LASmoons: Gudrun Norstedt

Gudrun Norstedt (recipient of three LASmoons)
Forest History, Department of Forest Ecology and Management
Swedish University of Agricultural Sciences, Umeå, Sweden

Background:
Until the end of the 17th century, the vast boreal forests of the interior of northern Sweden were exclusively populated by the indigenous Sami. When settlers of Swedish and Finnish ethnicity started to move into the area, colonization was fast. Although there is still a prospering reindeer herding Sami culture in northern Sweden, the old Sami culture that dominated the boreal forest for centuries or even millenia is to a large extent forgotten.
Since each forest Sami family formerly had a number of seasonal settlements, the density of settlements must have been high. However, only very few remains are known today. In the field, old Sami settlements can be recognized through the presence of for example stone hearths, storage caches, pits for roasting pine bark, foundations of certain types of huts, reindeer pens, and fences. Researchers of the Forest History section of the Department of Forest Ecology and Management have long been surveying such remains on foot. This, however, is extremely time consuming and can only be done in limited areas. Also, the use of aerial photographs is usually difficult due to dense vegetation. Data from airborne laser scanning should be the best way to find remains of the old forest Sami culture. Previous research has shown the possibilities of using airborne laser scanning data for detecting cultural remains in the boreal forest (Jansson et al., 2009; Koivisto & Laulamaa, 2012; Risbøl et al., 2013), but no studies have aimed at detecting remains of the forest Sami culture. I want to test the possibilities of ALS in this respect.

DTM from the Krycklan catchment, showing a row of hunting pits and (larger) a tar pit.

Goal:
The goal of my study is to test the potential of using LiDAR data for detecting cultural and archaeological remains on the ground in a forest area where Sami have been known to dwell during historical times. Since the whole of Sweden is currently being scanned by the National Land Survey, this data will be included. However, the average point density of the national data is only 0,5–1 pulses/m^2. Therefore, the study will be done in an established research area, the Krycklan catchment, where a denser scanning was performed in 2015. The Krycklan data set lacks ground point classification, so I will have to perform such a classification before I can proceed to the creation of a DTM. Having tested various kind of software, I have found that LAStools seems to be the most efficient way to do the job. This, in turn, has made me aware of the importance of choosing the right methods and parameters for doing a classification that is suitable for archaeological purposes.

Data:
The data was acquired with a multi-spectral airborne LiDAR sensor, the Optech Titan, and a Micro IRS IMU, operated on an aircraft flying at a height of about 1000 m and positioning was post-processed with the TerraPos software for higher accuracy.
The average pulse density is 20 pulse/m^2.
+ About 7 000 hectares were covered by the scanning. The data is stored in 489 tiles.

LAStools processing:
1) run a series of classifications of a few selected tiles with both lasground and lasground_new with various parameters [lasground and lasground_new]
2) test the outcomes by comparing it to known terrain to find out the optimal parameters for classifying this particular LiDAR point cloud for archaeological purposes.
3) extract the bare-earth of all tiles (using buffers!!!) with the best parameters [lasground or lasground_new]
4) create bare-earth terrain rasters (DTMs) and analyze the area [lasdem]
5) reclassify the airborne LiDAR data collected by the National Land Survey using various parameters to see whether it can become more suitable for revealing Sami cultural remains in a boreal forest landscape  [lasground or lasground_new]

References:
Jansson, J., Alexander, B. & Söderman, U. 2009. Laserskanning från flyg och fornlämningar i skog. Länsstyrelsen Dalarna (PDF).
Koivisto, S. & Laulamaa, V. 2012. Pistepilvessä – Metsien arkeologiset kohteet LiDAR-ilmalaserkeilausaineistoissa. Arkeologipäivät 2012 (PDF).
Risbøl, O., Bollandsås, O.M., Nesbakken, A., Ørka, H.O., Næsset, E., Gobakken, T. 2013. Interpreting cultural remains in airborne laser scanning generated digital terrain models: effects of size and shape on detection success rates. Journal of Archaeological Science 40:4688–4700.

LASmoons: Jesús García Sánchez

Jesús García Sánchez (recipient of three LASmoons)
Landscapes of Early Roman Colonization (LERC) project
Faculty of Archaeology, Leiden University, The Netherlands

Background:
Our project Landscapes of Early Roman Colonization (LERC) has been studying the hinterland of the Latin colony of Aesernia (Molise region, Italy) using several non-destructive techniques, chiefly artefactual survey, geophysics, and interpretation of aerial photographs. Nevertheless large areas of the territory are covered by the dense forests of the Matese mountains, a ridge belonging the Apennine chain, or covered by bushes due to the abandonment of the countryside. The project won’t be complete without integrating the marginal, remote and forested areas into our study of the Roman hinterland. Besides, it’s also relevant to discuss the feasibility of LiDAR data sets in the study of Mediterranean landscapes and its role within contemporary Landscape Archaeology.

some clever caption

LiDAR coverage in Molise region, Italy.

Goal:
+ to study in detail forested areas in the colonial hinterland of Aesernia.
+ to found the correct parameters of the classification algorithm to be able to locate possible archaeological structures or to document appropriately those we already known.
+ to document and create new visualization of hill-top fortified sites that belong to the indigenous population and are currently poorly studied due to inaccessibility and forest coverage (Monte San Paolo, Civitalla, Castelriporso, etc.)
+ to demonstrate the archaeological potential of LiDAR data in Italy and help other scholars to work with that kind of data, explaining basic information about data quality, where and how to acquire imagery and examples of application in archaeology. A paper entitled “Working with ALS – LiDAR data in Central South Italy. Tips and experiences”, will be presented in the International Mediterranean Survey Workshop by the end of February in Athens.

Civitella hillfort (Longano, IS) and its local context: ridges and forest belonging to the Materse mountains and the Appenines.

Data:
Recently the LERC project has acquired a large LiDAR dataset created by the Italian Geoportale Nazionale and the Minisstero dell’Ambiente e della Tutella del Territorio e del Mare. The data was produced originally to monitor land-slides and erosive risk.
The average point resolution is 1 meter.
+ The data sets were cropped originally in 1 sq km. tiles by the Geoportale Nazionale for distribution purposes.

LAStools processing:
1) data is provided in *.txt files thus the first step is to create appropriate LAS files to work with [txt2las]
2) combine areas of circa 16 sq km (still fewer than 20 million points to be processed in one piece with LAStools) in the surroundings of the colony of Aesernia and in the Matese mountains [lasmerge]
3) assign the correct projection to the data [lasmerge or las2las]
4) extract the care-earth with the best-fitting parameters [lasground or lasground_new]
5) create bare-earth terrain rasters as a first step to visualize and analyze the area [lasdem]

Second German State Goes Open LiDAR

The floodgates of open geospatial data have opened in Germany. Days after reporting about the first state-wide release of open LiDAR, we are happy to follow up with a second wonderful open data story. The state of Thuringia (Thüringen) – also called the “green heart of Germany” – has also implemented an open geospatial data policy. This had already been announced in March 2016 but must have gone online just now. A reader of our last blog article pointed this out in the comments. And it’s not just LiDAR. You can download:

It all comes with the same permissible license as OpenNRW’s data. This is open data madness! Everything you could possibly hope for presented via a very functional download portal. Kudos to TLVermGeo (“Thüringisches Landesamt für Vermessung und Geoinformation”) for creating an open treasure cove of free-for-all geospatial data.

Let us have a look at the LiDAR. We use the interactive portal to zoom to an area of interest. With the recent rise of demagogues it cannot hurt to look at a stark reminder of where such demagoguery can lead. In his 1941 play “The Resistible Rise of Arturo Ui” – a satirical allegory on the rise of Adolf Hitler – Bertolt Brecht writes “… don’t rejoice too soon at your escape. The womb he crawled from is still going strong.”

We are downloading LiDAR data around the Buchenwald concentration camp. According to Wikipedia, it was established in July 1937 and was one of the largest on German soil. Today the remains of Buchenwald serve as a memorial and as a permanent exhibition and museum.

We download the 15 tiles surrounding the blue one: two on its left, two on its right and one corresponding row of five tiles above and below. Each of the 15 zipped archives contains a *.laz file and *.meta file. The *.laz file contains the LiDAR points and *.meta file contains the textual information below where “Lage” and “Höhe” refer to “horizontal” and “vertical”:

Datei: las_655_5653_1_th_2010-2013.laz
Erfassungsdatum: 2011-03
Erfassungsmethode: Airborne Laserscanning
Lasergebiet: Laser_04_2010
EPSG-Code Lage: 25832
EPSG-Code Höhe: 5783
Quasigeoid: GCG2005
Genauigkeit Lage: 0.12m
Genauigkeit Höhe: 0.04m
Urheber: (c) GDI-Th, Freistaat Thueringen, TLVermGeo

Next we will run a few quality checks on the 15 tiles by processing them with lasinfolasoverlap, lasgrid, and las2dem. We output all results into a folder named ‘quality’.

With lasinfo we create one text file per tile that summarizes its contents. The ‘-cd’ option computes the all return and last return density. The ‘-histo point_source 1’ option produces a histogram of point source IDs that are supposed to store which flight line each return came from. The ‘-odir’ and ‘-odix’ options specify the directory for the output and an appendix to the output file name. The ‘-cores 4’ option starts 4 processes in parallel, each working on a different tile.

lasinfo  -i las_*2010-2013.laz ^
         -cd ^
         -histo point_source 1 ^
         -odir quality -odix _info -otxt ^
         -cores 4

If you scrutinize the resulting text files you will find that the average last return density ranges from 6.29 to 8.13 and that the point source IDs 1 and 9999 seem to encode some special points. Likely those are synthetic points added to improve the derived rasters similar to the “ab”, “ag”, and “aw” files in the OpenNRW LiDAR. Odd is the lack of intermediate returns despite return numbers ranging all the way up to 7. Looks like only the first returns and the last returns are made available (like for the OpenNRW LiDAR). That will make those a bit sad who were planning to use this LiDAR for forest or vegetation mapping. The header of the *.laz files does not store geo-referencing information, so we will have to enter that manually. And the classification codes do not follow the standard ASPRS assignment. In red is our (currently) best guess what these classification codes mean:

[...]
histogram of classification of points:
 887223 ground (2) ground
 305319 wire guard (13) building
 172 tower (15) bridges
 41 wire connector (16) synthetic ground under bridges
 12286 bridge deck (17) synthetic ground under building
 166 Reserved for ... (18) synthetic ground building edge
 5642801 Reserved for ... (20) non-ground
[...]

With lasoverlap we can visualize how much overlap the flight lines have and the (potential miss-)alignment between them. We drop the synthetic points with point source IDs 1 and 9999 and add geo-referencing information with ‘-epsg 25832’ so that the resulting images can be displayed as Google Earth overlays. The options ‘-min_diff 0.1’ and ‘-max_diff 0.4’ map elevation differences of up +/- 10 cm to white. Above +/- 10 cm the color becomes increasingly red/blue with full saturation at +/- 40 cm or higher. This difference can only be computed for pixels with two or more overlapping flight lines.

lasoverlap  -i las_*2010-2013.laz ^
            -drop_point_source 1 ^
            -drop_point_source 9999 ^
            -min_diff 0.1 -max_diff 0.4 ^
            -odir quality -opng ^
            -epsg 25832 ^
            -cores 4

With lasgrid we check the density distribution of the laser pulses by computing the point density of the last returns for each 2 by 2 meter pixel and then mapping the computed density value to a false color that is blue for a density of 0 and red for a density of 10 or higher.

lasgrid  -i las_*2010-2013.laz ^
         -drop_point_source 1 ^
         -drop_point_source 9999 ^
         -keep_last ^
         -step 2 -point_density ^
         -false -set_min_max 0 10 ^
         -odir quality -odix _d_0_10 -opng ^
         -epsg 25832 ^
         -cores 4
Pulse density variation due to flight line overlap and flight turbulence.

Pulse density variation due to flight line overlap is expected. But also the contribution of flight turbulence is quite significant.

With las2dem we can check the quality of the already existing ground classification in the LiDAR by producing a hillshaded image of a DTM for visual inspection. Based on our initial guess on the classification codes (see above) we keep those synthetic points that improve the DTM (classification codes 16, 17, and 18) in addition to the ground points (classification code 2).

las2dem  -i las_*2010-2013.laz ^
         -keep_class 2 16 17 18 ^
         -step 1 ^
         -hillshade ^
         -odir quality -odix _shaded_dtm -opng ^
         -epsg 25832 ^
         -cores 4
Problems in the ground classification of LiDAR points are often visible in a hillshaded DTM raster.

Problems in the ground classification of LiDAR points are often visible in a hillshaded DTM.

Wow. We see a number of ground disturbances in the resulting hillshaded DTM. Some of them are expected because if you read up on the history of the Buchenwald concentration camp you will learn that in 1950 large parts of the camp were demolished. However, the laser finds the remnants of those barracks and buildings as clearly visible ground disturbances under the canopy of the dense forest that has grown there since. And then there are also these bumps that look like bomb craters. Are those from the American bombing raid on August 24, 1944?

We are still not entirely sure what those “bumps” arem but our initially assumption that all of those would have to be bomb craters from that fatal American bombing raid on August 24, 1944 seems to be wrong. Below is a close-up with lasview of the triangulated and shaded ground points from the lower right corner of tile ‘las_656_5654_1_th_2010-2013.laz’.

Close-up in lasview on the bumbs in the ground.

Close-up in lasview on the bumbs in the ground.

We are not sure if all the bumps we can see here are there for the same reason. But we found an old map and managed to overlay it on Google Earth. It suggest that at least the bigger bumps are not bomb craters. On the map they are labelled as “Erdfälle” which is German for “sink hole”.

We got a reminder on the danger of demagogues as well as a glimpse into conflict archaeology and geomorphology with this open LiDAR download and processing exercise. If you want to explore this area any further you can either download the LiDAR and download LAStools and process the data yourself or simply get our KML files here.

Acknowledgement: The LiDAR data of TLVermGeo comes with a very permissible license. It is called “Datenlizenz Deutschland – Namensnennung – Version 2.0” or “dl-de/by-2-0” and allows data and derivative sharing as well as commercial use. It only requires us to name the source. We need to cite the “geoportal-th.de (2017)” with the year of the download in brackets and should specify the Universal Resource Identification (URI). We have not found this yet and use this URL as a placeholder until we know the correct one. Done. So easy. Thank you, geoportal Thüringen … (-:

LASmoons: Rachel Opitz

Rachel Opitz (recipient of three LASmoons)
Center for Virtualization and Applied Spatial Technologies
Department of Anthropology, University of South Florida, USA

Background:
In Spring 2017 Rachel Opitz will be teaching a course on Remote Sensing for Human Ecology and Archaeology at the University of South Florida. The aim of the course is to provide students with the practical skills and knowledge needed to work with contemporary remote sensing data. The course focuses on airborne laser scanning and hyper-spectral data and their application in Human Ecology and Archaeology. Through the course students will be introduced to a number of software packages commonly used to process and interpret these data, with an emphasis on free and/or open source tools.

Classification parameters and the resolution at which the DTM is interpolated both have a significant impact on our ability to recognize anthropogenic features in the landscape. Here we see a small quarry. More aggressive filtering and a coarser DTM resolution (left) makes it difficult to recognize that this is a quarry. Less aggressive filtering and a higher resolution (right) leaves some vegetation behind, but makes the edges of the quarry and some in-situ blocks clearly visible.

Goal:
The students will develop practical skills in applied remote sensing through hands-on exercises. Learning to assess, manage and process large data sets is essential. In particular, the students in the course will learn to:
+ Identify the set of techniques needed to solve a problem in applied remote sensing
+ Find public imagery and specify acquisitions
+ Assess data quality
+ Process airborne LiDAR data
+ Combine complementary remote sensing data sources
+ Create effective data visualizations
+ Analyze digital topographic and spectral data to answer questions in human ecology and archaeology

Data:
The course will include case studies that draw on a variety of publicly available data sets that will all be used in the exercises:
+ the PNOA data from Spain
+ data held by NOAA
+ data collected using NASA’s GLiHT platform

LAStools processing:
LAStools will be used throughout the course, as students learn to assess the quality of LiDAR data, classify raw LiDAR point clouds, create raster terrain and canopy models, and produce visualizations. The online tutorials and videos available via the company website and the over 50 hours of video on YouTube as well as the LAStools user forum will be used as resources during the course.

Aesthetic LiDAR Art in “REBIRTH”

Director Patryk Kizny shows us – once more – aesthetically pleasing visuals in his latest short “REBIRTH“. Already known for turning high-dynamic range (HDR) imagery captured in historic buildings into a stunning piece of art with his earlier short film “The Chapel“, Patryk returns to the screens with a larger project that integrates both LiDAR and HDR images. His new 14 minute short film can be watched online via this link. Below a few still images from the beautiful movie to make sure you really watch it … (-: … turn on the audio for narration and soundtrack. There is also a neat timelapse video about the “scanning of the temple” by EKG Baukultur with a Faro scanner.

rebirth_poster

LiDAR heights of burial mounds and cairns

[contributed by guest blogger Lars Forseth]

Archaeologists are increasingly finding ALS/LiDAR useful for making better surveys of archaeological sites and monuments. This is also done where these sites are in danger of being developed, and thus destroyed, see i.e (Risbøl 2011; Gustavsen et al. 2013). Norwegian archaeologists at several county councils and museums have detected unknown sites in woodland or areas previously not surveyed. LiDAR is now available for large areas, as the national mapping authority of Norway, Statens Kartverk, is using this data as a source for generating elevation and contour maps.

aerial view of a mound in North-Trøndelag

aerial view of a mound in North-Trøndelag

Working in North-Trøndelag county, my colleague Kristin Foosnæs at NTNU and I have embarked on a project to create a survey of the larger burial mounds and burial cairns of the county. North-Trøndelag seems to have an unusual large amount of such mounds larger than 20 meters (462 such are so far identified, far more than in any other county). We have gathered exact survey data for a sample of 2900 mounds/cairns. For these we have the exact polygon describing the area of the mounds. LiDAR and LAStools have been extensively used in the creation of this database.

the same  mound seen from the ground

the same mound seen from the ground

The height of the monuments however could only be gathered from the national monuments and sites database where the heights are stored as text. These were gathered by field surveys in the 1970s to 1990s. Then the only tools available to archaeologists for estimating the height were yardsticks or soil probes. Mapping the sites was done on aerial photos at a scale of 1:16.000. The height data gathered from the database is very variable in quality, which has to do with how they were generated. Mostly those that did the original surveys had to estimate the height of the monuments.

a first-return DSM and a ground-return DTM of the same mound

a first-return DSM and a ground-return DTM of the same mound

This summer I discovered that the lascanopy tool of LAStools could measure the min/max elevation for an area-of-interest. Using lascanopy I generated a csv report of elevations (min/max) within a polygon in a shapefile:

lascanopy -lof steinkjer.txt ^
          -keep_class 2 ^
          -lop tessem.shp ^
          -height_offset -1000 ^
          -centroids -min -max ^
          -o tessem.csv

Here I’m inputting a text file ‘steinkjer.txt’ with the list of LAS files to be queried and a shapefile ‘tessem.shp’ with the polygons of the mounds I want to know the height above the ground for. The output ‘tessem.csv’ looks like this:

index,min_x,min_y,max_x,max_y,centroid_x,centroid_y,min,max
0,616277.76,7108569.01,616290.33,7108581.37,616284.04,7108575.19,76.72,77.98
1,616292.40,7108572.37,616299.98,7108580.46,616296.19,7108576.41,77.83,78.96
2,616310.04,7108585.13,616320.96,7108596.96,616315.50,7108591.05,79.93,81.15
3,616714.65,7108371.35,616734.75,7108392.03,616724.70,7108381.69,83.47,86.73
4,616681.74,7108412.71,616699.80,7108429.61,616690.77,7108421.16,86.13,87.97
5,616672.13,7108436.30,616694.56,7108453.78,616683.34,7108445.04,86.55,89.19
6,616666.01,7108449.99,616696.89,7108475.04,616681.45,7108462.52,85.74,90.79
7,616665.14,7108471.25,616687.86,7108494.26,616676.50,7108482.76,86.81,90.35
8,616673.88,7108488.44,616691.35,7108505.91,616682.61,7108497.18,86.72,89.35
9,616695.43,7108602.90,616724.26,7108632.90,616709.85,7108617.90,81.18,85.27
10,617066.09,7108807.01,617080.97,7108819.01,617073.53,7108813.01,87.44,89.59
11,616010.62,7108764.98,616025.46,7108780.39,616018.04,7108772.68,88.3,90.78
12,621229.46,7111180.27,621246.66,7111197.69,621238.06,7111188.98,111.3,112.86
13,621196.44,7111192.72,621216.57,7111208.55,621206.50,7111200.63,110.18,112.08
14,621183.77,7111206.97,621206.39,7111226.65,621195.08,7111216.81,109.69,111.89

The resulting CSV file can be imported to QGIS with the centroid x/y as point location. In QGIS I can then generate a spatial join between the CSV file and the shapefile containing the surveyed mounds/cairns. Then using the field calculator on the table, I can compute the height as a difference of max and min elevation for the each mound/cairn. About 2600 of the 2900 monuments could get their height measured using lascanopy.

The results are shown in the three plots. These have been generated in R and ggplot2. The figure below shows the diameter plotted against the height gathered by lascanopy.

Diameter plotted against height

Diameter plotted against height

Height and diameter correspond to a large degree. One marked difference between mounds and cairns is that some of the larger mounds are lower than their expected height. This can have two explanations; one is that mounds are more likely to be affected by cultivation activities (i.e. they were plowed over by farmers) that have reduced their height. Mounds are more likely to be situated close to farms, while cairns are more likely to be sited along the coast or on hills.

histogram of diameter of monuments

histogram of diameter of monuments

The above histogram of the diameters of the monuments shows a skewing of the data towards the left. Mounds above 20 m of diameter are considered to be large, while those above 40 m are called “kongshauger” or “Kings mounds”. There are 19 such in North-Trøndelag. A normal – Gaussian – curve is fitted over the histogram.

histogram of heights of monuments

histogram of heights of monuments

Finally, the above histogram of the heights – as measured by lascanopy – for aproximately 2600 monuments. This shows that the maximum height lies at about 12.5 meters. The histogram of heights is again skewed to the left. The large mounds mostly seems to be above 20 m of diameter and above 4 m of height.

References:
Gustavsen, L., Paasche, K. 1964-, & Risbøl, O. 1963. 2013. Arkeologiske undersøkelser: vurdering av nyere avanserte arkeologiske registreringsmetoder. Oslo: Statens vegvesen.
Risbøl, O. 1963-. 2011. Flybåren laserskanning til bruk i forskning og til forvaltning av kulturminner og kulturmiljøer: dokumentasjon og overvåking av kulturminner. Oslo: NIKU.

locating German bunkers concealed by canopy

After accidentally finding Russian tanks in Polish forests I was curious to see if there was something else hiding under the forest canopy. Remember, I randomly picked a 500 by 500 meter LiDAR tile as example data to introduce a group of forestry students to LiDAR processing with LAStools during the ForseenPOMERANIA camp. After extracting ground points with “lasground.exe“, strange bumps appeared in the bare-earth hillshades generated with “las2dem.exe” for terrain that was supposed to be completely flat … they turned out to be Russian WW-II positions.

I met Achim when returning to teach the next two groups of students. His hobby is a mix between geo-caching and conflict-archaeology: locating old German bunkers based on approximate coordinates available in historic records and tourist maps and then mapping them precisely with GPS. Achim had a list of longitude/latitude positions as KML files where he was planning to search for known bunkers. I used “lasboundary.exe” to create polygonal outlines in KML format for all areas where we had LiDAR from the forestry project. With Google Earth it was easy to find overlaps between his target areas and our LiDAR coverage.

I extracted the ground points and created bare-earth DTMs of the relevant area with a LAStools batch processing pipeline of “lastile.exe“, “lasground.exe” and “las2dem.exe” and used “blast2dem.exe” to create a seamless hillshading with proper KML referencing (here is a tutorial for such a pipeline). What I found was pretty amazing.

At first glance it looks like a maze of little creeks that are running alongside the ridges of the hillsides but we know that water flows downhill and not “along-hill”. What we see is a network of WW-II trenches that are connecting the bunkers Achim is looking for. A closer look also reveals the likely location of those bunkers.

I placed a pin on each of them and exported their longitude and latitude coordinates for upload into Achim’s GPS device. The next day we set out to verify our LiDAR findings on the ground.

It was a rainy day. Walking through this maze of green and overgrown trenches from one moss-covered bunker ruin to the next felt oddly quiet and peaceful. Achim explained that these bunkers were originally built to defend the border with Poland – long before the Second World War broke out. Only when Russian soldiers were advancing on Germany after the collapse of the Eastern front, the young boys of the Hitler Youth were commanded to dig this network of trenches in order to fortify the bunker and stop enemy lines from gaining ground. After the war the Russians blew up all bunkers that were facing East so that their troops would not ever have to face them again.