- Why do we need a geoid-based datum?
- Who really need a new vertical datum?
- Why not delay the adoption of a new vertical datum?
- Do the elevations of benchmarks have to change?
- How can the confusion of having two vertical datums be minimized?
- Does the geoid-based datum represent mean sea level better than the levelling-based datum CGVD28?
- How do heights estimated by GPS and corrected using a geoid model agree with CGVD28 heights?
- How will I validate the precision/accuracy of my heights when I use a GPS/geoid approach?
- Can I still do surveys with spirit levelling and integrate them into the new vertical datum?
- Will I have to update elevations in my database or my topographic maps?
- How will I maintain compatibility between historical and new height surveys?
- What will be the name of the new vertical datum?
- Why a new name for the vertical datum?
- Will NAVD88 and the new datum in Canada coincide along the border?
2.0 Geoid Modelling / Transformation
- How Precise is the geoid-based vertical datum?
- How can the geoid model be further improved?
- How much will an orthometric height measured with GPS and corrected with a geoid model vary in time (h-dot and N-dot)?
- How is the geoid model validated?
- Without new levelling lines, how will it be validated in the future?
- How often will a new geoid model be published?
- To what 3D reference frame realization is the geoid based vertical datum referred to?
- What method can I use to convert my old CGVD28 elevations to the new datum (or the opposite)?
- I am confused with all these acronyms: NAD83, NAD83(CSRS), ITRF, CGVD28, NAVD88, NGVD29, IGLD85, GSD95, CGG2000, HTv2.0, GEOID03, etc. What is the difference?
- Why the height of benchmarks changes?
Answers - 1.0 User Concerns
Even though the current vertical datum (CGVD28) is very precise over short distances (e.g., 30 km), it includes significant distortions at the national scale. Its access is only available at benchmarks, which are mostly located in southern Canada. Furthermore, the maintenance of a levelling network is very expensive. Thus, the best alternative to levelling is geoid modelling. A geoid-based datum would be accessible by Global Navigation Satellite Systems (e.g., GPS) at any location across the Canadian territory.
Already, a large number of stakeholders rely on GPS as their tool of choice for accurate positioning and require a geoid model to convert their ellipsoidal heights to orthometric heights. In addition, an increasing number of stakeholders are conducting surveys in remote regions where the vertical datum is not accessible through traditional benchmarks.
NRCan has stopped maintaining the levelling network since 1996. It is important that NRCan implements a new datum before the current datum deteriorate to such a level that it might be difficult to have a smooth transition period between the old and new datums. Furthermore, it exists already a large community of GPS users, who require an accurate geoid-based vertical datum across the country.
Do the elevations of benchmarks have to change?
Unfortunately, CGVD28 includes significant distortions across the country. These distortions will be corrected in the new datum meaning that absolute heights may change by almost one metre in certain regions. For several regions, the change in heights will be less than a few cm. The local height differences will remain the same in the two datums.
Having two vertical datums during the transition period may bring confusion to some stakeholders. NRCan and the provincial agencies will indicate clearly the datum of the data when disseminating heights. Also, it is important for stakeholders to identify properly the datum used in their documents. Proper identification will remove a lot of confusions.
First, let’s mention that the mean sea level (MSL) is not a level surface. Just like land, oceans have permanent topography, albeit ranging from -2.0 to 2.0 m worldwide. CGVD28, which is constrained to a series of tide gauges across the country, represents well the MSL at these specific locations. However, these same constrains are also responsible in part for systematic errors in CGVD28. On the other hand, the geoid is a level surface, but it does not coincide with MSL along the coast. However, the geoid-datum will be near the MSL because it will be selected such as it represents the average of the coastal MSL for North America. The geoid-based datum will be about 38 cm above the MSL near Halifax while it will be about 17 cm below MSL near Vancouver. NRCan will make available a model showing the separation between the geoid and MSL along the Canadian coast.
CGVD28 contains systematic errors, which can reach close to one metre in an absolute sense. However, local GPS-corrected height differences will agree well with CGVD28.
The stations from the federal Canadian Base Network (CBN) and provincial High Precision Network (HPN) can be used by stakeholders to validate GPS/geoid procedure for the determination of accurate 3D positions. In addition, several benchmarks have 3D positions; however, the quality of these 3D positions may vary depending on the GPS epoch of observations. Also, one has to be careful about the stability of the benchmarks.
Yes, most benchmarks will have an elevation in the new datum. If there are no benchmarks within a reasonable distance from your project, you can install your own control stations by GPS in the project area and resume the work locally by levelling technique.
It will depend on the accuracy of your datasets. It might not be necessary if the changes are smaller than the error associated to your datasets. A conversion tool will be made available in order to convert heights between the two datums. If your datasets are local, the conversion can be as simple as adding a constant.
The compatibility between historical and new height survey can be maintained by conducting the first new survey in the two datums. This will allow you to determinate the relation between historical and new data. If the project is local, the conversion should be as simple as adding a bias to the old or new datasets.
The name of the new vertical reference system is the Canadian Geodetic Vertical Datum of 2013 (CGVD2013). The new vertical datum should be identified as follows: CGVD2013(CGG2013), where the name between parenthesis is the geoid model realizing the vertical reference system. If NRCan updates the geoid model in 2018, the vertical datum would be identified as CGVD2013(CGG2018).
The new vertical datum must be identified by a different acronym because CGVD28 and CGVD2013 are not in the same reference system, i.e., they do not have the same definition. In this case, we are not talking about levelling versus geoid modelling because these are only techniques to realize the vertical datum. Rather that CGVD28 is defined by MSL at five tide gauges across Canada while the new datum is defined by the equipotential surface (W0 = 62,636,856.0 m2/s2), which represents conventionally the mean potential of the coastal MSL for North America.
NAVD88 and CGVD2013 will not coincide along the Canada/US border. NAVD88 has a significant east-west systematic error, which indicates that MSL next to Vancouver is higher than MSL next to Halifax by 1.5 m. On the other hand, CGG2010 indicates a difference of about 0.55 m.
Answers - 2.0 Geoid Modelling / Transformation
The geoid-based datum will have a better absolute accuracy than CGVD28 across the country. Still, the geoid model will not be errorless, but no systematic errors should be larger than the decimeter (95% confidence) across the country. CGVD28 have systematic errors reaching close to the metre level. Overall, the accuracy of the geoid model will be approximately 3 to 5 cm. On the other hand, the relative precision of the geoid model will be comparable to spirit levelling. Naturally, the relative precision of the GPS-derived orthometric heights will also depend on the precision of your ellipsoidal heights.
Between the late 1980s and today, geoid models have improved quite significantly. We saw changes at a level of a few metres. Today, the theory takes care of many terms that were formally omitted or neglected; more gravity data are available from terrestrial and spatial techniques; and better Digital Elevation Models (DEM) are available. This new information is currently stabilizing the realization of the new geoid models. Furthermore, geoid modeling is gaining global acceptance as the future technique to define national or continental vertical datum. Thus, national and international academic institutions and governmental agencies are developing new techniques to achieve higher accuracy in geoid modelling. NRCan is keeping abreast with all new developments.
The Earth is a dynamic planet; it is always changing. Some changes can be quite drastic (e.g., landslide, earthquake) while others are can be more subtle (e.g., post-glacial rebound). When you have drastic events, the benchmarks can move significantly or be destroyed completely. In this case, you loose access to the vertical datum. On the other hand, the impact of a drastic event on the geoid is very small; thus, it is possible to take new GPS measurements and install new control stations immediately. Subtle changes are difficult to detect because they expend over a very large regions. These changes are usually not detected when conducting local relative surveys (e.g., levelling or differential GPS). However, GPS technique such as Precise Point Positioning (PPP) will show that the terrain can move by as much as 1 cm per year. This dynamic change of the topography will also bring a 10% change to the geoid. Thus, the long wavelength components of the geoid can change by approximately 1 cm every ten years.
A geoid model can be validated in two ways: error propagation and independent datasets. For the former, the challenge is to associate a realistic error model to the input data required in the determination of a geoid model. This internal accuracy can be too optimistic because it will not consider systematic errors and omissions. For the latter, GPS on BMs is the most common approach. It consists of comparing the geoid models (N) to geoid heights determine from GPS ellipsoidal heights (h) and spirit leveled orthometric height (H): h - H - N = ε. The discrepancies ε should be zero (or a constant) if each height would be errorless. The problem with this technique is the difficulty to disassociate errors from the geoid model, levelling data, GPS measurements and stability of the BMs. Other independent techniques for validation could be satellite radar altimetry and astro-geodetic deflections of the vertical.
Existing benchmarks (BMs) are here to stay for many more years and most of these BMs are fairly stable. Even if we do not conduct new levelling surveys, we still have a lot of levelling lines to validate geoid models for many more years. In addition, we can also validate the geoid model at tide gauges. By the time that most BMs will disappear, validation of geoid models by "GPS on BMs" will not be a priority.
We know that it is impossible to realize an errorless geoid model. The model will always be as good as the theory and input data. Certainly, these two elements will improve with time. It is hard to say at this time how often a new model will be published; however, we do not expect that the changes in geoid models will not be larger than what we saw with CGVD28 over the last 75 years. Furthermore, models will not be published at a higher rate than currently, which is approximately every five years.
A geoid model is determined from gravity measurements. Knowing that gravity points towards the center of mass, we assume that the reference frame is ITRF. Usually, we associate it to the latest ITRF realization (e.g., ITRF00). The geoid model is converted to NAD83 (CSRS) using a seven-parameter transformation (rotations, translations and scale). The epoch of the geoid model is determined from the observations period of the satellite data.
There are three approaches. First, you can conduct your own GPS surveys on a series on BMs in your area of interest and determine the separation between the two datums (ε = (h-N)new Hold). The advantage is that it will convert any local datums to the new datum. The disadvantage is that you must be able to determine accurate ellipsoidal heights in the proper reference frame. A second approach is to download benchmarks information in your area. Most BMs will have published heights in the two datums (CGVD28 and CGVD2013). Finally, a third approach is to use a grid shift file. This file will be produced by differencing geoid model CGG2013 and hybrid geoid model HTv2.0. HTV2.0 is a realization of CGVD28. The transformation will be fairly precise in regions with dense levelling lines, but the precision will deteriorate in regions with sparse or no levelling lines.
Traditionally, geodesy consists of two types of reference frames: a 2D horizontal network (latitude and longitude) and a 1D vertical network (height above mean sea level). Today, with the immersion of satellite positioning, we are now taking about 3D networks (latitude, longitude and ellipsoidal height).
- NAD27 and NAD83 are traditional horizontal networks while NAD83(CSRS) and ITRF are modern 3D networks. The horizontal components of NAD83(CSRS) are more precise than those of NAD83. NAD83, who was though to be at the Earth center of mass at the time, is actually off by approximately two metres. ITRF is a global reference frame with its origin is at the center of mass (±2 cm). There are several realizations of ITRF (e.g., ITRF97, ITRF00). These are new versions which are more precise than the previous ones. Coordinates in ITRF can be converted to NAD83(CSRS), or the opposite, using a seven-parameter transformation. WGS84 is an ITRF reference frame.
- CGVD28, NGVD29, NAVD88 and IGLD85 are traditional vertical networks. These networks are realized by spirit levelling. CGVD28 is the current vertical datum for Canada. NGVD29 is the former vertical datum for the USA. That datum was replaced in 1995 by NAVD88, which is a minimum constrain adjustment of the levelling data in North America. IGLD85 is a special vertical datum for the St-Lawrence Seaway and Great Lakes.
- GSD91, GSD95, CGG2000, CGG2005, HTV2.0 and GEOID03 are geoid models. A geoid model is an integral part to a 3D network to convert ellipsoidal heights to orthometric heights (heights above a vertical datum). GSD91, GSD95, CGG2000 and CGG2005 are pure gravimetric geoid models. Each new model is a better representation of the geoid. CGGxxxx is now the standard to identify gravimetric geoid models in Canada. HTv2.0 and GEOID03 are hybrid geoid models, i.e., they are distorted geoid models to represent a spirit-levelled datum. HTv2.0 is a representation of CGVD28. HT stands for Height Transformation. GEOID03 is a representation of NAVD88 in the USA. In Canada, we opted not to name our hybrid models geoid because these are distorted to represent a vertical datum, which includes systematic errors.
There are basically three reasons: 1) new reference system; 2) new realization of the reference system; and 3) the Earth is a dynamic planet.
- Changes due to a new reference system are rare because we rarely adopt a new reference system. For example, 3D coordinates are different between NAD83(CSRS) and ITRF because these two reference frames do not have the same origin (about 2 m apart).
- Changes due to a new realization of the reference system are more comment because all new observations will modify the reference frame. For example, control stations will have new updated coordinates based on more precise observations.
- Finally, the Earth is not static. There are earthquakes, landslide and post-glacial rebound to name a few natural changes to the planet. There are also some changes due to human intervention. For example, there is subsidence due to mining and oil exploration or construction of major hydro projects.