SDB in the Arctic and Beyond

Unlike many other remote sensing techniques, satellite-derived bathymetry is a passive remote sensing technique. So, instead of actively sending out a signal and measuring the intensity of the returned signal to derived distances, SDB uses the light of the sun. Throughout this blog we have explored the use of SDB in tropical areas with clear water and the sun shining overhead. Having a high sun angle will greatly improve how much light will penetrate through the water and how accurate the derived depth measurements will be. This is one of the many challenges faced by those conducting SDB in locations of high latitude, where the angle of the sun is considerably low. Using this technique to map remote and difficult to survey locations may become one of the most powerful applications of SDB. The Arctic, for example, has a limited amount of time during the year when ship surveys can be done due to ice coverage. But with increased exploration of this area for research on climate change, resource development and travel passages, it is important to have hydrographic surveys conducted and for ships to be provided with information of hazardous areas and shoreline depths. SDB can prove to be an important tool for researchers looking to explore the great North.

Here are just a few of the challenges faced when conducting SDB in Northern areas:

  • Low sun angle
  • Ice and clouds in satellite images
  • Turbidity
  • Sediment in the water

Researchers Chenier, Faucher and Ahola (2018) from the Canadian Hydrographic Service took three study sites in Northern Canada to test the implementation of SDB in these challenging climates.

The three study sites and associated problem to tackle were:

  1. Cambridge Bay, Victoria Island, Nunavut – assessing SDB in an Arctic environment
  2. Heath Point, Anticosti Island, Quebec – assessing SDB for updating hydrographic charts
  3. Havre-aux-Maisons, Magdelen Islands, Quebec – assessing SDB in heavy sediment waters

If possible, Chenier et al. (2018)  stress that to have the accurate SDB information, it is key to have good site data collected by surveys for calibration and accuracy assessment. The best method for surveying is multi-beam echosounder done close to the data on the image to be analyzed. Each of these sites had available multi-beam survey information and the Cambridge Bay site data was supported by LiDAR surveys (a remote sensing survey technique, most likely conducted by airplanes). All images underwent geometric and radiometric correction to evaluate the positional accuracy and account for the reflection from clouds or ice in the images. Two general models were used to determine depths from images, a band ratio model and a multi-band model. I have spoken about band ratio models in previous posts. It is the ratio of the light intensity reflected by two bands of visible light, most often the blue and green bands are used. Here three different ratios were used – Blue:Green, Coastal Band:Green, Blue:Yellow. The multi-band model works on the same principles, but considers each band to have a relationship with water depth rather than a ratio of two, and so uses a more complicated equation determine SDB depths. Below are the results of this study.

Table 1. RMSE statistics for band ratio and multiple-band SDB approaches.

From this table we can see that no method produces the same result. What else do we see:

  • RMSE is the root mean square error. Simply, this is how far the resulting value is from the line of best fit. This is called regression – the smaller the number, generally the more accurate the result.  
  • The smallest regression values for each method seems to occur at 4-6 meter depths for the Cambridge Bay and Heath Point Sites.
  • The regression values for the Havre-aux-Maisons site is similar for 0-2 meter, 2-4 meter depths and 4-6 meter depths.  
  • The site Havre-aux-Maisons likely can not achieve deeper than 6 meter depth measurements because it is an area with high levels of sediment in the water.
  • Cambridge bay probably had the the most accurate data to compare results with – three multibeam surveys from different years and LiDAR data. This site was able to determine depths up to 10-15 meters.
  • The multi-band model appears to generally give the best depth calculations.

Chenier et al. (2018) found that the results of their study showed that applications of SDB is promising, and results can be within a 0.5-1 meter error. This technique only has this level of accuracy for depths up to 10 meters, but this can have great impacts for navigation and safety in Arctic areas, and shows the potential for research in this field.

Satellite-derived bathymetry has come a long way since research on the topic first made its way into research in 1978, but there is still more to discover. Presenters at the recent Shallow Survey Conference in 2018 spoke about a Project Trident and the use of three different methods for SDB processing:

  1. Optically Derived Bathymetry
  2. Stereo-Photogrammetric Bathymetry
  3. Wave Kinetic Bathymetry

Fig. 2 – Project Trident from Ron Abileah, jOmegak Consulting. Shallow Survey Conference, 2018

The optically derived bathymetry is the processing method I have focused on in this blog and includes methods such as the band ratio models used in Chenier et al. (2018) research above. At best it can create a bathymetric surface with a 2 meter resolution. More research is pouring into the stereo-photogrammetric and wave kinematic processing methods as technology and research advances. Stereo-photogrammetric processing uses multiple overlapping satellite images taken at slightly different angles to derive depths. Wave kinematic bathymetry is a physics based processing method. Based on wave theory, this process method derives depths by analyzing two satellite images taken within a very short time of each other

Figure 2. shows that with a wave kinematic bathymetry processing method and validation from stereo-photogrammetric processing, a bathymetric surface with a resolution of 100 meters. By validating the results of each method with the results of the other two methods, a combined shallow water bathymetric surface can be produced. Using multiple processing methods will reduce the need for traditional ship surveys to validate the SDB derived depths. This will provide necessary knowledge of a shallow water environment before sending crews travelling to remote or dangerous areas. More work needs to be done to improve the resolution of these surfaces, but these techniques works well with sediment filled, turbid waters. Project Trident is currently using these combination bathymetric surfaces to map 5000 km of Brazil’s coast line, work that will valuable to ironing out the wrinkles of these newer processing techniques. Future research and applications of satellite-derived bathymetry will continue to propel our knowledge of our Earth and and it’s big blue ocean.

References:

Abileah, R. (October, 2018). Recent advances in exploiting the wavecelerity inversion methods for shallow water bathymetry. Presentation at the Shallow Survey 2018, St. John’s, N.L.

Chenier, R., Ahola, R., Sagram, M., Faucher, M. & Shelat, Y. (2019). Consideration of Level of Confidence within Multi-approach Satellite-Derived Bathymetry. ISPRS Int. J. Geo-Inf, 8(1), 48.


Advertisements

Maps by Hand

Being able to create accurate and precise maps of our world requires a lot of knowledge and years of study to become experts. But let’s take a moment to just appreciate the beauty of the world and how fun maps can be!

I credit the inspiration for this post to John Nelson and his blog, Adventures in Mapping. This blog is filled with information on mapping techniques, interviews, tips and fun maps, like these Lego Maps. One post in particular caught my eye, a tutorial on making maps by hand. Nelson used water colours, pen and coloured pencils to make an interesting bathymetric map. This looked like a fun Saturday activity, so I took out my coloured markers and did my best.

Bathymetric information retrieved from: https://www.oceangrafix.com/chart/zoom?chart=CHS4201

Although I made a few mistakes, mainly the island in the harbour that I though was a deep section, this was fun to make and I can’t wait to try again!

12 Algorithms of SDB

With so many recent publications on satellite-derived bathymetry, it is hard to wade through it all to find the best methodology. Thankfully, researchers at Pakuan University in Bogor, Indonesia saw the same problem and produced a comparative study of twelve algorithms to determine best practice for bathymetric mapping.

Manessa et al. (2017) recognized that the need for remote sensing techniques for mapping bathymetry arose because traditional surveys done by ships were costly, in time and money, and cover little area compared to aerial images.  And although ship surveys can map areas of much greater depths, mapping of shallow waters is critical for safety, navigation and environmental management, such as monitoring erosion along coastlines. Two locations of the Indonesian archipelago were selected for this study, Gili Mantra Islands and Menjangan Island of North Bali. The red dots in the images below are GPS makings showing where depth measurements were taken just off shore.

For this study, bathymetric data was collected by a ship survey using a single-beam echo sounder. This data was used for ground truthing and to compare the satellite-derived depths too. The depths measured by ship survey here were, “strongly affected by tide and wave” (pp. 129), as mentioned by Menassa et al. (2017). The tide effect was corrected by converting the depths to zero mean sea level (MSL), but depth measurements were not corrected for the wave effect. In the conclusion of this study, Menassa et al. (2017) strongly suggest that a wave correction by used when studying areas with strong waves.  

SPOT 6 images of the area were chosen to be use for this study because of the high spatial resolution (6 meters) of the multispectral images. They found depths up to 15 meters could accurately be determined from these images. Results of this study were less accurate for the Menjangan Island side in North Bali, likely because this area has a large reef, causing errors to occur in the single-beam echo sounder measurement of depth because of the different morphology and backscatter of the reef. More research may need to be done on satellite-derived bathymetry for coral reef areas.They also expect an error caused by high waves that occurred during the time of data collection that was not corrected for. Manessa at al. (2017), stress how important spatial resolution and environment conditions are on the accuracy of depth measurements.

Above all, one empirical method produced the most accurate results. The method of, Semiparametric Regression using Depth-Independent Variables and Spatial Coordinates stood out among the twelve as the best methodology for satellite-derived bathymetry in shallow waters. They also found that this method involves less assumptions about the data, leading to more accurate results. For example, the multiple linear regression model “assumed that water quality and atmospheric condition is uniform, and the number of bottom types is less than a number of band used” (pp. 132). The draw back of the semiparametric regression model is that this method requires more time to execute, but the complex statistical analysis used without the assumptions of the other methods provided the most accurate results for this study. 

References:
Manessa, M., Haidar, M., Hastuti, M. & Kresnawati, D. (2017). Determination of the best methodology for bathymetry mapping using Spot 6 Imagery: A Study of 12 empirical algorithms. International Journal of Remote Sensing and Earth Sciences, Vol (14), No. 2, pp. 127-136

Testing the waters with ArcGIS

Now that we know a little bit about how satellite-derived bathymetry (SDB) works, I wanted to get my feet wet and try it for myself!

The International Hydrographic Organization (IHO) and the Intergovernmental Oceanographic Commission (IOC) came together to create a Cook book for mixing up bathymetric data with different softwares. Contributions to this guide came from many experts in ocean mapping, such as the British Oceanographic Data Center, CARIS, and the Center for Coastal & Ocean Mapping among many others. The guide includes a step-by-step guide for using ArcMap to create a satellite-derived bathymetric map. Here’s a little bit about my experience with it.

I chose a test area of the Bahamas because the waters here are clear, there are little clouds obstructing the images and the contrast between the land and the water surrounding the shorelines is clear. There are some smaller islands in the area that would provide interesting bathymetry. Finding a satellite image with little to no cloud cover is important because this limits the amount of pre-processing required, such as atmospheric correction. If a satellite image is cloudy, the reflection of light can cause poor classification of data values. I was able to find a satellite image of the Bahamas with very little cloud cover from the Landsat 8 data set, but to my surprise the Cookbook says that, “the quality control on the images produced from the currently operational satellite is good enough,” (pg. 249). So I did not need to be concerned about radiometric correction, but a clear image is alway better.  For reference, here is the satellite image I selected:

This image was taken by the Landsat 8 satellite on January 18, 2019. As you can see, it covers a wide area and it was retrieved from a free data source, the USGS EarthExplorer. Landsat 8 can image the entire Earth in 16 days with a spatial resolution of 30 meters, and are free to download within 24 hours of the image being taken. Already it is clear that satellite-derived imagery is a cost-effective remote sensing technique for bathymetric mapping. But, as I learned from Manessa et al. (2017), spatial resolution of the images is very important for accurate depth calculations, and these images have a lower spatial resolution than the images used by Manessa et al. (2017) in their comparative study.

For the Landsat 8 image processing in ArcGIS, light bands 2, 3, and 6 were used. Band 2 is blue light, band 3 is green light and band 6 is shortwave infrared light. The shortwave infrared band is used to separate the land from water, and the bathymetric mapping is done by applying an algorithm, from Stumpf et. al (2003), on the blue and green bands. This is a linear ratio algorithm, taking the ratio of reflectance in blue light and green light. The advantage of this algorithm is that the “change in ratio because of depth is much greater than that caused by change in bottom albedo” (Manessa et al. 2017). This means that if there is a significant change in ratio, this algorithm will classify that as a change in depth, and different bottom materials that reflect light differently will not affect depth measurements because they have a similar ratio of light reflectance in the blue and green band.

The satellite-derived bathymetric map should be verified by comparing the results with the values of the depth soundings of a digital elevation model of the area, or with ground-truth data. For now, I am interested in getting a resulting image from ArcGIS rather than the accuracy of my result. If this method is to be used for anything other than creating a pretty looking map, a statistical analysis of the SDB results and the chart depth sounding references needs to be done. This is also important for determining if a change in the sea floor has occurred in the area. A comparison of satellite images from different years could be done quickly and with little cost to determine if there has been any changes in sea floor level that could affect safety in navigation through the area. As for the pretty looking map, here was my result:

Fig. 2 – Satellite-Derived Bathymetry in The Bahamas

I noticed that the orange area through the middle of this map has all been assigned the same depth. By looking at the nautical chart 4149 of the area, downloaded from the National Oceanic and Atmospheric Administration, I see that this area is labelled Tongue of the Ocean. This must be very deep water and the algorithm used did not properly classify these pixels as deep water, or there is another problem with the processing. Taking a closer look at the smaller islands would be beneficial for comparing the SDB depths to the charted depths.

Fig. 3 – Chart 4149 from NOAA

Some environmental factors that reduce the quality of satellite images are poor weather conditions bringing clouds and wind, waves, objects in the water such as ice, reefs or sediment, and sun angle. The tropical climate of the Bahamas makes it a great area to study satellite-derived bathymetry because low turbidity allows for sediment to settle and the waters off shore to be very clear. The angle of the sun affects how much light penetrates the water, and in turn affects how deep satellite-derived depths can be measured. In these warm areas, the angle of the sun can be 90 degrees in these areas allowing for light to reach greater depths.Generally these areas have great weather conditions, so retrieving a satellite image on a clear bright day can be a simple task. The challenge for researchers working with satellite-derived bathymetry is how to make this a realistic option to be used for mapping northern and arctic climates. These cold climate areas make it difficult for ships to do surveys because of ice coverage, allowing for a short time frame for surveys to be done by boat. The future of satellite-derived bathymetry will be developing the methodology to provide accurate results in northern climates. With the longest coastline in the world, researchers Chenier et al. (2018) studied SDB for improving Canadian hydrographic service charts, improving navigation and safety in the north. Coming up, I will discuss the efforts done to improve SDB in northern climates, the challenges faced here, and what is to come for satellite-derived bathymetry.

References:

Chenier, R., Faucher, M., Ahola, R.(2018). Satellite-Derived Bathymetry for Improving Canadian Hydrographic Service Charts. International Journal of Geo-Information, Vol(7), No. 306, pp. 1-15

International Hydrographic Organization, Intergovernmental Oceanographic Commission, The IHO-IOC GEBCO Cook Book, IHO Publication B-11, Monaco, Dec. 2015, 429 pp – IOC Manuals and Guides 63, France, Dec. 2015, 429 pp.

Manessa, M., Haidar, M., Hastuti, M. & Kresnawati, D. (2017). Determination of the best methodology for bathymetry mapping using Spot 6 Imagery: A Study of 12 empirical algorithms. International Journal of Remote Sensing and Earth Sciences, Vol (14), No. 2, pp. 127-136

Pe’eri, S., Tetteh, E., Marks, K. (2014). Updating Landsat Satellite-derived Bathymetry Procedure. IHO-IOC GEBCO Cook Book. Retrieved Feb. 18, 2019 from: https://www.hydro-international.com/content/article/updating-landsat-satellite-derived-bathymetry-procedure

What are the band designations for the Landsat satellites? Retrieved Feb. 18, 2019 from: https://landsat.usgs.gov/what-are-band-designations-landsat-satellites

Landsat 8 Instruments. Retrieved Feb. 18, 2019 from:https://landsat.usgs.gov/landsat-8

SDB with Ratio Algorithms

Techniques for determining shallow water depths from satellite images have developed significantly in recent years. Before we get to these new and cool techniques, let’s discuss some of the early research so we can see how far along we have come.

Early research on satellite-derived bathymetry recognized that depths could not easily be determined by the colour of the image (Lyzenga, 1978), propelling the development of techniques using multispectral images. This was first proposed by Polycn et al. in 1970 when they studies spectral signatures using remote sensing and found that using a ratio of absorption of light between two bands could be used to determine depth. A spectral signature is the variation of absorption or reflection of light by a material across a range of wavelengths. In the simplest sense, it was thought that a spectral signature could be used to assign depths to areas with the same signature.

This image from Zheng & DiGiacomo (2017) shows the spectral signatures of different bodies of water.

Image: Zheng & DiGiacomo (2017)

For example, we can see the dark green line is the spectral signature of the water off the cost of Hawaii, where there is a peak absorption of light in the range of 430-460 nm. Blue light has a wavelength of 450-485 nm, and green light has a wavelength of 500-565 nm. Water near Hawaii absorbs the most light at the lower wavelengths, so this water is very light blue in colour or could be very shallow. If we knew the true depth of this area, perhaps we could assign that depth to other water bodies with a similar spectral signature. You might be able to see the problem with this over simplification. A body of water could be very light blue in colour, but not necessarily shallow. According to Polcyn et al. (1970), the limitation of this simple technique is that different sea bed material reflect light differently and can change the spectral signatures for areas with the same depths. He proposed instead to use a ratio algorithm.

A ratio algorithm will take the spectral signature of a water body and compare the absorption of light at two different wavelengths. Polycn’s algorithm was based on the radiant and reflectance properties of light, but does not account for the effect of light scattering in the water (Lygenza, 1978), or for atmospheric absorption (Polcyn et al, 1970). Also not concerned was the spatial and spectral resolution of the images used. This is also affect the accuracy of your depth measurements.

From my research, it appears that ratio algorithms are better for determining relative depths, rather than true depths. But, I’d like to learn more. Manessa et al. (2017) have completed a study using twelve different empirical algorithms to determine depths surrounding islands in South East Asia. On my next post, I will discuss their research and what they found to be the best methodology for bathymetric mapping.

References:

Lygenza, D. (1978), Remote sensing techniques for mapping water depth and bottom features. Applied Optics, Vol (17), pp. 379-383.

Manessa, M., Haidar, M., Hastuti, M. & Kresnawati, D. (2017). Determination of the best methodology for bathymetry mapping using Spot 6 Imagery: A Study of 12 empirical algorithms. International Journal of Remote Sensing and Earth Sciences, Vol (14), No. 2, pp. 127-136

Polcyn, F. Brown, W. Sattinger, I. (1970). The measurement of water depth by remote sensing techniques. Willow Run Laboratories, The University of Michigan.

Zheng, G. & DiGiacomo, P. (2017). Uncertainties and applications of satellite-derived coastal water quality products. Progress in Oceanography, Vol (159), pp. 45-72.

What is Satellite-Derived Bathymetry?

Satellite images have been used for mapping Earth’s surface for many years, and with recent development in satellite technology, these high-resolution images are now being used to map the ocean floor. Satellite-derived bathymetry (SDB) has been a topic of research since the 1980s, but lately it is receiving more attention because it requires relatively low cost and investment compared to traditional surveys. In coastal areas where waters are shallow and there is a steady stream of marine traffic, mapping the sea bed is especially important for safe navigation. The ocean floor of these port areas can change quickly and often, and so a more efficient way to map these areas will become very important in the future of ocean mapping.

So, what is satellite-derived bathymetry? According to the UK Hydrographic Office it is:

  • A depth determined by measuring light intensities through processing of satellite images.
  • Originally based on the assumption that deeper areas will appear darker in images than shallow areas.
  • Can cover a wider area in an image than a survey vessel will, and in much less time with much less cost.

Satellite images are made up of many bands of light with different electromagnetic wavelengths. This means that satellites transmit light from the visible spectrum, infrared, near infrared, etc. As these bands of light reach objects on the earth’s surface, the objects will absorb and reflect wavelengths to a specific degree, creating a spectral signature. One method of determining depths with SDB is to use the ratio between two visible light bands, blue and green, which penetrate the ocean water to different depths and help to determine the depth of the sea floor. This seems great, a low cost and efficient solution to mapping frequently changing areas under water. But how can we trust satellite-derived bathymetry? How accurate are the current methods?

There are many limitations to satellite-derived bathymetry, a few of them being:

  • You first need to have a clear image of your area of study, one without clouds and little turbidity in the water.
  • Sea bed materials or underwater objects will have different reflective indexes, so a deeper area covered by material with a higher reflective index may reflect more light and appear to be shallower than it is. This means that water depth cannot be mapped based only on a scale of light vs dark areas of the image.
  • Light only penetrates to a depth of about 30 meters, and this depends on the clarity of the water being imaged. SDB cannot be used for deep ocean mapping.

SDB is being talked about often as an emerging technique in ocean mapping. New physics-based algorithms are being developed to improve the accuracy of SDB, which will allow for SDB to become a more trusted practice of mapping the ocean with satellites. Over the course of this blog, I will talk about the different uses and methods of determining depth with satellite-derived bathymetry. Stay tuned!

More Information From:

Chenier, R., & Faucher, M., & Ahola, R. (2018), Satellite-derived Bathymetry for Improving Canadian Hydrographic Service Charts. International Journal of Geo-Informaiton, Vol. 7(306).

CSPCWC. (2015, April 28). Satellite Derived Bathymetry. Retrieved from: https://www.iho.int/mtg_docs/com_wg/CSPCWG/CSPCWG11-NCWG1/CSPCWG11-08.7A-Satellite%20Bathymetry.pdf

EOMAP. (2018). Satellite-Derived Bathymetry for Coastal Monitoring Solutions. Retrieved from:https://sdbday.org/2018/05/04/coastal-monitoring-solutions-satellite-derived-bathymetry-techniques/

Hartmann, K., & Heege, T., & Wettle, M. (2017, May 4). The Increasing Important of Satellite-derived Bathymetry. Retrieved from: https://www.gim-international.com/content/article/the-increasing-importance-of-satellite-derived-bathymetry

Jegat, V., & Pe’eri, S., & Freire, R., & Klemm, A., & Nyberg, J. (2016). Satellite-Derived Bathymetry: Performance and Production. Canadian Hydrographic Conference.

Macnab, R., & Varma, H. (2008, January 19). Bathymetry from Space. Retrieved from: https://www.hydro-international.com/themes/satellite-derived-bathymetry

NOAA Office of Coast Survey. (2015, November 5). Studying the use of satellite-derived bathymetry as a new survey tool. Retrieved from: https://noaacoastsurvey.wordpress.com/2015/11/05/studying-the-use-of-satellite-derived-bathymetry-as-a-new-survey-tool/

Otte, C. (2016, July 14). Researchers look to the sky to peer underwater. Retrieved from: https://greatlakesecho.org/2016/06/14/researchers-look-to-the-sky-to-peer-underwater

Pe’eri, S., & Madore, B., & Nyberg, J., & Snyder, L., & Parrish, C., & Smith, S. (2016) Identifying Bathymetric Differences over Alaska’s North Slope using a Satellite-derived Bathymetry Multi-temporal Approach. Journal of Coastal Research, No. 76, pg. 56-63.

Said, N., & Mahumd, M., & Hasan, R. (2017). Satellite-derived bathymetry: Accuracy Assessment of Depths Derivation Algorithm for Shallow Water Area. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences.

Stumpf, R., & Holderied, K. (2003). Determination of water depth with high-resolution satellite imagery over variable bottom types. American Society of Limnology and Oceanography, Inc. Vol 48(1,2), pp. 547-556.