Comparison - NGA PGS to EVS Precision Vectors

Comparison - NGA PGS to EVS Precision Vectors (1:6,250)

I know a good deal about global shoreline vector files. Specifically, global shoreline vector files that are free, available to the public and relatively easy to load and work with. This post's focus is a comparison of NGA's Prototype Global Shoreline (PGS) to Enhanced Vector Shoreline (EVS) Precision shorelines. This is not my first comparison of these two shorelines and it probably won't be my last, but here it goes.

NGA PGS vectors too often wobble off of the imaged coastline unnecessarily. The NGA PGS vectors were derived from Landsat ETM+ imagery, the same imagery I use to create EVS Precision shorelines. The above image is a small portion of Somalia's shoreline. The Landsat image is cloud free and the shoreline is easily distinguishable. NGA PGS used a complicated formula to determine the shoreline, but I don't see why their vectors misplot by as much as 41 meters from the imaged shoreline. No, this doesn't happen all of the time, but it happens enough, that if I am working at very large scales (1:50,000 to 1:12,500), I must redo the shoreline. Their shoreline is too far from image shoreline true.

I've been doing EVS Precision mapping for over 9 going on 10 years and my opinion concerning NGA PGS has not changed. Don't get me wrong, NGA PGS is the most detailed global shoreline available, one that I recommend heartily. However, for my work, creating detailed island maps, it isn't good enough.

Enjoy!

Comparison - DCW, VMAP0 Shorelines and GYM's Mapping Opportunity

Shoreline Map Comparison - Yahoo, Google, MS and EVS Precison

Shoreline Map Comparison - DCW and VMAP0

This post shows that Google Maps, Yahoo Maps and MS Live Search Maps use DCW/VMAP0 global shorelines for their mapping, at least in the area of Rabul. That is 1:1,000,000 scale global map coverage. James Fee, a well-respected GIS professional, commented, "When it comes to precision, I think the efforts for GYM (Google, Yahoo, Microsoft) should be on the navigation/geocoding/routing and not shorelines." First, one must agree with James' statement that GYM's maps are intended primarily for navigation/geocoding/routing. I believe that their customers, global in reach, would subject GYM's maps to many other GIS uses beyond navigation/geocoding/routing. I know that there is an agressive effort within the GIS community to provide a richer, more powerful set of tools, allowing any of us the opportunity to explore more complex spatial issues using Google Maps and Google Earth as platforms of choice. Ultimately, the end user, faced with a spatial problem requiring small scale maps would either have to make their own map, pay to have a map made, use a substandard map in place or not use a map at all.

In addition, James seems to suggest that 1:1,000,000 scale is satisfactory for GYMs primary efforts of navigation/geocoding/routing. I understand his reasoning, but I disagree. Before teaching, I did extensive geocoding of environmental risks. Locational precision was critically important. 1:1,000,000 was an unworkable geocoding scale. To follow the flight path of an enemy aircraft using 1:1,000,000 is doable. However, to locate enironmental risks at that large of a scale is asking for problems. A misplotted environmental risk could well result in a lawsuit. Business locations was another area I worked in and locational precision determined an exclusive market and was a contractual promise to the franchisee.

James goes on to say, "For me the shorelines are somewhat irrelevant to the routing provided. If I do need good shorelines, I'll probably just be using the aerial imagery in the first place." Again, his reasoning seems to make sense, but I disagree. First, shorelines are not somewhat irrelevant, they are a dynamic boundary between man and water. A raster image of a shoreline is practical for some applications, but a vector defined shoreline offer one a wide range of mapping possibilities. If I am GYM's map guru, I want my customers to have the greatest mapping precision possible. I don't want to impose precision limitations when there is no need. EVS precision shorelines, on a global scale, requires money to fund the mapping effort and time to complete the project. It will be money well spent. The resulting shorelines and other feature/land use layers will allow the "G", the "Y" or the "M" to be the number one mapping destination for all users and that could generate lots of ad dollars.

Finally, vector maps allow for infinite zooming with no lose of detail. If the initial vectors are created from quality base imagery (i.e., Landsat ETM+) and the map customers are able to easily access these detailed vector maps, they can tackle complex spatial issues using GYM's hi-precision mapping. Thus a marketing boon to the company savy enough to provide the product.

Enjoy!

P.S. - After yesterday's post, I received a query from an individual about EVS precision mapping for a section of Canadian shoreline. He had used WVS/GHHS vectors and they were unsatisfactory for his project. I sent him NGA PGS vectors covering his target area. They looked okay at 1:250,000 scale. He loaded them and acknowledged they were a significant improvement over the WVS/GHHS vectors, but were angular and did not track a cloud-free shoreline as one would expect. I created EVS precision vectors for him. He compared them to NGA PGS vectors and immediately acknowledged their higher quality. He asked if I would go ahead and create EVS precision vectors for his entire study area. I will, but wouldn't it have been cool for him to go to one of the big three mapping portals and find this mapping not only available, but downloadable? I think so.

Comparison - Google Maps, Yahoo Maps, MS Live Search Maps and EVS Precision Maps

Shoreline Map Comparison - Yahoo, Google, MS and EVS Precison

Whose shoreline maps are currently the most accurate? Is it Google Maps? They have loads of bucks and a keen desire to put the best product into their sites. They must have the most accurate shorelines. Right? Nope, not them. Yahoo! Maps, the current darling of the stock market and Google "wannabe", has the same clunky shoreline used by Google. MS Live Search Maps has the prettiest maps. Lots of shaded hills and valleys, but clunky shorelines. For sure, MicroSoft must have the most accurate shorelines. For sure, not.

Would you believe, EVS Precision Maps has the most accurate shorelines compared to the big three map moguls. EVS precision mapping is derived from Landsat ETM+. Typically, it provides accurate shoreline depiction at scales between 1:100,000 to 1:50,000.

Rabul - MS Live Search Map

All map maker struggle with PPL (populated place) placement. The PPL data is available through NGA. I'm reasonably certain most of their data and locational information was derived from ONC (1:1,000,000) charts. Looking at the PPLs on the MS Live Search Map of the Rabul area, we see towns lined up in a grid. In the real world, this only happens in the midwest, not on New Britain Island in Papua New Guinea. One should replot these PPLs using Landsat ETM+ or hi-res imagery. The only thing certain about the above plots, is their locational uncertainty. But, back to shorelines.

Rabul - EVS Precision Maps

EVS Precision Maps provide the most accurate shoreline maps of the four being compared. The big three use, what appears to be DCW's 1:1,000,000 shorelines. 1:1,000,000 has it's uses, but when one zooms to 1:63,500 (1" to 1-mi), a shoreline that approximates reality is expected. EVS Precision Maps show that expected shoreline. Google Maps, MS Live Search Maps and Yahoo Maps do not.

Enjoy!

P.S. - I am aware of NGA PGS vector files, 24 tiles covering much of the world. They were derived from Landsat ETM+. EVS Precision Maps and NGA PGS should parallel each other. They do not. NGA PGS vectors used a digitizing algorithm to automate their work. Peculiar stutters, uncalled for strings of vectors and angular shorelines are often the norm. Also, NGA requested that water inundated areas (swamps, marshes, tidal flats) be excluded. This caused a great many problems in shoreline mapping. Also, the slightest cloud cover often thwarted their automated digitizing efforts. EVS Precision Maps consistently improve upon NGA PGS.

Thoughts - Mapping New Britain PP

I was explaining my mapping work to a person I met at a party earlier today. This individual asked all of the right questions ( What do you do? Where do you get your images? How do you digitize the shorelines? What do you do with your finished maps?) Poor guy! I answered every question with a detailed, lengthy and somewhat technically boring response. After my fifth lengthy reply, he excused himself and joined another group.

As I reflect on my exchange with this curious person, I am amazed at the breadth of my knowledge and convinced that I can be very boring. I really do know quite abit about digitizing shorelines. After a few thousand kilometers of continental shorelines and 9,000 island shorelines, I am able to digitize a precise shoreline (within the limitations of Landsat ETM+'s 14.5-meter resolution).

Early on, I had professional cartographers tell me my maps would be useful at 1:125,000 to 1:63,500. I have learned that my island shoreline maps are useful at scales well-below these upper limits. I typically post completed maps at 1:25,000 scale. After adding layers of information, they work well for an individual attempting to determine a relatively coarse view of an island. Landsat is not DigitalGlobe's hi-res imagery, but it does a great job at giving one a highly resolved broad view of a geographical area.

New Britain PP - Digitizing from Landsat S-55-05_2000 (1:19960)

So, how is it that I do what I do? In digitizing, that is. The above image shows a typical stretch of large island shoreline. An attempt to generate vector shorelines from Landsat ETM+ was done a few years ago. This data, called NGA PGS, is fairly good at defining a shoreline, but coastal swamps are excluded causing some very strange looking maps of islands. If the island or shoreline contains water-innundated shoreline, NGA PGS does not map it. They could have, but someone, somewhere thought it would be good to exclude it. My shorelines include these water-innundated areas. The red line in the above image is my completed EVS precision shoreline. Except for the final vector which I purposely placed in the sea, the shoreline is much more accurately delineated than NGA PGS, WVS, DCW or any of the other available global shorelines. I am not saying EVS precision shorelines are global yet. I am an optimist. Someone, somewhere, sooner than later will fund my mapping efforts and the result will be a global shoreline at EVS precision.

Enjoy!

Thoughts - When is a Shoreline a Shoreline? A Continuing Discussion.

Kyra Panagia - EVS Compared to NGA PGS

Comment posted by Anonymous:

Shorelines are usually created through stereoplotting or derived from surface models - rarely from monoplotting like you do. The problem with monoplotting is that your shorelines will be different depending on the tide, so your shorelines might very well be even more off in many places - especially at relatively flat areas and/or shallow water.

My response:

Anonymous, your observation concerning the use of stereoplotting and surface models providing an accurate approximation of a dynamic shoreline are well taken. If one had the imagery depicting the shoreline at various stages of tidal fluctuations during a set period of time, the methods you refer to would always provide the superior shoreline representation. However, after monoplotting 1,000s of kilometers of shorelines and comparing my EVS precision shorelines to NGA PGS precision shorelines - EVS shorelines track the Landsat ETM+ imaged shorelines more consistently then NGA PGS precision shorelines. My guess is that their method introduced unanticipated stutters, inconsistent shoreline identification and non-inclusion of water inundated areas.

EVS precison is not perfect, just better at tracking the visible shoreline more consistently then NGA PGS.

Comparison - EVS to NGA PGS, One More Time

Kyra Panagia - EVS Compared to NGA PGS

Enhanced Vector Shorelines are derived from Landsat ETM+ mosaics. When I carefully digitize an island shoreline using Landsat imagery as my base image, I create EVS precision vectors. This name describes an improved shoreline over World Vector Shorelines.

The NGA Prototype Global Shoreline claims to be a global vector shoreline derived from Landsat ETM+ mosaics, the same imagery I use. When I first learned of it, I thought to myself, "My work is done! MDA has produced the ultimate global vector shoreline. Well, they really haven't. It seems that their shorelines do not track Landsat's imaged shorelines. They are close, somettimes very close. But far too often, they are off, sometimes by 100's of feet.

It is my contention, if Landsat provides a clearly imaged shoreline, the vectors that describe that shoreline should be as close to the shoreline as possible. NGA PGS does a pretty good job, but when I compare their product to my EVS precision shorelines - mine are better at tracking the Landsat imaged shoreline.

Enjoy!

Comparison - WVS, SWBD, NGA PGS Global Shorelines and EVS Precision Shorelines

Three Vector Shoreline's Compared - WVS, SWBD and EVS Precision

One of my readers suggested I evaluate SWBD shorelines in comparison to other freely available shorelines. Previously, I compared these shorelines with my own EVS precision shoreline. I'll do it once more.

The WVS is widely available and manageable in size. There is another popular version of it called the GSHHS. It repairs varous WVS anamolies. In addition, it conveniently breaks the dataset into land, lakes, ponds and islands. At only 56MBs zipped, it is a very attractive option. As one can see in the two images, the WVS is the least precise of all large scale global shorelines. It is designed to be used at scales of 1:250,000 and above. For most of us designing projects for our classroom, employers or customers, this dataset does a fine job, it is free and the size is reasonable. It is when we need to do large scale precise shorelines that we require the next best global shorelines

Although the SWBD is a quality product and provides the end-user with precise SRTM water-defined shorelines, it tracks the water. The SWBD attempts to identify the place where the water meets the shore. That should be the shoreline, but it isn't. The top image shows the SWBD shoreline tracking the reef edges and ignoring the numerous motus within the reef. For us island mappers, the SWBD is not precise enough to identify the many shoreline features we require.

The same can be said for the NGA PGS shorelines. They are extremely precise, using Landsat ETM+ imagery as their base imagery. This is the same imagery I use to construct my EVS precision maps. I have the NGA PGS shoreline loaded into Marplot, a mapping package I use. The dataset is divided into 24 separate groups. My global coverage, using this dataset, is outstanding - except where clouds or other atmospherics obscure the shoreline or the shoreline consists of water-inundated areas (marsh, swamp, mudflats), which the NGA PGS product didn't identify. The cloud obscured areas on Landsat ETM+ imagery are validly poorly mappable to unmappable. The water-inundated areas are mappable, but the NGA contract called for the exclusion of these shoreline features, thus they were not mapped. I beg to differ. This is shoreline that must be mapped. There are massive river deltas NGA PGS mapped that depict shorelines that look nothing like the original Landsat ETM+ image, which is full of features that the NGA PGS product did not map. They just happen to be water-inundated. They are features you and I would describe as shoreline. Make no mistake, when the shoreline is cloud-free and not water-inundated, which most Landsat imaged shorelines are, this global shoreline dataset is outstanding. It's just all of those river delta islands that they didn't map that I have a problem with.

Four Vector Shoreline's Compared

However, my EVS precision shorelines are pretty outstanding as well. When one compares my hand-drawn shorelines to programmed color tracking shorelines (or many other automated processes), my hand-drawn shorelines track quite precisely. Something about using one's brain processing powers, manual dexterity and years of experience that consistently allows one to create more precise shorelines than the WVS, SWBD and NGA PGS products.

As I work with Landsat ETM+ imagery and learn more about the digitizing craft, I am certain software is available that could do what I do more accurately, faster and in much less time. But it wouldn't be as much fun! Would it?

Enjoy!