Dr Toby Waine, a lecturer in Applied Remote Sensing in the Soil and Agrifood Institute at Cranfield University, and Emma Garfield, R&D agronomist at G’s Growers Ltd in Cambridgeshire, detail a project that aims to introduce a real-time crop monitoring and management system across G’s land bank of more than 11,000 hectares
Satellite based monitoring systems have been widely used since the 1970s for the assessment of the use and management of land, but they have tended to provide only a general picture of crop production. As a result, the role of ‘remote sensing‘ (the principle of obtaining information from a distance without needing to tread the fields) in agriculture for individual farmers has been limited, and subject to scepticism from the agrifood industry.
Meanwhile, there have been dramatic improvements in digital image sensor quality and the availability of satellite images means that we now have access to near-daily earth observation capability at much higher resolution than before. At the same time, manned survey aircraft and drones (or unmanned aerial vehicles, UAVs) are offering another platform choices for collecting very high-resolution image data of crops from the air.
Low-cost IT now has the capacity to capture and crunch through vast amounts of visual data and provide a detailed and useful analysis for growers.
The Soil and Agrifood Institute at Cranfield University has been developing innovative digital classification approaches based on a combination of different types of remote sensing data alongside detailed information from the ground to create a more accurate picture of the extent of production, and also the likely levels of yield both now and for the future.
The growers and agribusinesses themselves are often the primary producers of these data; collected routinely as part of existing food production systems. The data is diverse – everything from hand augured soil samples, crop canopy measurements and weather data in the field to real-time fuel efficiency of tractors or yield data from combine harvesters sent wirelessly by telematics, via cloud based services. Through the application of informatics techniques such as machine learning and pattern recognition, new crop models can take into account many more agronomic variables at much higher data-volumes.
This information can then be analysed alongside the big picture data: satellite imagery of crop growth and development, weather data, topography and yield maps. In combination, it’s then possible to develop algorithms to determine the rules for the best variable rate applications: what varied inputs provide the best results for growers, prediction of yield, and other micro-perspectives such as soil degradation and traceability of food crops.
With an agri-informatics data-driven approach it is possible to have the basis for planning – and also for managing and supporting farmers’ livelihoods that can balance their immediate needs against those of long-term sustainability.
An independent producer organisation, G’s Growers has been working with Cranfield for several years as part of a strategic relationship. In the case of remote sensing, G’s didn’t have the in-house expertise to take advantage of the changing technologies and how it could be put to practical use.
G’s Growers has more than 20 grower members in the UK and Spain, supplying speciality crops to consumers across the UK, Europe and North America. The G’s land bank, estimated at more than 11,000 hectares, is in Norfolk, Suffolk, Cambridgeshire, Kent, West Sussex, Dorset, West Midlands and the Murcia region of southern Spain. It’s a large area – for these kinds of crops – which also involves a highly diverse mixture of land types. This means challenges for managing the crops, particularly due to the short crop cycles and small windows for any intervention from an agronomist or grower in response to problems.
A Knowledge Transfer Partnership was set up with G’s Growers and Cranfield to look at emerging remote sensing technologies and their application in the production of high value horticultural crops. A pilot study to test the performance of several commercial UAVs was carried out in nine lettuce and onion fields in Cambridgeshire.
Very high-resolution natural colour and near-infrared cameras were mounted on different platforms, including fixed wing and quad-copter aircraft. In order to cover large field areas, individual image frames were automatically ‘stitched’ together using specialist software using a computer vision technique called ‘Structure from motion’. Then for each field the digital numbers for each image pixel element are calibrated and converted into the digital information that is then used to make decisions.
The final interpreted images are presented to the growers either as maps, web services or straight to Google Maps on their tablets. These images provide an update on what’s growing on the ground in terms of the extent of the canopy and any indications of limited growth, of pests and disease. This information is cross-referenced against data on the variation of soil types and weather to build up a more accurate picture of all the factors involved and possible impacts on yields.
Monitoring also helps with risk analysis, also looking at the nature of the surrounding fields or area and how this might be affecting the crops. The data can then be used to target the necessary inputs – increased levels of fertiliser, or pest control intervention – with the aim of creating more consistency in yields across the fields. Access to this level of data helps with predicting performance and market yields, building up a picture over time of growth and performance and how variations are most effectively addressed.
Following on from the initial trial G’s Growers is extending the remote sensing approach to 30 fields, with the longer-term aim of implementing a real-time crop monitoring and management system across all its land bank. One of the main lessons from the trials has been the importance of speed of data delivery, especially with salad crops since they have a short growth cycle.
Making more timely and accurate management decisions depends on the next day delivery of analysed data. There has also been the issue of the most cost effective means of gathering data. UAVs need two people to operate them successfully, and are limited in terms of the weight of on-board equipment and sensors.
In the next phase of research G’s Growers is looking at using very light manned aircraft as an alternative. And as the technology advances further there is the potential for monitoring to include detailed diagnostics and automated treatment.
