Where is mean sea level, and why it matters at every stage of a mangrove project
Sea level - it’s a simple concept that almost everyone understands to some extent. It’s a convenient reference for measuring heights and altitudes. And for coastal ecosystems, where vegetation habitat is defined by inundation periods, storm surges can flood cities, and almost everything is continuously in a state of change, so it's critical to measure correctly to predict sea level rise risk, outcomes from extreme weather events, and plan coastal restoration and infrastructure projects. But with data sources from satellites, GNSS receivers and field measurements using different height datums, and sea level always changing with tides, it's not always so easy.
As an example, the image below shows contour lines from a high resolution lidar survey of a mangrove environment, with all trees removed to show the bare terrain underneath. Heights are all relative to a global ellipsoid, direct from the geodetic lidar measurements, which is why they are all negative. The only sites with mangrove growth were above -44.2m in altitude, with heights of -44.4 assessed as the marginal lower limit for the most water adapted species such as Avicennia. Lower than this was all mudflat and not viable at all for mangrove restoration. The difference between mangrove habitat and mudflat is only about 20 to 40cm, and this difference in height is what makes hundreds of hectares across the project unviable for mangrove growth, all directly resulting from the link between local topography, sea level and tidal range.
In this article, I will go over why this is so important, and how common height datums relate to meaningful sea level measurements.
Mapping local mean sea level reliably to site topography is the foundation of all coastal hydrological assessments, affecting:
Site feasibility: This is crucial for correlating site vertical height GPS measurements to local tide ranges. With this you can check at what tide heights (and inundation periods) natural mangrove species are establishing, and use this to inform zonation for planting or plan other interventions such as channel digging.
Current risks: This data is needed to investigate the effects of storm surges and the probability of flooding events in coastal communities or extended inundation and mangrove drowning events.
Future risks and forecasting: Over longer periods, this is a critical factor for sea level rise risk analysis and forecasting any migration or change in mangrove habitat areas, which can be key to building financial investment cases for projects.
On the 3rd point above, relative sea level rise (RSLR) is extremely important to assess the permanence of the carbon stored by the mangroves for any type of project, and is actually a requirement under Verra’s VM0033. It is a key part of Verra's AFOLU Non-Permanence Risk Tool (v4.2). The methodology requires project developers to determine how much land will be lost to sea level rise over the 100 year permanence period due to wetland migration, inundation, and erosion, requiring robust links between sea level rise predictions, local topography, and local mean sea level.
Mangroves, falling sea levels, and 40 million dead trees
The world’s largest recorded mangrove dieback event wasn’t manmade clearing, but actually driven largely by falling sea levels in the Gulf of Carpentaria in Australia in 2015. A specific combination of occurrences, including El Niño conditions and part of the moon’s 18.6 year “wobble” cycle, resulted not only in slightly lower mean sea levels for a period of 1 to 2 years, but also high tide levels around 20 to 40cm lower than normal. 40cm is much smaller than the uncertainty levels for any global digital elevation (DEM) product, smaller than the discrepancy between mean sea level and a geoidal model vertical datum (see below for an explanation), and yet enough to kill 7000 hectares of mangroves over 1000km of shoreline in northern Australia.
In this case, the cause is a natural fluctuation in sea levels and tides (with some component made worse through changing climatic conditions). This natural fluctuation (generally only a few centimetres in sea level), affects mangrove forest canopy extent across thousands of square kilometres in northern Australia in a long period cycle.
What does this mean for mangrove project sea level measurements? If your local value for mean sea level, and with that the estimates of tidal range, is off by a matter of centimetres, this can result in large planted areas which do not grow at all. Or, even worse, if this isn’t accounted for in long term risk models, this can result in diebacks of restored mangroves over larger areas at higher frequencies than the models predict (whether its from future sea level changes due to the moon’s cycle, sea level rise, or increased inundation extent of storm surges). For project developers, financers and insurers, a sound night’s sleep with the knowledge that these critical hydrological risks are well accounted for rests fully on knowing exactly where the sea level is.
LIDAR, tides and bare earth - oh my
Let me give an example, based on a true story which threw me down the rabbit hole of GIS datums and tide gauges, searching for water level references rather than a white rabbit. Let’s say you have a LIDAR derived digital terrain model (DTM) from an intertidal zone, and you want to draw a contour at the mean sea level height to clearly divide the lower tidal zones from the upper tidal zones. You’ve applied PPK (Post-Processing Kinematic) corrections to the resulting point cloud (deviations in the order of mm), using a base station positioned using PPP (Precise Point Positioning, reported height standard error of <3cm). You’ve verified the DTM using PPK corrected GNSS rover field measurements. With all of this, you now know, with pretty damn good certainty (<10cm), where the ground is.
But your lidar and GNSS measurements use the ITRF2020 reference frame, the default option in the coordinate reference frame box when you export the data (if this means nothing to you, see the explanation further on in this article). All the values are around -20m. So your high accuracy DTM right now does not map cleanly to the idea of “high” and “low” tides. You need to convert the ITRF2020 data to a local mean sea level.
After some quick googling, you see many articles describing geoidal height data as “relative to sea level”. Perfect, you just need to apply a geoidal correction to the DTM values (EGM2008 seems suitable). But when you look at your maps, you see a lot of mangroves growing in the lower tidal range (which doesn’t match with field visits), predictions of total site flooding at the upper tidal values and permanent water cover at sites you have personally walked over. The data is indicating the site today is already not suitable for mangrove habitat, and applying 50 year predictions of sea level rise just make it even worse.
Maybe this can be corrected using local tide gauge data. But no. The nearest tide gauge is 100km away, and when you look at the data, it doesn’t match field measurements at all. Also, the tide gauge data is only provided relative to historical mean sea levels, which you can’t map to your GNSS measurements. You then try a global tide model, but its predictions are all relative to a local mean sea level, which is exactly the issue in the first place.
So, to join all these data sources together, you need to answer a question that sounds deceptively easy. What is the local mean sea level?
Ellipsoids, geoids, and cows that aren’t spherical
Raw GPS and satellite measurements are uncorrected for their location on the planet, and are typically delivered relative to a global ellipsoid height datum. The ellipsoid is a simplified ellipsoidal representation of the earth (as it is not a perfect sphere), and using an ellipsoidal height datum is easier for satellite based measurements, as satellites themselves orbit around the planet’s centre of mass. The ellipsoid is smooth, geometric, mathematical, a “spherical cow” that simplifies analysis and avoids complexities of reality. For GIS applications, WGS84 is a common ellipsoidal height based reference frame alongside ITRF2020, which is also commonly used for satellite altimetry datasets or lidar point clouds. WGS84 and ITRF2020 (or the earlier ITRF2008) are roughly equivalent to within a few centimetres, so can be considered the same for most purposes but care should be taken if using altimetry products at the scale of centimetres or millimetres.
In reality there are no “spherical cows” however, and the earth’s shape is not an ellipsoid. It is far more valuable to have height measurements relative to a more locally meaningful datum, and for this, typically a geoid datum is used.
The geoid is generally defined as the shape the ocean would take due to gravitational forces alone, excluding the influences of tides, winds and currents, and so it is based on gravitational measurements of the planet. This means it is a “rough” good approximation for sea level, and most of the time measurements and heights reported relative to mean sea level (MSL) or above sea level (ASL) are heights relative to a geoid model datum. The planet’s “true” geoid is a very complicated shape that changes slightly over time due to geological forces, so typically the geoid is represented as a model. One of the more common models used as a datum is EGM96, a geopotential model from NASA and used in the shuttle SRTM dataset, however the newer and higher resolution EGM2008 (used for FABDEM) is also fairly common. The geoid differs from the ellipsoid by roughly +85m in Iceland to -106m in southern India.
For most terrestrial applications, geoid based heights are more than sufficient, but it is very important to note that the geoid is only close to, not exactly the actual mean sea level (or sea surface). The difference between the two is generally around 1 to 2 metres. And for intertidal applications, where differences in the order of 30cm can mean the drowning of mangroves or increased damage from storm surges, this difference becomes critical. The misconception that the geoid is exactly mean sea level is surprisingly common, but for coastal hydrology where an actual average sea surface height is the crucial factor, this misconception can cost money.
In fact, the picture gets even more complicated when considering “precise” (centimetre level) geodesic measurements. The earth itself flexes due to solid tidal forces, moving up to 40cm over the course of the day, and in coastal areas, the force of oceanic tides also influence ground level, which can change up to 5cm over the course of the day due to tide loading on the solid ground of the earth itself. Buildings themselves will tilt measurably over the course of a day as a result of this. Precise Point Positioning (PPP) services for GPS corrections actually apply corrections to account for this. This effect is generally balanced out when comparing with tide data, however it's important to be aware of if your analysis is assuming that the ground doesn’t move.
From datums to hydrology
Both ellipsoidal and geoidal datums can be considered geodetic, however sea level and tidal datums are technically hydrological. Generally, when one talks about mean sea level (MSL), one is thinking of the mean height of the sea. Water at rest should be level relative to gravity, and therefore level to the geoid. The sea however, is never at rest and is always changing due to tides, currents, wind, and many other factors, and so mean sea level generally refers to the mean level of the sea, averaged across the full tidal range and seasonal effects of winds and currents, shown in the figure below. It is very important to note however, that these effects are only averaged, not removed, and in many locations the averaged wind and current effects in the ocean result in mean sea surface heights over 1 metre different to the geoid. For satellite altimetry data, normally referenced to an ellipsoidal datum, the height of the mean sea surface (with tidal effects removed) is typically referred to as the Mean Sea Surface (MSS).
In general tide data is provided relative to a locally measured mean sea surface/level (often called MSL, but not to be confused with the geoid). This is a purely hydrological measurement, with the mean sea height averaged over a specific period of time and tide heights reported relative to this mean level. Crucially, this height is measured relative to land, so any changes in the land height of the sensor (coastal subsidence or geological activity) also affect the sea level heights measured. For this reason, many tide gauges also track their height relative to a geodetic datum, which is essential for comparing local tide gauge data to satellite altimetry measurements of sea surface height.
And so, for tying hydrological factors to topographical models and DTMs, you need to know a mean sea surface height relative to a geodetic datum.
This is where mean sea surface models come in. These are models of the tide averaged sea surface height (typically relative to the ellipsoid), generally based on satellite altimetry measurements, such as DTU25MSS (although be aware, DTU25MSS has versions referenced to the TOPEX ellipsoid and versions referenced to the WGS84 ellipsoid) . They can be a crucial connecting data point between a true mean sea level (to which tide heights, storm surges and sea level rise corrections can be applied) to a satellite or GNSS derived terrain model. A key challenge with many of these models however, is that they lose accuracy in complicated delta or riverine environments (exactly where mangroves are often most interesting), but they can be used to reduce the error in mean sea level (relative to a geodetic datum) to less than 50cm, much tighter than a purely geoidal model based measurement. Understanding this error and properly propagating it can help to significantly improve site assessments and forecasting, but for validation and tightening the error bars you’ll need some high quality field data.
Info box - Satellite altimetry For measurements of sea levels over large scales and remote locations, satellite altimetry is the key word to be googling. There are a number of satellites which provide precise (within centimetres) measurements of sea surfaces, generating large datasets which are used for improving and calibrating global tide models and mean sea surface heights. These satellites use either radar or lidar to measure the distance to the sea surface, and provide these measurements relative to a geodetic datum, typically a global ellipsoid, making it easy to compare these to GNSS measurements or satellite based DTMs. Some of the more commonly used datasets are from the ICESat-2 and SWOT missions. These are a key service for linking hydrological to geodetic measurements. |
Just get your feet muddy
While satellite data is great, especially over large areas, the best option however to link large scale DTMs to real local sea levels is just to take actual site measurements from your area of interest. Nothing can fully replace robust, scientific field data, and a few simple measurements can go a long way to calibrating and improving satellite data for coastal applications.
The large vertical errors in global DTM products are typically in the order of 1m or more (which is, once again, enormous and has big consequences for mangrove survivability), however with a few precise field GPS measurements, a calibration factor can be applied to the DTM over the site. When I say precise GPS however, I don’t mean standard uncorrected GPS (even that expensive Garmin won’t cut it), it needs to be survey grade with reported vertical errors in the order of centimetres to properly improve heights. Don’t think that survey grade needs a $5000 base and rover RTK kit, it's possible to get results approaching survey grade with $200 of hardware and a few extra tricks (it's a topic for another blog post, but reach out if you have questions). With this, you can take a few terrain height measurements across the site (preferably in flat and open areas) and correct and validate your satellite measurements.
Much more valuable than a global mean sea surface model are actual site tide measurements made relative to a geodetic height datum. If you are very lucky, there may be a local tide gauge close enough (less than 50km ideally, and within a few kilometres in more complicated delta or estuary environments) that already has a proper GNSS datum. You are probably not that lucky. You could set up your own tide gauge (making sure its position is properly GNSS referenced to within a few centimetres vertical accuracy), and run it for at least a month to estimate local mean sea level in a global reference frame. Off the shelf tide gauges are expensive, but again there can be some cheaper shortcuts, reach out and we can lend a system we have for a season or help with advice.
Getting out into the mud and taking either one (ideally both) of these measurements will significantly improve the accuracy and confidence in local hydrology measurements and predictions, and will pay off many times over the life of the carbon project. As in many cases, remote sensing is most powerful when combined with effective field work, a day with mud around the knees can align months of air conditioned office work with reality.
Mean sea level - present and future
So now, having understood the full picture, you have some ideas for how to find where actual local mean sea level sits on your space or drone derived topography data. One example workflow (and others are viable) could be:
Acquire a DTM over the area of interest, such as the FABDEM dataset, referenced to the EGM2008 geoid with root mean square error (RMSE) in the order of 2 metres.
Calibrate locally with survey grade GNSS measurements to bring local error from 2m to the order of centimetres.
Compare with global MSS product and validate the predicted MSL height with field measurements
AND/OR measure tidal range locally with heights referenced to a global GNSS datum with a tide gauge over a full tidal cycle.
You hopefully now have local mean sea level referenced to your topographical data. With this you can use the tide range (from local tide gauges or a global tide model calculated locally) to map out areas of low and high tide. You can compare this to local mangrove species distributions to improve knowledge on local mangrove zonation, informing future planting or restoration activities. You can use future storm probabilities to estimate the risks of storm surge and king tide flooding, either on vegetation or on local infrastructure. You have a solid base for an understanding of site conditions today and under various possible scenarios.
The next thing you are likely to need to do is to estimate the impacts of sea level rise. Is this as simple as adding the total sea level rise prediction at the location of interest from the IPCC AR6 sea level projection tool to the local mean sea level? Not quite. Over the time scales of sea level rise, several other factors need to be accounted for, including:
vertical land motion (with coastal land moving both down and up),
erosion and sediment accretion,
mangrove specific sediment accretion,
and even changing climatic conditions affecting local species diversity.
The topic of robust sea level rise risk assessment deserves its own blogpost, so I’ll save that for a future installment.
I hope this post however has managed to clarify some of the complications around the “simple” measurement of mean sea level, and can help others to implement rigour in their remote sensing based assessments of mangrove sites. The planet needs more mangroves, mangrove projects need to be less risky, and understanding mean sea level is one of the key foundations for mangrove site feasibility and risk forecasting. Satellite data alone is extremely useful, but to reiterate, it becomes most valuable when combined with robust field measurements to unlock its true value.
If you’re running a project and want to make sure your site feasibility and risk assessments are at the highest quality, reach out, and we can help you to make the most of local knowledge, field data, and engineering intelligence.