Using Sentinel-2 for crop monitoring

Sentinel-2 is the optical satellite of the Copernicus programme. It can be compared to Landsat, although it has a better resolution, of 10 to 20 meters. We’ll be using it for crop monitoring with simple vegetation indices.

Overview

Sentinel-2 is the high-resolution optical satellite of ESA and the EU. The images have a resolution of 10 to 20 meters, higher than Landsat, and, as always with the Copernicus programme, the data are free and open.
In this tutorial, we’ll download an image, make it look good, and create maps of vegetation indices to show the general health of crops and other vegetation. This is a basic tutorial, but even for those with experience in remote sensing, it is a good starting point for working with Sentinel-2 data.

Time to go through this tutorial and process an image: : 45 minutes
Short summary: We’re calculating vegetation indices from Sentinel-2 data to monitor crop health.
Want a more step-by-step approach? Follow along.

Getting the data and the tools

Although all Sentinel-2 data are available on the Copernicus data hub, they are also HUGE – 5 to 6 GB for an image. So in this tutorial we’re using a plugin in QGIS called
“Semi-automatic classification plugin”. It can download tiles of 100 by 100 km, instead of the entire image, plus it processes it for you, and has some nice other tools.
You can find a quick how-to on Youtube.

Open up QGIS, and under Plugins –> Manage and install plugins, search for the plugin name (“Semi-automatic classification plugin”), and install it. For me, the toolbar of this plugin appeared below the other toolbars in QGIS.

Semi-automatic classification plugin toolbar
Semi-automatic classification plugin toolbar

Visualize your image

Hover over the icons in the toolbar, and find the one that says “Tools” or “Semi-automatic classification plugin”. This will give you access to all the tools.

Semi-automatic classification plugin - download Sentinel
Semi-automatic classification plugin – download Sentinel

Go to the “Download Sentinel” tab, and follow along with the Youtube video, or do the following:

  • Enter your login information that you use for the Sentinel hub.
  • Press the PLUS sign under UL, and go to the map view in QGIS, and click on the upper left corner of your area of interest. Do the same for the lower right (LR) corner.
  • Specify the time period you want to search images for.
  • Click “Find Images”.

When the results return, all the granules (100x100km parts of the image) of an image covering your area of interest will be given. Play around with the “Display granule preview” to find the granule that covers your area of interest. When you’ve found it, select it, and click “Download granule from list”.

Be aware that “Pre process images” is ticked off in the lower right corner of the plugin window, so make sure to turn it off if you don’t want it preprocessed! For the purposes of this tutorial, I left it on.

When it’s done downloading, switch to the “Band Set” tab. Select band 2, 3, 4, 8A and 11, and click “Add rasters to set”. After you click “Create raster of band set (stack bands)”, the map in QGIS will display an RGB image:

True color RGB Sentinel-2 image
True color RGB Sentinel-2 image

This is just to make it easier to work – you can skip this step if you prefer.

Calculate vegetation indices for crop monitoring

Let’s move on and get some information from the image! We’re going to calculate the Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI), to check the vitality of the crops, and their moisture content. If you want to learn more about the myriad of vegetation indices out there, check out this paper.

In the plugin, switch to the “Band calc” tab. If you’ve saved the image as a stack, the formula for NDVI is like in this figure:

NDVI calculation in QGIS
NDVI calculation in QGIS

Otherwise, it is (Band8A-Band4)/(Band8A+Band4).

Press “Calculate”.

The result is displayed in grayscale.

NDVI results in grayscale
NDVI results in grayscale

To make the results easier to interpret visually, we’ll change the color palette.
Right-click on the NDVI image in the Layers panel on the left, and click “Properties”.

Image properties
Image properties

In the Style tab, choose “Singleband pseudocolor” as the Render type, tick off “Invert” in the Generate new color map box, and click “Classify” in the same box. Click “OK”.

NDVI results of crop fields
NDVI results

The NDVI image shows where vegetation is healthy and mature (red colors), while rivers and cities are shown in blue because of lack of vegetation. When you zoom in to a particular field, you can see variations in NDVI on the field, which indicates differences in crop health and maturity across the field. Farmers use this to finetune fertiliser spraying.

To get more information on what could be the problem, when faced with a lower than expected NDVI, you can use the NDWI, which will indicate drought stress in crops. This document shows how it is used for operational monitoring of European crops with the MODIS satellites.

To calculate the the NDWI from Sentinel-2, we’ll be using band 8A as the NIR band, and band 11 as the SWIR band. Otherwise, the calculation of the index is the same as that of the NDVI:

NDWI calculation
NDWI calculation

Or also: (Band 8A - band 11)/( band 8A + band 11)

Apply a pseudocolor palette to the NDWI results too.

NDWI results
NDWI results

Higher NDWI values, shown as red colors, indicate high water content and high vegetation cover.

When we compare the NDVI to the NDWI, we can see where vegetation has a high density (red in NDVI results), but a low water content (yellow in NDWI).

Which vegetation indices do you usually use for crop monitoring?

0 0 vote
Article Rating
Subscribe
Notify of
guest

This site uses Akismet to reduce spam. Learn how your comment data is processed.

21 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Simon
4 years ago

Very nice Article, Annekatrien. I really love your input about sentinel

Annekatrien
Annekatrien Debien
4 years ago
Reply to  Simon S.

Thanks Simon!

AFRAH
AFRAH
3 years ago

Dear Annekatrien Debien
I hope you are doing well,

Please just I was wondering if is there any tutorial to make stack layer for multi sentinel 2 images.

Your reply will be highly appreciated.
Best Regards,
Afrah

Norbert J.
Norbert J.
4 years ago

Realy very nice, Annekatrien! I love your input about sentinel too. I’m curious about your next blog post.

Annekatrien
Annekatrien Debien
4 years ago
Reply to  Norbert J.

Thanks Norbert!

dkaludjerovic
dkaludjerovic
4 years ago

Hello I am wondering is Band 8A 20 meters or 10 meters spatial resolution? I know that band 11 is 20 meter

Annekatrien
Annekatrien Debien
4 years ago
Reply to  dkaludjerovic
dkaludjerovic
dkaludjerovic
4 years ago

Ok, thanks for reference. I must say that reference is not current since 8a band in the most recent Sentinel 2 a download is 20 m resolution and 8a band is 8b in reference you mention. Thank you for helping me to clarify this, it will be useful to others

zsoooc
zsoooc
4 years ago

No, Band8A has 20 meter resolution.

dkaludjerovic
dkaludjerovic
4 years ago

Also, I am confused a little bit, in its original paper, Gao, B.-C. 1996. NDWI – A normalized difference water index for remote
sensing of vegetation liquid water from space. Remote Sensing of
Environment 58: 257-266., Gao is using 1240 nm but band 11 in Sentinel is 1610 nm. Can you explain it?

Annekatrien
Annekatrien Debien
4 years ago
Reply to  dkaludjerovic

I based my formula on this document – http://edo.jrc.ec.europa.eu/documents/factsheets/factsheet_ndwi.pdf – in which they use MODIS band 6 as the SWIR band, which is around 1630 nm. It’s not perfect, so I wouldn’t use it without testing first, but it gives an indication.

dkaludjerovic
dkaludjerovic
4 years ago

I got it. Thank you

SoDurval
SoDurval
4 years ago

My Sentinels login seems OK but when I clicked on the “Find images” button :

Error [50]: Internet error:

Anyone have this trouble ?

Thanks — So

P.S. Tested with QGIS 2.14.3-Essen / OS X 10.11.5

Zlata
Zlata
4 years ago
Reply to  SoDurval

Same here, same QGIS version on WIndows 10. Landsat dataset seem to be loading OK though.

Interpine
Interpine
4 years ago
Reply to  Zlata

Did you sort this? have the same problem

Bev Taylor
Bev Taylor
4 years ago
Reply to  Interpine

I’m getting the same error with QGIS 2.16.0. any solutions yet?

Interpine
Interpine
4 years ago
Reply to  Bev Taylor

yip I’ll email it to you. Also we are looking at having a half day training course on sential at the end of forest tech. Sarah

TacoChuck
TacoChuck
4 years ago
Reply to  SoDurval

I know this question is a bit old, but just in case it helps someone in the future, Luca posted a fix for this error and it worked for me:
http://fromgistors.blogspot.com/2016/06/sentinel-2-download-issues.html

zsoooc
zsoooc
4 years ago

Why did you used Band8A (red edge) instead of Band8 (NIR) for NDVI calculation?

Misbah Noureen
Misbah Noureen
3 years ago

Why did you used Band8A (red edge) instead of Band8 (NIR) for NDVI calculation? plz somebody ans it

johannes
johannes
3 years ago

Why did you used Band8A (red edge) instead of Band8 (NIR) for NDVI calculation?