Taiwan GPS and TWD67, TWD97, WGS84 coordinate transformations


This page should not be maintained by aging confused amateur me.

Modern Taiwan maps will only have the following coordinates, if any:

  1. TWD67
    1. longitude/latitude
    2. 2-degree transverse Mercator
  2. TWD97/WGS84
    1. longitude/latitude
    2. 2-degree transverse Mercator

How can I tell if a map is TWD67 vs. TWD97?

TWD67 examples:

If the above are off by about 828 meters then the map is TWD97.

TWD67, TWD97 confusion can delay rescues

If you mix up TWD67 and TWD97, your results could be off by a kilometer -- dangerous [Chinese].

So, these days, beside longitude and latitude, one must also mention "TWD67" or "TWD97" for the data point to be useful. TWD97 A responsible map publisher should print both grids...

2007/3 To sum it up, rescuers should check three places:


They report

Our map is

We should check
WGS84 WGS84 original X,Y
WGS84 TWD67 X-828m, Y+207m
TWD67 WGS84 X+828m, Y-207m
TWD67 TWD67 original X,Y

Recommendation to rescuers: just like medical workers still will test the patient's blood type before giving them a transfusion even though they insist they are type AB etc., we shouldn't trust the reporting public's vague notions of datums, if any. Nor should we trust various rescue teams with piles or PDAs of various maps. Therefore the helicopter should check in the above three places, which form a west-northwest - east-southeast line. During the process the coordination center should not "help" anybody by converting the original reported coordinates, lest confusion increase exponentially. Note however that I have no rescue experience and the above is all a guess as to how things should operate.

Basic facts for both TWD67 and TWD97

Latitude origin 0 degrees
Central meridian 121 degrees E
Scale factor 0.9999
False easting 250000 m
False northing 0 m

They are both 2 degree wide transverse Mercator projections. They differ in their ellipse.

Is there an origin (0,0)?

Go to the junction of the equator and 121 degrees east longitude, then point one's head west and go straight 250 km. When one stops one will be considerably above the earth's surface. I'm pretty sure that is the "origin".

If I go due east or north, will that increase TM2's E and N values?

Not exactly. Take a good look at a map. The only place where the longitude and latitude grid lines are parallel to the TM2 grid lines is 250 km E., i.e., 121 east longitude, that single line. All other lines are more or less askew.

Coordinate Transformation

Between latitude/longitude and 2-degree transverse Mercator projection

proj is free software:

GNU/Linux example, Dongjundashan TWD67: given x and y, compute longitude and latitude (and store them in variable LL):

$ LL=$(echo 258566.571 2613894.788 |
proj -I +proj=tmerc +ellps=aust_SA +lon_0=121 +x_0=250000 +k=0.9999)
$ echo $LL
121d5'2.255"E 23d37'42.655"N

Now convert back to x and y:

$ echo $LL |
proj +proj=tmerc +ellps=aust_SA +lon_0=121 +x_0=250000 +k=0.9999
258566.57 2613894.80

(If we were using TWD97 instead of TWD67, we would instead use:)

$ proj +proj=tmerc +ellps=GRS80 +lon_0=121 +x_0=250000 +k=0.9999

The ellipses we use can be seen with "proj -le":

Q: Where can I see the formula for latitude longitude to TM conversion?

A: Well, since it is Free Software, perhaps see the source code to proj ? Also note freegis.org .

GPSFUN has a comparison like:

Datum Reference spheroid major axis "a" flattening "f"
see note * see note * 6378388 1/297
TWD67 GRS67 6378160 1/298.25
TWD97 GRS80 6378137 1/298.257222101
WGS84 WGS84 6378137 1/298.257223563

* note: the names "hu-tzu-shan", and even "International 19xx", have been given many definitions here in Taiwan, so are dangerous to use. One must speak in terms of a, f !

"hu-tzu-shan" sometimes even means TWD67.

Online TM to lat/long transformation (for TWD67 to TWD67, TWD97 to TWD97 will be off by .0001 degree), [Chinese] but no source code, no decimals below 0.0001 degrees, no clear website name.

TWD97 vs. WGS84

TWD97 is very close to WGS84, at least where I live,

$ a="120d51'58.2\"E 24d10'54.2\"N"; echo $a #(just see what it looks like)

120d51'58.2"E 24d10'54.2"N

$ echo $a | proj +proj=tmerc +ellps=GRS80 +lon_0=121 +x_0=250000 +k=0.9999 -f %.5f

236402.76293 2675155.08596

$ echo $a | proj +proj=tmerc +ellps=WGS84 +lon_0=121 +x_0=250000 +k=0.9999 -f %.5f

236402.76293 2675155.08604

$ echo 2675155.08604 - 2675155.08596 | bc

.00008

Zones

gpsman author "M" == Miguel Filgueiras writes:

M> As far as I understand from your description these 2 degree zones centered from 115 to 125 will cover longitudes from 114 to 126. Does these cover all Taiwan territory?

yes, everything that Taiwan owns or still claims to own.

M> I also take it that these zones cover all the latitude range, that is, for each central meridian there is a single zone for all the possible latitudes in Taiwan.

Yes, and each has Y=0 at the equator, and X=250000 at the central meridian.

M> Another point: could you please confirm that the correct scale factor is k0=0.9999.

yes, that's right, 0.99990 And note that these 6 zones can be used with both the TWD67 or TWD97 datums.

I discovered many maps of Jinmen , at least as late as the 1990's still use an older system:)

$ proj +proj=utm +ellps=intl +lon_0=117
# intl a=6378388.0 rf=297. International 1909 (Hayford)

TWD67 -- TWD97 transformation: let's write a public version

(3/2008: Government's closed-source version: .PDFs, .ZIPs. Anyways, ignoring them we do it by hand:)

This is the best I can do at present, for all of Taiwan, with its few meters of error. (828: Odd, all the few formula I have below give approximately 841, why?)

Also john at thl.ncku.edu.tw 8/2003 contributed:

A = 0.00001549
B = 0.000006521
X67 = X97 - 807.8 - A * X97 - B * Y97
Y67 = Y97 + 248.6 - A * Y97 - B * X97
X97 = X67 + 807.8 + A * X67 + B * Y67
Y97 = Y67 - 248.6 + A * Y67 + B * X67
Perl implementation

2/2006: Now there is Perl Geography-NationalGrid-TW which perhaps solves many problems discussed by this page. 2007/10: xylonlat is an example program I wrote using it.

Minstrel made programs, 6/2004.

2007: note http://www.mobile01.com/topicdetail.php?f=130&t=374120

Vertical transformations

I'm not sure what the official Z axis transformation is for TWD67--TWD97.

Few open transformation programs

There are various programs here in Taiwan to convert from TWD67 to TWD97, but their authors or organizations are not willing to release them. Also most are for Micro$oft. Therefore, I wish there would be someone who understands coordinate transformations to step forth and under the free software license write a TWD67 -- TWD97 transformation program and also put its source code on the web. Who knows, perhaps in the future everybody will use ours!

Tools might only be for Windows.

From a fortran program I was given, I guess proj's cs2cs parameters might be

$ cs2cs +proj=tmerc +ellps=aust_SA \
+towgs84=-764.558,-361.229,-178.374,-.0000011698,\
.0000018398,.0000009822,.00002329 +lon_0=121 +x_0=250000 \
+k=0.9999 +to +proj=tmerc +datum=WGS84 +lon_0=121 +x_0=250000 +k=0.9999

The results and transfer.for's are close : for the whole island around x: -841, y: 207, (z: varies a lot). Maybe someone who understands this could lend a hand.

Note that the above is a seven parameter transformation different than:

dx, dy, dx, da, df; Garmin GPS

Chatting with M.H. Ho of Taiwan Garmin, in 2000 :

Q: I need dx,dy,dz,da,df for Taiwan Hu-tzu-shan datum (note: "Hu-tzu-shan datum" is Garmin's name for TWD67 )

A: For GARMIN GPS:

  • dx: -767 m
  • dy: -358 m
  • dz: -176
  • da: -18.00
  • df: -0.00081204

[Note from Eino Uikkanen.]

Q: Garmin's GPS units, at least to the year 2000, have the flaw that when we select "Taiwan Grid", the unit displays "T67". Note that that should be related to the "Hu-tzu-shan datum" [Garmin's name for TWD67]. How could one show a "67" when one selects WGS84? [The "67" has nothing to do with the grid. We use the same grid (X shifted 250000m etc.) for both TWD67 and WGS84 datums.] Of course we can still use the unit, it just looks bad.

A: Thanks for your letter. Indeed when one switches to WGS84, still showing T67 can create confusion. I will report this to our company. Thanks.

Perhaps Garmin still hasn't fixed this. Not only does it look bad, but it also creates danger on the trail.

2003.6.26 Cygnet said: After bugging them about it so long, new Garmins now say "TM2".

So, new Garmin models display:

Old/New Grid
Lat,Lon TM-2
Datum TWD67 N,E/N,E T67/TM2
WGS84 N,E/N,E T67/TM2

They no longer use the name of a datum for a grid. However one still can't tell what datum has been chosen from this display.

2004/3: Garmin products now have an additional "TAIWAN" datum now along with their original so-called "Hu-tzu-shan" datum. Apparently this new "TAIWAN" datum is supposed to be just a minor correction to their so-called "Hu-tzu-shan" datum. The instruction books apparently say this "TAIWAN" means TWD67, which would be apparently correct. (Why doesn't Garmin just name it directly "TWD67"?)

unofficial Garmin GPS cheap connector [Taiwan]

Regarding DA, DF, DX, DY, DZ (but beware of any 1924 based Hu-tzu-shan datums listed there).

Cadastral coordiantes

If in a land office you stumble upon an old grid that uses 20/11 meters per unit, that is the so called Cadastral coordinates.

Programs: cad2twd67

Turn TM or longitude/latitude into Taiwan map numbers

$ echo 120.85788 24.18347|./longlat2taiwanmapno
#1> 120.85788 24.18347
9521-2-025 9521-2-09 9521-2-NW
$ echo 9521-2-025|./taiwanmapno2longlat
#1> 9521-2-025
120.850 24.175

Taiwan Triangulation Points' Coordinates

TWD67 Taiwan Triangulation Points' Coordinates

Taiwan map curiosities

Orthophotomaps are labeled from 1-100 instead of the more systematic 0-99. See Chinese version.

Circles

Odd, on maps from the 1980's there is traces of a 1 km. diameter circular road centered at TWD67 213230 mE, 2676040 mN, (north of Fengjia Univ., Taizhong). Traces of an old cyclotron?

My friend guesses it might be a high frequency direction finder, like the ones in Linkou, TWD67: 2776.5 km N 289 km E, used in cooperation with the USA in coordination with others in Asia to triangulate signals from China. However the diameter of the largest Linkou one is only 1/3 km.

Daya Interchange distortion

Why isn't Taizhong's Daya Interchange on Zhongqing Rd. but instead on an extra side road? Looks so curious on a map. Doesn't that just increase the number of red lights? See Chinese version.

Etc.

The finder of 24N 121E reassures me he was using WGS84 and not TWD67...

Working with Coordinates and Units

bnpcoa2shp: Taiwan cadastral .coa, .bnp format conversion to Shapefile.


Up one level

Dan Jacobson

Last modified: 2008-03-26 06:47:24 +0800