Contents

Search (over 200 pages above & below)

Crop Agriculture

Course 3A - Fundamentals of Applying Satellite Imagery to Crops; Downloading Sentinel-2 Imagery for an Area of Interest.

INCLUDES IMPORTANT QGIS SYMBOLOGY INFORMATION.

The point of this course is to provide steps to acquire free satellite imagery, as well as to describe the proper fundamentals to understand and effectively apply satellite imagery to crops or vegetation, in general. Free QGIS software (QGIS download page on this website; opens in new tab) is used to show crop imagery examples. Information is included that describe how to use imagery in QGIS, but the principles shown apply to crop/vegetation analysis in other software, such as ArcGIS. NO EXPERIENCE IS NECESSARY.

The examples below show Sentinel-2 imagery (free, relatively new satellite imagery). The concepts also apply to Landsat imagery (the world's longest continuously acquired satellite imagery). Landsat & Sentinel-2 can be downloaded through links on the Free Satellite Imagery page or from pages on this website. With the combination of Landsat and Sentinel-2, it is now possible to acquire imagery of the same area every few days (cloud-willing).

 

1) DOWNLOADING FREE SATELLITE IMAGERY: WE RECOMMEND USING SENTINEL-2 IMAGERY FOR THIS COURSE (SENTINEL-2 EXAMPLES ARE SHOWN)

Although, you can download Sentinel-2 imagery from this website that can be used for this course, we recommend that you download Sentinel-2 imagery from the imagery source, Copernicus, for an area of your choice (described in next section). If you prefer to download Sentinel-2 from this website, you can download imagery from the Sentinel-2 Conversion to Surface Reflectance Tutorial (scroll down to download imagery). If you prefer Landsat, we recommend these Landsat 8 downloads.

 Details for Downloading Sentinel-2 Imagery for an Area of Interest

To download Sentinel-2 imagery from the European Space Agency (ESA) Copernicus viewer make sure you have decent internet speed (go somewhere that has free high-speed internet and use that, if necessary), then access the Free Satellite Imagery page (opens in new tab). When on that page, scroll down and click Copernicus, then click Open Hub. If you need to an account, click the person icon near the upper right, then click Sign Up - you will need to enter a small amount of information, then you are ready to download imagery for free.

FOLLOW THE COPERNICUS STEPS ON FREE SATELLITE IMAGERY PAGE TO LOCATE AND DOWNLOAD SENTINEL-2 IMAGERY FOR A PARTICULAR AREA. Keep in mind that Sentinel-2 original images are divided into tiles; tiles may vary in size from a complete square to a sliver (unlike Landsat, which are all complete scenes). Your job now it to try to determine which image/s has a clear view (free of clouds and cloud shadows) of your area of interest. To do this, click the Zoom to Product icon on each image (if there are more than one). The viewer will zoom to the extent of the image, and you should be able to see where your orange box or polygon is in relationship to the image extent. The small thumbnail image can give you an idea if the area of interest is clear, but you will probably want a more detailed view of the image. To get a more detailed view, click the View Product Details icon, the click the Quicklook image for it to expand. If it looks clear, close the Quicklook tab, then close the View Product Details window by clicking "X". You are now ready to download the image.

* IMPORTANT: IF ASSESSING CROP CONDITION, DOWNLOAD IMAGERY FOR A DATE RANGE WHEN THE CANOPY IS PREDOMINANTLY CLOSED (THE REASONS FOR THIS ARE EXPLAINED IN STEP 4).

 

2) DOWNLOADING AND INSTALLING FREE QGIS SOFTWARE

Click here to download Free QGIS from this website; the version has been scanned by Webroot antivirus software and had no malware detected. You need GIS software to process and apply the imagery. Course 1A on the Free Courses page guides you through the process of downloading and installing QGIS. Also download the OpenLayers Plugin as well as 7-Zip if you need extraction software. Run the 7-Zip program to install it and extract the OpenLayers Plugin. The OpenLayers Plugin enables you to have high-resolution imagery as a continuous background, so you can better locate areas within a field or your area of interest. You need fast enough internet speed to for the OpenLayer Plugin to work efficiently. If you cannot have a fairly fast internet connection, Course 1C on the Free Courses page shows how to download high-resolution imagery that can be used as a background area.

 

3) OPENING IMAGERY IN FREE QGIS

After satellite imagery is downloaded and decompressed, QGIS is installed, and your background imagery sources is established, you are ready to open and apply satellite imagery in QGIS software. Course 1D on the Free Course Listing page shows how to decompress the compressed downloaded imagery (with free software if needed). You will probably want to have guidance to open Sentinel-2 imagery, in particular, as the image folder is many folders deep into the downloaded folder. This is an easy process after you have learned it though. Keep in mind, you can decompress the entire downloaded folder or you can enter the folder by double-clicking on it and navigating to the image folder then copy an image, or images (by pressing the CTRL key and clicking on individual images), then pasting the file out of the folder to another location. The path to the image folder from a downloaded Sentinel-2 folder is as follows: double-click the compresses (zipped) downloaded folder > double-click the sole main folder that will appear > double-click the GRANULE folder > double-click the sole folder that will appear > double-click IMG DATA > the images will be listed there. For the example here, we are showing red (band 4; B04 in the image file name), NIR (band 8; B08 in the image file name), and SWIR (band 11; B11 in the image file name) band imagery. Many other of the bands have useful applications for crop agriculture and vegetation, in general. To follow along here, press the CTRL key, and left-click on B04, B08, and B11 to select them, then right-click on them and copy them, then back out of that folder and past them preferably in a new folder you have made or just inside the compressed folder so they are isolate

Add background imagery with the OpenLayers Plugin to start OpenLayers when in QGIS (needs to be downloaded from QGIS download page and extracted first), click Web (on the top menu), then hover the cursor over OpenLayers Plugin and you can select the type of high-resolution background imagery you would like (we recommend Bing Aerial). QGIS should align on the fly if you open satellite imagery fist (as described in the previous paragraph; the satellite imagery will have a different spatial reference than the OpenLayers Plugin background imagery, but the satellite imagery will adjust [and warp] "on the fly" to line up with the background imagery. If you are not using high (enough)-speed internet when using QGIS (for the Open Layer Plugin to work efficiently), download high-resolution background imagery (Course 1C; describes how to download high-resolution imagery for USA locations). 

THIS PARAGRAPH MAY APPLY IF YOU ARE DOWNLOADING HIGH-RESOLUTION BACKGROUND IMAGERY, INSTEAD OF USING THE OPENLAYERS PLUGIN. There may be an alignment issue if using satellite imagery and downloading high-resolution imagery (as described in Course 1C); the two types of imagery will have a different spatial reference and may not align "on the fly". If there is an alignment issue in QGIS after opening an image in QGIS, you may need to set the geographic coordinate system and projection to match that of the image. If this is new to you, simply apply the following steps: right-click on the image name in the Layers Panel and click Properties. In the Layers Properties window that appears, scroll down in the Properties window (near the bottom of the Layers Properties window) until you find Layer Spatial Reference System (ignore the "=" and "+" signs) to view the UTM zone and geographic coordinate system, should be WGS 84 , and record the information somewhere before you close the window. For example, you may record UTM zone 14 WGS 84. Close the Layer Properties window and click the globe icon near the bottom right of the main QGIS window. Make sure that the map projection and coordinate system is the same as that of the image. If not, click the box near the top to Enable 'on the fly' CRS transformation (OTF). You want these to be the same. You are now good to go. It is possible that two of the images you downloaded have different projections, in this case the pixels will have different spatial extents - that is okay, you can work with them in unison.

THE DEFAULT SYMBOLOGY FOR THE IMAGERY YOU OPEN AND VIEW WILL BE GRAYSCALE (BLACK TO WHITE, LOWEST TO HIGHEST) - THIS MAY SEEM UNSATISFACTORY AT THE FIELD-SCALE. FOR EXAMPLE, BAND 4 (RED) PIXELS MAY SEEM COMPLETELY BLACK THROUGHOUT THE FIELD, EVEN THOUGH THERE IS VARIABILITY. THIS OCCURS BECAUSE BAND 4 REFLECTANCE IS VERY LOW RELATIVE TO THE ENTIRE SCENE. PROPERLY SYMBOLIZING IMAGERY IN QGIS SO YOU CAN VIEW VARIABILITY, IS EXPLAINED IN STEP 4.

ALSO, KEEP IN MIND THAT SENTINEL-2 PIXELS REPRESENT TOP OF ATMOSPHERE REFLECTANCE, NOT SURFACE REFLECTANCE. THOUGH THERE ARE MORE APPLICATIONS WHEN IMAGERY IS CONVERTED TO SURFACE REFLECTANCE, THERE ARE STILL USEFUL APPLICATIONS FOR IMAGERY NOT CONVERTED. SENTINEL-2 CAN BE CONVERTED TO SURFACE REFLECTANCE USING THE FREE TUTORIALS ON THIS WEBSITE.

 

4) ANALYZING CROPS (AND VEGETATION, IN GENERAL) IN QGIS

When applying imagery to crops, it is important to keep in mind that canopy reflectance is lower than single leaf reflectance by a substantial percent due to many factors (such as leaf orientation and shadows), and that visible and infrared radiation can both provide important information about crop condition (Knipling, 1970; PDF downloaded from this website; opens in new tab).

The red, red edge (wavelengths between red and NIR where vegetation reflectance rises abruptly), near infrared (NIR), and short-wave infrared (SWIR) bands are the most useful for vegetation; although, green is also used for many applications (blue is used the least). NIR and red are typically used for crop condition and yield forecasting (higher NIR surface reflectance corresponds to better crop condition, while lower red surface reflectance corresponds to better crop condition). SWIR is used for plant water content quantification applications (lower SWIR reflectance corresponds to more water in the plant). See the About Crop Imaging page for a general background.

An important aspect to understand prior to discussing crop imagery, is that for soil, visible and infrared bands all largely positively correlate with each other (see the next graphic for examples). You want to make sure that these same relationship do not exist for the crop imagery you have selected (which can happen if you have select imagery early in the season; explained more later). While the level of surface reflectance for soil varies among bands, relatively higher and lower areas of soil surface reflectance for different bands are largely the same areas. Soil surface reflectance increases from blue to green to red to NIR to SWIR (though it can start to decrease at mid to larger SWIR bands).

The graphics below show Sentinel-2B imagery for soil, and shows the positive correlation between bands (brighter is higher relative reflectance). Access the Landsat 8 & Sentinel-2 Bands page for band and resolution details (resolution varies for Sentinel-2 bands, as can be seen below).

From left to right, below, is: Band 2 (Blue; 10-meter resolution; reflectance range = .0105), Band 3 (Green; 10-meter resolution; reflectance range = .0106), Band 4 (Red; 10-meter resolution; reflectance range = .0168). The area is about 10 acres (there is no field size limit for imagery) - individual pixels (either 10 or 20 meter resolution) can be seen at this scale.

From left to right, below, is: Band 5 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0148), Band 6 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0173), Band 7 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0187).

From left to right, below, is: Band 8 (Near Infrared; 10-meter resolution; reflectance range = .0192), Band 8a (Near Infrared Narrow; 20-meter resolution; reflectance range = .0209), Band 11 (Shortwave Infrared; 20-meter resolution; reflectance range = .0178), Band 12 (Shortwave Infrared; 20-meter resolution; reflectance range = .0134).

 

Crops/Vegetation Assessment

The next section shows how to locate areas of higher and lower crop condition and potential yield. Red, red edge (wavelengths between red and NIR where vegetation reflectance rises abruptly), NIR, and SWIR are the most common bands used for vegetation applications. The red band has more uses earlier in the season, after the canopy has closed enough but prior to a very dense canopy. Satellite pixels can extend to areas between plants where soil is visible (unless the canopy is closed) and, as a result, average in soil reflectance to pixel reflectance value - the canopy should be closed enough to properly detect plant condition, which means excluding enough soil reflectance (canopy needs to be largely closed). However, when the canopy becomes too dense, red reflectance can saturate low, meaning that values will be similarly low throughout the field even though distinct differences in crop (or vegetation, in general) condition exist. Unlike the red band (or any visible band), NIR can detect differences in a dense canopy. Therefore, if including the red band to detect crop condition within a field, it is vital that the canopy is closed enough whereby soil is not too visible, but not so dense that red reflectance has saturated. (Soil-adjusted vegetation indices have been developed to account for soil visibility, but it is better to simply avoid imagery where substantial soil is visible.) If red reflectance has saturated, just use NIR for a green field (as shown in the second visible band example below). Corn becomes challenging to assess with imagery after tasseling, due to the obscuring effect tassels have; but imagery can still be usefully applied to a corn field in the R-stages in certain ways. Unlike the red band, NIR can be used solely to detect crop condition variability and predict yield variability for green fields (such as soybeans) with a closed canopy.

For a largely green field with a predominantly closed canopy, red and NIR are negatively correlated (unless red has saturated too much), whereby healthier, more robust vegetation has lower red surface reflectance and higher NIR surface reflectance. A way to check for this is to open red and NIR images in GIS over a field of interest (images are by default classified with a grayscale in QGIS, where lower values are darker), and look to see see if areas of lower red values correspond to areas of higher NIR values (negatively correlate). This reverse from a positive red-NIR correlation to a negative correlation, reveals that the imagery is revealing crop condition (as opposed to soil patterns).

As mentioned at the end of Step 3, it can be difficult to see differences in pixel amounts when viewing red (and other visible) band imagery in GIS because values are so low (red surface reflectance in a healthy green soybean canopy is about 3%, blue is about 2%, green is about 5%, while NIR can be about 60%). In order to view variability in pixel values in QGIS, you can easily symbolize a raster so the grayscale classification is based on custom values that correspond to the upper and lower values of the field extent (the area will be stretched black to white). To do this in QGIS, first click the Identify Features tool (the icon on the top menu with a cursive "i" with a cursor on it) and click on various areas of the imagery to retrieve the lower and higher pixel values, then right-click on the image layer and select Properties to view the Layers Properties window, then select Style and enter those values as Min Max under Color Gradient and click OK. Modify the values until satisfied. YOU CAN BYPASS THIS CLASSIFICATION METHOD, BY EXTRACTING IMAGERY TO THE EXTENT OF A SHAPEFILE WITH THE CLIPPER TOOL (CLICK RASTER > EXTRACTION > CLIPPER).

The next graphic (same field as above, but later in season) shows Sentinel-2B imagery for a crop field with a largely closed canopy - it shows the negative correlation between red and NIR (unlike the positive correlation that exists for soil). It also shows the negative correlation between NIR and SWIR, as well as the positive correlation between red and SWIR. The symbology is stretched from lowest to highest value. Soil reflectance commonly ends up correlating to yield, where darker soil corresponds to higher crop condition and yield. The imagery below correctly shows crop condition; this can be determined because red and NIR reflectance now negatively correlate, unlike the soil imagery above where they positively correlate (the more robust the vegetation, the lower the red and higher the NIR reflectance is). SWIR reflectance correlates to vegetation water content whereby lower reflectance corresponds to more water in the plant. Typically more robust areas of vegetation have more water and, therefore, lower SWIR reflectance; so for a crop field, SWIR and visible bands positively correlate (unless visible bands have saturated too much [as shown in the last example], while NIR and SWIR negatively correlate (as is the case in the images below). 

* IF YOU DOWNLOADED SENTINEL-2 IMAGERY FOR A FIELD (AS SUGGESTED), OPEN BAND 4, 8, AND 11, AND SYMBOLIZED THE IMAGES IN QGIS AS PREVIOUSLY DESCRIBED, THEN SEE IF THEY MEET THE STANDARDS SHOWN IN THE NEXT GRAPHIC. *

The graphics below show Sentinel-2B imagery for vegetation for the same field as the soil imagery above, but for later in the season when the canopy is largely closed. It shows a negative correlation between red and NIR (as well as other bands).

* BANDS 6, 7, 8, AND 8A (RED EDGE AND NIR BANDS) ARE SHOWING CROP CONDITION (AND LIKELY FUTURE YIELD PATTERNS) WELL, WHERE THE LIGHTER THE SHADE THE BETTER THE CONDITION (POSITIVE CORRELATION WITH PIXEL VALUES). BANDS 4, 5, 11, AND 12 ( RED, RED EDGE, AND SWIR BANDS) ARE ALSO SHOWING CROP CONDITION (AND LIKELY FUTURE YIELD PATTERNS) WELL, WHERE THE DARKER THE SHADE THE BETTER THE CONDITION (NEGATIVE CORRELATION WITH PIXEL VALUES BECAUSE THE LOWER THE VALUES THE MORE WATER IN THE PLANT). *

From left to right, below, is: Band 2 (Blue; 10-meter resolution; reflectance range = .0071), Band 3 (Green; 10-meter resolution; reflectance range = .0089), Band 4 (Red; 10-meter resolution; reflectance range = .0155). The area is about 10 acres (there is no field size limit for imagery) - individual pixels (either 10 or 20 meter resolution) can be seen at this scale.

From left to right, below, is: Band 5 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0114), Band 6 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0221), Band 7 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0333).

From left to right, below, is: Band 8 (Near Infrared; 10-meter resolution; reflectance range = .0359), Band 8a (Near Infrared Narrow; 20-meter resolution; reflectance range = .0321), Band 11 (Shortwave Infrared; 20-meter resolution; reflectance range = .0172), Band 12 (Shortwave Infrared; 20-meter resolution; reflectance range = .0229).

 

You can locate relatively higher and lower areas of crop condition and potential yield (or vegetation condition, in general, as long as the canopy is predominantly green and closed) without converting imagery to surface reflectance just by simply viewing higher and lower areas of pixels values (whatever those relative values may represent). You can identify locations on a field by adding a shapefile point in QGIS to mark locations and view them over the high resolution imagery, then locate that area by site in the field or export the shapefile and enter it into a GPS-based device. It becomes much more difficult to produce a map in GIS based on imagery. You may be able to find guidance for this online. Products available in the Store here, can be based on one or many images and can be custom made based on any request.

If you want to calculate indices, such as NDVI or WDRI (which is our recommended index; described below), you need to convert imagery to surface reflectance. See the pages in the Landsat & Sentinel-2 Surface Reflectance Guides drop-down menu above to convert to surface reflectance (starting with the About page). For vegetation assessment purposes, if NIR and red are negatively correlating well (like above) and if the reflectance variability is reasonable, we recommend using the Wide Dynamic Range Index, where α = .1 (Gitelson, 2004; PDF downloaded from this website; opens in new tab). (If NIR and red are not negatively correlating well and red values have saturated [as shown in the graphic below], do not use an index - use NIR solely.) For this index, simply multiply NIR surface reflectance by .1 so that NIR is more equal with red surface reflectance (NIR surface reflectance is more than 10 times greater than red reflectance for a green closed canopy). The index can be written as: ([NIR x .1] - red) / ([NIR x .1] + red). Converting to surface reflectance and calculating indices will produce amounts that can be correlated to important values, such as yield amount and leaf area index. It is important to note that indices will not produce map that appear meaningfully different (a map calculated with TOA reflectance or Landsat digital numbers) will appear very similar to one produced with surface reflectance. Also, map produced based on the Simple Ratio (SR) (NIR/Red) will look very similar to a map produced with other indices that use NIR and red, such as NDVI ([NIR - Red] / [NIR + Red]). So if you do not want to covert to surface reflectance, simply divide NIR by Red to produce a map that shows areas of relatively higher and lower crop condition.  

Below is a Simple Ratio map (about 10 acres; pixels are 10 x 10 meters) for the previous images developed with TOA reflectance (the downloaded values). The map looks very similar to a SR map developed from surface reflectance, and looks very similar to a NDVI or WDRI map developed with TOA or surface reflectance.

 

If you want to assess plant water content, use the Normalized Difference Water Index (SWIR - NIR / SWIR + NIR). An advantage to using this index is that NIR TOA reflectance is very similar to NIR surface reflectance, and SWIR TOA can be used as surface reflectance (for all practical purposes, there is no scatter). Sentinel-2 resolution increases to 20 meters for SWIR bands. For this index, we recommend using band 8a (NIR narrow) because it also has 20 meter resolution with band 11 (SWIR). However, bands 8 and 12 can also be used in combination. An NDVI map is shown below developed with TOA; brighter values correspond to higher plant water content. (Looks identical to map based on NIR/SWIR).

NDWI Map for Previous Images

(shows areas of higher and lower plant water content; about 10 acres, pixels are 20 x 20 meters)

 

Imagery Later in Season when Visible Bands have Begun to Saturate are Shown Below (notice the decreased reflectance range).

The graphics below show Sentinel-2B imagery for vegetation for the same field as the vegetation and soil imagery above, but for later in the season when the canopy is full. It does not show the same level of negative correlation between visible and NIR bands (blue and green correlation has diminished the most; the red has diminished to a noticeable, but lesser extent) or reflectance range (max - min). THE REFLECTANCE RANGE IS 3 TO 4 TIMES LARGER FOR NIR BANDS THAN VISIBLE BANDS. Visible band reflectance has saturated more than the previous image, which is evident by both the lower reflectance range and more incoherent patterns - the red band still has a reasonable, but lower, visual correlation to NIR (the red band can saturate as the the green and blue have shown below, and can have a similarly low negative correlation with NIR reflectance). Be cautious when including visible bands in an analysis when these types of relationships appear. In this particular situation, do not include blue or green - if you are going to include red in an index, also use Band 6, 7, 8, and 8a, solely, for analysis. Also, because SWIR bands keep a high correlation and variability, you can apply those bands along with NDWI to map crop water content patterns (the lower the SWIR values, the more water in crops; the higher the NDWI values, the more water in the crops).

* BANDS 6, 7, 8, AND 8A (RED EDGE AND NIR BANDS) ARE SHOWING CROP CONDITION (AND LIKELY FUTURE YIELD PATTERNS) WELL, WHERE THE LIGHTER THE SHADE THE BETTER THE CONDITION (POSITIVE CORRELATION WITH PIXEL VALUES). BANDS 11 AND 12 (SWIR BANDS) ARE ALSO SHOWING CROP CONDITION (AND LIKELY FUTURE YIELD PATTERNS) WELL, WHERE THE DARKER THE SHADE THE BETTER THE CONDITION (NEGATIVE CORRELATION WITH PIXEL VALUES BECAUSE THE LOWER THE VALUES THE MORE WATER IN THE PLANT). *

From left to right, below, is: Band 2 (Blue; 10-meter resolution; reflectance range = .0058), Band 3 (Green; 10-meter resolution; reflectance range = .0068), Band 4 (Red; 10-meter resolution; reflectance range = .0059). The area is about 10 acres (there is no field size limit for imagery) - individual pixels (either 10 or 20 meter resolution) can be seen at this scale.

From left to right, below, is: Band 5 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0060), Band 6 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0162), Band 7 (Vegetation Red Edge; 20-meter resolution; reflectance range = .0193).

From left to right, below, is: Band 8 (Near Infrared; 10-meter resolution; reflectance range = .0203), Band 8a (Near Infrared Narrow; 20-meter resolution; reflectance range = .0214), Band 11 (Shortwave Infrared; 20-meter resolution; reflectance range = .0107), Band 12 (Shortwave Infrared; 20-meter resolution; reflectance range = .0110).

 

Start with these steps, and you will independently start to learn the utility of satellite imagery.

 

Reference

Gitelson, A.A. 2004. Wide dynamic range vegetation index for remote quantification of biophysical characteristics of vegetation. Journal of Plant Physiology 161; pp. 165–173.

Knipling, E.B. 1970. Physical and physiological basis for the reflectance of visible and near-infrared radiation from vegetation. Remote Sensing of Environment 1; pp. 155-159.