Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 Newly acquired high resolution Lunar Reconnaissance Orbiter Camera (LROC) images allow accurate determination of the coordinates of Apollo hardware, sampling stations, and photographic viewpoints. In particular, the positions from where the Apollo 17 astronauts recorded panoramic image series, at the so-called “traverse stations”, were precisely determined for traverse path reconstruction. We analyzed observations made in Apollo surface photography as well as orthorectified orbital images (0.5 m/pixel) and Digital Terrain Models (DTMs) (1.5 m/pixel and 100 m/pixel) derived from LROC Narrow Angle Camera (NAC) and Wide Angle Camera (WAC) images. Key features captured in the Apollo panoramic sequences were identified in LROC NAC orthoimages. Angular directions of these features were measured in the panoramic images and fitted to the NAC orthoimage by applying least squares techniques. As a result, we obtained the surface panoramic camera positions to within 50 cm. At the same time, the camera orientations, North azimuth angles and distances to nearby features of interest were also determined. Here, initial results are shown for traverse station 1 (northwest of Steno Crater) as well as the Apollo Lunar Surface Experiment Package (ALSEP) area.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 During the nominal phase of the Lunar Reconnaissance Orbiter (LRO) mission the Lunar Reconnaissance Orbiter Camera (LROC) acquired images from a near-circular 50 ± 15 km polar orbit [Vondrak et al., 2010]. The LROC system consists of a Wide Angle Camera (WAC) and two identical Narrow Angle Cameras (NACs) and provides global (75 m/pixel) multispectral coverage as well as high-resolution (0.5 m/pixel) monochrome close-up views of the Moon, respectively [Robinson et al., 2010]. Stereo images with substantial overlap are acquired from adjacent orbits providing the means to derive Digital Terrain Models (DTMs) and orthoimages.
 These high-level topographic data products represent a detailed depiction of the Moon and are of great value to ongoing science and exploration analyses. Furthermore, the qualitative and quantitative progress in acquisition and processing of remote sensing data also allows reanalysis and interpretation of data returned from previous lunar missions, e.g. from Apollo landings. Returned rock samples, astronaut observations, in-situ measurements, and surface photography represent an invaluable and unique record of a non-terrestrial surface. Until now, however, our knowledge of the coordinates of sampling sites and instrument locations relies on estimates from surface images and astronaut descriptions.
 Accurately tying these unique surface photographs to modern high-resolution image maps and DTMs not only enables realistic, multisource and multidimensional data visualization, but also improves the geologic and cartographic context of the features captured in the historic “on site” imagery. Precise geometric reconstruction of the moment of image acquisition (camera position and orientation) is indispensable for accurate identification and mapping of sample sites, ALSEP components, surface features, and topography captured in Apollo images.
 This study focuses on the Apollo 17 landing site, which is located along the south-eastern rim of the Serenitatis basin. This last Apollo lunar mission is characterized by the longest traverse path (30 km), the most recorded images (>2,200), and the largest sample mass returned (nearly 120 kg) [Wolfe et al., 1981]. We used panoramic images taken by the astronauts during their three Extra-Vehicular Activities (EVAs), which then and now served as the prime resource for the determination of astronaut positions. Based on high-resolution 0.5 m/pixel LROC orthoimages we obtain precise selenocentric body-fixed coordinates of astronaut and surface feature positions. Like all LRO archival data these coordinates are given in the Mean Earth/Polar Axis (ME) Reference System, with the z-axis being the mean rotational pole and with the prime meridian (0° longitude) defined by the mean Earth direction [NASA, 2008].
2. Image Data Sets
2.1. LROC WAC DTM
 The LROC WAC is a multispectral camera with two separate optics for the visible (VIS) and ultraviolet (UV) spectrum, imaging onto different sections of the same CCD array (1,024 × 1,024 pixel). Two bandpass filters in the ultraviolet (321, 360 nm) and five filters in the visible (415, 566, 604, 643, 689 nm) subdivide the array into seven sub-frames of 16 lines (summed to 4) and 14 lines, respectively [Robinson et al., 2010].
 The WAC instantaneous field-of-view (IFOV) is 5.15 arcmin for the VIS optics. From 50 km orbit altitude this results in a ground resolution of about 75 m/pixel. When operated in the seven-band (color) imaging mode, only the central 704 pixels/line of the visible bands are read out. The ground track swath of a WAC color image is typically about 60 km wide and 300 km long, depending on the orbital beta angle. Within one month, the WAC provides near complete coverage of the Moon. From images taken under different lighting conditions a global 100 m/pixel basemap was derived, suitable for morphological studies [Speyerer et al., 2011].
 Additionally, a near-global 100 m/pixel DTM, called the Global Lunar DTM 100 m or GLD100, was derived from about 69,000 stereo models from one visible band (604 nm). High quality stereo observations (33° stereo angle and 52% overlap at the equator) are provided by images from adjacent orbits, as well as between images taken on consecutive months. GLD100's mean vertical accuracy was quantitatively compared with the Lunar Orbiter Laser Altimeter (LOLA) [Smith et al., 2011] global altimetric map; the vertical match between the two data sets is better than 20 m (for more details, see Scholten et al. ).
 In this study a 75 km × 75 km subset of the GLD100 covering the Taurus-Littrow Valley, including the hills and massifs surrounding the Apollo 17 landing site, was used for initial positioning purposes.
2.2. LROC NAC DTM and Orthoimages
 From 50 km orbit altitude LROC NAC images have a pixel scale of 0.5 m, which allows meter-scale assessment of regions explored by the astronauts. Both LROC monochrome CCD line-scan imagers, NAC-L and NAC-R, are aligned side-by-side with a small overlap to provide a wider field-of-view (FOV) of two by 2.85° in the cross-track direction. A combined NAC-L/R-mosaic covers a ground track swath of about 5 km and typically 26 km in length.
 To acquire stereo images from adjacent orbits, LRO is slewed up to 30° off nadir to either side in the across-track direction. For stereo image processing we used an LROC-adapted version of the German Aerospace Center (DLR) photogrammetric processing system, which has operationally been applied to Mars Express High-Resolution Stereo Camera (HRSC) data [Gwinner et al., 2009] and other stereo imagery. This system is characterized by area-based image matching within stereo model overlap, 3D forward ray intersection, and a final interpolation of a DTM grid. Two stereo pairs from different mission phases were used to generate high-resolution NAC DTMs of the Apollo 17 landing site, providing elevations with a relative accuracy of a few decimeter.
 LRO was initially placed in an elliptical orbit (45–190 km altitude) for a three months commissioning phase. From this orbit LROC obtained a 1.4 m scale stereo set (M104311715L/R, M104318871L/R) of the Apollo 17 landing site. This stereo pair (12° stereo angle, 80% overlap) was used to generate a 4.0 m raster DTM [Oberst et al., 2010], which covers the area traversed by the astronauts. The DTM ranges from 30.425°E to 30.910°E in longitude and from 18.762°N to 21.143°N in latitude, including parts of the North, East, and South Massifs.
 From the nominal mission phase, a 0.5 meter pixel scale stereo pair (M113751661L/R, M113758461L/R) providing a stereo angle of 35° was used to generate a 1.5 m/pixel raster DTM. This stereo model covers an area of 3.2 km × 3.2 km in the vicinity of the landing site, including the ALSEP area, station 1, and station 9 (see white outlines in Figure 1). To assess the relative accuracy of this DTM, ten, evenly distributed LOLA tracks (2,206 height observations) containing altimetric crossover solutions [Mazarico et al., 2011] were accurately registered to the DTM [Gläser et al., 2010]. It showed a standard deviation of the elevation differences of the LOLA profiles and the corresponding DTM elevations of ±40 cm on average, after a mean vertical offset of 37.7 m had been accounted for by shifting down the NAC DTM.
 Based on these DTMs, the NAC images of the landing site were orthorectified maintaining their original image ground scales. We chose (azimuthal) stereographic map projection, placing the origin at the approximate position of the astronaut, respectively, to preserve directions from the origin of the map.
 The ALSEP's radio transmitter, which is housed in the central station, served as the control point for accurate referencing of the orthoimages. The central station, which could be identified in the images, is one of the nine most accurately known positions on the Moon. Along with four laser ranging retro-reflectors (LRRR) and four other radio transmitters, it constitutes the ME-reference frame on the lunar surface. Their positions were determined from precise Earth-based measurements, i.e. Lunar Laser Ranging (LLR; ongoing experiment with increasing accuracies) and Very Long Baseline Interferometry (VLBI). VLBI observations on radio transmission from the five ALSEPs, monitored between 1972 and 1974, resulted in relative coordinates with uncertainties of about 10 m horizontally and about 30 m vertically [King et al., 1976]. Absolute ME-coordinates of the ALSEP transmitters were determined by tying the VLBI network to the LRRRs, which have absolute accuracies of about 1 m. So far, the most accurate estimate of ME-coordinates of the Apollo 17 ALSEP transmitter (20.19209°N, 30.76492°E) was given by Davies and Colvin .
2.3. Apollo 17 Surface Photography
 The Apollo 17 astronauts G. Cernan and H. Schmitt took more than 2,200 photographs to document and support their scientific investigations of the landing site and along traverses. The astronauts explored the valley with a battery-powered Lunar Roving Vehicle (LRV) enabling them to extend the range of their surface activities. They visited the bases of the South Massif (EVA 2) and the North Massif (EVA 3), covering a distance of about 30 km in total. Several sampling stops along their traverse path were included to collect lunar material to return to Earth.
 At nine pre-planned, major sampling stops at geologically interesting sites, the so-called “traverse stations”, the astronauts performed systematic documentary and panoramic photography [Wolfe et al., 1981].
2.3.1. Hasselblad Panoramas
 The Apollo 17 astronauts used two modified Hasselblad electric data cameras (60 mm focal length, 70 mm film) while performing EVAs. The cameras were typically attached to the chest of the astronauts' spacesuits while they were photographing. To enable photogrammetric measurements, these cameras were accurately calibrated and contained high-precision reseau plates, i.e. a glass plate situated immediately in front of the film plane with 25 measuring crosses etched at 10 mm intervals [Kammerer, 1973].
 The astronauts acquired single frames, partial panoramas, full panoramas, and stereo pairs along their traverse path and at sample stations to aid in post-flight geological analysis. Additional pictures were taken to illustrate deployment of the ALSEP and for other operational purposes [Batson et al., 1981]. A complete index list of all surface and orbital photographs taken during the Apollo 17 mission can be found in the Apollo 17 Photo Index [NASA, 1974]. Furthermore, many of the pictures taken on the lunar surface were digitized by the Johnson Space Center (JSC). Starting in 2004 they scanned the well preserved, original Hasselblad films at 4,096 × 4,096 pixels per image. At reduced image sizes of 2,340 × 2,350 pixels and along with supplemental information such as time of image acquisition stated as Ground Elapsed Time (GET; corrected for the delay in launch), these scans are available to the general public, e.g. from the Apollo Lunar Surface Journal (ALSJ) (http://www.hq.nasa.gov/alsj/frame.html).
 This study concentrates on full 360° Hasselblad panoramas, which were routinely recorded at each of the traverse stations and at the ALSEP area. The panorama sequences were achieved by slightly turning and changing the aiming direction of the camera each time a picture was taken, providing a 360° view from a single point. (Comment: Because the camera was attached to the front of the space suit the focal point is assumed to have moved in circle rather than about a point.) A complete panorama mosaic consists of at least 15 overlapping individual frames. Assembled panoramas are available from the ALSJ or the Apollo Image Atlas (http://www.lpi.usra.edu/resources/apollopanoramas/).
2.3.2. Apollo Traverses Maps From Apollo Era
 Up to the present-day the most widely known reconstruction of the astronaut traverse path is the Apollo 17 Traverses Lunar Photomap (scale 1:25,000) published by the Defense Mapping Agency Topographic Center (DMATC) in 1975. Based on the Hasselblad panoramas taken at the traverse stations the traditional surveying method of three-point resection was used to locate the camera positions on a photograph from the orbit of the Apollo capsule. Therefore azimuth angles of three or more surface features, which were identified in the panorama as well as on the vertical photograph, were measured in the panorama and plotted on tracing paper as lines radiating from one point. The tracing paper was then placed over the image map so that each ray intersected the appropriate feature. The point from which the lines radiated was identified as a panorama station [Batson et al., 1981].
 All 9 traverse stations were mapped using this method, the accuracy was estimated to be better than 10 m. Detailed data from Very Long Baseline Interferometry (VLBI) provided by Salzberg  was used to fill the gaps between the astronauts' stops, i.e. to map the tracks of the LRV between the previously located traverse stations.
 The lateral accuracy of this historic image map is relatively low, as the photograph from Apollo orbit had not been orthorectified. For example, a comparison of the coordinates of the Lunar Module (LM) derived from the traverse map to the ones given by Davies and Colvin  reveals an error of 59 arcsec in latitude and 48 arcsec in longitude. This corresponds to a mean positional error of the LM of about 625 m.
3. Precise Positioning of Apollo 17 Panorama Stations
 Our improved positioning of the Apollo astronaut locations was based on measurements from the Hasselblad panoramas in conjunction with LROC NAC orthoimages. Similar to Apollo era analyses, as described in section 2.3.2, the panoramas provided direction angles to three or more lunar landmarks, such as large boulders, crater rims, or anthropogenic objects.
 Then, the same features were identified in LROC NAC orthoimages, providing their positions with respect to the ME-reference frame. Feature identification was facilitated by superimposing contour lines on the orthoimage. It helped to isolate the appropriate features correctly by limiting the number of potential matches to the ones realistically visible from the astronaut's point of view.
 By a weighted least squares adjustment the observed network of angular directions was optimally fitted to the identified features (reference points) in the LROC NAC orthoimage. As a result, we obtain the most probable location of the astronaut while taking the images, as well as a statistical assessment of the accuracy, that was achieved. The functional model of this resection adjustment is given by the fundamental direction observation equation (1).
angular direction from the astronaut's position I to a surface feature J (observation)
residual in the observed angle
North azimuth of a surface feature J
orientation angle toward North (unknown)
xi, yi .
coordinates of the astronaut's position I (unknown)
xj, yj .
coordinates of a surface feature J (treated as unknown quantities/observations)
 Every single angular direction Dij measured in a panorama yields one observation equation, setting up an over-determined, nonlinear equation system. Furthermore, the coordinates of the reference points were treated as unknown quantities and were integrated as additional observations. This enabled the detection of possible gross errors and provided corrections and accuracies for all coordinates of the network. For the definition of the datum we used all of the reference points, placing no external constraints on the measured angles (free network adjustment). To apply a weighted least squares adjustment we assigned standard deviations to the angular measurements (±0.4°) and the orthoimage coordinates (±1 m).
Equation (1) was linearized and solved using a first-order Taylor series approximation, which required initial approximations for the unknowns (astronaut position, orientation angle, and reference point positions). Two different methods, described in section 3.2, were used to acquire these initial values.
3.1. Analysis of Hasselblad Panoramas
 We used assembled panoramas from the ALSJ, which displayed views of slightly more than 360°. This had the beneficial effect that some features were displayed twice at the vertical image borders. By setting the horizontal distance (number of pixels) between those features to 360° we accurately determined the horizontal IFOV for each panorama we used. This allowed us to derive angles from the panorama.
 In the case of geometric inconsistencies in these panoramas, e.g. duplications within adjacent pixels, or if the available panorama proved to be too distorted, we used the original single frames. Knowledge of the camera focal length (60 mm) in conjunction with the nominal reseau positions in the digitized images, we determined the horizontal resolution and the FOV of each frame; the average scale was 1,144 pixels per inch and a FOV of 46.8° (≈ 0.02°/pixel).
 Angles measured within the overlapping areas of the single frames showed small differences, up to 1°–2°. These differences are caused by the camera's attachment to the astronauts' chest, causing small changes in perspective from image-to-image. These discrepancies were adjusted within the least squares adjustment, incorporating all measured angles.
3.2. Initial Camera Positioning
 The determination of an astronaut position using least squares techniques requires adequate initial values for the unknowns: position and the angular orientation toward North. Regarding values for initial orientation, ephemeris data and time of acquisition were used. By means of the solar azimuth angle measured from the Hasselblad panorama, the complete network of angles was pre-oriented toward North.
 Initial coordinates for the unknown camera position were determined in two steps by visual alignment. As a rough, first approximation, we took advantage of the historic traverse map by semi-transparently overlaying it onto an LROC NAC orthoimage. The traverse map layer was then distorted to match projection and size of the orthoimage (see Figure 1). The locations of the depicted traverse stations were used as a first approximation for the camera position, which was further refined in a second step. Depending on the different stations, this coarse approximation came as close as 20–40 m to our final, least squares adjusted results.
 For surface images with no position information we followed a different initial control method. Directions to targets on the horizon, e.g. to local peaks and the summits of adjacent mountains, as well as to the Sun were measured in an assembled 360°-panorama taken at traverse station 1 (see Figure 2). This network of angles was plotted on a cartographic reconstruction of the subset of the 100 m raster DTM GLD100. To facilitate the identification and positioning procedure, heights were represented by colors and contour lines to clearly display the local topography. Additionally, the positions of the summits were derived from the DTM (see Table 1) and labeled with crosses. Visually fitting the network of angles to the GLD100 model provided a first, 100-m-scale approximation for the location of traverse station 1 (see Figure 3).
Table 1. GLD100 Derived Positions of (Selected) Summits of the Apollo 17 Site
Elevations are referred to a zero vertical datum of the mean lunar radius of 1,737.4 km [Archinal et al., 2011].
Family Mountain (fm)
Peak E-n (pEn)
Hill F (hF)
North Massif (nm)
East Massif (em)
Bear Mountain (bm)
South Massif (sm)
 The initial, first order camera positions were further refined by ties to the LROC NAC images. The observed directions to features in the vicinity were plotted onto the LROC NAC orthoimage using the approximate values for position and North azimuth. The network of angles was then “moved and turned” on the orthoimage until its location, by visual inspection, best matched the appropriate features. The coordinates of the origin of the network of directions served as starting values for the least squares adjustment.
4.1. ALSEP Area
 The precise positions of two panorama stations near the ALSEP were determined. Both pans were recorded by H. Schmitt during EVA 1 (12 December 1972), capturing the ALSEP instruments, Geophone Rock, the LM, and other landmarks, each from a different perspective.
 Standing between the Geophone Rock and the Geophone 3 instrument (part of the seismic sounding experiment), the first color panorama ‘Geo 3 Pan’ was taken at about 03:33 UTC (120:40 GET) consisting of frames AS17-147-22544 to -22562 [Wolfe et al., 1981]. An assembled version of ‘Geo 3 Pan’ from ALSJ was used to measure angular directions to 10 features, such as the Geophone Rock, a crater rim, a rock next to the Geophone 3 instrument, and six other rocks (see Figure 4; for the use of acronyms, see Table 2). The panorama showed distortions of a few degrees right of Geophone Rock. Therefore, single frames were used to measure relative angles within that area. Plotting the network of measured angles on the LROC NAC orthoimage, immediately showed high consistency between the observed angles and the appropriate features in the orthoimage (see Figure 5). Hence, a subsequent least squares adjustment only improved the initial camera location in the order of a few decimeters.
Table 2. Used Acronyms
rock next to Geophone 3 (ALSEP instrument)
central station (ALSEP instrument)
right crater rim
Radioisotope Thermoelectric Generator
a boulder named after Hasselblad frame #20773
 We obtained ME-coordinates of the camera location of ‘Geo 3 Pan’ with a mean point error of 30 cm (relative), which is within the pixel size (see Table 3). At that spot, heavily disturbed regolith can be seen in the orthoimage revealing activity of the astronaut.
Table 3. ME-Coordinates of Apollo 17 Panorama Stations
Relative Accuracy (m)
Geo 3 Pan
B&W ALSEP Pan
Station 1 B&W Pan
 A black and white panorama, the so-called ‘B&W ALSEP Pan’, was taken about a quarter of an hour later at 03:49 UTC (120:56 GET), after changing the film magazines and walking 54 m in northwest direction crossing the ALSEP area. It consists of frames AS17-136-20683 to −20710. The available panorama mosaics of the ‘B&W ALSEP Pan’ had large geometric errors, so three single frames (136–20701, −20703, −20704) were used instead (see Figure 6). From the combined least squares adjustment calculation, including13 measured angles to 8 different features, e.g. ALSEP's central station and Radioisotope Thermoelectric Generator (RTG) (see Figure 7; one target was omitted for better viewing), we assessed the relative accuracy of the adjusted camera location to be better than 40 cm. The ME-coordinates of both ALSEP panoramas, ‘Geo 3 Pan’ and ‘B&W ALSEP Pan’, are listed in Table 3.
4.2. Traverse Station 1
 After deployment of the ALSEP instruments, the crew drove about 1.2 km in southeast direction to their first major sampling station. They stopped at the so-called “Station 1 Crater”, a small crater 12 m in diameter, which is located about 185 m from the northwest rim of Steno Crater. At 05:26 UTC (122:33 GET), standing between the LRV and the southern rim of Station 1 Crater, H. Schmitt obtained the required station panorama (‘Station 1 B&W Pan’), which included 33 frames: AS17-136-20744 to −20776.
 Four single frames (136–20752, −20754, −20772, −20774) were used to derive angular directions to 19 features such as large-size boulders, crater rims, and the LM (see Figure 8; for better presentation only selected features are included). The least squares adjusted position, located 10 m from the southern rim of Station 1 Crater, has a relative accuracy of 0.5 m (see Table 3).
4.3. Mapping of Apollo 17 Landing Site Area
 Determining the exact position and orientation of the camera allows measurement of precise coordinates of any feature within surface panorama – as long as it is resolved in a high-resolution orthoimage. Using LROC NAC images from the nominal 50 km orbit, object matching is limited to features comparable in size (or larger) to the ground pixel scale of 0.5 m. Prominent objects captured in the surface photographs were located in the orthoimages and their identity was verified within a least squares fit. We determined object positions relative to the camera location by North azimuth and distance as well as their absolute ME-coordinates. Locations of selected features imaged in the three panoramas used in this study, e.g. the LM, RTG, and the Geophone Rock, are listed in Tables 4 and 5. Their relative positional errors were assessed by the least squares fit to be better than 25 cm.
Table 4. ALSEP Station: Adjusted Positions of Selected Features
North Azimuth (°)
Geo 3 pan
B&W ALSEP pan
Table 5. Station 1: Adjusted Positions of Selected Features
North Azimuth (°)
Station 1 B&W Pan
5. Summary and Outlook
 High-resolution images (0.5 m/pixel) provided by the LRO mission allowed us to carry out a new, detailed cartographic investigation of the Apollo 17 landing site. As part of this mapping project we precisely determined positions on the lunar surface from where astronauts obtained panoramic images with their calibrated Hasselblad.
 The astronauts' positions while taking two ALSEP panoramas as well as a panorama at traverse station 1 were determined within 0.5 m (LROC NAC pixel size). For this purpose, weighted least squares adjustment was applied to the angular directions observed in assembled panorama mosaics or in the original Hasselblad frames. Precise ME-coordinates of the camera positions as well as their individual accuracies are presented. Additionally, prominent surface features such as large boulders, crater rims, and astronaut equipment, e.g. the LM, the ALSEP central station, and RTG, captured in the Hasselblad images were identified in LROC NAC orthoimages. After the free network adjustment the relative point error of the adjusted features were assessed to be better than 0.25 m. Selected object positions are given as absolute ME-coordinates as well as relative to the adjusted camera location by their North azimuth angle and distances.
 Applying this technique to all of the 9 traverse stations will enable creation of a new, digital Apollo 17 Traverse Map based on half-meter-scale LROC imagery. In this context, the possibility of the positioning of single frames or partial panoramas recorded by the astronauts along their traverse path will be investigated, to provide additional anchor points for mapping the VLBI-based tracks of the LRV. An improved cartographic framework of the landing site supports further detailed exploration and scientific investigation of the Apollo EVAs and lunar surface operations. Researchers will have the ability to derive accurate coordinates of in-situ observations and rock samples, for cross-reference with complementary orbital data sets.
 So far, the identification and positioning of prominent features is limited by the resolution of the orthoimage. Further analysis of stereo images or triplets, i.e. images of the same object from two or three different astronaut positions, will allow for the integration of all suitable features to the network adjustment, regardless of their sizes. By measuring two (or three) different networks of angular directions, feature positions are determined by ray-intersection of the appropriate direction vectors. In addition, most recent LROC NAC images obtained from lowest altitudes of about 21 km will support our improved mapping of the landing site.
 Furthermore, the least squares adjustment of angles measured in the overlaps of single Hasselblad frames allows for an improved reconstruction of those panoramas, which were found to suffer from geometric distortion.
 This work has been supported by a grant (FKZ 50 OW 0902) from the German Aerospace Center (DLR), Bonn. J. Oberst and I. Karachevtseva have been supported by a grant from the Ministry of Education and Science of the Russian Federation (agreement N 11.G34.31.0021 on 30.11.2010). We thank the NASA Lunar Reconnaissance Orbiter project and the LROC Science Operations Center team. We also thank three anonymous reviewers for their helpful comments.