Archive by Author | The Zooniverse

Apollo 12 – a new challenge is set!

Dear MoonZoo aficionados,

Our next surveying exercise will be centred on the Apollo 12 landing site.

Your previous and successful endeavour saw hundreds of thousands of craters and interesting features noted in the region of the Apollo 17 landing site in the Taurus-Littrow valley. Here we witnessed a chaotic and highly scarred terrain, squeezed between tall mountains and crossed by a deep fault (the Lee-Lincoln Scarp): a rather complex geological setting. Indeed, the landing site was selected based on its geological diversity, with the aim of collecting pre-Imbrian age highland material, mare basalts, and igneous products from potential volcanic edifices.

Apollo 12 – Commander Pete Conrad is working at the equipment bay of Lunar Module ‘Intrepid’ on the Ocean of Storms (©NASA)

Apollo 12 – Commander Pete Conrad is working at the equipment bay of Lunar Module ‘Intrepid’ on the Ocean of Storms (mission patch and image ©NASA)

Now we are turning our attention to the Apollo 12 landing site, and from 9 May all the Moon Zoo images relate to this site. In November 1969 Apollo 12 landed within a vast lunar mare (lava plains) region called Oceanus Procellarum (Ocean of Storms), and in particular an area baptised as Mare Cognitum (Known Sea), so called given that it had already been visited by three unmanned lunar missions (Luna 5, USSR, Surveyor 3 and Ranger 7, US). The landing region was estimated to be younger than the Apollo 11 site based on kilometre-size craters census (2.37 times fewer craters). In the following years, returned sample analyses (i.e. Stöffler and Ryder, 2001; Barra et al., 2006) estimated ages of 3.58 ± 0.01 and 3.80 ± 0.02 Gyr (both Late Imbrian Epoch), for Apollo 11, against 3.15 ± 0.04 Gyr for Apollo 12, (Eratosthenian Period). It will be very interesting to compare these direct age estimates with your high resolution/volume crater count survey, AND also compare them with the results from the Apollo 17 blitz (samples’ age: 3.75 ± 0.01 Gyr).

Obviously, as before we are also going to harvest data generated by the Moon Zoo users regarding bouldernyness and shape of the noted craters in order to build a fuller picture of the impact record in the region. As it happens, the lunar science team based at Birkbeck/UCL, UK, has been looking at the Apollo 12 region for quite sometime, both in terms of geological mapping and analysis of returned samples. We are particularly interested in the different lava flows found in the region and the mapping of small craters; the associated boulder distribution will be employed to estimate the different ages and thickness of these lava flows. Your Moon Zoo measurements of the Apollo 12 site will therefore be greatly appreciated, and they will potentially be incorporated in future scientific publications.

So, let’s start this new and exciting journey together: I will keep you posted on both results from previous efforts (A17, etc.) and the ongoing ones. Go and explore!

Barra F., et al., 2006. 40Ar/39Ar dating of Apollo 12 regolith: Implications for the age of Copernicus and the source of nonmare materials, Geochimica et Cosmochimica Acta 70, 6016-6031.

Stöffler D. and Ryder G. 2001. Stratigraphy and isotope ages of lunar geologic units: chronological standard for the inner solar system. Space Science Reviews 96: 9-54.

Dr. Roberto Bugiolacchi
Moon Zoo science lead
Birkbeck, University of London

Thank you Moon Zoo!

The MoonZoo science team would like to extend a gigantic thank you to all 20,627 users who contributed in counting craters (and more!) relating to the Apollo 17 landing site (Taurus-Littrow)!

Let’s ponder on some astonishing numbers: to date, around 8.5 million craters in total have been marked by MoonZoo citizen scientists, with around 670,000 (~8%) relating to the A17 region (from 21 selected NAC images, Figure 1); further, 3.3% (22,063) of these craters have been classified as containing boulders and 6.9% (45,893) were found to be non-circular.

Figure 1. On the right we see the MoonZoo users crater input. Different colours relate to different NAC images basemaps. On the left we see the A17 landing site (red dot) and the astronauts exploration paths and stations.

Figure 1. On the left we see the MoonZoo users crater input. Different colours relate to different NAC images basemaps. On the right we see the A17 landing site (red dot) and the astronauts exploration paths and stations.

Our next step is to compare your input with the ‘expert’ count looking to validate and quantify your contributions.  The ‘expert’ in question is a professional lunar scientist who has published research including the statistical occurrence of impact craters on planetary surfaces. The logical assumption is that given a more or less constant collision rate of interplanetary bodies (asteroids and comets), a surface will carry the record of impact products (craters and pits) as a function of time, i.e., from the time of resurface (maybe a lava mare flow) the scarring would be proportional to the length of exposure.

As most things in geology, this scenario is true but with caveats… : first, the resurfacing by lava flow or ejecta mantling might have only partially buried ancient craters, or, more probably, only the smaller ones, thus skewing the crater-size statistical record; crater rims erode with time, even on an airless body like the Moon, at a rate of around 0.06-1 cm per million year. This might not seem much, but in the lunar chronology scale, measured in billions of years, this factor becomes significant; in reality, the biggest source of uncertainty is represented by secondary craters: most impacts generate coherent distal ejecta that, when landed, produce smaller craters virtually indistinguishable from space-born ones. And this is fractal, i.e. scaled: big impacts will generate hundreds of smaller craters that will overlap with similar ones from nearby big impacts…

The hard reality is that there are no cast-iron methods to establish the origin of each excavation (although it has been advocated that a secondary crater might be somewhat shallower in comparison to a similarly-sized primary one). So, an ‘expert’ becomes so by developing a ‘sense’ or instinct on what ‘feels’ a statistically significant crater against one that is not. This approach is more akin to ‘artistic interpretation’ than ‘hard’ science, but qualitative investigation of certain geological features is an acceptable compromise when a physical method is either not yet available or even impossible to develop.

These considerations do not stop the development of alternative methodologies though; indeed, we are working closely with a research group at Manchester University which is building an automated pattern recognition software of circular features (and others) based on theoretical models, and actual data: ‘expert’ counts, AND MoonZoo users’ data.

Now, whatever approach brings us closer to a reliable crater counting method this cannot be easily accomplished by even a troupe of crater-counting planetary scientists: the 8.5 million craters noted by the MoonZoo community would have taken years to harvest otherwise!

So, what is going to happen now? Well, the ‘expert’ and pattern recognition software data will be compared with the MoonZoo output, uncertainties and limitations of all approaches established and, hopefully, develop a method that will represent the basis for ‘trusting blind’ the MoonZoo craters stats. In practice this will translate into something like “MoonZoo crater data are consistent with other methods for crater of sizes ‘x’ to ‘y’, in images with resolution higher than ‘z’ meters, and illumination of ‘n’ degrees or higher”.

Ultimately, the crater statistics (Cumulative Crater Frequency) plotted against known crater accumulation functions (i.e. Neukum, 1983, 2010) give us an estimate of the age of the lunar region. Using these data from landing sites allows for comparison with returned samples whose age has been established in the laboratory.

Figure 2. Age estimates based on estimated crater frequency distribution against crater size (diameter)

Figure 2. Age estimates based on estimated crater frequency distribution against crater size (diameter)

Our next journey will focus around the Apollo 12 landing site, in Mare Cognitum. The geology of this region is radically different from the Apollo 17 and it should serve as a perfect complement to our work so far. Elsewhere my colleagues will discuss and introduce the region in more detail, including ulterior scientific reasons behind the choice of this landing site.

We shall keep you informed of all further developments and new projects, and, once again, thanks for your patient and enthusiastic contribution to planetary science!


Michael G.G., Neukum G., Planetary surface dating from crater size-frequency distribution measurements: Partial resurfacing events and statistical age uncertainty, Earth and Planetary Science Letters, 2010, DOI: 10.1016/j.epsl.2009.12.041.

Neukum G., Meteoritenbombardement und Datierung planetarer Oberfl�chen. Habilitation Dissertation for Faculty Membership, Univ. of Munich, 186pp, 1983.

Dr. Roberto Bugiolacchi
Moon Zoo science lead
Birkbeck, University of London
University College London (UCL)

December 15: Measuring the regolith thickness at the Apollo 17 site

By  Ian Crawford
(Department of Earth and Planetary Sciences, Birkbeck College)

 Estimating the thickness of the unconsolidated lunar regolith is one of the major scientific objectives of Moon Zoo. This is because understanding the thickness of the regolith in different regions of the Moon will address a number of important scientific questions. For example, as regolith thickness increases with time, measuring the regolith thickness in areas which have not been dated by returned samples will help provide additional surface age estimates. Conversely, measuring the regolith thickness on surfaces with well-determined ages (such as the Apollo landing sites) will help us determine the regolith accumulation rate. Improved global regolith thickness maps will also provide important information for future exploration of the Moon, including the quest to identify future lunar resources.

There are three ways in which studies of small craters can be used to estimate regolith thickness. The first is to determine the minimum size of craters which have excavated blocks of bedrock (i.e. boulders) from below the regolith layer (Fig. 1).  If the crater dimensions are known, then an estimate of a maximum depth of excavation can be estimated as about one-tenth of the diameter.

Figure 1. LROC image of a boulder-covered bench crater. The crater has formed in a basaltic regolith close to the Apollo 12 landing site. The impact has punched through the thin regolith cover and into the harder rock, excavating large blocks that have covered the surrounding surface. This example is 130m in diameter, so the regolith here must be less than about 13m deep. By determining the maximum size of craters in this area which have not excavated boulders the actual depth of the local regolith can be determined. (LROC image M114104917L/ASU/NASA).

The second method relies on identifying flat floors or benches within a crater, which also indicates that a crater has penetrated an overlying regolith layer to a stronger layer beneath. Figure 1 again provides an example. For features like this a simple expression has been derived which estimates the regolith thickness from the ratio of the bench diameter to the overall crater diameter. For the example shown in Figure 1 this indicates a regolith depth of about 6 m, consistent with the upper-limit of 13m estimated from the presence of boulders around the rim.

The third method is more subtle, and exploits the process of impact gardening, whereby rocky surfaces are disaggregated and overturned by meteorite impacts, thus destroying the record of previous impact cratering events. The equilibrium diameter is identified when the cumulative number of craters seen on the surface is less than the number actually produced, and can be recognized as a change in slope in a graph which plots number of craters in a given area as a function of their size. Because the number of craters buried under new regolith depends on the regolith thickness, measuring the equilibrium diameter gives a guide to the latter.

In order to test these different methods it is necessary to apply them to areas where the regolith thickness has been directly measured. However, this can only be done at the small number of Apollo landing sites where seismic measurements of regolith thickness were conducted. By far the best estimates have been provided by the Apollo 17 Lunar Seismic Profiling Experiment (LSPE). For this experiment the astronauts deployed eight small explosive packages during their traverses around the Taurus-Littrow Valley (Fig. 2) which, when detonated, provided seismic signals for detectors setup close to the Lunar Module.

Figure. 2. One of eight explosive packages deployed by the Apollo 17 astronauts to provide data for the lunar seismic profiling experiment which measured the thickness of regolith in the Taurus-Littrow Valley. The Apollo 17 LRV is in the foreground and the lunar module, where a geophone detector array was deployed to collect the signals, in the middle distance about 300 m away (NASA)

By measuring the time taken for the seismic signals to travel from the explosive packages to the detector, geophysicists were able to determine the thickness of both the regolith layer and the underlying lava flows at the Apollo 17 landing site. The results are shown in Fig. 3.

Figure. 3. Subsurface structure under the Taurus-Littrow Valley, as determined by the Apollo 17 seismic profiling experiment. The numbers indicate seismic wave speed in meters per second. Yellow represents the lunar crust, which outcrops locally as the South Massif (“LM impact” schematically indicates where the Apollo 17 Lunar Module ascent stage was crashed into the South Massif to provide an additional seismic data point). The green layers indicate the thickness of basaltic lava that has flooded the valley to a depth of about 1.4 km. The thick black line shows the regolith layers (inset). (Image adapted from a paper by M.R. Cooper et al., published in Reviews of Geophysics and Space Physics, Vol. 12, pp. 291 – 308, 1974).

Five separate layers were identified below the surface of the Taurus-Littrow valley:

(i)  The topmost layer, 4 m deep with the very low seismic wave speed of 100 m/s, is interpreted as being due to the local regolith.

(ii)  Beneath the regolith is a layer with a velocity of 327 m/s, which is still too low for solid rock. It may be due to more consolidated regolith, or possible highly fractured lava.

(iii)  At a depth of 32 m the velocity rises to 495 m/s, and this is interpreted to be the fractured and/or vesicular top of the lava flow filling the valley.

(iv)  At a depth of 390 m the velocity rises to 960 m/s. This is interpreted as being due to a more coherent basalt unit.

(v)  Finally, at a depth of 1.4 km the velocity rises sharply to 4.7 km/s, and this is interpreted as being due to crustal bedrock underlying the lava layers.

The deeper layers are too deep to be probed by craters found in the MoonZoo images, although the presence of a lava layer at a depth of about 30m is consistent with the excavation of basaltic blocks from 300-400 m diameter craters in the valley floor. Where MoonZoo can really help is to confirm that the seismic boundary at a depth of 4m (which will be probed by craters about 40 m across), and to determine whether the underlying layer is more consistent with fractured basalt or compact regolith.

In order to address these issues, we need MoonZoo users to look carefully at craters in the images of the Apollo 17 area, determine their sizes accurately, and note the presence of boulders around the rims and/or interior benches or flat floors. Don’t worry that scales are not provided on the MoonZoo images (this is deliberate to avoid the possibility of biasing the results), but users may be sure that the sizes and morphologies of all thecraters in these images are relevant to the task in hand.

 Ian Crawford is based in the Department of Earth and Planetary Sciences, Birkbeck College, London, and is a member of the MoonZoo science team. This blog article is based on a longer article published in the December 2012 issue of the Royal Astronomical Society journal Astronomy and Geophysics.


The Scientific Legacy of Apollo

By  Ian Crawford
(Department of Earth and Planetary Sciences,
Birkbeck College, London)

Fig. 1. One of the last two men on the Moon: Harrison Schmitt stands next to a large boulder at the Apollo 17 landing site in December 1972. (NASA).

This December marks 40 years since the last human beings to set foot on the Moon, Gene Cernan and Harrison “Jack” Schmitt of Apollo 17, left the lunar surface and returned safely to Earth. In the three and a half years between Neil Armstrong’s ‘first small step’ in July 1969 and the departure of Cernan and Schmitt from the Taurus-Littrow Valley in December 1972, a total of twelve astronauts explored the lunar surface in the immediate vicinity of six Apollo landing sites.

Fig 2. The Apollo landing sites. Note their restriction to the central part of the nearside – there is a lot more of the Moon to explore! (Image: NASA).

The total cumulative time spent on the lunar surface was 12.5 days, with just 3.4 days spent performing extravehicular activities (EVAs) outside the lunar modules. Yet during this all-too-brief a time samples were collected, measurements made, and instruments deployed which have revolutionised lunar and planetary science and which continue to have a major scientific impact today.

Fig. 3. A view across the Apollo 17 landing site in the Taurus-Littrow Valley. The Apollo 17 Lunar Roving Vehicle is in the foreground, and the Lunar Module is in the middle distance about 300 m away. The black box in the foreground is one of eight explosive packages deployed to provide data for the lunar seismic profiling experiment which measured the thickness of regolith and the underlying lava in the Taurus-Littrow Valley (NASA).

In their cumulative 12.5 days on the lunar surface, the twelve Apollo moonwalkers traversed a total distance of 95.5 km from their landing sites (heavily weighted to the last three missions that were equipped with the Lunar Roving Vehicle), collected and returned to Earth 382 kg of rock and soil samples, drilled three geological sample cores to depths greater than 2 m, obtained over 6000 surface images, and deployed over 2100 kg of scientific equipment.

Fig 4. Jim Irwin next to the Apollo 15 LRV with the 4.6 km high Mt Hadley in the background; note the sample bags attached to the rear of the LRV (NASA).

These surface experiments were supplemented by wide-ranging remote-sensing observations conducted from the orbiting Command/Service Modules.

Fig. 5.The Scientific Instrumentation Module (SIM) bay of the Apollo 15 Command/Service Module (CSM). On Apollo 15 the SIM included mapping cameras, a laser altimeter, and ultraviolet, X-ray and gamma-ray spectrometers (NASA).

 Probably the greatest scientific legacy of Apollo has resulted from analysis of the 382 kg of rock and soil samples returned to Earth. One of the key results has been the calibration of the lunar cratering rate. Only by comparing the density of impact craters on surfaces whose ages have been obtained independently by laboratory analyses of returned samples is it possible to determine the rate at which meteorite impacts have created craters on a planetary surface. Analysis of the Apollo samples (supplemented by those obtained by the Soviet Union’s three Luna robotic sample missions) has enabled this to be done for the Moon, which remains the only planetary body for which such a data-set exists. Not only has this facilitated the dating of lunar surfaces from which samples have yet to be obtained, but it is used, with various assumptions, to estimate the ages of cratered surfaces throughout the Solar System from Mercury to the moons of the outer planets.

Another important result of Apollo sample analysis by seo services uk has been the evidence provided for the origin of the Moon. In particular, the discovery that lunar materials have compositions broadly similar to those of Earth’s mantle, but that the Moon is highly depleted in volatiles compared to the Earth and has only a small iron core, led to the current view that the Moon formed from debris resulting from a giant impact of a Mars-sized planetesimal with the early Earth. It is very doubtful that we would have sufficient geochemical evidence usefully to constrain theories of lunar origins without the quantity and diversity of samples provided by Apollo, and indeed these samples are still being actively exploited for this purpose.

Fig. 6. The current theory of the Moon’s formation from debris produced by a giant impact on the early Earth is largely based on the geochemical analysis of samples collected by the Apollo missions (image: Wikipedia Commons).

Beyond this, the Apollo samples have been vital to our understanding of the Moon’s own geological history and evolution. While lunar geology may at first sight appear to be a relatively parochial area of planetary science, it is important to realise that the Moon’s surface and interior retain records of planetary processes which will have occurred in the early histories of all the terrestrial planets, such as the formation of cores and crusts. In all these respects the Moon acts as a keystone for understanding the geological evolution of all the rocky planets.

Fig. 7. Fragments of Apollo 12 soil sample 12023 at the Lunar Sample Laboratory at the Johnson Space Center, being selected for a study of lunar volcanism in 2009. Forty years after they were collected, Apollo samples like these are still being used for scientific investigations (photo: I.A. Crawford)

In addition, Apollo samples of the lunar regolith have demonstrated the importance of the lunar surface layers as an archive of material which has impacted the Moon throughout its history. These include records of solar wind and cosmic ray particles, and meteoritic fragments. Extracting meteoritic records from lunar regolith samples is especially important for planetary science as it potentially provides a means of determining how the flux and composition of asteroidal material in the inner Solar System has evolved with time.

Last, but not least, the Apollo samples have been used to calibrate remote sensing investigations of the lunar surface. The visible, infrared, X-ray and gamma-ray spectral mapping instruments carried by a host of recent orbital missions to the Moon have produced a wealth of information regarding the chemical and mineralogical nature of the lunar surface. Although these orbital missions post-date Apollo, the reliability of their results largely depends on their calibration against known compositions at the Apollo landing sites. Without the ‘ground truth’ provided by the Apollo samples, it would be difficult to have as much confidence in the results of these remote sensing measurements as we do.

 In addition to study of the Apollo samples, many other areas of scientific investigation were also performed by the Apollo missions, especially geophysical investigations of the Moon’s interior. Key results included the discovery of natural moonquakes and using them to probe the structure of the crust and mantle, geophysical constraints on the existence and physical state of the lunar core, and measurements of the flow of heat from the Moon’s interior. Although these data are over thirty years old, advances in interpretation means that they continue to give new insights into the interior structure of the Moon. For example, only last year an apparently definitive seismic detection of the Moon’s core, and strong evidence that, like the Earth’s, it consists of solid inner and liquid outer layers, was made by a re-examination of Apollo seismic data.

Fig. 8. Apollo 14 seismometer deployed on the lunar surface; the silvery skirt provided thermal stability. These instruments, also deployed at the Apollo 12, 15 and 16 landing sites, constituted the Apollo passive seismic network which remained active until 1978 and yielded valuable data about the interior of the Moon (NASA).

Looking over the totality of the Apollo legacy, I think one could reasonably make the case that Apollo laid the foundations for modern planetary science, certainly as it relates to the origin and evolution of the terrestrial planets. Arguably, the calibration of the lunar cratering rate, and its subsequent extrapolation to estimating surface ages throughout the Solar System, could alone justify this assertion. If one also considers the improvements to our knowledge of lunar origins and evolution, and the records of solar wind, cosmic rays and meteoritic debris extracted from lunar soils, it is clear that our knowledge of the Solar System would be greatly impoverished had the Apollo missions not taken place.

 However, it is also clear that Apollo did little more than scratch the surface, both literally and figuratively, of the lunar geological record. With only six landing sites, all at low latitudes on the nearside, it is clear that much remains to be explored. Therefore, as we pass the 40th anniversary of the last human expedition to the Moon, there are good scientific reasons to start planning for a return.

Fig. 9. Artist’s concept of astronauts supervising a drill on the Moon. Returning humans to the lunar surface later in the 21st Century would facilitate larger scale exploration activities than was possible with Apollo, and will further increase our knowledge of lunar and Solar System evolution (artwork: NASA).

Ian Crawford is based in the Department of Earth and Planetary Sciences, Birkbeck College, London, and is a member of the MoonZoo science team. This blog article is based on a longer article published in the December 2012 issue of the Royal Astronomical Society journal Astronomy and Geophysics.

Computers counting craters

By Anthony Milbourne (Birkbeck College)

In this blog I would like to talk about automated crater counting, which as the name suggests, is the use of computers to identify and count craters.  Computers are very good (and getting better) at many things, but they are still not as good as humans at many important classes of problems.  The human brain is amazingly powerful (you knew that) and the latest generation of super-computers have only just reached the same order of magnitude of computing power.  In other words, there are currently only a handful of machines in the world that can truly rival the human brain in terms of raw processing-power.  Humans are particularly good at pattern recognition and computer scientists have been trying to create computer programs to do this for a long time, but the results generally look rather pathetic compared to humans.  One application of pattern recognition is crater identification, which I will talk about below.

In general there are two approaches to pattern recognition: designed algorithms and machine learning.

A designed algorithm is a set of very specific instructions that allow the computer to solve a specific problem.  It would be like programming a car to drive from point A to point B by defining exactly how much to accelerate or brake at what points and exactly how much to turn the wheel when.  If you know the car will always be running on the same road then this approach is reliable and predictable, but it’s not very flexible.  If a pedestrian steps into the road they are in trouble because the car will take no account of the changed environment, and if the car has to go on a different road it will be close to useless.

In machine learning the computer is still given detailed instructions, not on how to solve a specific task, but on how to learn what works well to solve the task.  This would be like giving the car-driving program instructions on what the brake, accelerator and steering wheel did and then letting it experiment until it found a route from A to B.  Obviously, this requires a training phase, where the algorithm crashes a lot (let’s hope the car wasn’t expensive), but eventually it figures out some general principles about driving and is able to deal with a certain amount of change in its environment.

Designed algorithms are safe and predictable; they don’t need training and are often easier to implement and faster to run, but they are inflexible.  Machine learning may be better at the job and will certainly be more flexible, but it is tricky to train and you may end up training it to take short-cuts that you didn’t want:

CHT circles

CHT circles: The points (in red) on the circle (in black) each create a ring of votes (in blue) around themselves. Where vote rings overlap the votes combine (more intense blue) and the greatest vote value indicates the centre of the circle. The radius of the vote rings is determined by the radius of the circle being searched for.

An example of a commonly used designed algorithm is the Circular Hough (pronounced like rough) Transform (CHT).  We assume that the image is taken from above and that the vast majority of craters will be roughly circular.  The program then uses a CHT to look for circular patterns of a set radius (perhaps repeating many times for different radii).  A CHT essentially takes each image pixel that represents an edge and uses it to generate a vote for all possible circles that the edge could be on.  The centres of all these possible image circles form a circle of votes around the edge point.  If you do this for every edge point then the votes tend to build up at points that really are the centres of circles (it’s easier to see in the diagram).

Of course, this is much harder in practice as the images have a lot of noise and other features that generally confuse the algorithm, so various people have come up with various ways of improving it and making it faster.  There are of course many other types of designed algorithm, but I won’t bore you with all of them.

An example of a machine learning approach is to use a neural network.  This is a program that tries to simulate a simplified model of how the brain works.  It consists of a large number of ‘nodes’ which are connected to each other by links of varying strength.  The nodes are normally arranged in layers, and each node combines the inputs from nodes in the preceding layer and sends the result out to the nodes in the next layer.  The nodes in the first layer act as input points and the nodes in the last layer as the output.  The strength of the links can be varied in order to change the behaviour of the network, and this is done during training when the error from each training run is used to adjust the link strengths.

Simplified neural network

This is a very simple (too simple to be useful) example of a neural network. A set of values are passed to the input layer (in green) and an output is generated by the output layer (in purple). What happens in between is determined by the connection strengths, which are the result of training.

Neural networks are deceptively simple in concept but are very powerful and can end up spotting trends that are not clear to humans, or that are too complex or nuanced to implement easily as a designed algorithm.  However the number of nodes needed to achieve anything useful is normally large, so figuring out what is actually happening inside the network is not practical.  For this reason you can never be quite sure that the network, given new or unexpected data, won’t do something crazy!

Again, this is not the only way of implementing machine learning, but it gives an idea of the way this sort of system works – trained rather than designed.

In general, machine learning approaches take more processing power than designed algorithms, so in most cases a pipeline is used.  First, a quicker, more predictable, designed algorithm is used to select areas of interest (potential craters), and then a machine learning approach is used to sort the real craters from the noise.

Hough Transform, on an HRSC image of Mars

Hough Transform, on an LROC image of the Moon (TOP) The result of running an algorithm, based on the Hough Transform, on an HRSC image of Mars. The smooth terrain and crisp crater rims produce fairly good results, although there are still a few errors, some of which are glaringly obvious to a human. (Image: modified from HRSC/ESA). (BOTTOM) The result of running the same algorithm, on a tile from the Moon Zoo database. The degraded rims and noisy background confuse the algorithm which finds lots of craters, but almost all are in the wrong place! (Image: modified from NASA/GSFC/ASU)

The images at left are an example of the kind of output that can be achieved by automated crater recognition (this one is based on Hough Transforms), and the problems with it.  This is far from the best algorithm available, and other researchers have developed much more accurate programs, but they all suffer to a greater or lesser degree from image noise.  The first image shows how accurate an algorithm can be in a clear image with little noise.  It misses many smaller craters and there are a few false positives (which are somewhat surprising to a human eye), but in general it finds the rims of the most obvious craters very accurately.  The second image features degraded crater rims, a lumpy surface and sub-optimal illumination.  The result is that the same algorithm does very badly at spotting craters.  This is not surprising; even a human would have to look harder at the second image, but the algorithm performs so badly that it is arguably not worth using, and this is the sort of image where humans really are the only (reliable) show in town at the moment.You might think that automated crater counting would be a direct competitor to crowd sourcing efforts like Moon Zoo, and in some cases you would be correct, but it can also be used as a complementary technique.  This could be done by using moon Zoo crater identifications as the areas of interest and then running an algorithm to find the exact location of the crater rim, or using an algorithm to spot Moon Zoo data which has been entered by mistake, or by users who didn’t understand the task.

Most interestingly, in my view, is the idea that algorithms are just another type of user.  Some algorithms are not great at spotting craters, but some human users are a bit variable too.  Admittedly, the best humans are far better than the best algorithms, but the best algorithms are probably better than the worst humans, so they fall within the quality spectrum that Moon Zoo (and the rest of the Zooniverse) already deals with.  They probably won’t be much good at spotting unusual objects and they certainly won’t be much fun on the forums, but perhaps we might one day be working with algorithms as our (less able) peers.

By Anthony Milbourne (Birkbeck College)

Notched Cavities or INA-like?

User placidstorm posted some Ina-like formations in the Dark Halo Craters thread in December 2011.

I think these look more like “notched cavities” than INA-like formations – see end of posting for references to these features.
These cavities appear to be associated with lava which covers a lava tube or rille, part of which later collapses leaving depressions. I don’t know if that’s what has happened here but this whole region is full of interesting features.

The LPOD Lunar Photo of the Day for 10 August 2007 mentions this area:

Quote from: lpod

A feature, previously unknown to me, is the degraded linear rille segment that extends westward from between Arago Beta and Manners. A similar short but more subtle rille is nearly perpendicular to the Sosigenes Rilles between Beta and Sosigenes A. These two rilles must relate to structures that are now covered by the Tranquillitatis lavas – perhaps whatever is under Lamont. Linear collapse troughs just north of Sosigenes A is evidence for a buried lava tube, another feature from the past of this mare area.

An overview of the features that placidstorm found. This area is in Mare Tranquillitatis north of Sosigenes A crater at approximate coordinates, latitude = 8.1 longitude = 19.0.
Left image: M177508146LE  Right image: M177508146RE

Closeup of one of the “bays”.

The image below gives some context to this feature – the area inside the orange oval is the approximate site of the images above. It shows that there is a rille running underneath the feature and this could help explain how the “notched cavities” were formed. The “feature” crossing the rille is a secondary crater chain.
The Lunar Networks article referenced below the image is well worth a read but does contain a mistake where it refers to “linear rille system in the northeast Mare Tranquillitatis” it should say “southwest”.

Full two kilometer width segment of LROC NAC frame M146858595LE as shown on Lunar Networks 16 Sep 2011
[NASA/GSFC/Arizona State University]

More information about INA features will be found here: TLP Project – INA-like Features and here: INA Images
“Notched Cavities” are discussed here: TLP Project – Notched Cavities in Lava

Strange Whites

Forum member Ewan, otherwise known as Dynamo Duck, found some strange looking craters on the South-East edge of Mare Crisium at Latitude: 12.275° Longitude: 62.1034°. They appear to have unusually high albedo floors and possibly some ejecta of the same high albedo material.

Searching around the area I found some more of these “strange whites.”

Here are two.
And closer…

And then they were everywhere.

At the moment they remain a puzzle. They look like bench/concentric craters but why are the floors so “white”? What is this high albedo material and why are there so many of these craters in this region of Mare Crisium?

While we work out the answers here are some NACs for you to peruse showing the “strange whites” under different illumination.


Hadley Rille

Continuing the rille theme from last week, Hadley Rille has been a bit of a forum feature this week. Forum regular kodemunkey sent me a interesting couple of NAC (Narrow Angle Camera) images he had come across while exploring a “wandering rille”. Here they are:

 .. M144612571RE

He thought he might have spotted a volcanic vent but wasn’t sure. This turned out to be the start of the Hadley Rille-Apennine region shown here on the ACT-REACT Quick Map.

We can now check the topography of any feature we see using the new ACT-REACT line tool to produce a map of surface elevation. The “vent-like” trench kodemunkey spotted on the  NAC images is Běla, an elongated crater thought to be either a collapsed magma chamber or a volcanic vent. Hadley Rille itself is thought to be a either a volcanic vent or a collapsed lava tube. There is more about the volcanic nature of the region in this LROC article Layers near Apollo 15 landing site. Using the ACT-REACT tool produced the following plot of Běla revealing a vent or tube -like “V” shaped dip in the terrain:

Newcomer to the forum Dynamo Duck asked if we could view the tracks around Hadley Rille from the Apollo 15 mission. Here is the location of the Apollo 15 landing site and a map of the various routes and tracks. Click on the images to enlarge.

From LROC article “Follow the Tracks”

The latest batch of NACs taken from a lower orbit, 25-30 kilometres above the surface, show these tracks – but you need a keen eye. Here’s what you are looking for:

NASA/GSFC/Arizona State University

The new NACS are now available to examine! Enjoy.


Jules is a volunteer moderator on the Moon Zoo forum

Exploding Boulders!

We are used to seeing boulder tracks on Moon Zoo and often come across (or actively go hunting for!) the boulder that caused them. Usually we find something like these large intact boulders having come to rest at the end of their tracks.

highlighted by placidstorm and kodemunkey

Moon Zoo team member Dr Anthony Cook recently sent me this picture of two boulder tracks in Schiller crater:

In this case the boulders are far from intact and appear to have “exploded” at the end of their journeys. What might have caused these boulders to fracture and fragment? One theory Tony suggested was that due to being under tension the boulders might have fractured before they rolled down the slope and that the movement further weakened them. Then over time the extreme temperature variations between lunar day and night could have fragmented the weakened rocks resulting in the appearance we see in the image.

Tony said:


I’m a bit puzzled though why the one on the top left has rock debris so far away from the centre. The boulder that looks like a skull rock on the bottom right has debris a lot closer to it, that could simply be explained by bits falling off as one would expect from the explanation above.

An alternative theory is that the boulders did roll down the hill intact, but were of sufficient size, area and age to be impacted by later meteorites, and these high velocity impacts split the rocks into many pieces. However, as Tony points out, the chances of this happening to two large rocks next to each other seem a bit remote.

Here is the NAC image M109502471L and the LROC article “A Recent Journey.”

In order to study this process in more detail we need more examples. So if you find any exploded (or partly exploded) boulders please post them on the forum here.

A new void in the melt?

One of the many types of features we are looking out for on Moon Zoo are the Lava tube skylights – ceiling collapses in lava tubes in regions which have been subjected to lunar volcanism.

Marius Hills Lava Tube Sky Light –65 metres wide
~ Mare Ingenii Tube Sky Light –130 metres wide

These pits or caves would provide ready made shelters for any future manned missions. There’s more information in these LROC news articles: Marius Hills Pit – Lava Tube Skylight? and How Common are Mare Pit Craters?.

Forum member JFincannon started discovering similar looking features in non-volcanic regions and called them “collapsed voids.” These appear to be holes in impact melt possibly as a result of the melt cooling rapidly and cracking. However the regular round morphology of these “voids” still has us puzzled. There are more details in JFincannon’s blog post  Potential Caves and Sink Holes in Copernicus Crater

Here are some of his Copernicus finds:


There was much discussion on the forum about how these voids might have formed but their origins are still unresolved. However, they are clearly not craters. They do not display typical crater morphology, having flat, boulder-filled bottoms and very distinctive sharp “rims” without the familiar crater walls and without any ejecta.

Impact melt can crack in spectacular ways as this LROC article featuring Giordano Bruno shows: Fragmented Impact Melt. JFincannon referred to another source describing collapsed features in melt ponds: Lunar Caves in Mare deposits imaged by the LROC Narrow Angle Cameras which stated that:

“Collapse features over probable lava tubes within mare (skylights) may provide points of ingress to larger “trunk” cave passages. Collapse features over areas of melt pond drainage suggest additional sublunarean voids. Both types of cave offer intriguing exploration and habitation opportunities.”

We thought JFincannon’s latest potential impact melt void was similarly intriguing and worth highlighting. It is much smaller than previous examples at around 24m diameter and is situated at latitude 72.468 : longitude -31.393 in Philolaus Crater. As JFincannon points out, this far north the Sun never gets higher than 18 degrees above the horizon. This means the bottom of any pits are unlikely to be illuminated making visual confirmation impossible. It also makes spotting them at all very challenging as shadows at this latitude are very long and black. However, this latest candidate does bear all the hallmarks of a potential void which JFincannon describes as:

“… a sharply dark area surrounded by a lighter, grayer one.  In these images, the crater-like feature has a steep enough inner slope to brighten this Sun facing side, while the darkness does not seem to be due to a hill or raised crater rim. Also, other craters around it seem shallower. So it could be a deep small crater or a pit.”

Here it is:

Other views can be found in M168399883RC and M170754606LC