The Cartographic Project


Nowadays, few people spend time in creating maps, except thematic maps. Who, try to enter into competition with the governemental or private organizations speciallized in cartography?
The Tibet case is special and my undertaking seems to be isolated. Formerly closed to the West, Tibet was very little known, even by the Tibetans themselves who do not have any map at their disposal. Recently, the implantation of communism by the Chinese made sources of information that could be at our disposal disappear. This country four times larger than France consists of vast highlands populated with nomads living outside our civilization and small villages where new Chinese immigrants have settled.


Tibet is now scattered in five autonomous provinces of the Chinese Republic, and the Tibetans are in a minority in four of them.
For the Chinese, the Tibet boundaries are those of the TAR (or Tibet Autonomous Region), and for the rest of the world, regions where are populations live who belong to the same culture, same religion, and speak the Tibetan language. It must be noticed that before 1959, the Tibetan state was under the Dalai-lama leadership. We have at our disposal a wide literature concerning Tibet. Throughout it, about several ten thousand toponyms are mentioned, but no serious map can locate these names. If one says that 2000 monasteries and holy places were destroyed by the Red Guard in the seventies, no one can provide a list of them, or locate them on a map. In the near future, we hope that some Tibet specialists will write both a description and a list of such holy places.
A huge work on Tibetan heritage preservation is sorely needed. If Tibet specialists have numerous sources of information in the religious and historic domain, it is not the case in geography.
The Tibet cartography project fits well with the Tibet cultural preservation. A certain amount of experience is required to achieve such an enterprise. After attending courses in classical Tibetan, I was involved in the Central Asia geography studies in 1967, and noticed quickly the lack of information sources and maps.


Yangtse headwaters

In 1969, when I received the first space pictures from the NASA St Goddard Flying Center, I was realising the wide interest of these pictures in designing maps giving an exact relief. I started researches on the Yangtse springs with the first pictures of the Tibetan region made by Cooper, and the results were published in the Royal Geographical Society magazine. In addition, I elaborated a method that selects color informations (by equi-density) using a new type of film manufactured by Agfa for the Cancer research (Agfa-Contours film). This process permited to reveal the wet surfaces, the vegetation and the orography. Nevertheless, there was no application of this method with the introduction of multiple spectrum video cameras within the Landsat satellites that gave a result very close to it. Finally, I undertook drawing maps upon paper, but I could not pursue this work further, aware that it would require a team of several people for a long period. For all these reasons and others I gave up achieving this project.


Five years ago, just retired, I became aware of the progress of the data processors power so that it was possible to take up the work again. I was really motivated by the fact that no satisfactory map of Tibet had yet been published.
The most recents maps were=
- The International Aeronautic TPC map at 1500 000 scale, that is a credible reference source for the longitude and latitude information. If some details seem to be well identified on the map, it is not the case for the reliefs and valleys which are really approximatives. In addition, roads and toponyms, mostly provided by the Chinese authorities, are quite fanciful.
- A copy of the Tibetan map found in China and published by T.I.N. in London with numerous Tibetan toponyms, generally not mentioned by occidental travelers, and apparently suspicious. The rivers and the new district limits are drawn, but the relief is not present and the village locationss are inaccurate.
- A soviet map serie 1500 000 scale and 1150 000 scale (available since 1999), the first one giving acceptable contour levels but with few toponyms in cyrillic without interest.
- A small Chinese atlas (2000) with a map per district. Roads are only goods, toponyms are numerous and in chinese, without any interest except you real chinese. TIN' map is better.
- Gyurme Dorje Handbook map is quite good for a traveller into Tibet but the placename's number is limited.
- >General road maps available in tourism bookseller's shops and quite unsatisfactory.

2- Project planning
I had numerous NASA pictures and I decided to transfer directly the satellite picture to a bottom of map.
After receiving support and some advice that proved to be usefull, from the ISTAR Company at Sophia-Antipolis, I started effectively. The beginnings were encouraging and I persisted in my endeavour. In the first step, considering the data processors' limited speed, I could design maps in a reduced size. Later, I lost considerable time when I had to redesign the maps on larger surface areas.
At the beginning, I was lucky enough not to meet any difficulty while assembling the first two pictures that were used for the map background. Indeed, as one works on small surfaces (until 0.5 degree, or less than 60 km), one can consider that the given pictures are at the indicated scale, the image is linear and has no disruption, and consequently results are satisfactory. When I wanted to go further, I met the largest difficulties for assembling the pictures conveniently.
The different stages that resume these difficulties are described below.
This work can be shared into three parts.
The first one concerns the design of map background bases on satellite pictures, the second one the dressing of the map background and the third one printing.

3- The satellites sources
Landsat pictures from NASA are sold at a reasonable price and will permit to create a bottom of map with a adequate definition. However the resolution (30m per pixel) cannot allow one to see any details, roads, cities or villages site. An alternative would be to use the pictures from the French satellite Spot, using a 10m per pixel resolution but costing each 200 times the price of Landsat picture ...(that means a total cost of 9x200 = 1800 more expensive).

4- Map design
My first difficulty consisted in gathering all necessary Landsat pictures that would offer a global covering of the Tibetan region. This required about a hundred pictures. NASA delivers a list of the Landsat pictures available with cloud blanket indications, that incites one to select only the maps with zero cloud rate. However some pictures are far from good enough. Sometimes the background is pale without any relief due to an insidious mist all over the region. And if the snow is present everywhere, the picture will be widely white. This induced me to buy many more pictures than initially scheduled.
The Landsat pictures, at 1/1.000.000 scale, look like parallelograms of 180km side. Considering the Earth rotation, the polar orbit and the satellite moving diection, this parallelogram is leaning to the right. The delivered picture has been digitized with a resolution of 600 pixels per inch, that gives sides of about 4300 pixels as well as a resolution of 120 pixels per inch at 1/200.000 scale, that fits perfectly with the ink jet printer performances at the present time.

4-1 Problems linked to the use of satellites pictures
The corrections done by NASA on the pictures are well known and developed in specialized literature.
This consists of radio-metric corrections with detector equalization, radio-metric and geometric corrections following the flight system parameters. NASA specifies for this set of pictures, a resolution and differential of less than 1%. If such differentials may seem negligible, they remain important when some 4300 pixels side images are manipulated, because the differentials can reach the value of 43 pixels. These images will have to be re-processed, in order to become a useful product, considering there was no additional information at our disposal. The goal consisted in minimizing the differentials to a value equal or less than 1/1000 enabling a correct picture assembly. Finally, we must make sure that these images superimpose perfectly the International Aeronautic map drawings.
The accurate measure of the picture size revealed multiples disruptions. Not only the scale changes everywhere on the surface, but the sides are not really straight. Moreover, when we try to assemble 4 pictures from the same corner, the sum of the four angles gives a value between 358,5 and 361,5 degrees. This is not suitable. After correcting the pictures to be assembled by 2, they cannot be assembled by 4. I looked for a rigorous method to solve this issue. At this level, I requested advice without results.
The French IGN proposal was simple = move the image details to known positions. This advice was not applicable to regions where no exact datas are available. Only the International Aeronautic map TPC drawings will be used as marks by default.
table
Any method correcting the images by means of risky corrections is not a convenient method. I preferred using a method considering all the pictures and doing successive corrections that reduced defaults progressively until the maps could be perfectly assembled, with some detailed locations similar to those of the International Aeronautic TPC maps ones.
Important to notice the differences given by the projection of a spherical surface on a plane surface, showed that the scattering is quite easy to master for the Tibetan areas, considering the resolution used for assembling the pictures and the scale chosen for the maps. So, this point was not an additional difficulty.
4-2 The defaults identification
I operated in the following manner, to evaluate the corrections to be made.
Each picture was measured carefully, and the values were registered.
The corner and sealing defaults were identified and registered on a paper, describing the entire pictures. It showed differences of the same type over 90% of the pictures. The faults are multiple the scale is not uniform throughout the image surface. The image details at the bottom left of the pictures are larger than the neighbouring ones, the sides are not straight, and so on. Finally the parallelogram lean varied from one picture to another.
4-3 Corrections done for image dimensioning.
All the corrections issued on the picture dimensioning were carefully registered.
The image was treated in four stages
- Scale widespread and straight sides of the picture;
- Length correction of the picture sides;
- Angle correction of the parallelogram;
- Leaning correction of the picture.
4-3-1 Scale widespread and straight sides
Left and right pictures are covered by nearly 20% of the width of the picture. But on the covering zone, the image on the right is more expanded than the one on the left but only on the horizontal size. Also one will be narrowed while the other will be expanded, in order to correct the default. The treatment is applied on most of the pictures, and after that, no additional drawback has to be identified.
4-3-2 Length correction of the picture sides
Pictures are going to be assembled by 2. But they do not. Sides and angles are measured again and side lengths are modified in such a way that the angle modification is going to improve the angle default.
One checks that the two sides' details fit perfectly. From there, the pictures are assembled per pair. The corrections remain moderated within the 1%.
4-3-3 Angle correction of the picture tops
A new set of angle measurement is then done for the whole pictures and the differences with 360 degrees are registered. Sometimes more than the half of the initial difference has been cured, but a last modification is required because the sum of the four angles needs to give a value between 359,9 and 360,1 degrees to enable correct pictures assembly. We notice that these positive and negative differences are distributed quite well on the whole, indicating a right physical situation. The method used is not rigorous. It consists in increasing or reducing one of the four concerned sides for a peak, and concerns two neighboring pictures as well.
Once the corrections are executed, the pictures are assembled in group of 4 or 6. To go further, it is advised to reduce the resolution in order to use a file size less than 100Mo.
The agreement between the image and the International Aeronautic map drawings can then be checked. If a perfect fitting is checked along 1500km of longitude, it is not the case for the latitude. The angle default still remains.
4-3-4 Lean correction of the pictures
After checking the results, an angular or lean error of 25/1000 was observed for all the maps. This error higher than the 1% suggested by NASA, had to be corrected and it was the last dimensional correction operation impacting the map background design.
4-3-5 Chromatic corrections
4-3-5-1 The cleanness, the luminosity, and the contrast of each image were corrected afterwards, in order to have the transition undetectable from one picture to another. This operation is executed when assembling four pictures.
4-3-5-2 The satellite images showed a south lightening on the reliefs, even though classic maps were used to be lighten the North-West. So the black and white aspect observed on the images was reversed.
After this operation, some valley bottoms needed to be made clearer.
4-3-5-3 A new correction seemed to be necessary. In fact, some dark parts of the image were not representing the projected shadows of the mountains, but some wet and widely irrigated zones. So, a good knowledge of the region was needed to make the necessary corrections without making errors.
4-3-5-4 The image is then converted to a sepia color, in order to give the best result for a map background.

5 Observations concerning the map background
Concerning the Center and the West of Tibet, the sepia color is the color of the landscapes in winter and spring when the pasture land grass has not appeared yet. The image shown on the computer screen gives far more details than the one observed upon any other kind of map, this Tibet vision looks like a real trip. It seems quite possible to rebuild the landscape from the pictures taken from space. This allows me to introduce to you Sven Hedin who undertook a reverse approach.
At the beginning of this century, the cartography service Survey of India designed Tibet maps from the notes of the Indian pandits who crossed the country, revealing blank spaces between the valleys. A little while later, the famous Swedish traveller Sven Hedin went all over the vast highlands in the West and North -West of Tibet taking many pictures and doing many drawings. When he came back, his assistants worked very hard trying to reconstitute the reliefs from the drawings. The results were disputed at the time. Nowadays, these results do not seem to be comparatively serious to the ones obtained from the satellite pictures. In fact, it is difficult to find some details of his travelss due to hazardous interpretations. Finally he made the same mistakes as the Indians pandits for the latitude/longitude measures. The mistake is approximately about a quarter of degree in longitude and that seems too much.


samplea
Concerning Tibet pictures taken by the Landsat satellite, most of the images are well contrasted.
We easily notice mountain ranges followed by valleys with the orography mystery. This will give enough rich map background, where we can identify very easily the tracks to join the valleys, the rivers, and so on. However some ranges are covered with snow in the South - East of Tibet, and consequently the relief details are ignored.

6 Map background dressing, means the addition of the rivers, the tracks, the roads and the toponyms
6.1 Concerning the rivers, Tibet appears to be like an unlimited succession of valleys and mountain ranges. It would be difficult to know whether a watershed has a permanent trickle or not. Most of the valleys I crossed, has dried watersheds that become torrents during the rainy period. Anyway I will give sufficient details to get the best visibility on the map. The reader will easily complete on his own the details from the relief.
6.2 The tracks are passages used by caravans and not accessible by vehicles. If the mountain passes are well identified, the way from one valley to another at mid-altitude can use several tracks. Travellers tell about tracks that can be well distinguished on the satellite pictures. I only kept the most commonly used tracks on the maps.
However with the details given on the map, the reader is able to notice the different possible ways to go from one valley to another. The method used (map background based on a satellite picture) visualizes all the details of an adjacent valley before going in.
6.3 Concerning the roads, those shown as "highways" on the map are simply two way roads where buses can run. The new roads were built by the Chinese in a straight manner between two curves of the valleys. Ancient road locations are difficult to discern. In the past they crossed fields and villages. The main roads are of poor quality except around Lhassa or the main cities. The position of the road in the valley is not always accurate where information does not exist. Rope bridges exist no more and modern metallic bridges have been built.

7 Toponyms
The choiced scale can locate the villages in a target of one hundred meters and raises the problem about their localization. Like for roads and tracks, we cannot know if they are on the left or the right side of the valley. Happily, the position in the middle or at the entrance of a small valley will give a more precise position The villages are often hidden in the landscape. If some villages cling to the hill flanks and can be seen from a distance in some south-west regions, at the center and the west of the country villages composed of low houses with closed walls, melt easily into the uniform brown landscape. If the village is not located near a road where Chinese military vehicles are driven, but in a separate valley, the roofs are decorated with lungtas, or little multicolor flags covered with prayers. In addition it is not certain that the villages crossed one century ago, still exist.
Like any serious work in cartography, a permanent presence and control of human activity is necessary to create maps with 100% of reliability. The village concept is not always respected in so much as a new place-name represents a simple building located along the road, or a nomads camp, or a group of 4 or 5 farms at a place where the valley is enlarged, or at the crossing of two valleys. There is often confusion between the name of the village, of the river, of the bridge, of the mountain, or of the monastery which are all the same. One should notice that two different names might be attributed to the same village, on the way out and at the return of a trip in a valley, or by two different travellers. Such a situation can be confusing.
Generally ancient Dzong where located on a sharp hill or boulder. The village on the foot is not a convenient location for a modern town. New place-names are located either besides the former ones or at the place of a nearby former village (noted with a parenthesis). New prefectures are named with the Chinese suffix "Xian" and the canton names are formed with the Chinese suffix "Qu". Those new toponyms mean only that the staff of the prefecture or the canton are located here and are not used by local people.
Some villages can be found on the map that may be either destroyed, moved, or replaced by new towns. So, I wrote down
- the former village toponyms that seem suspicious before 1990 and appear in italic New Roman font;
- the toponyms belonging to the 1990-1959 period appear in New Roman font;
- the new toponyms (post-59) appear in Arial font with the former name in parenthesis with a New Roman font = Dingri Xian (Shekar Dz); Dingri Qu (Dingri Dz)
- Concerning villages without any known name, their localization is placed on the map.
This work would not be complete if I did not give all my information sources. That is why I adjoined to the maps, an index referencing the bibliography and localisation on the maps.
Concerning the place-names in Tibetan language, I compiled the names most commonly used in the classical literature about Tibet. Recently the TIN' map covered with Tibetan place-names has reached us. On this map, the most common names are not written correctly, either through negligence, or deliberately. Concerning the new toponyms placed on this map, if some have been confirmed by refugees, the others seem to be ignored. Some Chinese names are also included. Concerning the choice of toponyms, there is real work to be done on the land, in order to clarify the obscure points and the questions pointed out about the subject.
Several sets of maps are prepared depending on whether one looks for the toponyms with Tibetan types, Wylie transliteration, or classical Latin types.
Maps can be supplied with their glossary. Here is a sample:

glossary

8- Map printing
The techniques used by printers are inconvenient.
The usage of ink jet printers available in the computer market is easier. In so far as maps cannot be considered as conclusive products, they will need many successive restricted printings.
Concerning the map size, the 1/250.000 scale has been chosen at the beginning to print maps with 1x1 degree in longitude and latitude, considering that the sheet size of the printers is limited to the A2 size (or 42x60cm). This scale will permit to include all the toponyms of the Tibetan center valleys. Later, the 1/200.000 scale will be used when new printers will be available and when printing cost become reasonable.

9- Projects
A Tibetan cartography office is located in the French Riviera area. It is supposed to gather all the necessary information for map design. The office works closely with the Amnye Machen Institute of Dharamsala and several Universities in Europe and United States. The work already done and which remains to be achieved is considerable. Any voluntary help would be appreciated.
First, the office objective is gathering more information than trying to broadcast them. More than 15.000 toponyms files have been created as well as 62 maps at 1/250.000 scale covering only Central Tibet. The office works with a target of about one map per month (80 maps end of 2002). Every map is supplied with a corresponding index indicating the toponym origin. The Tibetan territory coverage will need more than 150 maps.
In 2001, the arrival of new Chinese maps (covered with Chinese signs) may encourage us to pursue the work. These maps include the names of villages and of industrial or administrative sites. They must be converted in Latin or Tibetan signs to be legible. Names of rivers, mountains and holy places are generally missing. An update of our mapping will be difficult as these maps
- change many well known historical tibetan place-names to new ones;
- bring to us new village names;
- but unfortunately half of them are located in areas where there is no human life.
Only a ground study will permit to us to improve the existing work.
It is expected to collect some information from young foreign tourists or trekkers who visit Tibetan valleys every summer. We can even talk about adventurous initiatives a private trip permits to bring back accurate descriptions about the locations visited (as the Indians Pandits did at the end of XIXth century for the India Survey interest).



The Tibet Cartography Office' web site is located at http//www.tibetmap.com (Cl.andre@wanadoo.fr)