SICE — Snow and Ice Products on Demand
Operational Sentinel-3 Snow and Ice Products Within EDC
Areas with permanent snow and ice cover, namely those in the Arctic, play a crucial role in maintaining a global energy balance by reflecting as much as 90 % of the radiation from the sun back into space. Doing so, the Earth absorbs less heat, keeping it cooler. As snow and ice physical properties (such as grain size) change, so does the percentage or solar radiation they reflect (albedo). Additionally, large scale melting of ice sheets lowers global albedo by increasing the amount of dark surfaces (oceans), as well as increases global sea levels and disrupts ocean currents and arctic ecosystems. Monitoring the properties of snow and ice is thus important in monitoring of glacial ecosystems, as well as predicting the progression of climate change.
In this blog post, we explore the algorithm developed for the purpose of monitoring the properties of snow and ice, called SICE (Operational Sentinel-3 snow and ice products). We look into the importance of polar ice cap monitoring, dive into SICE data and use it to look at Southern Greenland snow and ice changes through recent years. Finally, we look into how you can access already processed SICE products, as well as how you can order them on demand with just a few clicks within the Euro Data Cube.
About SICE
SICE algorithm processes Sentinel-3 satellite data to retrieve snow and ice properties on all glaciated and fully snow covered areas. Its processing chain is automated and open source. The resulting snow and ice products include retrieval of snow albedo, optical grain size, snow quality, Sentinel-3 band data and cloud identification. SICE algorithm has been used to process several locations already, and some of the results, such as planar albedo, can be obtained directly from the SICE Dataverse.
Before SICE, snow and ice properties were retrieved using MODIS satellite. Researchers have shown that SICE algorithm is better at retrieving these data than MODIS and just as good as PROMICE data retrievals. Read about SICE data retrieval algorithms in detail in the official specifications document.
Within Euro Data Cube, Sentinel Hub was used to reproduce the SICE processing chain and offer data on demand to EDC users. The results are identical to those obtained with the original processing chain (save for the lack of a cloud detection layer in Sentinel Hub data). Explore some of the Sentinel Hub’s already processed SICE data using the PolarTEP viewer.
EDC SICE Data
EDC SICE algorithm produces 9 GeoTIFF files, all in spatial resolution from 1200 to 300 meters. The products are described below.
- Diagnostic Retrieval
- Snow Grain diameter
- Snow specific surface area
- Shortwave Broadband Planar Albedo
- Shortwave Broadband Spherical Albedo
- Top of the Atmosphere Reflectance (OLCI band 1)
- Top of the atmosphere Reflectance (OLCI band 6)
- Top of the Atmosphere Reflectance (OLCI band 17)
- Top of the Atmosphere Reflectance (OLCI band 21)
DIAGNOSTIC RETRIEVAL
This product holds information on snow quality — specifically whether snow is clean or polluted, and whether property retrieval was possible. Pollution refers to snow impurities, such as dust, soot, algae or liquid water, and is only recognized when relatively heavy (above 1ppmv). Diagnostic retrieval is a classified dataset with 8 classes, with the first 4 classes retrieved (meaning they hold information on whether snow is clean or polluted) and the last 4 classes indicating a specific error. The colors in this case were chosen by us.
To simplify the visualization of diagnostic retrieval, we grouped values together to display clean snow (values 0 and 7) as blue, polluted snow (values 1 and 6) as grey and values of not retrieved classes as white.
ALBEDO
Albedo is a parameter every surface has. It is the ratio of reflected light compared to total sunlight that falls on the surface. Another way we can think of it is as reflectivity of the surface, which is not self illuminated. Low albedo means that less light is being reflected, or, the other way around, the more sunlight the surface reflects, the higher albedo it has.
Albedo depends on the following factors:
- wavelength of light — materials absorb and reflect specific wavelengths.
- the angle under which sunlight reaches the surface, also called the incidence angle — at certain angles, light is more likely to be absorbed than at others.
- texture of the surface — smoother surfaces are more reflective.
- thickness of the material
- Structure of the material — determines how easily the light can penetrate the surface
See the global albedo map from January 2017 on the image below. We can see that water with its very low reflectivity is completely dark, indicating low albedo, while snow covered arctic areas are completely white, thus indicating a high albedo. We can also see high albedo in deserts. This is due to sand color being quite light and the general surface of sand dunes relatively smooth.
When large ice caps melt, the area previously covered in snow is converted into water, lowering global albedo. The oceans can then absorb more solar radiation, heating the planet up. This in turn increases ice thinning and melting, producing a positive feedback loop. Albedo of snow covered areas is not constant, but changes with time, for example, due to seasonal snow cover differences, due to presence of melting water, changes in snow grain size, presence of algae, ice cap melting, etc.
Shortwave Broadband Albedo Products in SICE
In SICE, Shortwave broadband relates to the shortwave infrared (IR) wavelength used in the calculations. The values of the albedo products in SICE generally range between 700–2400 nm. 700 nm indicate that low values belong to the red part of the light spectrum and 2400 nm indicate that high values belong to the IR part of the light spectrum. Note that albedo is only calculated for areas 100 % covered by snow. A lot of glacial areas on Greenland’s coast thus lack albedo calculations.
SICE offers two shortwave (sw) broadband (bb) albedo products:
- Planar albedo (albedo_bb_planar_sw)
- Spherical albedo (albedo_bb_spherical_sw)
Planar albedo is an albedo of a flat surface, illuminated by sunlight coming from a single direction, while spherical albedo is the albedo of a surface when illuminated by sunlight that comes from all directions and angles (diffuse illumination). Spherical albedo is considered to be an indicator of Earth’s radiation budget and thus a key driver in global climate systems.
SNOW GRAIN PRODUCTS
Snow is composed of tiny ice crystals, that accumulate into snow grains on the ground. The snow grain size impacts how much radiation is reflected or absorbed, as larger grains increase the path length of incoming light within ice crystals, which increases the probability for the light to be absorbed. When snow absorbs solar radiation it heats up, turning into liquid on the surface, which flows between snow grains and increases the optical grain size. Wet snow grains are thus similar to large snow grains, both increasing the surface albedo. We know that infrared reflectance responds to snow grain size, so we can use infrared satellite bands to discern it.
Because snow grain shape is highly variable, snow grain size is a subjective measurement that is hard to quantify in a standardised way, and its quantification varies from one study to another. For this reason, scientists came up with the metric called specific surface area, which is an objective measurement to define the physical structure of the snow pack. The specific surface area of snow is calculated as the surface area per unit mass, essentially representing the ice–air interface per unit mass.
SICE algorithm returns two snow grain products:
- Snow grain diameter (grain_diameter)
- Snow specific surface area (snow_specific_surface_area)
Snow specific surface area is a measurement of snow grain size in m2kg-1, while Snow grain diameter is measured in millimeters.
SENTINEL-3 OLCI BANDS B01, B06, B17, B21
Sentinel-3 OLCI has 21 optical bands, which are all accessible via Sentinel Hub. SICE algorithm returns four of these bands, which were deemed the most useful for snow and ice property retrieval. The files containing these products are named r_TOA_01, r_TOA_06, r_TOA_17 and r_TOA_21, with TOA indicating top of the atmosphere reflectance.
Below is an RGB composite of bands B01, B17 and B21, returning Southern Greenland in pleasing colors. Properties of the three bands are combined together to visualize the surface in various color shades. See how it works within Polar Tep Viewer.
Use-Case Example: Greenland Ice Sheet
Let’s look at how Greenland snow and ice parameters changed between April 2017 and April 2022.
The following three images display changes in clean vs. polluted snow cover. The first two images show diagnostic retrieval on April 2017 and 2022, respectively, and the third image displays changes in clean snow cover between these dates. Magenta colors indicate clean snow reduction and green colors indicate its increase. White color indicates that no change in clean snow cover occured.
The following change visualizations were created using a multi-temporal RGB evalscript with 2017 data in the red and blue channels, and 2022 data in the green channel.
As we can see, clean snow cover was reduced drastically on the west side of Greenland since 2017, replaced instead by polluted snow. This could be a consequence of melting, if the amount of liquid water in the snow increased (remember, presence of liquid water is one of the possible causes of snow pollution for this product).
As seen on the images below, we observe a significant albedo reduction on the west side of Greenland, as well as a general mild reduction across central Greenland. The albedo reduction is similar to the reduction of the clean snow cover above, which can be explained by the fact that clean snow generally has a brighter color and smoother surface (thus a higher albedo) than polluted snow. Albedo likely decreased due to ice melting and presence of melting water as well.
On the image below we see a striking grain diameter increase since 2017. The bright green section on the East can be attributed to glacier melting in April 2022, moving chunks of ice into the ocean that weren’t present in April 2017. General grain diameter increase is in large part consistent with albedo decrease from the first two images.
Let’s look at two specific Greenlandic glaciers up close. Snow cover reduction between 2017 and 2022 is clearly visible from Sentinel-2 natural color images.
By comparing diagnostic retrieval values between 2017 and 2022, we see a dramatic decrease in clean snow, as seen in magenta colors on the image showing differences between the two dates on the right.
Snow grain size increased significantly in 2022, as seen by vivid green color on our comparative image below.
Due to glaciers not having 100 % snow cover, albedo visualizations have gaps in between. By comparing images from 2017 and 2022, we see a drastic drop in albedo in the area, which we can attribute to increased grain size and decrease of clean snow area.
To figure out how much exactly these parameters decreased over Southern Greenland, we used Sentinel Hub’s Statistical API to observe the mean albedo and grain size values for 6 April dates from 2017 and 2022. The chart below shows a downtrend for mean spherical albedo, and an uptrend for snow grain diameter. In fact, the two look almost perfectly inverted. We can see that the year 2020 was an outlier, showing an increase in albedo and decrease in optical grain size, which can be attributed to thicker snow cover that year. Despite the increase in 2020, chart indicates an alarming trend in disappearing snow cover through the recent years.
Order on Demand SICE Data With EDC
Browsing freely available data is great, but sometimes we need data for a specific location and date, which isn’t yet available for download. When this is the case, you can order SICE processing for any location and date using the EDC’s Insights on demand functionality. You can purchase SICE data either in the EDC Browser or within the PolarTEP viewer, as soon as you have an EDC account and allocated at least 10 credits.
You will need the following specified:
- polygon of interest
- date (any date since April 2016 — the date of Sentinel-3 launch )
- resolution (can be any number between 300 and 1200)
Important note: SICE won’t be retrieved when solar zenith angle is lower than 75 degrees, as that means the area is in darkness and we can’t retrieve optical imagery. This is common over polar regions in winter (remember that in some regions, the area is in darkness for several months). If you order data on such a date, your GeoTIFFs will lack SICE calculations. We suggest to generally only order data between May and September. Note also that SICE algorithm is best suited for large scale polar and arctic regions.
When clicking purchase, you will see a summary of your order, as well as an estimation of how many credits will be used. Credits depend on your area of interest and resolution. The algorithm isn’t too costly — the whole Greenland in 300 m resolution for a day costs approximately 20 credits, which corresponds to roughly 20 EUR.
To learn more about Insights on demand and how to use the functionality, check out the following blog posts:
