The drone detective

How De Rust Estate uses infrared imaging as a cost-effective tool to identify and address stress in orchards. By Anna Mouton.

Is there an affordable tool for rapid water-stress detection in orchards? This question set Karin Clüver, production director at De Rust Estate in Elgin, on a path that led to a research project and her MSc at Stellenbosch University’s South African Grape and Wine Research Institute (SAGWRI) under the supervision of Profs Carlos Poblete-Echeverría and Melané Vivier.

Her results confirmed that infrared imaging can map water stress in commercial apple orchards.

“Thermal cameras have existed since 1969, but infrared imaging technology keeps improving,” she says. “And we know the basic principles of how water-stressed plants become warm — I wanted to find out how to measure tree temperature and apply the information.”

Water-stressed and warm

Transpiration is the process whereby trees take up water through their roots and lose it through their leaves. Evaporation at the leaf surface is the primary driver of transpiration. Each water molecule lost to the atmosphere is replaced by another pulled up against gravity. The water molecules entering the roots effectively join the back of the transpiration queue.

Insufficient soil moisture stresses trees because the transpiration queue slows or stops. Plants respond by closing their leaf stomata to slow or stop evaporative losses.

Stomatal closure has two significant downsides. The first is that plants need open stomata for effective photosynthesis and respiration. The second is that plants use transpiration to cool themselves — evaporation is associated with heat loss.

Trees that can’t cool themselves will become hotter than the ambient air temperature as they absorb solar radiation. Potential performance detriments include reduced fruit size and increased fruit-quality issues such as sunburn.

A new tool in the box

Growers aim to optimise irrigation scheduling to avoid over- and under-watering. De Rust Estate already employs technologies such as soil-moisture probes, weather stations, and remote sensing to inform their scheduling decisions. “Infrared imaging doesn’t replace any of these,” says Clüver. “It’s just another tool in your toolbox.”

She explains that infrared imaging provides temperature data for calculating the crop water-stress index (CWSI). This index provides a relative measure of transpiration that is normalised for different environmental conditions.

The biggest benefit of CWSI is that it’s a real-time window into the tree’s stress. “The advantage of having your own drone is that it offers the opportunity to do on-demand flights,” says Clüver. “You can capture the canopy temperatures during the warmest part of the day when the difference between stressed and non-stressed trees is highlighted.”

Most remote-sensing satellite and drone systems provide the normalised difference vegetation index (NDVI). This index is based on measurements of reflected visible and near-infrared light. Healthy plants absorb mostly visible but not near-infrared light, so the amount of visible and near-infrared light reflected correlates to the amount of vegetation.

Chronic water stress is one potential reason for a reduction in vegetation. However, Clüver points out that NDVI is a lagging indicator. It measures the plant’s longer-term structural rather than its shorter-term physiological response to stress.

A bonus of thermal imaging is showing the entire orchard’s status. “We use soil-moisture probes, but a probe is in one location in a block,” says Clüver. She thinks drone technology is a cost-effective tool for expanding the narrow focus of soil-moisture data.

CWSI validation in apple orchards

“We wanted a rapid tool to supply actionable data, in near real-time so we could solve a problem today,” says Clüver, “but it has to be grounded in science.”

Her search for relevant scientific information eventually led her to Poblete-Echeverría, the Digital Agriculture Research Group coordinator at Stellenbosch University. Poblete-Echeverría has extensive experience in infrared imaging and previously worked on proximal- and remote-sensing projects in apple orchards.

Clüver discovered that research was needed to validate the CWSI in South African apple orchards. “The more I thought about it, the more I realised I could take on the challenge. I ended up pursuing an MSc,” she says. “Luckily, SAGWRI accommodates people without a viticulture background as one of their focus areas is precision agriculture.”

She spent two seasons developing and validating a protocol for measuring water status across an apple orchard using the CWSI calculated from infrared drone images. Her research site was a full-bearing commercial Cripps Red block at De Rust.

The experiments compared water-stressed trees to well-watered trees. Water stress was induced by deficit irrigation. The CWSI for the sample trees was compared to stem water potential, a recognised method for assessing water stress. Data was collected from 26 December to 13 March in the second season.

Clüver found that the CWSI and stem water potential were strongly correlated, indicating that canopy temperature does reflect water stress. However, establishing the traditional reference temperatures for calculating CWSI is labour-intensive and time-consuming, as wet and dry reference leaf surfaces are used to normalise temperature measurements for current environmental conditions. This is impractical, especially for drone imagery.

She explored alternative reference methods and materials, including artificial reference surfaces, to speed up and simplify the CWSI application in commercial orchards.

The reference temperatures and the temperatures of alternative and artificial surfaces were strongly correlated across a wide range of environmental conditions, offering a more practical and efficient method of calculating CWSI in an orchard.

In addition to conducting research, Clüver has been exploring how the drone can contribute to everyday orchard management. Drone images have provided clues to solving several problems at De Rust Estate.

Closed cases

The case of the missing water

The drone detective has helped the De Rust team home in on several irrigation issues. Figure 1 shows an example where the drone image revealed a hot patch in the middle of a Braeburn block. Further inquiry uncovered an obstructed irrigation pipe.

Figure 1

The drone image in Figure 2 has a red-hot area on the left. When De Rust production manager Jacques van Dyk went to investigate, he discovered that an unknown suspect had closed the automated irrigation tap supplying the hot section.

Figure 2

“We’ve long been trying to find a way to see whether every tap opens,” comments Clüver. “We’ve even looked into putting a pressure gauge on each, but that’s expensive and only gives you partial information.”

Other drone images have revealed design flaws in the irrigation systems of some blocks. Too-high pressures in some areas led to micro-sprinkler misting and poor water deposition, whereas other regions of the same block were over-irrigated.

The case of the overheating Grannies

Clüver noticed that one orchard had entire rows that were much warmer than the others. These turned out to be cross-pollinating Granny Smith trees. De Rust covered the Grannies with draped nets — the results are visible in Figure 3.

Figure 3

Repeated temperature readings at different times of day show that temperatures under the nets never exceed the general air temperature outside the orchard. In contrast, exposed Grannies become much warmer, putting them in the danger zone for sunburnt fruit.

“There is as much as 10 °C difference in canopy temperatures in the open and under draped nets during the hottest part of the day,” says Clüver

She confirms that the infrared drone images are not affected by the nets. The drones can measure canopy temperature through the nets while flying above them.

The case of the sunny soil

De Rust has fields, orchards, and vineyards, providing an opportunity to measure how different vegetation types modulate temperature. In one instance, an open field that had just been sowed and had no vegetation adjoined a vineyard block with a winter cover crop.

Drone imaging showed that soil-surface temperatures in the open field exceeded those of the vineyard floor by up to 20 °C.

This suggests that perennial cover crops in orchards could help mitigate extreme temperatures, allowing trees to maintain photosynthesis for longer on hot days.

Putting all the tools to use

Every physical tool — whether a hammer or a spanner — has strengths and weaknesses and is best for specific uses. The same can be said for digital tools. The best strategy when fixing problems is to have a well-stocked toolbox.

De Rust combines NDVI data from satellite images with heat maps of flowers and fruit from ground-based visual imagery to map fertiliser and spray applications. They employ GPS-enabled variable-rate spreaders and sprayers based on these maps.

During harvest, pickers log every filled bin in the orchard on a cell phone, generating a heat map of yield. For example, the map showed yield variation of 47–79 tonnes per hectare in one Braeburn orchard. Yield variability is likely associated with other variability, such as in fruit size, colour or maturity.

“If you want to farm pome fruit successfully, you need a uniform orchard,” says Clüver. “We have all these layers of data, but is it possible to see the full picture? To identify what causes variation, and what you can do about it?”

In one Royal Beaut block, the flower, yield-prediction, and yield heat maps all showed a steep drop in productivity from one side to the other. The less productive side was also warmer. However, the pattern didn’t match the irrigation design. The cause was different soil types, and the issue was resolved through variable phosphate applications.

Soil variability in a different block led to warmer areas owing to lower water-holding and 2–3 days less buffering capacity. De Rust targeted the warmer areas with mulch to reduce evaporative losses.

Infrared images are not only handy for figuring out solutions to existing problems. When expecting heatwaves, Clüver uses drone images proactively to prioritise the warmer blocks for pulse irrigation.

Drones in practice

Clüver’s drone has a visible and an infrared camera. It can be controlled manually from a handheld console or programmed to carry out a mission autonomously. You don’t need a licence to fly your own drone on your own property.

Infrared light isn’t visible to humans. The drone’s software converts the infrared data to an image by representing temperatures in different colours as specified by the user.

For ongoing monitoring at De Rust, the drone goes on a mission to collect a continuous image set. Clüver combines the set into a single orchard image with open-source software. Van Dyk can access the images on his mobile phone to tweak his irrigation schedule and troubleshoot problem areas.

“I can fly the drone over our 140 hectares in three hours,” says Clüver. Overseas, fully autonomous drones are already being deployed in orchards. These operate themselves, heading out from stations inside the orchard to complete their daily data-collection missions. Unfortunately, free-living drones would likely be poached to extinction in South African orchards.

Infrared images are much smaller files than visible images, so they are quick to process. Clüver only processes the visible images if she spots a potential problem in infrared. Nonetheless, results are available within less than 24 hours.

“There are enormous advantages if we can apply precision farming to use water more efficiently and monitor the results in almost real-time,” says Clüver. “The drone lets you see with other eyes, and it’s a relatively affordable technology that anyone can use.”