16 December 2021
Remote Sensing in the potato sector
REMOTE sensing is becoming an increasingly important technique in agriculture, with many applications in the potato sector, both for routine crop agronomy and the assessment of field trials.
Some of the uses of remote sensing in potatoes include phenotyping of trials or for roguing of seed potato crops; identifying areas of crop stress caused by pests and diseases, such as potato cyst nematode, blackleg, virus or blight; for crop measurement and yield estimation; and for informing decision making, such as variable rate fertiliser applications.
Remote sensing is the science of gathering data on an object or area without making physical contact. Aircraft, satellite and drone (UAV)-based cameras and sensors are used to measure reflected and/or emitted electromagnetic radiation. This information, often captured as images, can then be analysed to extract
additional, valuable data which can be mirrored in a GIS environment for spatial visualisation and mapping.
Depending on the type of data capture used, it is possible to automatically detect disease stress or phenotypical differences between crop varieties, and where stress is present, a lot of ongoing work is focused on the extent to which different forms of crop stress can be distinguished, for example, pre visual presence of blackleg or virus in individual potato plants using different spectral signatures.
Paul Brown, Remote Sensing Expert with FERA Crop Science, said it is important that remote sensing is carried out at the correct resolution for the aims of the project. For example, when researching disease stress in potato plants (which are typically 50 to 100 cm in diameter), enough spectral information per plant
must be collected as possible.
Each sensor collects a separate dataset and making sure these datasets spatially overlap as accurately as possible is imperative when conducting this type of research, for example, using drone GCP targets with highly accurate network RTK GNSS systems to tie the datasets together, allowing these different imagery datasets
to be accurately overlaid in desktop software.
Different sensors allow a range of different data to be collected, for example, using different plant
health indices to identify stresses and remove the spectral noise of background information, such as soil, from any further analysis.
Data can also be used to train neural net learning algorithms which can, with sufficient data, determine
relationships between the sensor data and disease stress, which can then be assessed using the ground
Remote sensing can highlight clear differences at an individual plant level. This research requires extensive
ground truthing to train and check the results obtained from any machine learning algorithms developed. FERA's
studies show these methods are able to detect a comparable virus rate in seed potatoes as a visual inspection.
A drone or UAV equipped with a multispectral sensor can survey afield much quicker than traditional
inspection methods and therefore potentially provide a lower cost alternative to identifying fields with
relatively high prevalence of infection or confirming low prevalence.
Remote sensing technology has already delivered significant advancements in precision agriculture and crop uniformity. Advances in image resolution, sensor technologies and computation analyses continue to provide greater management insight, crop productivity and environmental outcomes.
Source: Fera Science Ltd