One of the most striking and valuable developments in aerial photography has been the use of stereoscopic views. Pairs of pictures, taken with a considerable separation in their points of view and studied later by the aid of the stereoscope, show an elevation and a solidity which are entirely wanting in the ordinary flat aerial vista. Often, indeed, these attributes are essential for detecting and recognizing the nature of objects seen from above. Stereoscopic aerial photography has been justly termed “the worst foe of camouflage.”

Principles of Stereoscopic Vision.—The ability to see objects in relief is confined solely to man and to a few of the higher animals in whom the eyes are placed side by side. When the eyes are so placed they both see, to a large extent, the same objects in their fields of view. Owing to the separation of the eyes the actual appearance of all objects not too far away is different, and it is by the interpretation of these differences that the brain gets the sensation of relief. Thus in Fig. 149 the two eyes are shown diagrammatically as looking at a cube. The right eye sees around on the right-hand face of the cube, the left eye on the left-hand face of the cube. The two aspects which are fused and interpreted by the brain are shown in the lower diagram.

Stereoscopic views or stereograms, made either by photography or, in the early days, by careful drawing, consist of pairs of pictures made of the same object from two different points. For ordinary stereoscopic work these points are separated by the distance between the eyes, approximately 65 millimeters or 2¾ inches. These two pictures are then so viewed that each eye receives its appropriate image from the proper direction, whereupon the object delineated stands out in relief.

Fusion of the two elements of the stereoscopic picture can take place without the assistance of any instrument, if the eyes are properly directed and focussed, but this comes only with practice. Holding the stereogram well away from the face the eyes are directed to a distant object above and beyond, in order to diverge the axes. Then without converging, the eyes are dropped to the picture, which should spring into relief. It is necessary in moving the eyes from the distant object to the near stereogram to alter their focus somewhat, depending on how near the stereogram is held; and the success of the attempt to fuse the images depends on the observer's ability to maintain the eyes diverged for a distant object while focussing for a near one. Near-sighted people (on taking off their glasses) fuse the stereoscopic images quite easily, since their eyes do not focus on distant objects even when diverged for them. Transparencies are easier to fuse than paper prints, but in any case where a stereoscope is not used the separation of image centers should not be more than that of the eyes.

Stereoscopes.—The easier and more usual method of fusing the stereoscopic images is by a stereoscope. The simplest form consists merely of two convex lenses, one for each eye, their centers separated by a distance somewhat greater than that between the eyes. Their function is to bring the stereogram to focus, and, by the prismatic action of the edges of the lenses, to converge the lines of sight which pass through the centers of the two pictures to a point in space in front of the observer. The two lenses should be mounted so as to provide for the adjustment of their separation to fit different eyes and print spacings. The most common form of stereoscope is that designed by Holmes, for viewing paper print stereograms (Fig. 150). It has prismatic lenses of an appropriate angle to converge pictures whose centers are three inches apart, instead of the lesser distance appropriate to stereograms intended for fusing without an instrument. No adjustment is provided for varying the lens separation, but the print can be moved to and fro for focussing.

Another form of stereoscope, one of the first produced, is the mirror stereoscope (Fig. 152), now used extensively for viewing stereo X-ray pictures. It consists of two vertical mirrors at right angles to each other, with their edge of contact between the eyes. The two prints to be studied are placed to right and left, an arrangement that permits the use of prints of any size. The convergence point is controlled by the angle between the mirrors. The Pellin stereoscope (Fig. 153) utilizes two pairs of mirrors in a way to permit the use of large prints. The prints are, however, placed side by side on a horizontal viewing table, which avoids certain difficulties of illumination met with in the simpler mirror form. The box form of stereoscope (Fig. 151) using either prisms or simple convex lenses, is particularly adapted for viewing transparencies, although the insertion of a door at the top provides illumination for paper prints. The Schweissguth design (Fig. 154) is intended primarily as an aid to selecting the portions of the prints to be cut out for mounting. The platform on which the pictures rest is composed of two long rectangular blocks, on which are plates of glass raised sufficiently to permit the prints to be slid underneath. The space between the blocks allows the unused portion of the photograph to be turned down out of the way. Prints of any size can thus be moved about until the proper portions for stereo mounting are found. Either block can be moved in its own plane and also to and from the eye, whereby two prints of somewhat different scales can be fused.

The Taking of Aerial Stereograms.—The normal separation of the eyes is altogether too small to give an appearance of relief to objects as far away as is the ground from a plane at ordinary flying heights. In order to secure stereoscopic pairs it is therefore necessary to resort to a method originally employed for photographing distant mountains and clouds. This is to take the two pictures from points separated by distances much greater than the interocular separation—by meters instead of millimeters—corresponding to the positions of the eyes on a veritable giant. In the airplane this is accomplished by making successive exposures as the plane flies over the objective, at intervals to be determined by the speed, the altitude and the amount of relief desired (Fig. 155).

An all important question which arises immediately is: What separation of points of view shall we select? If the exposures are too close together there will be little relief; if too distant the relief will be so great as to be unnatural, even offensive. Obviously we cannot here establish a criterion of natural appearance, since the natural appearance to ordinary human eyes is devoid of relief. We may, however, define correct relief as that obtained when the apparent height of elevated objects is right as compared with their extension or plan.

In order to secure this condition it is necessary, first, that each element of the stereoscopic pair be correct in its perspective. This is fortunately an old photographic problem, already well understood. Its solution is to view the photograph from a distance exactly equal to the focal length of the camera lens. Since the normal viewing distance is not less than 25 centimeters, lenses of this focal length at least are requisite for correct perspective. Secondly, it is necessary for correct relief that the two views be taken with a separation equal, on the plane of the plate, to the separation of the eyes, or 65 millimeters. If d is the interocular distance, a the viewing distance, identical with the focal length of the lens used, and A the altitude, then D, the distance between exposures, is given by the relation—

For a = 25 centimeters, D
A = 6.5
25, approximately ¼, or the interval between exposures must be a quarter the altitude. With a 50 centimeter lens this becomes ⅛, and so on. These figures show the fallacy of the suggestion sometimes made that we take stereoscopic pictures by two cameras placed one at the extremity of each wing.

When lenses of more than 25 centimeters focal length are employed, the stereoscope should be one capable of throwing the convergence point farther away than the customary 25 centimeters. In the simple lens type of instrument we can do this by bringing the centers of the lenses closer together, and by making the focus agree with the convergence point by adjustment of the distance between lenses and stereogram. If enlargements are used they should be treated in all respects as originals made by lenses of the greater foci corresponding to the scale of the enlargement.

When all the conditions are covered, the appearance presented in the stereoscope is that of a model of the original object at a distance a, and a
A times natural size. If pictures are made at exposure intervals less than those indicated for correct relief, they show insufficient relief. This does not, however, give an unnatural effect, because anything between no relief and “correct” relief appears natural with large objects which are not ordinarily seen in relief by eyes not Brobdignagian. Conversely, stereograms made with too large exposure intervals show exaggerated relief. Yet this is often no objection. It is indeed rather an advantage if we wish to bring objects of interest to notice. Consequently, so long as the exaggeration of relief is not offensive, the permissible limits of exposure interval are pretty large. Actually, the eye tolerates such great deviations from strictly normal conditions that satisfactory stereoscopic effects are obtained for pictures viewed at very different distances from the focal length of the taking lens, and with the axes of the eyes parallel or even diverging, although there is some strain whenever focus and convergence points differ. On the whole, therefore, it may be said that the conditions above laid down for correct relief are only a normal, to be approximated as nearly as is practicable.

Having established the correct relation of taking points for stereos the next problem is how to determine these when in the plane. The simplest way is by means of a stereoscopic sight. This consists essentially of two lines of sight (fixed by beads, crosses, or other objects), inclined toward each other at the angle determined by the ratio of the ocular separation to the focal length of the lens. If the back sight is made a single bead or cross, the rest of the stereo sight will consist of two beads or crosses, separated from each other by the ocular distance of 65 millimeters, and distant from the back sight by the focal length of the lens (Fig. 157). The first picture is taken when the object is in line with the forward pointing line of sight, the second when it lies along the backward pointing one. Like other sights, the stereoscopic sight may be attached either to the camera, or if this is fixed in position, to any convenient part of the plane. A very simple sight for vertical stereoscopic photography consists of an inverted V painted on the side of the fuselage, so that the eye can be placed at the vertex and sighted along either leg.

The common method of determining the space between exposures is by the time interval. If V is the speed of the plane, and t the desired time interval, we have, from the last equation—

If A = 2000 meters, d = 65 millimeters, and a = 25 centimeters, and if the plane is traveling 200 kilometers per hour, the time interval must be—

At 1000 meters altitude the interval will be half this, and so on in proportion. If the pictures are taken with a 50 centimeter focus camera, and are hence to be viewed at 50 centimeters convergence distance instead of at 25, the time will again be halved. These relations are clearly shown in the diagram (Fig. 156). Here the left-hand portion shows how to find the stereoscopic base line at each altitude for each focal length; while the right-hand portion shows how to translate this into time interval for any plane velocity. The Burchall slide rule (Fig. 130) shows another way to arrange these data in form for rapid calculation.

Plates used for stereoscopic negatives should be at least twice as long as the ocular separation, if correct relief is desired, and the full size of the stereoscope field is to be utilized. This relation follows at once if we consider that we wish to cut from each negative a rectangle 65 millimeters wide, and that the image of the target has shifted 65 millimeters between exposures. If the plate is larger than this there is opportunity to select the view, or to pick several. If the plate is smaller the elements of the stereogram must be narrow strips. This, however, holds only for contact prints.

The ordinary English practice in making stereo negatives is to take successive pictures with an overlap of 60 to 75 per cent. This practice is probably dictated by the 4 × 5 inch plate, since 60 per cent. overlap on 4 inches means a separation of just over an inch and a half instead of 2¾, but it leaves 2½ inches of picture common to the two negatives. With ¾ overlap the common portion is 3 inches, which permits of cutting 2¾ inch prints, and allows some latitude for irregular motion of the plane or for chance error in calculation of intervals. Data on the basis of ¾ overlaps for a 4-inch plate are shown in connection with Fig. 155 which shows in diagrammatic form the variation of exposure interval with height, together with other points of interest.

Elevation Possible to Detect in Stereoscopic Views.—Can the actual difference in elevation be discovered by the use of stereoscopic views? An approximate idea may be obtained from the following considerations: Suppose we have two small point-like objects, one above the other, such as a street lamp globe and the base of the lamp pillar. In a view taken from directly overhead these will be superposed, and so will not be capable of separation. But, as the point of view is shifted sideways, the two objects separate, until a point is reached where they can just be distinguished as double. When this condition holds for either picture of the stereoscopic pair it will be possible to obtain stereoscopic relief.

Now the separation which can just be distinguished is commonly assumed to be one minute of arc. This angle corresponds to about 1
3400 the distance from the eye to the object. If the object is assumed at a distance a from the face, and on a line with one of the eyes, which are separated by the distance d, then (all angles being small) the object must be of height a
d times the horizontal distance which corresponds to one minute. For 25 centimeters' viewing-distance this quantity is about 4, so that the least perceptible elevation is 4
3400 or about 1
900. The stereogram having been made under conditions giving correct relief, this fraction is also the fraction of the altitude of the plane when the photograph was taken which may be detected. An object as high as a man (6 feet) should be visible as a projection in a stereoscopic view taken at 6 × 900 = 5400 feet. This relation—1
900—holds (irrespective of the focal length of the lens), as long as the conditions for correct relief are maintained.

Stereoscopic Aerial Cameras.—Cameras for aerial stereoscopic photography need in no way differ in construction from those made for mapping or spotting, provided only they permit exposures to be made at short enough intervals. The addition of special sights, as already discussed, constitutes the only real difference between single view and stereoscopic aerial cameras. But even without such sights ordinary aerial cameras are applicable to stereo work by the usual procedure of determining the exposure spacing by time.

One scheme employed for taking low stereos, where the interval is only two or three seconds, is to mount two cameras in the plane, exposing them one after the other at the correct interval. Another method which has been tried with success is the use of a double focal-plane shutter in a single lens camera (Fig. 157). The two shutters are side by side, with their slots parallel to the line of flight. To take a stereo negative we expose first the shutter nearer the tail of the plane, and then the other, after an interval which can be calculated from the speed and altitude, or, better, determined by a stereoscopic sight. The two views are thus obtained on a single plate. Prints from these negatives are transposed right and left, and, if the prints are viewed in an ordinary stereoscope, have to be cut apart and transposed for mounting, or else this may be done to the negatives.

In this connection attention may be drawn to an alternative method of viewing stereograms, which may be used on transposed prints—a method which needs no instrument, and so has sufficient advantage to even warrant mounting ordinary stereoscopic pairs in the transposed position for observation. This method consists in crossing the optic axes, in the fashion illustrated in Fig. 158. A finger is held in front of the face in such a position that the left stereogram element and the finger are seen in line by the right eye; the right element and the finger by the left eye. The proper position is found by alternately closing each eye, and advancing or retracting the finger. Then both eyes are opened and converged on the finger tip, which is thereupon dropped, leaving the picture standing out in relief. An opportunity to try this method is afforded by Fig. 159.

Stereo Obliques.—The theory of making oblique stereo pictures is identical with that of other stereos. The only problem peculiar to obliques is that of making the exposures at short enough intervals apart. This problem is due largely to the fact that oblique views are ordinarily taken from low altitudes, for the purpose of “spotting” particular objects, rather than for mapping the gross features of an extended area. The same problem of how to secure a short exposure interval is met with when we attempt to take vertical stereos from a low altitude, but as already discussed, it is much preferable from the pictorial standpoint that pictures of definite small objectives be made obliquely.

Another reason for taking stereo obliques from points but little separated is of some interest in connection with the discussion above given of “correct” and “natural” relief. When the relief is “correct” the object appears, as already stated, to be a small model in its true proportions, standing at the convergence distance. When the eyes are converged to a small object 25 to 50 centimeters away all objects beyond are hopelessly transposed and confused. This does not happen when we look at large distant objects, since their background is at a distance effectively but little beyond them. As a result, when a stereo oblique is made in “correct” relief of such an object as the Washington monument with buildings beyond, the confusion of the background presents an appearance entirely contrary to our visual experience with objects as large as the neighboring buildings are known to be. This effect may be avoided by choosing a uniform background such as grass, or by taking the pictures very much closer together, at the expense of “correct” but at a gain in “natural” relief.

Stereo obliques can of course only be made with any facility by laterally pointing cameras. From the calculations already given it appears that a “correct” stereo oblique of an object 500 meters away will mean exposures only two or three seconds apart, too short an interval for any of the ordinary plate-changing and shutter-setting mechanisms; and the case is even worse should less relief be desired. One solution of this problem has been the use, already mentioned, of two cameras mounted together, either side by side or one over the other, with separate shutter releases. Both releases may be controlled by the observer, using a sight, or else pilot and observer may work in harmony as has been recommended in the English service, where the pilot releases one shutter and the observer counts time from the instant he sees the first shutter unwind and releases the second.

A very satisfactory apparatus for the taking of stereo obliques consists of a 10-inch focus hand-held camera (Fig. 157), provided with a two-aperture focal-plane shutter. The right-hand half of one curtain aperture is blocked out, the left-hand half of the other. The first pressure on the exposing lever exposes one-half of the plate, the second the other. A stereoscopic sight of the type already described is placed on the bottom. To make an oblique stereo negative the camera is held rigidly by resting the elbows on the top of the fuselage and the first exposure is made when the object comes in line with the rear sight and the leading front sight. The eye is then moved so as to look along the line of the rear sight and the following front sight, and when the object is again in alinement the second pressure is given the exposing lever. Fig. 159 shows a stereo oblique made by this camera. The elements are transposed right and left, and the stereogram may be viewed by crossing the optic axes as shown in Fig. 158, or the two pictures may be cut apart and remounted.

The Mounting of Aerial Stereograms.—The first step in making the printed stereogram is to select two pictures taken on the same scale, but from slightly different positions. These may be two chosen from a collection made for other purposes, or else a pair taken at distances calculated to fit them for stereoscopic use. The next step is to mark the center of each picture, either with easily removed chalk or with a pin point. They are then superposed, and afterward carefully moved apart by a motion parallel to the line joining their centers when superposed. The final step before mounting is to mark out and cut the two elements, their bases being parallel to the line of centers, their horizontal length the distance between the optic axes of the stereoscope (or as near this as the size of the prints will permit). They are then mounted on a card, with their centers separated by approximately 65 millimeters. The right-hand view is the one showing more of the right-hand side of objects, and vice versa. This process of arranging, cutting, and mounting is shown clearly in Fig. 160. In this case the stereoscopic elements lie symmetrically about the line joining the centers of the original prints. This is not necessary, as they may be selected from above or below this line so long as their bases are parallel to it. A simplification of this method consists in superposing the two prints, laying over them a square of glass of the size to which they are to be cut, then turning it so that a side is parallel to the line of centers, and cutting around it through both prints with a sharp knife. The principle and results are of course the same with both methods.

If large numbers of stereoscopic prints are required it is necessary, for economy of time, either to photograph a finished stereogram and make prints from this copy negative, or to set up special printing machines. Under the general discussion of printing devices a stereoscopic printer is described (the Richard) in which the two negatives are placed so that stereo prints can be got by two successive printings on one sheet of paper.

Uses of Stereoscopic Aerial Views.—Attention has already been called to the characteristic flatness of the aerial view. Neither the picture on the retina nor that on the photographic plate affords any adequate idea of hills and hollows. Unless shadows are well defined, small local elevations and depressions cannot be distinguished from mere difference in color or marking. Even in the presence of shadows it is often only by close study that differences of contour are noticeable. But with stereoscopic views these features stand out in a striking manner. Taking our illustrations from military sources, we may note the use of stereoscopic pictures to detect undulations of ground in front of trenches (Fig. 161). They reveal the hillocks, pits, small quarries, streams flowing behind high banks, and other features which make the attack hard or easy. Commanding positions are shown, the boundaries of areas exposed to machine-gun fire, and the defilades where the attackers may pause to reform. Concrete “pill boxes” are located in the midst of shell holes of the same size and outline, and can be differentiated from them.

Railway or road embankments and cuts can be detected and studied to extraordinary advantage in stereoscopic pictures. Thus what appears to be a mine crater on a level road, easily driven around, may be a gap blown in an embankment, a serious obstacle indeed. Bridges, observation towers and other elevated structures jump into view in the stereoscope when often they have entirely eluded notice in the ordinary flat picture. Once presented in relief, camouflaged buildings or gun emplacements, however carefully painted, are ridiculously easy to pick out.

Practical peace-time applications of stereoscopic views can easily be foreseen following the lines of war experience. Such, for instance, would be the study of proposed railway or canal routes. A series of stereograms would obviate the necessity of contour surveys, at least until the exact route was picked and construction work ready to start.

Apart from their utilitarian side, however, stereoscopic views have very great pictorial merit. Stereoscopic pictures of cathedrals, public and other large buildings, have often great beauty, and afford opportunities for the study of form given by no other kind of representation, short of expensive scale models. They may very well lead in the near future to a revival of the popularity of the stereoscope.

Impression of Relief Produced by Motion.—An appearance of solidity can be obtained in moving pictures by the simple expedient of slowly moving the camera laterally as the pictures are taken. As an illustration, if the moving picture camera is carried on a boat while structures on the shore are photographed, when these are projected on the screen they appear in relief, due to the relative motion of foreground and background. As relief of this sort is not dependent on the use of the two eyes, it demands no special viewing apparatus. This idea has been utilized to a limited extent in ordinary moving picture photography by introducing a slow to-and-fro motion of the camera, but this can hardly be considered satisfactory, since this motion is so obviously unnatural.

In moving pictures made from the airplane the normal rapid motion of the point of view is ideal for the production of the impression of relief in the manner just described. For instance, in moving pictures of a city made from a low flying plane, the skyscrapers and spires as they sweep past stand forth from their more slowly moving background in bold and satisfying solidity. In fact, such pictures probably constitute the most satisfactory solution yet found of the vexing problem of “stereoscopic” projection. No better medium can be imagined for the travel lecturer to introduce his audience to a foreign city than to throw upon his screen a film made in a plane approaching from afar and then circling the architectural landmarks at low altitudes.

Oblique aerial photographs if on a large enough scale are even easier to interpret than are ordinary photographs taken from the ground, since they practically preserve the usual view, and add to it the essentials of a plan. With verticals, however, this is far from the case. In them all natural objects present an appearance quite foreign to the ordinary mortal's previous experience of them. This may be easily demonstrated by taking any aerial view containing a fair amount of detail and trying systematically to identify each object. A necessary preliminary to doing this accurately is acquaintance with and study of the ground photographed, or of similar regions, and of objects of the same character as those likely to be included.

The interpretation of military aerial photographs is of such importance, and has become such an art, that it is the function of special departments of the intelligence service. Extended courses in the subject are now given in military schools. This instruction must cover more than the interpretation of aerial photographs as such. General military knowledge is essential, so that not only may photographed objects be recognized, but the significance of their appearance be realized. Whether attack or retreat is indicated; whether a long range bombardment is in preparation, or a mere strengthening of local defences.

The natural difficulties of interpreting aerial views are enormously increased by the unfamiliar nature and frequently changed character of the military structures, and particularly by the attempts made to conceal these from aerial observation by selection of surroundings and by camouflage. The small scale of the photographs, in which a machine gun shows as a mere pin point, adds to the uncertainty, with the net result of making interpretation a task of minute study and deduction worthy of a Sherlock Holmes.

Little detailed information on interpretation can be profitably written in a general treatise, partly because the illustrations available are of a highly technical military character, partly because original photographs instead of halftone reproductions are practically imperative for purposes of study. Nevertheless some general instructions, applicable to any problem of interpretation, may be given, as well as a few illustrations, drawn from military sources, which will serve to show the detective skill necessary.

First of all it is important that the print or transparency be held in the right position. The shadows must always fall toward the observer; otherwise, reliefs will appear as hollows and hollows will show as hills. The reason for this is that the body ordinarily acts as a shield, preventing the formation of shadows except by light falling toward the beholder. Thus in Fig. 162 the slag heap looks like a quarry when the shadows fall away from one. The necessity for proper direction of shadows is, it may be noted, in conflict with the ordinary convention for the orientation of maps—at least in the northern hemisphere. A city map, made by sunlight falling from the south, presents its shadows as falling away from the observer, when it is mounted with its north point at the top, as is customary. As a consequence buildings in aerial photographic mosaics of cities occasionally look sunken instead of standing out.

The relation between the shape of the shadow and the object casting it must be well learned. This is a part of the training of every architectural draftsman, but the appearance of shadows from above has not heretofore been a matter of importance. The difference between high and low trenches, between cuttings and embankments, between shell holes, occupied or unoccupied, and “pill boxes,” must be detected largely from the character of the shadows. Which elevations and depressions are of military and which of merely accessory nature, whether this black dot is a machine gun or a signaling device, whether that dark spot is an active gun port or an abandoned one—these are all matters of shadow and of light and shade study. Several illustrations of these points appear in Figs. 163, 164 and 165.

Shadows may be used to get exact information as to directions and magnitudes. If we know the time of day at which a picture is taken, the direction of the shadows will give the points of the compass. A chart for doing this is shown in Fig. 166. The length of a shadow is a measure of the height of the object casting it, and the exact relation between the two dimensions is determined by the day and hour. Fig. 167 embodies in chart form the values of this relationship for all times of the year and day, while Fig. 168 shows the kind of picture in which shadow data could be utilized to great profit.

Minute changes, both in light and shade and in position, must be watched for with great care. Naturally growing foliage and the cut branches used for camouflage differ in color progressively with the drying up of the leaves. Hence a mere spot of lighter tone in a picture of a forest, especially if the picture is taken through a deep filter, becomes instant object for suspicion. The complete study of any position calls for photographs of all kinds—verticals, obliques, and stereos. Stereoscopic views are the worst foe to camouflage. A bridge painted to look like the river beneath is labor thrown away if the stereo shows it to be a good ten feet above the real river!

A few illustrations of the more ordinary and obvious objects whose detection is the subject of aerial photography are shown in accompanying figures. Fig. 169 pictures a typical trench system, with barbed wire. The trenches show as narrow castellated lines, from which run the zigzag lines of communicating trenches, saps, and listening posts. The minute pockmarks behind the main trench lines are shell holes and machine gun pits. The barbed wire shows as double and triple gray bands, intricately criss-crossed at strategic points. Another form of defence, intended for the same purpose as the barbed wire of the western front, is that furnished by overthrown trees in forest regions. Fig. 170 reveals a mountain fortress surrounded by a zone of felled trees, and indicates in striking manner the value of the information a single aerial photograph may furnish to an attacking force. Fig. 123 shows on a comparatively large scale opposing trench systems in which a natural obstacle—a river—separates the adversaries. Nicks and dots indicate machine guns to the skilled eye, and several rectangular structures are revealed as concrete buildings which have survived unscathed the shell fire which has obliterated, and caused to be rebuilt, nearly every other element of the trench system.