Datum Shifts - or ways to transform between datums...(centimeter level differences), except the...

35

Transcript of Datum Shifts - or ways to transform between datums...(centimeter level differences), except the...

Page 1: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’
Page 2: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Datum Shifts - or ways to transform between datums

• Modern datum shifts (NAD83(HARN/96/2007/2011, ITRF, WGS84) use the transformation equation X’ = Ti + sR Xi we’ve been talking about between datum families (NAD83xxxx to ITRFyyyy or WGS84zzzz)

• Transformations from legacy datums in NAD27, NAD83(86) use a different, grid-based method (highly empirical)

• Transformations within NAD83xxxx (e.g., NAD83(1996) to NAD83(2007) aren’t all published, can use and average local offset for a small area, or sometimes grid-based solution

• ArcMap doesn’t apply actually shift anything between post 1986 NAD83 datums, although it will tell you they conflict

Page 3: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Transformation among “Families” of Modern Datums

Page 4: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’
Page 5: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

We can only use our equation when the parameters have been estimated, that is, we know Ti, s, and R

We do use it for different versions of WGS84/ITRF, and for conversion between WGS84/ITRF and NAD83(CORS96) and later NAD83

We don’t use it within NAD83, or from the old NAD83(86) to WGS84 or ITRF

Transformation Between Datums

X’ = Ti + sR Xi

Page 6: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Transformation Between Modern Datums• Unfortunately, plate tectonics and earth rotational variation means we can’t assume R

matrix is constant….. NAD83 and ITRF address this differently.

• NAD83 datums specify a fixed North American plate, so Rx is “constant”, points don’t “drift” much through time. Coordinates have changed through time, but in non-constant ways, NO datum transformation among NAD83(xxxx), (e.g., NAD83(1996) to NAD83(2007)) - Ignore because shifts are small, or use local Bench Marks to calculate a local shift

• ITRF/WGS84 set “average” continental drift to zero, so R matrix changes with time. There are “standard” R and DR for among years of ITRF/WGS84 (e.g. ITRF00 to ITRF01).

• Datum transformations published for NAD83xx to ITRFyy (e.g., Snay readings). Since ITRF points drift, most precise transformations use R and DR, but drift effects small for most areas (except in tectonically active areas like California)

• Sometimes must apply several successive R matrices in a path from one Datum to another

• For older datums (NAD27, NAD83(1986)), a 3D transformation won’t work (they weren’t fit to a globe) - use a grid or regionally-fit conversion

Page 7: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

WGS84/ITRF Datum Transformations?

WGS84 and ITRF were aligned (unified) in 1992, 1994, 2000, and 2008, and likely will be in the future

Most versions are equivalent for nearly all GIS work (centimeter level differences), except the first WGS84

ITRF are fit every year or so, and the T and R matrices published

Page 8: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

X’ = Ti + sR Xi

XITRF2008 = Ti + sR XITRF2000

For ITRF/WGS84, there are proper datum transformations

Page 9: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Transformations from ITRF/WGS84 to NAD83(CORSxx)?

Find relevant s, R and T, apply our equation, e.g.,

XITRF2008 = Ti + sR(t) XNAD83(CORSxx)

We often need to do this transformation, because we collect GPS/GNSS data in WGS84, work in NAD83, and the shifts are often near a meter

Page 10: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Transformation Between Datums

XNAD83ITRF89 = Ti + sR XITRF96

Page 11: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

A First Complication: Continents Drift

We can estimate how fast, and slightly modify the R matrix to reflect changes in the translation, rotation, and scale

XITRF2008 = Ti + sR(t) XNAD83(CORSxx)

The translation and rotation matrices change with time

Page 12: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Greenwich Meridian

Page 13: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

NAD83(CORS96) Velocities Relative to ITRF

Page 14: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Datum Transformation Parameters (R, T) fit for ITRF96 to NAD83(CORS96)

T, v values depend on time since 1997

Page 15: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

We use our usual equations, but the coefficients of the R matrix are time dependent

Modify our R by their velocities and time

- note that ArcMap/Projection doesn’t implement this - uses a fixed R

Page 16: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Complication number Two:We Sometimes Need Multiple Datum Transformations

If I collected data in ITRF06, and want to convert to NAD83(CORS96), how do I?

Transform ITRF2006 x,y,z coordinates to ITRF96, then ITRF96 x,y,z to NAD83(CORS96)

R, T matrices not calculated for all versions of NAD83xxxx to all versions of ITRFyyyy

R, T available from NAD83(CORS96) to ITRF1996

R, T available from all ITRF to all other ITRF

Page 17: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’
Page 18: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

How  Big  is  the  Shift  from  NAD83(96)  to  WGS84(1150)  or  ITRF00?  

in  Minnesota,  about  a  4  feet

http://www.ngs.noaa.gov/TOOLS/Htdp/Htdp_transform.html

First

Then

http://williams.best.vwh.net/gccalc.htm

Page 19: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Datum Shifts - or ways to transform between datums

• Modern datum shifts (NAD83(HARN/96/2007/2011, ITRF, WGS84) use the transformation equation X’ = Ti + sR Xi we’ve been talking about between datum families (NAD83xxxx to ITRFyyyy or WGS84zzzz)

• Transformations from legacy datums in NAD27, NAD83(86) use a different, grid-based method (highly empirical)

• Transformations within NAD83xxxx (e.g., NAD83(1996) to NAD83(2007) aren’t all published, can use and average local offset for a small area, or sometimes grid-based solution

• ArcMap doesn’t apply actually shift anything between post 1986 NAD83 datums, although it will tell you they conflict

Page 20: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Why is NAD83 from 1986 to present so complicated?

Datum shift is large, up to 2 meters for NAD83(1986) to the other NAD83 datums, e.g., NAD83(CORS96)

First local fit by states with NAD83(HARN) - then different versions of NAD83 datums fit in 1993, 1996, 2007, and 2011

Didn’t use Earth centered approach for NAD83, errors aren’t globally predictable - NO R or T exists!

In ArcGIS, though, Datum Transformation is worse than useless - R matrix values all 0 for NAD83(86) to NAD83(96) - implies they are equal when they’re not.

What to do if you have older, high-quality data to bring forward, or newer data to “take backward?” Use data sheet, local shift

NAD83(86) is pre-satellite, and imprecise

Page 21: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

2nd column pg. 1, start of pg. 2:

This means a datum transformation is only exactly valid for the exact point in time (epoch) used in development

Page 22: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’
Page 23: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Conclusion - NAD83 since CORS96 are basically equal

Page 24: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

http://www.ngs.noaa.gov/NGSDataExplorer/

Page 25: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’
Page 26: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’
Page 27: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Dx = -11.5 cm, Dy = 19.5 cm

So add -11.5cm to all 1986 x coords to convert to 2011 add 19.5cm to all 1986 y coords to convert to 2011

Shift from NAD83(86) to NAD(2011) in St. Paul:

Page 28: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

NAD83(CORSxx) been fit at least five times (HARN, 1993, 1996, 2007, 2011).

No datum transformation published between fits (no R, T matrix). Points recalculated for US control network, and HTDP software provided to estimate drift between epochs (measurement times)

Most GISers don’t bother with datum shifts,(outside of California): T is 0, s is 1, and shifts usually less than 5 cm (2 inches) over a decade

How to transform between NAD83(CORSxx) datums? e.g., NAD83(CORS96) to NAD83(2011)?

Page 29: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

NAD83(CORS96) Velocities in NAD83(CORSxx) Datum

Page 30: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

e.g., ArcGIS provides “bookeeping” transformations, with T = 0, R= 0 rotations from NAD83(CORS96) to NAD83(2007), so the datum transformation does nothing in those cases

Real shifts between NAD83xxxx versions are small most places, < 5 cm

What about areas with big shifts?

Use HTDP, or NGS control sheet to estimate x,y,z or latitude/longitude offset, and apply these as simple addition/subtraction to all data coordinates

Page 31: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’
Page 32: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Dx = 0.9cm, Dy = 1.4 cm

So NEAR the St. Paul Campus, add 0.9cm to all 2007 x coords to convert to 2011 add 1.4cm to all 2007 y coords to convert to 2011

Shift from NAD83(2007) to NAD(2011) in St. Paul:

Page 33: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

If  not  near  a  benchmark,  use  NADCON  program  (Windows  download  or  online)  

Shift  from  NAD83(86)  to  NAD83(96)?  in  Minnesota,  about  a  foot

Page 34: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

To reiterate: We often don’t use datum transformations among NAD83 datums, e.g., NAD83(CORS96) to NAD83(2011) - no R, T matrix estimates, xy shifts are small

For modern NAD83 datums, versions HARN and 1993 through 2012, consider them equivalent, or if you must, use a benchmark to calculate a local shift

For NAD83(1986), use a benchmark/local shift, or use the program NADCON to calculate a local shift

Page 35: Datum Shifts - or ways to transform between datums...(centimeter level differences), except the first WGS84 ITRF are fit every year or so, and the T and R matrices published X’

Transformation Between Datums• For modern Datums (NAD83 post 1993, WGS84, ITRF), use the X’ = T +sRX equation

described over past three weeks to go between NAD83 and ITRF/WGS84

• Unfortunately, plate tectonics and earth rotational variation means we can’t assume R matrix is constant….. NAD83 and ITRF address this differently.

• NAD83 datums specify a fixed North American plate, so Rx is “constant”, points don’t “drift” much through time. Coordinates have changed through time, but in non-constant ways, NO datum transformation among NAD83(xxxx), (e.g., NAD83(1996) to NAD83(2007)) - Ignore because shifts are small, or use local Bench Marks to calculate a local shift

• ITRF/WGS84 set “average” continental drift to zero, so R matrix changes with time. There are “standard” R and DR for among years of ITRF/WGS84 (e.g. ITRF00 to ITRF01).

• Datum transformations published for NAD83xx to ITRFyy. Since ITRF points drift, most precise transformations use R and DR, but drift effects small for most areas (except in tectonically active areas like California)

• Sometimes must apply several successive R matrices in a path from one Datum to another

• For older datums (NAD27), a 3D transformation won’t work (they weren’t fit to a globe) - use a grid or regionally-fit conversion