Data Cookbook: Agricultural Monitoring

Discover how to utilize  GRUS-1 satellite images to monitor crop health and growth

Download the satellite image dataset from our Axelglobe platform.  Need a refresher? Here’s the AxelGlobe User Manual for your reference.

1A. For this exercise, we will be using a dataset from Demo Brazil. Four tiles are selected (Cell ID: S21056450, S21056451, S21066450, and S21066451) for three dates ( Mar 10, May 11, and Aug 6). All in all, there should be 12 selected and downloaded GRUS Tiles. We will be utilizing Multispectral Product for this agricultural monitoring use case.

Learn how to merge or clip your areas of interest. Check out our Getting Started Manual for more details. 

2A. To easily manage and analyze our data, we will merge each of the four-tile dataset per date of acquisition using a GIS software. There should be three image outputs for this process: one merged imagery per date (2020/03/10, 2020/05/11, and 2020/08/06)

Note: The default band rendering of QGIS is: Red Channel – Band 1, Green Channel – Band 2, and Blue Channel – Band 3. The reason why the images are not displayed in their true color form is because the channels are not matched with their correct Band designation yet. 

Visualize your data in their True Color and False Color composites. Our Getting Started Manual has a step-by-step information on how you can achieve this.

3A. True Color Composite. Also called “Natural Color Composite”, this consists of three visual bands: Red, Green, and Blue. The image is displayed in its true color as it resembles closely what can be seen by human eyes. For GRUS-1 Multispectral product, Band 3 corresponds to Red Band, Band 2 to Green Band, and Band 1 to Blue Band.

3B. False Color Composite. Compared to True Color Composite, False Color composite imagery shows changes in area which cannot be easily seen using True Color imagery (Red, Green, & Blue) by visualizing them through different band combinations (Red Edge or Near Infrared). The False Color imagery shown below made use of Bands 5 (NIR), 3(Red), and 2(Green) in the Red, Green, and Blue channels, respectively.

Band rendering settings for True Color and False Color Composites are shown below as used in this manual.

Manage and monitor crop growth and health using vegetation indices derived from our satellite images.

4A. Vegetation Indices. Generally, Vegetation Indices (VI) are indicators of photosynthetic activity which can be linked to the growth and health of plants. High values correspond to healthy vegetation, and low values correspond to degradation. For this data cookbook, we will be using two of the most used VIs – Normalized Difference Vegetation Index (NDVI) and Normalized Difference Red Edge Index (NDRE).

4B. Raster Calculator Function. To generate NDVI and NDRE image derivatives, we need to use the Raster Calculator  tool in QGIS. To access the tool, just navigate to the Raster Menu and click on the Raster Calculator tab. 

Notice that the raster bands are automatically separated per dataset, as shown below, for direct computation.

4C. NDVI Computation. To compute for NDVI, we need to use Bands NIR (Band 5) and Red (Band 3). The formula for NDVI is: (NIR-Red)/(NIR+Red) or (Band 5 – Band 3)/(Band 5 + Band 3). Ensure that the bands you are using for each NDVI computation are from the same data period. Do this for all three datasets.

The raster formula used for the NDVI computation is shown below:

4D. NDRE Computation. To compute for NDRE, we need to use Bands NIR (Band 5) and Red Edge (Band 4). The formula for NRE is:  (NIR-Red Edge)/(NIR+Red Edge) or (Band 5 – Band 4)/(Band 5 + Band 4). Do this for all three datasets. The raster formula used for the NDRE computation is shown below:

4E. Visualization. Navigate to Symbology tool by right clicking the layer > Select Properties > Select Symbology. Set the Render Type to  Singleband Pseudocolor. For Color Ramp, you can select from the choices using the arrow button. In this visualization procedure, Spectral color ramp is used.  Click Classify, then Apply button. To close the dialogue box, click OK.

Final symbology settings and visualization result are shown below:

4F. Time Series Visualization Comparison. To compare the datasets from different time period, we visualize them in a similar min-max NDVI/NDRE value. From their NDVI values, we set -0.27 as minimum and 0.79 as maximum for each NDVI imagery.

From their NRE values, we set -0.25 as minimum and 0.53 as maximum for each NDRE imagery.

Time-series comparison for NDVI and NDRE Analysis are shown below. NDVI is the most commonly used VI to estimate crop health in the earlier growth stages of crops while NDRE is for later-stage growth. To interpret the values, the higher the value of NDVI or NDRE, the healthier or more robust the vegetation is in the area at a particular growth stage.

4G. Zonal Statistics. To monitor the farmland areas  more comprehensively, we can analyze the VI values per crop parcel. The dataset used in this step was obtained in Getting Started Manual: Data Management. By quantifying the mean VI value for each parcel through time, we can efficiently monitor and track vegetation health and growth stage. 

(a) Single Process. To start with the processing, navigate to the Processing Toolbox > Raster Analysis > Zonal Statistics. If Processing Toolbox is not shown as default, just press Ctrl + Alt + T or simply click the gear icon.

For the Raster Layer, select the images you would extract the VI values from. For Vector Layer, select the parcel which would serve as the bounding area for VI statistics computation. Zonal statistics tool has a number of selection for statistics computation.  Since we are only interested in the mean values, we only select Mean  among the options. The output column prefix represents the name of the column attribute for the computation + the selected statistics to compute. To avoid confusion, we set a similar name to the raster layer.

(b) Batch Process. If users need to process multiple images, the Batch Process tool provides a more suitable approach. From the Zonal Statistics Tab, click the Run as Batch Process button to prompt the processing window. 

Raster Layer – Under this column, click on the ellipsis on the second row and select all the datasets you want to include in the batch zonal statistics processing. Once these are selected, click OK. The selected layers should automatically appear under the same column.

Vector Layer Containing Zones – Under this column, click on the ellipsis on the second row, choose Select from Open Layers, and then select the shapefile which will be used as zones. For this example, this would be the Test_Parcel. Since the same layer will be used across all datasets, click Autofill to automatically load it for all rows. Note: If the vector layer is not yet imported in QGIS, just choose Select Files instead of Select from Open Layers.

Output Column Prefix – Under this column, indicate the name of the output attribute for each raster layer. To avoid confusion, we set a similar name to their corresponding raster data. Statistics to Calculate -  Under this column, click the ellipsis, select Mean, click OK and then proceed with the Autofill.

The final parameters setting for the Batch Process Zonal Statistics are shown below:

(c) Result. To check the values extracted, right click on the vector layer Test_Parcel, and then click Open Attribute Table. This contains all the information generated from the Zonal Statistics tool where the row represents the parcel or the region of computation and the column contains the computed statistics value from the raster layer where these were extracted from.

(d) Visualization. To view the parcel values per column attribute, right click on the vector layer Test_Parcel, select Properties, navigate to Symbology tab. Within this window, select Graduated. For the Value, choose the data you want to visualize. For this example, we choose 0310_NDVI. Select the Color Ramp, click Classify, then hit the OK button. 

Shown below are the final symbology settings and the final output.

Last updated: June 2021