launched to Mars in 1996), there was the MOLA altimeter, projected with the main goal of evaluating the height of surface features of the Martian surface. During its mission, MOLA collected high‐resolution topographic data, producing a global grid with a vertical resolution of 30 cm per pixel and a horizontal resolution of ~460 m per pixel at the equator (Smith et al., ).
In 2001, the THEMIS experiment onboard the Mars Odyssey spacecraft collected visible and infrared (IR) images of the Martian surface. THEMIS is an imaging system that combines a 5‐wavelength visual camera with a 9‐wavelength IR camera. The instrument has a resolution of ~19 m per pixel in the visible range and of ~100 m per pixel in the thermal IR (Christensen et al., ). Using the THEMIS IR images acquired during the first 7.5 years of the Mars Odyssey mission, Edwards, Christensen, and Hill () and Edwards, Nowicki, et al. () created new daytime and nighttime global mosaics of the Martian surface with a spatial resolution of 100 m per pixel. To date, this mosaic represents the highest‐resolution and best‐quality global data set of the Martian surface.
We combined this THEMIS daytime IR mosaic with MOLA topographic data to improve the capability to recognize and map fluvial structures over the use of the single THEMIS or MOLA data set. The mix between these data sets produces a stronger approach to mapping and analysis of the features displayed on the Martian surface. Where THEMIS data resolution were not sufficient (for small‐scale systems and/or valleys with a very high level of erosion), we also used data collected by CTX, the Context Camera of the NASA's Mars Reconnaissance Orbiter launched in 2005. The CTX images have a resolution up to 6 m per pixel (Malin et al., ).
Valleys, identified in THEMIS daytime IR and/or CTX images, plus topographic MOLA data, were manually mapped as vector‐based polylines within the open source QGIS (Quantum Geographic Information System) software (Athan et al., ). To analyze the selected data sets and produce the map, we used an equidistant cylindrical projection at low latitudes, and sinusoidal and polar stereographic projections at high latitudes. We looked for topographic troughs that showed morphological features suggesting a fluvial origin such as a network structure. Each network was mapped joining segments showing a hierarchical structure that starts at the outlet and proceeds upstream bifurcating at many points.
In our work, we preferred a manual approach rather than automated procedures for three main reasons:
- By means of automated routines, one can end up mapping every single depression on the planet. The user has then to check and manually correct the map deleting every false detection (Hynek et al., ). This, in our view, can result in an even longer procedure than the manual approach. This problem of false detections is particularly evident in complex regions characterized by different faults. For example, as already acknowledged by Hynek et al. (), on Alba Patera volcano, discriminating between channels carved by water flow and those that instead have a faulting origin may be challenging even using a manual approach;
- in our case, manual mapping uses not just topographic data but also THEMIS imagery with a higher spatial resolution (100 m per pixel) plus, when necessary, CTX data. The use of multiple data sets allows a better interpretation of the observed features;
- mapping any features using computer analysis of digital data represents a tricky problem because, as acknowledged by Luo and Stepinski (), even the best algorithms lack human understanding of the problem and are, alone, unable to perform such an analysis considering a broader context.
For all these reasons, we adopted a manual approach even though it also has its problems and limitations. These include a high level of subjectivity in valley identification and a potential loss of valleys in regions that do not show a strong incision on the surface. In fact, the identification of these features can be affected by variations in the albedo and by the quality of the data (Hynek et al., ). As done in previous manual approach (Carr, ; Carr & Chuang, ; Hynek et al., ), we tried to bypass this problem combining imagery and topographic data. For that reason, in this work, we used THEMIS daytime IR mosaic, plus CTX data for highly eroded systems, combined with MOLA data.
Our approach of manual mapping is similar to those first applied by Carr () and Carr and Chuang () and, more recently, by Hynek et al. (). During the mapping procedure, we used the same criteria followed by those authors: We searched for sublinear, erosional channels with a branching network morphology characterized by the presence of small branches upslope and a size that increases downstream (Carr & Chuang, ; Hynek et al., ). In order to be included in our catalog, a feature has to
- reflect the actual topography (as verified by longitudinal profile) and
- show V or U‐shaped cross sections (as verified by cross section profile).
Using QGIS, we stored the data as vector polylines in a shapefile containing the whole data set. The attribute table contains for each valley the coordinates of the system, the type of valley, and other data useful also for further analysis, like the total length and the estimated age (see below).
In the literature, the fluvial systems observed on the Martian surface have been described with several terms (as, e.g., valley networks, outflow channels, and runoff channels). To avoid confusion, in our work, the mapped valleys were divided, on the basis of their morphology, into six different groups here defined as follows: (1) valley networks (well‐developed and branching systems), (2) longitudinal valleys (systems featured by a long main branch and few tributaries), (3) valleys on volcanoes (small valley networks and single valleys located on volcanoes), (4) valleys adjacent to canyons (small tributaries of huge channels such as Kasei Vallis and of large crustal fractures such as Valles Marineris), (5) single valleys and valleys segments (systems with one or at most two tributaries), and (6) small outflow channels (systems with a morphology very similar to that of the outflow channel but at smaller scale). The outflow channels are huge channels carved by catastrophic events characterized by a sudden release of abundant liquid water. They can have maximum width around 100 km and lengths between 1,000 and 2,000 km, while the depth is usually greater than 1 km (Baker et al., ). Outflow channels start in fragmented and amassed chaotic terrains as a result of the sudden release of water from the subsurface probably due to the rapid melting of Martian permafrost subsequent to volcanic activity (Masursky et al., ). We decided to not include large outflow channels in our map because the duration of the water flow in these structures was ephemeral, so from a paleoclimatic perspective, the largest of these structures are not indicative of favorable conditions to the relatively stable presence of liquid water on the surface of the planet.
The distinction between the above reported groups of valleys was based on the analysis of the drainage pattern structures. These patterns, in fact, can be diagnostic since they are determined by the climate, the topography, and the composition of the rocks. Consequently, we can see that a drainage pattern is a visual summary of the characteristic of a particular region both geologically and climatically.
We note, however, that a drainage feature cannot always be unambiguously assigned to one of the six categories above mentioned, and, in fact, there are some systems that show ambiguous features. Ma'adim Vallis (20°S, 183°W), for example, has network characteristics in its upper reaches but resembles to an outflow channel for part of its length (Irwin et al., ). This suggest that the valley probably formed by a combination of large flood and slow erosion. Also, Mawrth Vallis (20°N, 15°W) has also mixed features of both outflow channels and networks.
Based on the interpretations of Baker et al. (), we included the Ma'adim Vallis in the group of longitudinal valleys, while we consider Mawrth Vallis a small outflow channel. In fact, Ma'adim Vallis shows a more branched structure and a long main branch. Mawrth Vallis consists mainly in a wide and long main branch, and it is also located close to the outflow channels around Chryse Planitia (Figure ).
Enlarge this image.Fig. 1. Examples of the different types of valleys mapped in the present work: (a) Warrego Valles, valley networks system located at 42°S 93°W; (b) Ma'adim Vallis located around 22°S 177°E, typical example of longitudinal valley; (c) valleys on a volcano centered at 31°N 150°E; (d) valleys adjacent to Valles Marineris (the image is centered at 7°S 82°W); (e) group of single segments located around 35°S 159°E; and (f) Mawrth Vallis at 22°N 20°W, example of small outflow channel. In all the images the base map is a grayscale THEMIS daytime IR mosaic at 100 m per pixel.
In this section, to better understand our choice, the different types of mapped valleys are described in more details.
Most of the cratered uplands of Mars show a dissection mainly due to networks of branching valleys with many tributaries (Hynek et al., ), having a width of no more than a few kilometers but with a length of hundreds or even thousands of kilometers (Carr, ). First observed in 1972 by the Mariner 9 mission (Masursky, ), these fluvial structures are known as valley networks.
Valley networks have a dendritic drainage pattern, which is very common on Earth and is characterized by the presence of many branches that usually are much shorter than the main trunk. This characteristic makes the dendritic drainage pattern very efficient because the branched structure allows water to flow quickly from the source or sources downslope (Carr, ).
Martian valley networks often originate near drainage divides (Craddock & Howard, ) and, especially during the peak‐activity period, showed an interconnection across great distances (see, e.g., Fassett & Head, , and Irwin et al., ). Owing to their prevalent morphologic features, such as meandering branch and densely dendritic forms, these valleys seem to be close analogs to terrestrial fluvial valleys (Carr, ). In addition, very often these fluvial structures are related to possible paleolakes (Fassett & Head, ). From this point of view, their formation may imply a runoff origin.
However, as reported in section 1, their morphology is sometimes ambiguous, and the interpretation of their features can be difficult. For example, most of them show steep walls and have V‐shaped cross sections in the upper reaches, while rectangular or U‐shaped in the lower reaches. Usually, the valleys start with stubby, alcove‐like terminations (Carr, ). These features led several authors to conclude that valley networks on Mars were originated mainly by groundwater flow (Carr, , ; Goldspiel & Squyres, , ).
In addition, groundwater‐driven valley erosion does not explain alone many other features of these valleys, such as the dendritic and high‐order tributaries (Hynek et al., ). In any case, even if some valleys originated from groundwater discharge, some sort of precipitation‐based recharge is required. In fact, some authors (Goldspiel & Squyres, ; Gulick, ) suggested that subsurface water reservoirs would need to be refilled several times to produce the observed eroded volume.
However, the morphology of these valleys may not be pristine. It is likely that after the incision of the valley, its structure changed owing to resurfacing processes such as aeolian deposition, dust and volcanic mantling, and impact crater ejecta. These phenomena may have caused a removal of the smaller tributaries of several Martian valleys (Craddock & Howard, ; Hynek & Phillips, ; Irwin et al., ; Irwin & Howard, ). In addition, it is important to observe that it is usually rare to observe this kind of features entirely produced by a single process. So it is possible that a single valley is the result of different geomorphological processes. Each one of these processes can modify the system's original morphology. The degree of modification is usually related to the nature of the environmental change. Sometimes, these changes can have an impact so dramatic to totally erase or hide the precedent morphology of the surface (Gulick & Baker, ). In this point of view sometimes it can be complicated to asses a valley origin. In any case the formation of valley seems to require, at least, some periods when rainfall was possible, and water was stable or metastable at the planet surface (Craddock & Howard, ; Hynek et al., ).
Longitudinal valleys (or sinuous valleys) are much larger than valley networks. They are narrower and more sinuous than outflow channels and have lengths of hundreds of kilometers and widths of a dozen kilometers (Baker et al., ). Unlike the outflow channels, these types of valleys never start from areas of chaotic terrain. Different hypotheses have been advanced about their origin. Some researchers believe that these channels formed by runoff (e.g., Masursky, ); many other authors, instead, claim that these valleys have been generated by basal sapping processes (e.g., Baker et al., ).
Basal sapping is subdivided into ground‐ice sapping or groundwater sapping. The latter is observed when the collapse of the terrain is triggered by the flow of water that would gradually erode the above ground until it collapses (Craddock & Maxwell, ). In the ground‐ice sapping process, the sublimation of the ice would have caused the collapse of the ground.
As discussed above, the valleys originating from runoff processes have typical V‐shaped cross sections, while those generated by basal (groundwater) sapping generally show U‐shaped transversal profiles. In Figure it is shown an example of longitudinal valley, Ma'adim Vallis, with its long and wide main branch.
Many major Martian volcanoes show the presence of valleys. These valleys have a radial pattern where the streams flow from a central point, close to the center of the caldera. Unlike the dendritic systems where the pattern is determined by the direction of the slope of the land and the stream flow in more or less the same direction, in the radial pattern the flow has multiple directions starting from the highest point.
According to some authors (i.e., Gulick & Baker, ), many Martian valleys located on volcanoes had a two‐stage evolutionary sequence based at first on surface runoff processes and then on groundwater sapping, which brought to an expansion of these features. However, several lava tubes and channels of volcanic origin have been observed, for example, on Alba Patera or around the caldera of Apollinaris Patera. This is an indication that lava flows could have played an important role in the channels formation (Gulick & Baker, ). However, Gulick and Baker () analyzed the fluvial features observed on six Martian volcanoes and concluded that, even if lava flows and volcanic activity could have initiated the formation of many of the valleys, then fluvial processes have long modified the original morphology with fluvial imprints, which have obscured the primitive valley features.
Many small and short valleys have been observed on outflow channels and on crustal fractures of the Martian surface. We included in this category the small tributaries of the huge outflow channels, like Kasei Vallis, and small valleys located on crustal fractures such as those of the fault zone of Valles Marineris. In general, the valleys located on both of these structures display theater‐headed tributary terminations and primitive, angular, dendritic‐drainage patterns. These features provide a strong argument for a sapping origin (Baker et al., ; Kochel, ). In fact, sapping valleys often originate along terrain fractures and joints in the bedrock, resulting in a rectilinear and controlled pattern (Baker et al., ). A sapping pattern is often simple and few integrated like those observed for these valleys. However, in some cases, it is possible to see small adendritic valleys locate at the edges of Martian canyons. For example, according to Mangold et al. (), some valleys located on Valles Marineris show a network system, whose heads seem to be randomly located on the plateau, that are characteristic of a surface runoff derived by rainfall.
Many single small segments, apparently disconnected one to another, are present on the Martian surface. In our map these segments were classified as single valleys and valleys segments. Those systems are defined as a group of single segments located very close to each other. These segments do not belong to any of the categories mentioned above and probably were once part of more integrated systems that were then eroded. In these cases, it is difficult to recognize their original morphology and to include them in one of the previous groups.
To the category of the outflow channels belong channels with a very large range of size: As already mentioned, the largest is Kasei Vallis and it is over 400 km across and 2.5 km deep, but other outflow channels may be less than 1 km across (Carr, ). We remind that we have not mapped large structures such as Kasei Vallis or Ares Vallis but we introduced, instead, in our map the category of small outflow channels, that is, channels that have a similar morphology of that of the real large outflow but on a smaller scale. In fact, since the length of the main branch of the mapped small outflow channels varies in the range between 120 and 1,000 km, these channels are shorter than standard outflow channels. In addition, their morphology is close to that of the longitudinal valleys, but their origin could have been similar to that of the large outflow channels. In fact, most seem to have originated by eruption of groundwater due to impacts, by volcanic or tectonic events, or by catastrophic drainage of lakes (Carr, ). In any case their presence across the whole planet surface indicates the abundant presence of water and water ice in the Martian crust (Carr, ). For example, we classified Mawrth Vallis as small outflow (see Figure ).
With respect to the previous global maps (Carr, ; Carr & Chuang, ; Hynek et al., ; Luo & Stepinski, ; Scott et al., ) data of higher image quality (new THEMIS mosaic plus CTX data) and topographic information (MOLA data) allowed us to map these structures at a finer scale, and this led us to identify a greater number of tributaries for several systems, to remove false positive and to identify several new systems.
In Figure some examples of mapped systems in comparison with the map of Hynek et al. () are shown. Using the global THEMIS mosaic with the resolution of ~200 m per pixel (Figures a, c, and e) only few parts of the system are visible, while, thanks to the new global mosaic (100 m per pixel of resolution), it has been possible to see in more detail the valley morphology and to recognize the presence of other small tributaries (Figures b, d, and f). In all cases we can observe that in our map, each system appears more developed with a higher number of tributaries.
Enlarge this image.Fig. 2. Examples of valleys mapped by Hynek et al. () using THEMIS data with a resolution of ~200 m per pixel (a, c, and e) compared with the same features mapped in the present work with the THEMIS daytime IR mosaic at 100 m per pixel (b, d, and f). In (a) and (b) valleys centered around 32°S 162°E (in Terra Cimmeria) are shown; in (c) and (d) one can see some small system located around 41°S, 43°E; finally, in (e) and (f) some valleys at 25°S, 7°E are shown.
We identified new tributaries for 919 systems and 204 new small systems in total. At the same time, different segments (about 360) of the Hoke & Hynek () map have been removed because they were revealed to be false detections.
The complete map of Martian valleys is shown in Figure where the different types of valleys are reported with distinct colors.
Enlarge this image.Fig. 3. Mapped Martian fluvial systems in different colors: Valley networks (black), single valleys (red), longitudinal valleys (blue), valleys on volcanoes (green), valleys adjacent to canyons (yellow), and small outflows (pink). The background map is a grayscale MOLA topographic mosaic from high (white) to low (gray). The map projection is equidistant cylindrical at low latitudes and sinusoidal and polar stereographic at middle to high latitudes.
As shown in Figures and and reported in Table the majority of valleys present on the planet surface are single diffuse segments (single valleys or disconnected systems of them) apparently not connected each other. This is probably due to the fact that, as reported in section 4.5, once these single segments were part of more integrated systems that there were then highly eroded.
Enlarge this image.Fig. 4. Percentage of the total length of the mapped Martian valleys grouped by typology. Inset: detail of the distribution of the last four classes.
| Type | Number | Total length (km) | Percentage (%) |
| Valley networks | 1,638 | 537,491 | 69 |
| Single valleys and valleys segments | 2,037 | 143,270 | 19 |
| Valleys on volcanoes | 10 | 36,824 | 5 |
| Longitudinal valleys | 22 | 29,539 | 4 |
| Small outflow channels | 14 | 18,493 | 2 |
| Valleys adjacent to Canyons | 51 | 7,942 | 1 |
| Total | 3,772 | 773,559 | 100 |
It is important to note that all the single segments associated to a specific epoch of the Martian history and located very close to each other were grouped together in a single multipart feature thus forming a system of single valleys. In this view, the number reported in the third row of Table is not the total number of each single segment; instead, it represents the total number of groups of single valleys not only spatially but also temporarily very close to each other. In fact, as discussed in the following section, to improve the information associated to our global map, we assigned an age to each valley assuming that a valley is at maximum as old as the terrain on which it has been carved (Carr, ; Hynek et al., ) and by checking the recent and detailed global geologic map produced by Tanaka et al. ().
Using a similar approach, all the valleys present on a single volcano were mapped together in one multipart feature. The number of valleys of this kind reported in the fourth row of Table represents the number of volcanoes incised by fluvial valleys (except for the valleys on Apollinaris Patera, which are divided into two different groups associated to different periods of Martian history). In the same way (see seventh row of Table ), mapping small valleys associated to outflow channels, we grouped all the valleys connected to a crustal depression or an outflow channel in relation with their age (in this way for one outflow we can have different groups of valleys according to their age).
Global mapping of the surface of a planetary rocky body is a unique and useful tool to study and estimate the spatial and temporal sequences of the geologic processes that operated on that surface. In addition, the combination of different kinds of maps (albedo maps, thermal inertia maps, fluvial valley maps, and geologic maps) allows a more complete and thorough analysis and understanding of those processes. For this reason, we tried to improve our data set combining the information concerning the age of the Martian terrains obtained from the geologic map of Tanaka et al. () with our global map of Martian valleys in order to have an estimation of the maximum age of these features. This method is a rough estimation, but it allows us to assign an age also to those valleys that are too small for age determination from superposition of impact craters and to have an idea of the global distribution of these valleys in relation with their ages (Carr, ).
Previously, other geologic maps, showing the age of the various Martian terrains, have been produced and used for such kind of analysis. The first geologic map was realized using Mariner 9 images with a resolution of 1–2 km per pixel and creating a photomosaic of them at 1:25,000,000 scale (Scott & Carr, ). Then thanks to the Viking Orbiter data, a series of three local 1:15,000,000‐scale maps were generated with a resolution ranging from 100 to 300 m per pixel (Greeley & Guest, ; Scott & Tanaka, ; Tanaka & Scott, ).
Carr () and Hynek et al. () used the Viking data‐based geologic map (Greeley & Guest, ; Tanaka & Scott, ) for the age evaluation of their mapped valleys. The improvement in the quality and resolution of orbital topographic and imaging data allowed a more detailed analysis of the Martian surface. Using MOLA, (460 m per pixel resolution), THEMIS (100 m per pixel mosaic), and where necessary CTX data (5–6 m per pixel), Tanaka et al. () produced a new global geologic map at a 1:20,000,000 scale. The Martian surface has been divided into different units. An age range is assigned to each map unit on the basis of the Martian chronostratigraphic period (Noachian, Hesperian, and Amazonian—Scott & Carr, ). The latter has been subdivided into eight epochs (Early, Middle, and Late Noachian; Early and Late Hesperian; and Early, Middle, and Late Amazonian—Tanaka, ), based on stratigraphic relationships and crater‐density determinations. The boundaries of each epoch have been defined using crater‐density analysis and chronology functions (Hartmann, ; Hartmann & Neukum, ; Ivanov, ; Neukum & Ivanov, ; Tanaka et al., ; Werner & Tanaka, ).
Combining the age information extrapolated from the geologic map of Tanaka et al. () with our map, we obtained the results shown in Figures and . We have grouped all the mapped valleys in three large categories corresponding to the three epochs of Martian history (Noachian, Hesperian, and Amazonian). In addition, we included two other groups: Noachian‐Hesperian and Hesperian‐Amazonian. In the first case (Noachian‐Hesperian), the majority of the valleys are located in the units defined by Tanaka et al. () as “Hesperian and Noachian basin unit,” “Hesperian and Noachian highland undivided unit,” and “Hesperian and Noachian transition unit”; for these valleys we do not have a precise indication of the epoch because, according to the authors, they span the entire Noachian‐Hesperian epochs. The valleys of the category Hesperian‐Amazonian, instead, belong to the so‐called Amazonian and Hesperian impact unit; also, in this case we do not have a precise indication of the age of the terrains in which the valleys are incised because, according to the map of Tanaka et al. (), they span the entire Hesperian‐Amazonian periods.
Enlarge this image.Fig. 5. Distribution of the mapped Martian valleys grouped by typology and by epoch. Same color code as Figures and . Inset: detail of the last four classes.
Enlarge this image.Fig. 6. Mapped Martian fluvial systems in different colors according to various age classes: Noachian (red), Noachian‐Hesperian (yellow), Hesperian (black), Hesperian‐Amazonian (pink), and Amazonian (blue). The background map is a grayscale MOLA topographic mosaic from high (white) to low (gray). The map projection is equidistant cylindrical at low latitudes and sinusoidal and polar stereographic at middle to high latitudes.
It is important to note that, with this method of age estimation from the youngest unit that a valley cuts, we obtain for each valley an indication about its age that represent a maximum value. The determination of the true relative age of a valley is instead based on crater counting (Fassett & Head, ; Hoke & Hynek, ; Tanaka, ). So far, some of the larger valleys systems have been dated using crater counting statistics assigning in this case an age that represents a minimum value of the actual age of the valley.
In addition, the maximum age assigned with this approach is affected by the uncertainties in the geologic mapping. As already mentioned, in some cases it is difficult to assign ages to extensive geologic terrains (Carr, ; Tanaka, ), so for these units we do not have a clear indication of their age.
However, even though these intrinsic uncertainties, this method can give us a good indication of the valley maximum age distribution and it is also useful for the age estimation of many small systems for which it is not possible to perform a careful analysis of the superimposed impact craters (Carr, ; Hynek et al., ). Hitherto the global geologic map of Tanaka et al. () is the most accurate planetary dating of the Martian surface.
Considering the same ages breakdown, we can compare the results here obtained with those reported by Carr () and Hynek et al. (; Table ). Compared to the map of Hynek et al. (), we obtained that 94% of the mapped valleys lie within Noachian terrains, while 4% are carved in Hesperian terrain and only 2% in the Amazonian terrain.
Comparison of Valleys Age Obtained in Our Work and for the Map of Carr () and Hynek et al. ()| Valley's age | Carr, (%) | Hynek et al., (%) | This work (%) |
| Noachian | 90 | 91 | 94 |
| Hesperian | 5 | 6 | 4 |
| Amazonian | 5 | 3 | 2 |
Table shows that the progressive improvement in the resolution of the photographic data leads to an increase in the number of Noachian valleys identified by Carr () and Hynek et al. (). According to these previous studies, the majority of the mapped valleys are carved in the Noachian terrains, while the youngest units (corresponding to volcanoes, as well as crater and canyon walls) show a poor dissection. The results here obtained confirm what previously suggested: The rate of valleys formation declined with time. There was an intense and widespread formation of valleys in the Noachian period but then the process slowed progressively and became more and more restricted to areas of steep slopes, such as canyon or crater walls, and/or to volcanic regions characterized by high heat flows (Carr, ). In Hesperian and Amazonian terrains outflow channels are more likely than valley networks (Carr, ; Hynek et al., ).
Fluvial valleys cuts Martian volcanoes of all ages: As already acknowledged by Gulick and Baker (), this suggests that likely in some isolated regions of the planet surface, there were active water flows until middle to late Amazonian. This means that the formation of the valleys on volcanoes can be related to local climatic changes due to volcanic activity and hydrothermal circulation (Gulick & Baker, ). As it happens on Earth, it is possible that locally warmer environments develop along regions of hydrothermal activity and this induces enhanced precipitation in these areas (Gulick & Baker, ). Another alternative scenario, proposed by Hynek et al. (), attributes the formation of valleys on volcanoes to periods characterized by a temporarily warm and wet climate characterized by rainfall.
Geological resurfacing events occurred after the formation of these valleys (Hynek et al., ) may have drastically changed the morphology of the latter, making difficult the identification of several systems. For example, owing to processes of mantling and terrain softening (Kreslavsky & Head, ; Mustard et al., ; Squyres & Carr, ) is more and more complicate to recognize and map valleys located in the southern hemisphere at latitudes higher than 30°. As a result, one can consider the visible geographic distribution as an underestimation of the original one.
If we consider in more detail all the valley networks and analyze their spatial distribution in relation with their total lengths and their ages, we obtain the results shown in Figure . In this case we considered all the eight epochs: Early, Middle, and Late Noachian; Early and Late Hesperian; and Early, Middle, and Late Amazonian. In some cases, we also observed well‐developed systems in two of these distinctive and consecutive epochs.
Enlarge this image.Fig. 7. Geographic distribution (with latitude—upper panel; with longitude—lower panel) of the mapped Martian valley networks with age and total length. The coordinates correspond to the central point of each mapped system.
All the valley networks are more or less evenly distributed with longitude, even though a greater concentration to the east latitudes is visible (see in Figure ), while regarding the latitude, as already noted by other authors (e.g., Hynek et al., , and references therein), these valleys are mainly located between 60°S and 20°N with few of them between 20°N and 50°N. The oldest valleys are mainly located in the southeast of the planet surface and are entirely Noachian.
It is also important to note the visual absence of fluvial valleys (and in particular of valley networks) in the northern lowlands, and this implies that the Martian planetary dichotomy had important effects upon the distribution of valley networks. The observed distribution could be easily explained if, when the valley networks formed, an ocean was present in the northern hemisphere. This hypothesis is suggested by a number of geological observations that show inter alia the presence of putative shorelines along the boundaries between lowlands and highlands (see Di Achille & Hynek, , and references therein) and is supported by the results of a recent work that suggests that valleys and ocean could have been coeval (Citron et al., ).
As previously discussed, the Martian valley networks density is greater in the Noachian terrains. This supported the hypothesis widely discussed (Carr, ; Hynek et al., ) that most valleys originated during this phase of the Martian history. Then the activity decreased drastically, but few valleys formed in the subsequent period.
The valley eroded volume is an important parameter in the investigation of the geologic history of early Mars. The volume of the eroded valley material is, in fact, related to the amount of removed sediments and consequently to the amount of water required to carve the valley. Starting from the produced global map, we selected a subset of the mapped valleys that includes a large part of the most developed Martian fluvial systems: 63 valleys with a main branch longer than 150 km and with a total length greater than 600 km (see also Orofino et al., ). We estimated, in terms of order of magnitude, the total eroded volume of these valleys, using the instruments of the QGIS toolsets.
Each fluvial valley manually mapped in our global map has been then approximated through polygonal shapes. In particular, a series of polygons were manually traced with the vertices located along the entire route of the fluvial system, following the valley outer walls visible in the THEMIS daytime IR images coupled with topographic MOLA data (Orofino et al., ). Using this approach, the area of the valley was approximated by that of the large polygon obtained by joining of the various smaller polygons.
To estimate the volume of each valley, a digital elevation model was created using MOLA data and the previously obtained polygonal shapefile. From this, the volume was evaluated considering a reference plane at the highest level of the terrain in which the valley is incised and the valley bottom surface and taking into account the slope of the valley. Furthermore, according to Hoke et al. () and Luo et al. (), the infill of material after the end of valley network formation was considered negligible compared to the total volume of the valley.
For the whole chosen sample of 63 valleys, we obtained an eroded volume of 3 × 1014 m3. In order to estimate the contribution to the total eroded volume of the remaining valley networks we at first analyzed the distribution of the total length of these valleys obtaining the results shown in Figure . From this distribution, we obtained a mean value of 190 km and a median of 150 km.
Enlarge this image.Fig. 8. Distribution of the total length of the mapped Martian valley networks with a total length smaller than 600 km.
Analyzing a sample of valleys with a total length around the average value (in particular between 180 and 200 km), we obtained a mean volume of 2 × 1010 m3. If we multiply this volume for the number of the remaining 1,575 valley networks mapped (assuming that this dimension is representative of all the remaining valleys), we estimated a total volume of 3 × 1013 m3. We can conclude that the volume of the rest of the valley networks mapped is negligible with respect to the volume of the 63 longer fluvial systems. Therefore, we can assume that the volume of the latter valleys is representative, in terms of order of magnitude, of the total eroded volume since the remaining valleys give only a small contribution.
We can now compare our value with those obtained by previous works. In a recent study Rosenberg and Head () used a fluid/sediment flux ratio function obtained from terrestrial empirical data, to estimate the water volume needed to carve the observed valley networks. To do that, they used the volume estimation of only seven largest valleys previously studied by Hoke et al. () and assumed that the rest of the valley networks give a negligible contribution to the total global volume. They thus obtained a sediment volume of 1 × 1013 m3.
With the development of many morphological image processing techniques, other estimations of valleys' volume have been obtained (Jung et al., ; Luo et al., , , ; Rodriguez et al., ). Luo et al. () developed a technique called progressive black top hat transformation method. The latter allows to estimate the minimum cumulative volume of water required to produce the observed excavation through the evaluation of the depth of each valley pixel. The estimation of the total eroded volume obtained by these authors, using both the topographically derived and manually digitized valley networks, is (3.0 ± 1.4) × 1014 m3 (Luo et al., ). This value is in good agreement with that obtained in this work, even if the two approaches differ in the procedure: of evaluation of the valley area.
In our work, as well as in that by Hoke et al. (), this area was outlined using a manual drawing. As acknowledged by Luo et al. (), the manual procedure can be more accurate even though it is time‐consuming and requires a direct human intervention to outline the boundaries of the valley, while the progressive black top hat method evaluates the valley area and the eroded volume automatically. Both volume estimates, ours and that obtained by Luo et al. (), are 1 order of magnitude larger than that obtained by Rosenberg and Head (). These authors based their estimation on the eroded volume of eight large systems previously studied by Hoke et al. (), considering negligible the contribution of the rest of valleys. For this reason, their value is an underestimation of the real total volume.
In this work we updated previous global maps of Martian valleys using the recent THEMIS mosaic with a resolution of ~100 m per pixel plus CTX data (~6 m per pixel) and MOLA data (463 m per pixel). The used mosaic is the one with the best resolution presently available.
These data sets along with a manual approach and a thorough analysis allowed us to better map those features at a fine scale and consequently to identify new tributaries for a considerable number of systems, along with some small valleys not previously mapped, and to remove false positives. To improve our map, we associated to the mapped valleys an attribute table containing valuable information including the following: central coordinates, total length of the systems, and an estimated maximum age. For the latter, we combined the data obtained by our map with those of the map of Tanaka et al. (), which is considered, to date, the most detailed dating of the surface of Mars. In this way, we obtained an indication of the maximum age of each valley since we assumed that a valley is as old as the terrain on which it has been carved (Carr, ; Hynek et al., ). This assumption is a good choice to have a global idea of the age distribution of these valleys, and it is also a way to assign an approximative maximum age to valleys that are too small for age determination by means of crater counting techniques (Carr, ).
Furthermore, we selected a subset of the mapped valleys that includes systems with a main branch longer than 150 km and with a total length greater than 600 km (Orofino et al., ). The map produced allowed us to extract digital elevation model for these 63 Martian valleys and to determine their volume. On the basis of the volume obtained for this sample of valleys, we estimated the total eroded volume associated to whole mapped valleys. Analyzing the total length distribution of the remaining valleys, we provided an estimation of the total eroded volume of the planet surface that is in good agreement with that previously obtained by Luo et al. ().
The works done so far have many potential future developments. Using the produced global map and the results here obtained, it will be possible to perform analyses of the correlation between the valleys' maximum ages and their eroded volume. We are also planning to evaluate the drainage densities of the mapped systems and to compare the values obtained with those of the previous manual maps (Hynek et al., ; Luo & Stepinski, ). In addition, it will be useful to compare our data set with other data sets of the surface of Mars in order to better characterize these valleys, constraining their origin and their connection with the lithology and the potential habitability of the planet. To allow further analysis our updated global map is released to the scientific community on Zenodo (Alemanno & Orofino, ) and is also included in the Open Planetary Map platform (Manaud et al., ).
We thank Gaetano Di Achille for the fruitful discussions and comments. Vincenzo Orofino acknowledges the TAsP and Euclid INFN Projects. We are also grateful to Angelo Rossi and Nicolas Manaud for integrating our global map of Martian valleys in the Open Planetary Map system at
© 2018. This work is published under http://creativecommons.org/licenses/by-nc-nd/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Orofino, V 2
; Mancarella, F 3
1 Dipartimento di Matematica e Fisica “E. De Giorgi”, Università del Salento, Via per Arnesano, Lecce, Italy; Now at Institut d'Astrophysique Spatiale, CNRS, UMR‐8617, Universite Paris‐Sud, Orsay Cedex, France
2 Dipartimento di Matematica e Fisica “E. De Giorgi”, Università del Salento, Via per Arnesano, Lecce, Italy; INFN, Sezione di Lecce, Via per Arnesano, Lecce, Italy
3 Dipartimento di Matematica e Fisica “E. De Giorgi”, Università del Salento, Via per Arnesano, Lecce, Italy