Use of Unmanned Air Vehicles for early, real time detection of extreme weather events in vineyards

Abstract

This project developed and demonstrated technology to measure and map whole vineyard microclimate at a sub-metre scale in 3D using atmospheric acoustic tomography (AAT) coupled with long wave infrared thermography. The AAT system was based on a commercially available unmanned aerial vehicle (UAV) fitted with sensors that measure the sound of the UAV as it flies autonomously over the vineyard. A two-dimensional array of microphones deployed throughout the vineyard also measure the sound generated by the UAV as it flies. Accurate, 3D, concurrent, continuous observation and visualisation of air, vine and ground temperatures, and wind speeds across vineyard blocks of around 10Ha is now possible. High-resolution plant and heat stress factors and thermal maps of frost patterns can thus be derived as a function of geographic and temporal variation across a vineyard.

Summary

This report represents the final submission to Wine Australia for the project entitled, “Use of Unmanned Air Vehicles for Early Detection of Extreme Weather Events in Vineyards”. The project was co-funded by Wine Australia and the Australian Government Department of Agriculture, Water and Environment as part of its Rural R&D for Profit program.

The project, which had a strong technology focus, has resulted in the developed of a transportable, unmanned aerial vehicle (UAV)-based meteorological monitoring capability for observing frost and heat- prone vineyards with high accuracy and at unprecedented levels of resolution. The technology is designed to measure microclimates across a vineyard block over a growing season, or even multiple seasons, which would assist vineyard management decisions for protecting against crop damage resulting from weather- related stress.
The research allows users to accurately estimate and visualise microclimates across vineyard blocks in 3D, permitting computation of plant stress factors like evapotranspiration (ET) and crop water stress index (CWSI). Using such parameters, users can now assess differences in microclimatic conditions within a region or vineyard and optimise irrigation, which is normally applied uniformly across a block. This would deliver water and energy savings.

Similarly, during frost events the technology can provide high resolution thermal maps of the surface and air temperatures surrounding the vineyards as a function of geographic and temporal variation. This permits evaluation of the effectiveness and/or need for different frost mitigation strategies, such as frost fans or sprinklers. As the technology is still in its infancy, the full value of the microclimatic information it offers is still being assessed. However, initial conclusions suggest frost mitigation strategies based on air temperature measurements alone may be sub-optimal.
In addition to the design and development of the technology and lessons learnt therein, several heat and frost stress events were observed and continue to be examined. Feedback from these analyses informed the technology’s potential application space and commercialisation considerations. However, the project would benefit from more analysis of the data sets acquired: this will be an ongoing task.

The technology is based on a commercially available Matrice 600 unmanned aerial vehicle (UAV) that was fitted with acoustic sensors that measured the pressure field (noise) generated by the UAV as it flew autonomously. Micro-sensors capable of synchronously recording meteorological variables such as wind velocity, temperature, pressure, and relative humidity were also fitted to this UAV. Two-dimensional arrays of microphones were deployed across two 10Ha vineyards: one at Wynn’s, Coonawarra, the other at Jacob’s Creek in the Barossa Valley. Each ground sensor also synchronously measured the pressure field generated by the UAV and matched them to those measured onboard the aircraft. From the correspondence relationships variations in sound speed for the signals propagating between the UAV and the ground sensors were computed as the aircraft overflew the vineyards. Using a technique commonly used in medicine, archaeology, and remote sensing known as tomography—which determines the inner properties of the observed medium—temperature and wind velocity profiles above and surrounding the vineyards were then computed in 3D.

In conjunction with this UAV, a second Matrice 600, equipped with a camera capable of high spatial and temporal resolution thermal imagery, was also flown over the vineyards, and the temperatures of the ground and vines calculated from the long wave infrared (LWIR) measurements. A technique known as structure from motion (SfM) was applied to the LWIR images to create 3D thermal point clouds of the vineyard. Finally, the 3D thermographic measurements were fused with the 3D tomographic information and the micro-climates around the vineyard visualised at unprecedented levels of accuracy and resolution. Several severe frost and heat events were observed.

Furthermore, using the thermographic and tomographic data, together with estimates of solar irradiance and object emissivity (obtained during the LWIR sensing operations), evapotranspiration levels and crop water stress indices were computed as a spatial function of vineyard geography over a day. The technology also enabled visualisation of the impact of devices such as frost fans and irrigation strategies. Ultimately automatic dissemination of information gathered using this technology in near real time would enable growers to respond more quickly, cost-effectively, and precisely to severe weather events, which will become increasingly common through climate change.

In addition to the above, a technique for identifying and classifying vine properties such as row width, height, cover-fraction, and missing segments was developed. The genesis of this algorithm was as a bi- product of the need for the tomographic-thermographic data fusion routines to uniquely identify the thermographic (LWIR radiation) properties of the vineyard, i.e. the need to automatically separate vines from inter-row material as they have different emissivity. The algorithm offers users the ability to rapidly, efficiently and non-destructively visualise plant vigour as a spatial function of vineyard geography: the information may be integrated into decision support tools to improve management practices.

The algorithm uses a sequence of overlapping aerial images obtained from visible and long wave infrared cameras carried by the UAV and SfM to extract the underlying topography of the surface terrain. The 3D point clouds were then classified in terms of their hue, saturation, surface temperature and height relative to this surface topography, and the vine and inter-vine material discriminated from one another. The accuracy of the algorithm in terms of its ability to identify vine properties was evaluated using field measurements and was shown to be very effective.