Many satellite crop monitoring systems operate within the visible spectrum. This means that they are ‘blind’ in cloudy weather, resulting in large gaps in the observational data produced and making it difficult to monitor agricultural crops. With CropSAR, VITO has developed technology that monitors the status of agricultural fields regardless of weather conditions.
For a number of years now, VITO has provided farmers with precision-agricultural technologies via online applications such as WatchITgrow and mapEO that better enable them to monitor their land and crops and thus increase their yields. The data behind the technology is deliverd by remote sensing sensors on aircraft, drones or satellites. The advantage of the latter is that they are constantly orbiting around the Earth and cover large areas of land surface. However, one disadvantage of satellite crop monitoring is that observations are frequently interrupted by cloud cover.
This the case, for example, with the Sentinel 2 satellites within Copernicus (the European Commission’s earth observation programme). Each location on the Earth is observed at least once every five days by one of these satellites, producing large amounts of objective data on crop growth and productivity.
Farmers, authorities and insurers
‘The problem is that a significant proportion of the land area is often hidden under cloud cover’, according to VITO’s Kristof Van Tricht. ‘This is inconvenient, particularly in periods of rapid crop changes (for example during spring growth). As a result, farmers using WatchITgrow can miss key stages in the growth cycle of their crops.’ The gaps in the data also create difficulties for other users such as authorities and insurers. ‘If a heavy storm has just passed and it is still cloudy, it is impossible to determine how much damage crops have sustained.’
Small gaps can be ‘closed’ by extrapolating the measurements and observations. However, this solution does not extend to larger gaps. That’s why VITO introduced a new approach three years ago: combining visual observations with radar data generated by other satellites that also cover the Earth’s entire surface (the Sentinel 1 satellites). Van Tricht: ‘Radar waves pass diagonally through the clouds. A small part of the beam is reflected onto the ground, through both vegetation and soil, meaning that the signal can tell us something about the status of the agricultural crops. However, these reflected waves are not easy and straightforward to interpret.’
Deep learning
Nevertheless, this is precisely what VITO Remote Sensing has achieved thanks to hypermodern AI technology such as deep learning. ‘We have a huge supply of images of cloud-free land area, produced using both optical and radar waves’, explains Van Tricht. ‘We “fed” these into a deep neural network, an algorithm that then searches for patterns and links. Once the algorithm was trained, we applied it to the gaps in the visual observations caused by clouds.’ This proved to be a success, as the algorithm was able to accurately close the gaps.
The ‘CropSAR’ technology has been operational since spring 2019, also within WatchITgrow, the VITO Remote Sensing online information platform that provides farmers with a smooth and efficient agricultural plot monitoring solution. ‘The new technology ensures continuous crop production monitoring, regardless of cloud cover’, according to Van Tricht.
