Crop Monitoring using Satellite Imagery — Part I

An Introduction to the approach.

Satellite imagery

We, a group of Computer Science undergraduates of the University of Colombo School of Computing were able to perform a study on how to perform Crop Monitoring using Satellite Imagery. Due to the technological advancements around the globe, wide usage of technology is being used to assist & improve the day to day as well as industrial activities.

The agricultural sector plays a huge role in any country. As the growing generation it is our duty to overcome the limitations and complexities in activities of such importance. Crop Monitoring is such an activity that is very much required for the farmers and agricultural product industries to maintain a stable position in the agricultural sectors.

Satellite Image Analysis

Why Satellite Imagery?

At present, there are people who use drones to spray the fertilizers over the crops and drone images to monitor their growth. Images generated by these drones can be decided by us. Because the drones can be controlled when and where to fly and capture. These images can be clear and very useful compared to Satellite images. Because we cannot control Satellites but accept the images they provide with the respective resolution.

With the above discussion, you may want to choose a drone or any other hand-controlled device that generates images for you. But, this can cause a great cost when performing a research on how to use these images for crop monitoring. Therefore, it was always better for us to choose a Satellite image provider to collect the images as the test data.

Another advantage is that these Satellites provide images with different sets of band frequencies. Most of these frequencies that can be specifically used for our study are being described as we go on.

Image from Landsat 8 Satellite

What is a VI?

VI stands for Vegetation Index and, it is calculated by considering the different property variations that influence the crop growth. There are different VIs that can be used for different purposes in Crop Monitoring. These VIs are calculated using different band values that influence crop health and growth.

Bands commonly refer to the different layers of an image. When we consider the satellite imagery and data, these bands are also called ‘channels’. These band values denote the relevant channel output captured by the satellite. In Satellite images, the bands that represent the reflection of different light frequencies of different wavelengths (such as Visible red light, Infrared light, Visible blue light, etc.) are used to calculate and generate the Vegetation Indices (VIs).

Process of capturing Band data

There are many VIs generated and improved throughout the years. We have listed down some of the most useful ones for the study and, below mentioned are the formulas, features and, applications of them.

NDVI

Vegetation health is calculated using the concept of the leaves absorbing a high amount of Visible red light and reflecting a high amount of Near Infrared light (NRI). This occurs during the photosynthesis performed by the plants. Therefore, this can help to identify the areas with proper vegetation. Basically, these values range from -1 to +1 and have different value variations denoting different meanings.

Process during photosynthesis

Negative values — Water surface, Man Made rocks, Basically dead plants or Inanimate objects.

0.1– 0.2 — Bare soil

Above 0.5 — Healthy Vegetation

0.1– 0.5 — Sparse Vegetation (Vegetation with large special distances in between)

NDVI = (NIR — Red) / (NIR + Red)

Some other value variations with more definitions are shown in the image below. And more of these values hold similar meanings.

NDVI value ranges

Dried spots can be identified with the range for ‘bare soil’ and, it can also denote the fire hazardous areas, which means that these areas may catch fire. So, these areas should be given more water.

NDVI Time series can be very useful to identify the Crop growth process. By considering and comparing the calculated NDVI values of a particular area for a particular time range in the past year and similarly in the current year, the differences in the growth can be monitored. Similarly, when we consider a particular season and the values show less vegetation than it should be, then it means that the area needs more concern.

By using different fungicides on different areas and by comparing the NDVI values over the time, the fungicides that help in proper crop growth can be identified.

EVI (Enhanced Vegetation Index)

As the name suggests, EVI is considered to give out more enhanced results. Because NDVI only considers the Red light and NIR which means, the atmospheric influences & the background signals can affect the values and produce inaccurate results. Background soil can clearly affect the NDVI values during the early stages of the crops.

Here, 0.2–0.8 value range denotes Healthy vegetation.

According to the search made, EVI is more suitable to monitor the areas with large chlorophyll (Eg: Rain Forest) but not for mountainous areas.

EVI = 2.5 * ((NIR — Red) / (NIR + (C1 * Red) — (C2 * Blue) + L))

Here the coefficients C1 & C2 can be used to correct for aerosol scattering and L adjusts the value for soil & green backgrounds. And these values can be adjusted accordingly when monitoring. But usually C1=6, C2=7.5, and L=1

SAVI

SAVI stands for Soil Adjusted Vegetation Index. It means that the NDVI values are more adjusted to overcome the background soil signals. Due to this reason, it is more likely useful to monitor the growth of Young Crops and Sparse Vegetation. This means the areas exposed to more soil background can be monitored.

SAVI = ((NIR — Red) / (NIR + Red + L)) * (1 + L)

Here for high greenery area, L=0 and values are equal to NDVI and for low green vegetation L=1.

ARVI

ARVI stands for Atmospheric Resistant Vegetation Index. This is also the NDVI values adjusted for atmospheric influence by considering the values of Blue light wave-length. As this has the ability to filter out the blue influences, Tropical Mountainous regions can be monitored considering ARVI values.

ARVI = (NIR — (2 * Red) — Blue) / (NIR + (2 * Red) — Blue)

GCI (Green Chlorophyll Index)

GCI is the value calculated by estimating the chlorophyll amount in leaves. Useful for monitoring impacts on Pesticides on plant health.

GCI = (NIR) / (Green) — 1

SIPI (Structure Insensitive Pigment Index)

This indicates the loss of Chlorophyll amount in the leaves that leads to a conclusion that it has been affected by a related plant disease.

SIPI = (NIR — Blue) / (NIR — Red)

NBR (Normalized Burn Ratio)

NBR values usually indicate the burnt areas. Mostly, they are used to detect an active wildfire and to monitor the survival plant rate. This helps to analyze the cost incurred due to the effect of fire.

SIPI = (NIR — SWIR) / (NIR + SWIR)

NDSI (Normalized Difference Snow Index)

NDSI values are used to identify the snow covered in the area without the influence of clouds. Values greater than 0.4 normally indicate that there is snow in that area.

NDSI = ((Green — SWIR) / (Green+ SWIR))

Conclusion

In this introductory article, you would have got a brief understanding on how VIs calculated using the band values from satellite images are being useful for crop monitoring. But you would have clearly seen that only some of the VIs are provided with value ranges. This may lead to a confusion on how to use these values provided by different VIs for proper crop analysis.

The next article that briefly explains the Image processing performed by us would produce the relevant answer for the confusion.

Stay tuned for the next article for more updates on the study.

References

6 Spectral Indexes To Make Your Vegetation Analysis Complete

NDSI: Normalized-Difference Snow Index For Agriculture