Prologue
This is the first part of a blog post series dedicated to processing Sentinel-2 multiband imagery in ArcGIS Pro. I was recently playing with raster functions and other features in Pro and I found out that it is suprisingly easy to produce outstanding maps, mostly basemaps, derived from raw Sentinel-2. So, I decided to share my enthusiasm and of course my results with you.
These series include three blog posts which cover:
- Part One: making all sorts of crazy combinations of the Sentinel-2 bands, to produce impressive multiband imagery and to point out specific phenomena;
- Part Two: calculating spectral indices out of the multiband imagery, to further leverage the Sentinel-2 capabilities, in order to produce meaningful land cover layers;
- Part Three: combine elevation data with the land cover layers to produce high-quality, realistic basemaps.
I understand this might appear too obvious for experienced Remote Sensing and Earth Observation professionals, or too hard for people not familiar with raster processing, but ArcGIS Pro makes it really easy and creative!
Above all, the purpose is to enhance cartographic procedures and to provide fresh ideas for making better, more impressive and more real maps!
Understand the Data
Sentinel-2 is a mission of two twin satellites, promoted by the European Space Agency, travelling around the Globe, completing a full circle every ten days. The two satellites are flying in the same orbit but phased at 180°, which means that they co-operate to provide freshly captured imagery every five days. One is called S2A and its twin is called S2B!
Their flying path is happening above a grid of granules, which separates the surface of the Earth to 56,984 tiles! Each tile has a unique identifier code with a specific spatial extent. The coordinate system of the grid is the Universal Transverse Mercator, which divides the Earth’s surface into 60 zones. Each tile inherits the coordinate system of its corresponding UTM zone (e.g. UTM Zone 34N).
Both satellites are equipped with a powerful, high-resolution MultiSpectral Instrument (MSI). For every tile, that instrument samples thirteen different spectral bands (Table 1), which are saved in a huge database, in JP2 format, and are freely distributed to scientists and professionals and to anyone who wishes to observe the Earth!
These thirteen bands, together with a bundle of associate files, comprise the Sentinel-2 Products. There are two types of Sentinel-2 Products, broadly available to the public, the Level-1C and the Level-2A. The first type, also called L1C, is the unprocessed image, which the satellite captured, and features the top-of-the-atmosphere reflectances. The second type, also called L2A, is a processed copy of the same image and features the bottom-of-the-atmosphere corrected reflectances.
Every tile is saved with one unique name, a long but meaningful one, which provides useful information. That name follows a well-defined Naming Convention and it looks like this S2A_MSIL1C_20210115T091331_N0209_R050_T35TLF_20210115T104747.SAFE
The tile with the name of the example above tells me:
- S2A: I was captured by the first Sentinel-2 satellite,
- MSI: with its MultiSpectral Instrument,
- L1C: I was stored, without any processing, so I feature the top-of-the-atmosphere reflectances of the ground of the tile I sampled,
- 20210115T091331: I captured this imagery on January 15, 2021, at 09:13:31 am,
- N0209_R050: I was travelling along my Relative Orbit with number 050 and I was processed with PDGS Processing Baseline 02.09,
- T35TLFL: I was flying over and eventually sensed the tile with the unique identifier 35TLFL,
- 20210115T104747: I ceased sensing on January 15, 2021, at 10:47:47 am.
Every Sentinel-2 product is available to download in a zipped folder. Within this folder you may find the various spectral bands. Every spectral band represents one specific wavelength of the Electromagnetic Spectrum. Table 1, below, is a simplified table presenting the basic parameters of each band.
| Band Number | Band Name | Spatial Resolution (m) | Central Wavelength (nm) | Bandwidth (nm) |
|---|---|---|---|---|
| 1 | Coastal aerosol | 60 | 443 | 20 |
| 2 | Blue | 10 | 494 | 65 |
| 3 | Green | 10 | 560 | 35 |
| 4 | Red | 10 | 665 | 30 |
| 5 | Vegetation red edge | 20 | 704 | 15 |
| 6 | Vegetation red edge | 20 | 740 | 15 |
| 7 | Vegetation red edge | 20 | 781 | 20 |
| 8 | NIR 1 | 10 | 834 | 115 |
| 8A | Narrow NIR or NIR 2 | 20 | 864 | 20 |
| 9 | Water vapour | 60 | 944 | 20 |
| 10 | SWIR – Cirrus | 60 | 1375 | 30 |
| 11 | SWIR 1 | 20 | 1612 | 90 |
| 12 | SWIR 2 | 20 | 2194 | 180 |
One major thing to consider, for the purposes of map making, is the spatial resolution of each band. The NIR, Red, Green and Blue bands have a spatial resolution (cell size) of ten meters, which is awesome! Some other bands have spatial resolution of twenty meters, while others sixty meters.
Note, also that the band numbers of Sentinel-2 products, which correspond to a specific band name, hence a specific wavelength, are not the same with those of other remote sensing products, such as Landsat etc.
For an in-depth description of the mission’s products and algorithms, as well as details on the MultiSpectral Instrument (MSI) and its performance, I highly recommend that you read the Sentinel-2 MSI Technical Guide, as well as the Sentinel-2 User Handbook, the official guides provided by the European Space Agency.
Get the Data
The first step I always take, before start searching the internet for data, is to define the area of my case study, or in Sentinel-2 terms, to identify in which tile this area falls into. I have downloaded the Sentinel-2 Tiling Grid in KML format and converted it to a feature layer in ArcGIS Pro, using the KML To Layer (Conversion) tool. Having the whole World in tiles is much easier to understand which is the one I am looking for.
The tile I have selected to work with for the purposes of these blog post series is the 35TLF, located at the Northeastern part of Greece (Picture 1).
Next step is to download that tile from either Earth Explorer or Copernicus Open Access Hub. Both platforms have a user-friendly interface and provide a number of filter parameters to narrow down my selection, until I find the specific product I am looking for.
I am not going to spend lines describing how to search and download Sentinel-2 products from these platforms, because I assume you will find one of the thousands of freely available articles and videos on the internet on how to do such a thing. Tom Patterson has also very recently written a Sentinel-2 Photoshop Tutorial, where he also explains how to get the data.
I will only briefly point out the most important filter parameters that I have to use in order to perform a better, faster and most effective search:
- The area of my case study, which I may graphically define in both platforms, as well as to explicitly write the identifier code of the particular tile I am searching for (recommended);
- The date range from where the platforms should search; I most of the times provide a narrow date range, considering the scope of my study; for instance, if I want to depict the snow-covered mountain tops of Crete, I will select dates between December and March; On the other hand, If I want to depict the pick of the olive trees cultivation, I will select dates between September and November;
- The Sentinel-2 product type, which may be either a L1C, or a L2A;
- The cloud coverage, which rarelly is 0%, but it is crucial to find imagery with the least possible clouds.
Both Earth Explorer and Copernicus Open Access Hub will return a number of product options according to my searching criteria and eventually I will download the one that is the most suitable to my needs.
For these blog post series I downloaded from Earth Explorer the Sentinel-2 tile with name S2A_MSIL1C_20210115T091331_N0209_R050_T35TLF_20210115T104747.SAFE. The product comes in a zipped folder, which I further unzip and store in my project’s folder. Within the unziped folder, I have to click a couple of subfolders until I find the one that contains the Sentinel-2 bands. Normally the bands are located in a path similar to this GRANULE\L1C_T35TLF_A029076_20210115T091334\IMG_DATA.
Having stored the bands in my project’s folder and knowing their location, I can now merge them in one Band Composite in ArcGIS Pro and do all sorts of crazy stuff. So, let the creative process begin!
The Band Composite
In ArcGIS Pro I create a new map with the name Band Combinations. I click on the Add Data button and I navigate to the folder I have stored the Sentinel-2 bands. I don’t have to add all thirteen bands, so I select only those that I need for my project. Particularly, I select the bands B02, B03, B04, B08, B08A, B11 and B12 (Picture 2) and I click OK.
The selected bands are added on the map. If I right click on one of them and open its properties, I can read all the necessary parameters like pixel depth and type and spatial reference.
All bands appear in a singleband grayscale mode, so I have to merge them in one single multiband colorful raster layer. I will do this with the Composite Bands function.
So at the Imagery contextual tab I click on the Raster Functions button to open the Raster Functions pane (probably my most favourite pane in Pro!) I expand the Data Management group and I select the Composite Bands function.
At the General tab of the Composite Bands Properties I select 16 Bit Unsigned as the Output Pixel Type and at the Parameters tab I add the bands on the Rasters lists and for Cellsize Type I select Min Of.
The reason I selected Min Of is that the bands B02, B03, B04 and B08 have a spatial resolution of 10 meters, while the bands B08A, B11 and B12 have a spatial resolution of 20 meters (see Table 1), so I definitely prefer the output composite to have 10 meters resolution.
Another very important thing to consider is the order of the bands in the Rasters list. Sentinel-2 has its own bands numbers, but these numbers are not preserved in the output band composite in Pro. The band numbers in Pro are defined by their order in the Rasters lists of the Composite Bands function. So, it’s up to me to define that order, according to my needs. Table 2 exhibits my preferred order of the input Sentinel-2 bands and their correspondence with the output ArcGIS Pro bands.
| Sentinel-2 Bands | Band Name | ArcGIS Pro Bands | Resolution (m) |
|---|---|---|---|
| B04 | Red | Band 1 | 10 |
| B03 | Green | Band 2 | 10 |
| B02 | Blue | Band 3 | 10 |
| B08 | NIR 1 | Band 4 | 10 |
| B8A | NIR 2 | Band 5 | 20 |
| B11 | SWIR 1 | Band 6 | 20 |
| B12 | SWIR 2 | Band 7 | 20 |
In ArcGIS Pro, in the list of the Parameters tab, I use the Move Up and Move Down arrows to rearrange the order of the input Sentinel-2 bands, according to Table 2, and I click Create New layer (Picture 3).
The newly created multiband image is added on the map and on the contents pane. I no longer need the singlebands of Sentinel-2 so I remove them from the contents pane.
If I now right click on the multiband image and open its properties dialog I may see at the Band Metadata group that indeed it contains all the seven bands I merged (Picture 4).
Band Combinations
Band composites are useful for illustrating specific phenomena on a map, to facilitate the process of photo-interpretation and to better observe and understand an area of interest. I can combine the bands of the multiband image in all possible and meaningful ways to produce various bands composites, each one of them having their own purpose and use.
This can be achieved at the Symbology pane by selecting which band will participate at the Red, Green and Blue channels of the multiband image. Brightness, Contrast and Gamma corrections, as well as the appropriate Stretch type are also key for a better visualization of the band composite.
There are numerous different band combinations, each one of them having a certain scope. Here I will only illustrate the most popular and frequently used.
Natural Color (B04 – B03 – B02)
The Natural Color band combination is the most classic one, where the Red, Green and Blue bands of the multiband image are assigned to the Red, Green and Blue channels, respectively. They call it Natural for the reason that the colors of the bands, used at the composite (Table 3), fall into the visible range of the electromagnetic spectrum, or in simple words, our human eyes can see them.
| Sentinel-2 Bands | Band Name | ArcGIS Pro Bands |
|---|---|---|
| B04 | Red | Band 1 |
| B03 | Green | Band 2 |
| B02 | Blue | Band 3 |
So, in ArcGIS Pro I select the Band Composite at the Contents pane and I open the Symbology. By default RGB is selected at the Primary Symbology and Bands 1, 2 and 3 of the image correspond to the Red, Green and Blue channels (Picture 5).
To enhance the appearence of the composite I select Standard Deviation for Stretch Type, Cubic for Resampling Method and I adjust the values for Brightness, Contrast and Gamma parameters to -4, 12 and 0.7, respectively.
The composite has a spatial resolution of 10 meters, which means it can scale up to 1:20,000 and even more, before start losing information.
Natural Color is similar to the World Imagery or Bing Aerial or Google Satellite. I can apply all the classic photo-interpretation techniques. Natural Color derived from Sentinel-2 has two great advantages though. One is that it is orthorectified, thus it is suitable for digitization or measurements, and two that it is being updated every five days, hence it is great to track frequent changes.
Color Infrared (B08 – B04 – B03)
Color Infrared is the most common and well-known false color composite. I can produce it by assigning the NIR, Red and Green bands of the multiband image to the Red, Green and Blue channels at the Symbology pane, respectively. The correspondence between the bands of Sentinel-2 and those at the band composite in ArcGIS Pro is illustrated on Table 4.
| Sentinel-2 Bands | Band Name | ArcGIS Pro Bands |
|---|---|---|
| B08 | NIR 1 | Band 4 |
| B04 | Red | Band 1 |
| B03 | Green | Band 2 |
In False Color Infrared areas covered with vegetation appear in shades of red. Generally, deep red hues indicate broad leaf and/or healthier vegetation, while lighter reds signify grasslands or sparsely vegetated areas. Urban and bare areas appear in cyan, while soil appear in shades of brown. Ice, snow and clouds are white or light cyan. This is a very popular band combination and is useful for vegetation studies, monitoring drainage and soil patterns and various stages of crop growth (Picture 6).
Agriculture (B11 – B8A – B02)
The Agriculture band composite can be produced by assigning the SWIR 1, NIR 2 and Blue bands of the multiband image to the Red, Green and Blue channels at the Symbology pane, respectively. The correspondence between the bands of Sentinel-2 and those at the band composite in ArcGIS Pro is illustrated on Table 5.
| Sentinel-2 Bands | Band Name | ArcGIS Pro Bands |
|---|---|---|
| B11 | SWIR 1 | Band 6 |
| B8A | NIR 2 | Band 5 |
| B02 | Blue | Band 3 |
As the name implies, this band combination is useful for monitoring agricultural crops. Bright green represents vigorous, healthy vegetation while non-crops, such as mature trees, appear in a muted green. Coniferous forests appear as a dark, rich green, while deciduous forests appear as a bright green. Sparsely vegetated and bare areas appear brown and mauve. Snow and ice appear in vivid light blue (Picture 7).
It is worth to note that the band combination of the Agriculture band composite is similar to the Vegetation Analysis (B11 – B8A – B04) combination, which is useful for vegetation studies, and is widely used in the areas of timber management and pest infestation.
Healthy Vegetation (B8A – B11 – B02)
The Healthy Vegetation band composite can be produced by assigning the NIR 2, SWIR 1 and Blue bands of the multiband image to the Red, Green and Blue channels at the Symbology pane, respectively. The correspondence between the bands of Sentinel-2 and those at the band composite in ArcGIS Pro is illustrated on Table 6.
| Sentinel-2 Bands | Band Name | ArcGIS Pro Bands |
|---|---|---|
| B8A | NIR 2 | Band 5 |
| B11 | SWIR 1 | Band 6 |
| B02 | Blue | Band 3 |
Healthy vegetation appears in shades of reds, browns, oranges and yellows. Soils may be in greens and browns, urban features are white, cyan and gray, bright blue areas represent recently clearcut areas and reddish areas show new vegetation growth, probably sparse grasslands. Clear, deep water will be very dark in this combination, if the water is shallow or contains sediments it would appear as shades of lighter blue (Picture 8).
It is worth to note that the Healthy Vegetation band composite is similar to the Land/Water (B8A – B11 – B04) a band combination good for picking out land from water, where land appears in shades of orange and green, ice stands out as a vibrant magenta color, and water appears in shades of blue.
Atmospheric Penetration (B12 – B11 – B8A)
The Atmospheric Penetration band composite can be produced by assigning the SWIR 2, SWIR 1 and NIR 2 bands of the multiband image to the Red, Green and Blue channels at the Symbology pane, respectively. The correspondence between the bands of Sentinel-2 and those at the band composite in ArcGIS Pro is illustrated on Table 7.
| Sentinel-2 Bands | Band Name | ArcGIS Pro Bands |
|---|---|---|
| B12 | SWIR 2 | Band 7 |
| B11 | SWIR 1 | Band 6 |
| B8A | NIR 2 | Band 5 |
This combination involves no visible bands and it provides the best atmospheric penetration. Vegetation appears in blue colors, bare soil in shades of brown and yellow. Water appears in very dark shades or black, thus coast lines and shores are well defined (Picture 9).
This band combination can be used to find textural and moisture characteristics of soils and it can be also useful for geological studies.
Snow and Clouds (B02 – B11 – B12)
The Snow and Clouds band composite can be produced by assigning the Blue, SWIR 1 and SWIR 2 bands of the multiband image to the Red, Green and Blue channels at the Symbology pane, respectively. The correspondence between the bands of Sentinel-2 and those at the band composite in ArcGIS Pro is illustrated on Table 8.
| Sentinel-2 Bands | Band Name | ArcGIS Pro Bands |
|---|---|---|
| B02 | Blue | Band 3 |
| B11 | SWIR 1 | Band 6 |
| B12 | SWIR 2 | Band 7 |
Thick ice and snow appear in vivid red or red-orange. Vegetation will appear greenish and soil will appear bright cyan in the image. Clouds appear white, so this combination of bands is especially useful for distinguishing clouds from snow (Picture 10).
Epilogue
Thank you for reaching so far. In the second part I will describe how I transform this band composite to a meaningful land cover layer in ArcGIS Pro, which will serve as a pretty basemap.
In the meantime, I would be very happy to know that anyone got inspired by part one here and actually produced a map with Sentinel-2. I would also be happy to receive any feedback or suggestions for improvement.
Kindest regards from Crete, Greece
Spiros
Sources
- Sentinel-2
- Sentinel-2 Imagery
- Sentinel Explorer
- EOS Land Viewer
- List of band combinations in Sentinel 2
- Sentinel Hub
- Sentinel Hub Playground
- Sentinel 2 Bands and Combinations