Agric101: How Complicated Is Precision Agriculture?
Updated: Mar 3
Drone technology is becoming easier to afford, operate, and deploy both industrially and casually. Initially built for military purposes, we now have dozens of drone varieties available in the consumer market, ranging from a couple of thousand bucks, for what is essentially a flying toy, to several lakhs for a more competent drone. Drone engineers have had great success in acclimitzing the market to such concepts as automated flying for content creation but with the rise of precision agricultural sensors being mounted on to drones, how long will it take for this technology to be adopted by end users? How complicated is it to understand and deploy, and are its costs prohibitive from the outset?
The State of Agricultural Drone Technology
Despite the COVID-19 pandemic, the agricultural drone market is expected to grow from USD 1.2 billion in 2020 to USD 5.7 billion by 2025 indicating that adoption of this technology by larger players in developed economies is a no-brainer. At the highest end of the agricultural industry hierarchy, the use of multispectral analysis using handheld and fixed systems is an industry standard; mounting a multispectral sensor on a drone is viewed as a logical evolution of the tech. Perhaps the most critical use of this technology is to ascertain the crop count, and to map arable land.
Major drone manufactures DJI and Parrot have already brought multispectral analysis into the prosumer market. DJI modified their enourmously successful and criticall acclaimed Phantom 4 quadcopter, replacing the standard RGB camera with a 6 sensor module. In 2019, a Sri Lankan university had published a tender for the supply of multi spectral enabled agricultural drones, which gave me the impetus to delve into this type of drone use case. To me, this tender was a clear sign that Sri Lankan academics were aware of the technology, actively studying its use in Sri Lankan agriculture, & were keen to be on the cutting edge of drone technology.
(DJI Phantom 4 Multispectral - sensors top row left to right [ Red Edge sensor, Near Infrared sensor, Green Light Sensor ] sensors bottom row left to right [ Visible Light (RGB) sensor, Red Light sensor, Blue Light sensor ]
End User Adoption In Sri Lanka
Some Sri Lankan firms stand on the cutting edge, using agricultural drone technology in their process work flows. Hayleys has been using DJI's agricultural drone platform for spraying their crops with drones, & Ceylon Biscuits has recently invested in a multi spectral enabled quadcopter. This is a sign of good things to come. What I have found is that, for those companies which hire agricultural scientists or specialists, making the case for agricultural drone use is easy; there is someone present to interpret the data & make decisions based on the information.
However, to a layperson the data would make no sense whatsoever. Let me illustrate this for you with some visual aids.
Below is a section of a map of an estate. The first picture will be easy to understand, & we can deduce things from this picture; there are coconut trees on this estate, the soil is mostly clay, & there seems to be some pathway or shape to the property.
This is a section of an orthographically rectified survey map of an estate in Sri Lanka. This RGB (red-green-blue) picture shows us what the property looks like using the visible spectrum of light.
The picture below is the same section rendered in normalised difference vegetation indexed colours.
Now the picture becomes difficult to 'read'. We are unable to deduce significant information from this picture...unless we knew more about what the colours meant, & perhaps how the data set was processed.
Using the same data set capture by a multispectral drone, below is the same section of estate but rendered in visual atmospheric resistance index colours.
If you look closely, you'll find some subtle differences between this picture & the previous one, but what do these differences tell us? What information is communicated by this data set?
It's clear that these data sets aren't immediately legible to us lay folk. But, to those agriculturalists who know their VARI from their SAVI, interpreting these multispectral data sets would yield valuable insight into how crops are growing on site.
Vegetation Index (VI) - this is a spectral calculation of two or more bands of light that highlights vegetative properties.
Normalised Difference Vegetation Index (NDVI) - this algorithm is designed to detect differences in green canopy area, emphasising the green colour of a healthy plant. This is a reliable & popular VI to deploy.
Enhanced Normalised Difference Vegetation Index (ENDVI) - this algorithm better isolates plant health indicators as it takes into account the plants absorption rate of blue light & its reflectance of green & near infrared light.
Visual Atmospheric Resistance Index (VARI) - This allows for estimations of how much vegetation is in a particular area. Initially designed for satellite imagery, to differentiate between man made structures, soil, & vegetation
Soil Adjusted Vegetation Index (SAVI) - This minimises the brightness of the soil beneath a canopy & emphasises the vegetation data set. It'll allow for more accurate readings of just the vegetation, excluding soil.
This is just the tip of the iceberg, but If you'd like to know more about agricultural data sets, stay tuned for my follow up article for this that deep dives into the world of multispectral agriculture.
Sri Lankans default to this when evaluating any new Technology; how much is it going to cost?
Drone technology has been getting cheaper, with all things accounted for, consumers can buy a new drone each year, which implements ever increasingly complex functionality without breaking the bank. This holds true for enterprise drones too. A DJI system may set you back 5 to 7 thousand USD & a Parrot system might be around 3 to 5 thousand USD, but I'm fairly confident that the cost of these out-of-the-box agricultural solutions will only decrease over time.
Cheaper alternatives do exist. Mounting multispectral sensors with well designed payload systems on to drones like the Phantom 4 Pro, would shave thousands of dollars of the initial capital cost of deploying precision agriculture on site.
It's also important to understand that the hardware & software are two separate entities to be costed in. Whereas DJI does have a multispectral analysis tool, it is quite expensive. Pix4D has a much cheaper option with much more functionality under the hood. I'll be comparing Pix4D Fields & DJI Terra soon so sign up for blog updates!
To go to the cheapest end of the spectrum, finding a service provider who can conduct the flight & furnish you with the data seems to be the most cost effective way to test the waters with both the work flow & the output data set.
The bottom line remains, if you are unable to interpret the data in a meaningful way that will lead to tangible changes on site that conclude in a higher yield or better overall estate management, this technology isn't going to suit you.
If however you are able to interpret the data & make decisions based on its conclusions, precision agriculture is going to be pretty hard to stay out of. As the costs of this tech come down each year, you'll either have to adopt with the herd or adopt via services from external operators.