For the past 1000 years or so, farming has advanced by thinking bigger: fields have got larger, farm machinery has become more powerful and crop inputs have proliferated. But this trend may finally be reversing, as farmers start once again to think small.
This shift is being driven by the rise of precision agriculture, which involves planting the right crop or applying the right treatment in the right place at the right time. Rather than applying fertiliser, pesticides or water to an entire field or crop at regular intervals, farmers only apply them to sections of the field or crop that will benefit from treatment. This saves money and also helps to protect the environment as there is less excess fertiliser or pesticide to run off into streams and rivers. And rather than planting the same crop variety over the whole field, farmers can plant different crop varieties or even different crops in different areas, taking advantage of variations in soil conditions.
All this will require a change of mindset. Instead of treating a field and the crop growing in it as a large, uniform entity, farmers will need to treat it as a collection of much smaller sections, each with its own requirements for irrigation, pesticides and fertiliser, and each suited for growing particular crop varieties. Ultimately, this approach could see farmers giving each specific crop plant individual attention.
Studies have shown that thinking small can still raise yields, with a seven-year study on precision agriculture in Brazil reporting a 14% rise in maize yields and a 10% rise in soya bean yields. As such, farmers are increasingly adopting at least some aspects of precision agriculture, especially farmers in the US, Canada, the UK, Germany, Japan and Australia, and it is now estimated to be involved in over 10% of global food production.
Several large agricultural companies are betting on this trend continuing, after making major moves into precision agriculture technologies. In 2013, agricultural machinery company John Deere signed agreements with BASF, Dow AgroSciences and DuPont Pioneer to develop integrated precision farming solutions, with the first solutions from the agreement with BASF due to appear in 2015. Meanwhile, since 2012, Monsanto has acquired both Precision Planting, which develops novel planting technologies, and The Climate Corporation, which has developed a sophisticated platform for analysing agricultural data.
‘One can imagine there would be a future where every aspect of land management and farming operations is informed through real-time data collection, analytics and decision support tools,’ says John Hamer, investment director at Monsanto Growth Ventures. ‘All of this could roll up under a modelling scenario that optimises farm inputs and outputs and improves land management value over the long-term.’
In truth, farmers have always known that their fields are not uniform. For a start, hardly any fields are completely flat, causing different regions to have different amounts of moisture, as rain runs off slopes and collects in dips. The composition of the soil will also vary across the field, in terms of whether it’s sandy or clay-like and the concentration of organic matter and nutrients such as phosphorus. As a consequence, crop plants will always tend to grow better in some areas of a field than others.
Precision agriculture is essentially a way for farmers to put this knowledge onto a firmer footing, by analysing the variation in their fields and crops using a range of different sensing techniques. They can then combine this data with other forms of data, such as weather forecasts, to help determine what and where they should plant; when and where they should apply inputs; and even when they should harvest the crops.
Although precision agriculture has been around as a concept since the 1980s, only in the last few years has technology become advanced enough to realise it properly. Take satellite remote sensing, in which spectral data in images of the Earth’s surface taken by orbiting satellites reveal a range of information about the crops and soil. This takes advantage of the fact that different compounds and materials absorb electromagnetic radiation at different frequencies, with biological materials tending to absorb at visible and infrared frequencies. So chlorophyll absorbs visible light at red wavelengths; other plant pigments such as carotenoids tend to absorb green wavelengths; and soil components tend to absorb near-infrared (NIR) wavelengths.
Analysing light reflected from the surface of the Earth and captured by satellites can identify whether or not any wavelengths are missing due to absorption by material on the surface, thereby revealing information about it. So the presence of visible light at red wavelengths in an image implies a low concentration of chlorophyll in the crop plants, because these wavelengths haven’t been absorbed, which in turn suggests poor growth and a need for fertiliser.
On top of this, the total amount of NIR light reflected from a field provides a measure of the amount of biomass, while the amount of reflected infrared light indicates the temperature of the plants and thus whether they are receiving enough water. So by using all this spectral data, farmers can potentially determine a great deal of information about their fields and crops.
Satellites have been used for remote sensing in agriculture since the 1970s, when scientists used the spectral data collected by the satellite Landsat 1 to distinguish between fields growing maize and soya beans. It wasn’t until the early 1990s, though, that they began using satellites for precision agriculture applications, such as estimating spatial patterns in soil organic matter.
The agricultural information generated by this early remote sensing was not particularly detailed, both because the satellites were only able to monitor a few wide bands of wavelengths and because their spatial resolution was fairly low. Landsat 1 could only collect spectra from four bands – green, red and two infrared bands – at a spatial resolution of 80m, meaning that one pixel in the resultant image covered a fairly large area of a field.
This low spatial resolution means that any absorption will come from many different crop plants, each potentially with different requirements for inputs, and also from weeds and exposed soil, making the spectral data difficult to interpret. Even those satellites used for the first precision agriculture applications, such as Landsat 5 launched in 1984, were only able to collect spectra from seven bands at a spatial resolution of 30m.
Since then, both the spatial and spectral resolution have improved dramatically, although not usually in the same satellite. So the WorldView-2 satellite, launched in 2009, has a resolution of just 50cm while collecting spectra from eight bands. This kind of spatial resolution is detailed enough for each pixel to contain a single plant, reducing interference from other plants or the soil.
In contrast, the EO-1 satellite, launched in 2000, uses a technique known as hyperspectral imaging to collect data at wavelengths from 400nm to 2500nm at 10nm intervals, meaning 210 separate bands, but only at a spatial resolution of 30m. Nevertheless, the ability to monitor such fine spectral bands also means less interference, because the specific wavelengths absorbed by different materials can be distinguished from each other. For example, a red band centred at 697nm can indicate crop leaf area, while a NIR band centred at 970nm can indicate crop moisture levels.
The detailed information that can now be produced by modern satellites is one of the main drivers behind the growth of precision agriculture over the past few years. Companies that own satellites, such as Airbus Defence and Space, now offer to supply farmers with a range of real-time information on their fields, including soil composition, crop distribution and moisture levels, based on the spectral data collected by their satellites.
Satellite remote sensing does still have several inherent limitations, prime among them being an inability to see through clouds. But satellites are not the only option for this kind of remote sensing, because it can also be performed by planes. The advantage of using planes is that they can be fitted with the latest imaging technology, such as hyperspectral imaging, allowing detailed spectral data to be collected at finer spatial resolutions than can be obtained with satellites.
The disadvantage of planes is that paying for one to fly over your fields repeatedly can be fairly costly, but a cheaper option could soon become available in the form of unmanned aerial vehicles (UAVs). Such UAVs are easier and cheaper to operate than a conventional plane, and can also fly slower and lower over a field, obtaining higher resolution data. Already, several companies in the US, such as Aerofarming, use UAVs to conduct remote sensing of fields, while studies conducted by researchers at Kansas State University have shown that a single UAV can collect spectral data from a 640 acre field in just 20 minutes.
Such remote sensing can even be conducted on the ground, where it is known as proximal remote sensing. Imaging sensors can now be mounted on farming machinery such as tractors and sprayers, while portable versions can simply be carried through a field by hand. These ground-based sensors are able to produce the same kind of information on moisture and chlorophyll concentrations as satellite and aerial remote sensing but in much finer detail. They are also able to produce information about plants and soil that can only be obtained at close quarters, such as identifying plants suffering from a fungal infection.
Various other types of data can also be obtained about fields at ground level. Some of these data are automatically produced by the latest farm machinery, such as chemical application rates and crop yields during harvesting. Other sensing instruments can be towed along behind the machinery, such as instruments that measure the electrical conductivity of soil. Conductivity measurements are primarily used to determine variations in soil texture, as clay particles conduct more electricity than silt or sand. But they can also reveal information about a range of other factors, including organic matter content, water holding capacity and even compaction.
So farmers now have access to a wide range of detailed data from various different sources, potentially allowing them to monitor their crops one plant at a time. In order to do this, however, they need to be able to bring all these various sources of data together, as well as combining them with external sources of information such as weather forecasts and optimum growing conditions for specific crop varieties. They then need an easy and convenient way to display, analyse and interpret the data, transforming it into useful information about what should be planted where, and how and when it should be treated.
This is the next frontier in precision agriculture, and the first such data management and processing systems have recently started to appear. One such system is being developed by agricultural engineers at Purdue University in the US, who are developing apps to analyse and display data on tablet computers and smartphones.
‘We set out to create a collection of free, easy to learn and use, task-specific, mobile apps that could run on the devices the farmers and operators already had in their pockets,’ explains former team member Jonathan Welte, who now works for the 360 Yield Center, a US developer of soil and crop analysis systems. ‘We have the ability to collect data from underground, from the machinery and crops, from just above the ground with UAVs and from satellite images. Farm owners and managers don’t want to spend their time sitting in front of their computer using expensive and complicated software trying to make sense of all their data. They need something that quickly educates their decision in the location where they’re likely to make that decision.’
In addition, several commercial data management systems have also recently been launched. In 2013, Du Pont Pioneer released Field360 Select, a web-based subscription service that combines field data with real-time agronomic and weather information. Earlier in 2014, Monsanto introduced FieldScripts, which uses field data to determine what combination of crop hybrids and agronomic practices will produce the highest yields.
Even with access to all this information on individual crop plants, however, farmers still need to plant seeds and apply inputs using manually-operated machinery designed for large fields, meaning they only have so much control over where the seeds are planted and the inputs applied.
Here too, though, they may need to start thinking small, as several companies and research groups are looking to replace large machines with lots of small, GPS-guided robots that can plant seeds in specific places and treat individual plants. This could eventually lead to highly automated farming operations, in which data management systems use data from different sensors to highlight exactly where inputs need to be applied and then communicate this information to autonomous robots that go off to apply them.
Subsistence farmers were able to treat each crop plant individually because their growing areas were so small; modern technology will allow tomorrow’s farmers to do the same thing on a much larger scale.
Jon Evans is a freelance writer based in West Sussex, UK