How to use Sentinel Hub's EO Browser to scout for ice by satellite
Using satellite images to scout ice can be super helpful. The data that’s now freely available is absolutely incredible. But if you’re not up-to-speed on how to acquire and interpret the images, they can be a little confusing at first. Thus, I've written this primer.
This has come up on the Nordic skating email lists a bunch of times over the years, but the information is a little scattered. Luc Mehl has some relevant and helpful resources, but I think what’s appropriate to big trips in Alaska (or sea ice off the Swedish archipelago, for that matter) is not quite the same as what’s needed for smaller bodies of water in the northeast.
In what follows I use Lake Winnipesaukee as my main example, because I’ve been on it recently and have had the chance to compare the images to what I found in real life, and because it’s got a good diversity of conditions at the moment to illustrate some of what we can look for in satellite scouting.
- Getting the images on your desktop
- Visualizing the data
- Examples: Lake Winnipesaukee; Lake Sunapee
- Getting the images while on the ice
- Additional information
- Addendum: An even easier, but less useful, method
NB: Click any image to view it larger. There's lots of detail to examine in them.
Caveat: The process outlined here is not the only way to do it, it’s just the way I’m doing it currently.
BIGGER CAVEAT: Looking at pretty pictures on the internet is no substitute for first-hand knowledge. Read reports, talk to people, test the ice, make your own observations. Anything else is not enough.
1. Getting the images on your desktop
The data we want comes primarily from the Sentinel-2 satellites (thanks Europe!), which provide frequently updated (at least every 5 days), very detailed (up to 10m resolution, which is amazing), and freely available images (free as in beer).
There are a few ways to get it; the overall best is to use the Sentinel Hub EO Browser. Click on “Start exploring!” or go directly to the app.
This service is free to use and does not require an account. You just need to agree to the terms of service (the usual stuff, and free usage has to be for non-commercial purposes). You can create an account, which gives you some additional functionality (for example, saving pinned maps and downloading high resolution maps), but you definitely don't need to. When you first open the EO Browser, it’ll give you links to the user guide and to a nice overview tutorial.
This website requires an up-to-date browser and a fast internet connection. If you’re using an old browser, or an old operating system, or a very old computer, or a very slow connection, it might not work. It's loading and processing huge amounts of data.
NB: The order of the following steps matters. If you're having trouble, make sure you're doing things in the order listed here.
The EO Browser opens up to a map and the Discover tab of the control panel. In the top right, you’ll see a search box labeled “Go to place”. Type a location in. For example, type “Lake Winnipesaukee”. Then select the best matching item from the drop-down that appears:
Once you see the map update to show the correct location, choose the types of satellite images you're interested in, using the control panel on the left. Make sure “Sentinel-2” is selected (it should be checked by default), then also click the checkbox next to “Harmonized Landsat Sentinel” to select it as well. Then click the “Search” button on the control panel:
A list of results will appear. They’re sorted chronologically, with the most recent satellite passes at the top. Each result will be labeled either “Harmonized Landsat Sentinel (HLS)” or “Sentinel-2 L2A”, then a date, then a time (in UTC, which is 5 hours ahead of Eastern Standard Time), then a percentage of cloud cover. You can either hover your pointer over a result in the list, and the swath it corresponds to on the map will turn blue (all others will be a pale cyan), or you can click on an area on the map and the result it corresponds to on the list will be selected. (Or if it corresponds to multiple results, you’ll be allowed to choose between them.)
The simplest thing to do is to start at the top and work backwards in time until you find what you need. Best case scenario would be Sentinel-2 imagery from today with a low cloud cover percentage.
In this example, since both the latest HLS data and the latest Sentinel-2 data (which are kind of the same; see the question in part 5 about the difference) are from 2 days ago, I’m going to choose the Sentinel-2 result to get better resolution. (HLS images more often but at a slightly lower resolution.)
Click the green “Visualize” button on the result:
You should something like this:
If clouds are covering the spot you’re interested in, click the “Discover” tab in the control panel (almost the top left corner) and choose a different result from the list.
2. Visualizing the data
Now that you have something to look at in the EO Browser, what do you look for?
A short primer within the primer: How satellite sensor data becomes images on your screen
Each satellite captures different bands of information, using different sensors. Those bands of information are turned into the red, green, and blue of images that our eyes can process. So the simplest case are the sensors that capture visible light — ie red, green, and blue — because those are just displayed as received, and show us what the satellite saw.
Slightly more complex is using non-visible light sensors to create false color images. For example, the common infrared satellite images (“False color” in the EO Browser “Visualize” tab) are created by replacing the normal visible light blue color with what the near-infrared sensor sees (and leaving the visible light green and red data alone).
More complex, and more useful, yet is using multiple different bands of information and doing calculations with them. I have a lot more to say about this, but I’ll save it for another day.
With all that in mind, let’s take a look at the images. If you select a Sentinel-2 data source, you should see the first list of available visualizations; if you select an HLS data source, you’ll see the second:
Luckily, the visualizations that matter to us for scouting ice are available from both satellite data sources.
“True color” is just a standard satellite image (red, green, ablue visible light). Zooming in with this visualization selected shows the resolution difference between the two data sources (Sentinel-2 vs HLS). This is the Leavitt Beach plate we’ve been skating this past week:
Both are good enough, but the HLS data is slightly pixelated. No big deal for our purposes (though it can be a problem for very small bodies of water). From here on I’m going to use the Sentinel-2 data, since it’s slightly clearer. Everything should apply to the HLS data as well.
The next visualization option in the list is “False color”, which is visible red and green light plus near infrared light in place of visible blue light. I’ll quote Jamie’s recent description of how to interpret this:
Bluish gray and bluish white areas are the most likely to be skateable. These areas have remained frozen long enough to have been snowed on, and then the surface was subjected to a thaw-and-refreeze cycle. Bluish white indicates older and thicker ice than bluish gray.
Grayish black areas are slightly less likely. They consist of black ice that's old enough to have cracks in it - the cracks give it the grayish appearance - but it might not be skateable yet. To make an assessment, you'd need to look at the date the satellite image was taken, and how cold it's been since then.
Areas that appear solid black or solid white are probably either open water or snow-covered ice. Unless you have information to the contrary (like an eyewitness report) I would steer clear of them.
The “Moisture index” (NDMI) visualization sounds promising, but it’s meant for vegetation. Sentinel Hub recommends using the “NDWI” (water index) visualization for water bodies instead.
Each pixel in an NDWI image is on a spectrum from green to white to blue (as illustrated in the screenshot), where green means vegetation, blue means water, and white means not water or vegetation. How do we interpret this? More on that in the part 3 examples.
“NDSI” (snow index) is very simple: bright blue is snow (or rather “snow”, which very much includes ice), everything else is not:
“SWIR” (short-wave infrared) is another false color visualization, but this one uses short-wave infrared, narrow infrared, and red. It looks much the same as the “False color” infrared visualization (well, except different colors), but it tells us a bit more. Again, more on this in the part 3 examples.
Let’s continue with Winnipesaukee for a moment, but zoom out one notch. And let’s start by looking at what’s changed in the past few weeks. First is 1/17, then 1/27 (1/22 was cloud covered), then 2/1, all in infrared false color:
You can see ice coverage growing, shrinking in places, and changing character, all of which makes sense given the weather we’ve had.
Zooming back in to the area just north of Leavitt Beach and looking at the 2/1 image, we see:
I’ve circled in yellow a lead that was at least partially covered in black ice on Monday (2 days before this image was taken). It was thin but some skaters crossed it, very carefully.
You can see how different newly forming primary ice looks from the thick snow ice that these plates are composed of, but it’s hard to tell the difference between black ice and open water (which the area to the north definitely was).
Switch to NDWI (water index) and the difference jumps out:
Switch to NDSI (snow index) and the distinction is even clearer:
Switching between visualizations like this is very powerful. No single visualization can tell you everything you’d like to know, but by comparing multiple different kinds you can get a clearer picture of what’s going on.
Jamie’s post to the NH Nordic Skating list on 1/27 (the archives are only viewable by group members) about satellite images and how to interpret them has more detail and is a better example than any I’ve got here. Go read that.
In that email, he was describing the images from 1/17 and how they compared to what he saw on the lake that day. On 1/29 he and a group skated down the entire eastern shore of the lake, and found decent conditions. On 1/31 I tried to skate on the lake, following their route, and was stymied by a wide variety of bad conditions. Let’s look at what happened, from the satellite perspective.
1/17, pretty much all black ice. Promising! But possibly not quite ready for skating. Hard to say without getting out there and ground truthing.
1/27, following Jamie’s guidance, looks like gray ice! Much more definitive. But is it good ice?
1/27, following Jamie’s guidance, looks like gray ice! Much more definitive. But is it good ice?
And are those clouds covering the eastern shore? Hard to tell. Switching to NDSI (snow index) makes it clear: no, the clouds aren’t hiding anything:
But how much variation between ice and snow covered ice is there? It’s a little hard to tell using the infrared image, so let’s look at NDWI (water index):
Much clearer. Easy to see now how the group was able to skate the eastern shore, and why that specifically was relatively good quality.
2/1, the day after I found much worse conditions (back to infrared false color):
The eastern shore looks whiter than on 1/27, but it’s still hard to tell exactly what’s going on. Take a look at the NDWI (water index) image and comparing it to the NDWI image from 1/27 makes it clear: what was ice with areas of snow is now much more evenly covered with snow:
And here's a side-by-side split-screen comparison of the NDWI visualizations from 2/1 (on the left) and 1/27 (on the right), to make the difference even clearer. (I'll explain how to use the comparison tool in another article).
Again, impossible to say for sure what’s going on without getting out there. But by comparing multiple visualizations, you can get a better sense of what’s going on and see how conditions are changing over time.
4. Getting the images while on the ice
Okay, you’ve done your research, you’re headed out to skate, but how do you use the satellite images while out on the ice?
The most straightforward way is just to keep using the EO Browser website. It works great on mobile browsers:
But though thatt’s great for viewing the satellite data visualizations, you’ll have to jump back and forth between that and whatever other maps you’re using, and or whichever app you’re using to track your skate.
If you use Gaia GPS, there are two map layers you can use: "Fresh Sat – Recent" and "Fresh Sat – Cloud Free". Here’s roughly the same view using the Fresh Sat – Recent layer, showing the gaps between ice plates clearly:
These use the visible light version of the HLS (Harmonized Landsat Sentinel) data, which has slightly lower resolution but updates more often. The “Recent” layer shows the most recent pass, regardless of cloud coverage; the “Cloud Free” layer shows the most recent pass that had less than 20% cloud cover. The map tiles have datestamps on them.
If you use Caltopo, you can make use of the “Sentinel Weekly” layers, and you can choose from several different visualizations. They call it “weekly” even though it’s 5-daily for simplicity. You can also choose specific passes from the menu.
You can also download images from the EO Browser, for saving to your phone or for printing. If you’re not logged in (remember, accounts are free for non-commercial use), you can only download a basic image, which looks like this:
To get that, click the little image icon on the right side of the screen, then click the green “Download” button. You'll probably also want to check the "Show legend" option.
If you’re logged in, you can download higher resolution images more suitable for printing:
I’m sure there are other good options as well, these are just the ones I know about and use!
5. Additional information
Q. Why the EO Browser and not the Sentinel Hub Playground?
A. The Playground is designed to always show results, even if that means stitching together images from a wide range of dates. Since ice conditions change so rapidly, we always want to be looking at the latest images (or if not, to explicitly know that we’re not). I find it easier to achieve this using the EO Browser, but YMMV.
Q. What’s the difference between Sentinel-2 and Harmonized Landsat Sentinel (HLS)?
A. By combining Landsat (8 and 9) and Sentinel-2 data, we get images that are updated every 2-3 days, but have a lower resolution (30m as opposed to 10-20m). (More on how this works.)
Q. Why only use Sentinel-2 / HLS and not Sentinel-1, Landsat, or MODIS (etc.)?
A. The combination of resolution and sensor data. Sentinel-2 gives us the best resolution with the information we need for ice scouting. HLS gives us good enough resolution with the information we need. Sentinel-1 gives better resolution, but not helpful information. The others have much worse resolution.
Q. How often do the images get updated?
A. Depends on exactly where you’re looking, due to the way the orbits overlap. The Sentinel-2 images typically update every 5 days (but some areas effectively more often, due to orbital overlap); the HLS images typically every 2 days.