Use of satellite data in agriculture has a much richer potential than just creating task maps. The satellites help farmers to make better, faster and deeper assessments of their crops’ status. And at the same time, it can help them to document their activities and events through monitoring.
The European Copernicus programme creates incredible opportunities for analysing the Earth’s surface. The original name was Global Monitoring for Environment and Security, which hints at the kind of sensors that are on board and gives away its principal mission. Despite that, the Copernicus programme is of great value to agriculture.
Around 30 years ago, my professor handed me four big magnetic data tapes containing a single Landsat scene to analyse. I spent three months on ‘preprocessing’ before I could even look at the satellite photo. Today, within 24 hours after acquisition, imagery is analysis-ready and available to the whole world.
What seems to be constant over all these years, however, are the high expectations of what can be done with satellite images in agriculture. This goes hand-in-hand with the poor ability of the satellite services industry to manage these expectations.
Text continues underneath image
Most (or all) services for farmers focus on making application maps or task maps: digital location-based instructions for computer-controlled machinery. This is one of the big promises of precision farming: you only give a plant the application it needs, so differences in crops require different treatments – in this case in application intensity. This seems to have become the holy grail for satellite data (and also for drone imagery, soil mapping or previous-year yield data).
The satellite reveals local differences in growing conditions. Its birds-eye view on fields and crops can indeed help a great deal in identifying zones of crop stress and translate that into an application map to relief that stress. And this task map is also good news as we envisage future field operations to be done by robots, or in a more immediate future by contractors and drivers who lack the knowledge about a particular field.
The advantage of satellite data is that it can provide recent images of the crop at affordable prices – or even for free. The big question is, of course: can the stressed state of the crop be attributed to a single cause throughout the field?
There are three ways of working with satellite data: mapping, measuring and monitoring. For mapping, we look for patterns and we can identify and classify objects. In agriculture, crop classification was the champion mapping application for many years, and in particular it was used by scientists and administrations. Not for farmers – they already know which crops they have on their land.
The second way is using the satellite as a measuring device. Scientists have built complex models to determine the properties of the surface. Pragmatists have translated these to indicators lsuhc as the well-known Normalised Differential Vegetation Index (NDVI), which shows how accumulated biomass absorbs and reflects different parts of the spectrum. We can map an NDVI image out again and look for patterns and classifications.
The third way is monitoring. Mapped values are systematically compared over time and used to demarcate changes, evolutions, discrepancies, etc. In fact, when monitoring we look at the Earth’ surface as a dynamic system and we can observe activity and behaviour. In agriculture: sowing and harvesting (even grassland mowing) and irrigation, etc. Or, if you look on a larger time scale, you can identify changes in land use – which are of major interest in terms of climate change.
Again, there are many things that you don’t need to tell a farmer, who is the person doing – or ordering – these events in the first place. But this is very useful for other actors in the agricultural domain, including administrators, insurance companies and food processors. For farmers, other types of monitoring are more useful, for example detecting anomalies and seeing how their crops are growing. And, of course, we can map out these events again to close the ‘data’ loop: mapping, measuring and monitoring.
Text continues underneath image
The frequent overpass of satellites brings exciting new opportunities. The Copernicus programme has several different satellite types, which were designed for a different monitoring purposes. For farmers the most interesting ones are the Sentinel-1, which measures microwave reflections, and Sentinel-2, which measures reflected solar radiation. Both have a designed six-day revisit, thanks to two identical satellites working together.
Besides Copernicus, other satellite constellations can also be used for monitoring. The most sensational one is Planet, a commercial alternative to Copernicus, which operates hundreds of very small satellites, allowing almost daily monitoring of the whole Earth. And there are more to come.
When we see stressed crops, can we identify the cause? The NDVI is a measure for accumulated biomass, so if we see in-field differences these must relate to the environmental conditions. Fifty years of Earth Observation reveals a wealth of algorithms and relations between NDVI and different stresses.
Often the relationship between an observed stress and NDVI is quite good. But, without additional knowledge, it’s not easy to tell if an observed lower NDVI (or any other vegetation index) is due to water stress, nutrient stress, unfavourable soil conditions, competing weeds, pests or diseases – or it might even be caused by a malfunctioning sowing machine.
The most important thing to do with satellite data is to verify in the field what causes the observed vegetation stresses. Combining data from different sources, for instance a soil map, a height map, weather data and agronomic knowledge, will certainly help farmers to narrow down the list.
Satellite-assisted crop scouting is one of the most direct and useful applications for remote sensing, but it’s not often used. And, by identifying the cause, farmers get better actionable insights about their crop and field – as well as better insights into how satellite imagery is helpful. And, of course, if the cause is e.g. nitrogen shortage, the satellite data provides a good source for drawing up a variable application map.
Text continues underneath image
When presented in the right way, data can help farmers to organise their field visits, find spots where growth is suboptimal, and identify causes. And the data can help farmers to make better choices. But farmers are still releuctant to let the data make decisions. Even systems based on artificial intelligence or deep learning will not help until we train those systems with a range of observed causes.
To summarise, the use of satellite data in agriculture has a much richer potential than just creating task maps. The satellites help farmers to make better, faster and deeper assessments of their crops’ status. And at the same time, it can help them to document their activities and events through monitoring. If every farmer were to use satellite data in this way, it could even provide proof to the food processing industry or to administrations that certain activities took place – or didn’t.
Tamme van der Wal
Tamme van der Wal is a part-time researcher at Wageningen Environmental Research, working on data science solutions for agricultural performance. He is partner at AeroVision and founder and co-owner of Agri-Dataservices BioScope.