Industries / Food, farming and forestry / Agriculture and farming / Nut and fruit orchards
Pollination
Pollination is getting pollen from flower to flower so the trees set fruit and nuts: planting varieties that pollinate each other, bringing in bees, and pollinating by hand or by machine where bees fall short.
5
Milestones achieved, from 1894 to 2023.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1894
Pears need a second variety
In 1891 and 1892 Merton B. Waite of the US Department of Agriculture covered pear blossoms with paper bags to keep insects out. Bartlett, Anjou, Clapp's Favorite and Winter Nelis set almost no fruit inside the bags, and Bartlett would not set fruit with its own pollen, only with pollen of another variety. A self-sterile tree like this flowers well and still bears nothing unless pollen comes from a different variety. Waite traced the poor crops of the Old Dominion Fruit Company's orchard of about 22,000 Bartlett trees near Scotland, Virginia, to this: the only Bartletts that fruited well stood around two Clapp's Favorite trees planted there by mistake. His bulletin told growers to avoid solid blocks of one variety and to make sure there were enough bees within 2 or 3 miles (about 3 to 5 km).
LessonA tree that blooms heavily can still fail if no compatible pollen reaches it.
Credit
Merton B. Waite, USDA Division of Vegetable Pathology
1990
Mason bees take over Aomori
Aomori Prefecture grows nearly half of Japan's apples, about 500,000 t a year on about 25,000 ha (about 62,000 acres). Growers there were told to pollinate apples by hand from the 1930s, and in 1959 about half the prefecture's apple trees were pollinated by hand. Honeybees served 40% of the apple area in 1978. From 1981 growers turned to the hornfaced bee, Osmia cornifrons, a solitary bee that most growers house in bundles of hollow reed stems set out in the orchard, and by 1990 it pollinated over 60% of the apple area, by the count of the prefecture's apple experiment station.
Credit
N. Sekita and M. Yamada, Aomori Apple Experiment Station
2001
Apples pollinated by hand
In Maoxian County, Sichuan, China, apple farmers pollinate each flower by hand across a valley about 60 km (37 miles) long, with every member of the family taking part, Uma Partap, Tej Partap and He Yonghua reported in 2001. They found that yields had fallen for lack of pollination and that beekeepers would not rent out colonies because the orchards were sprayed so heavily with pesticides, although renting bees would cost about one eighth as much as pollinating by hand. A 2012 follow-up by Uma Partap and Tang Ya called it a case of the failure of pollination services.
LessonSpraying that drives the bees away leaves people to do the bees' work.
Credit
Uma Partap, Tej Partap and He Yonghua; Uma Partap and Tang Ya
2008
Almonds that pollinate themselves
Most almond varieties cannot set nuts with their own pollen, so orchards alternate rows of two or more varieties and rent bees to carry pollen between them. Independence, a self-fertile almond bred by Zaiger Genetics, has been sold since 2008. By 2016 it took one quarter of all new almond acres in California, and Dani Lightle, a University of California farm advisor, suggested one hive per two acres (about 1.2 per ha) for it instead of the usual two to two and a half hives per acre (about 5 to 6 per ha). In Australia, mapping by Land IQ found that self-fertile varieties made up more than 25% of almond trees planted in the three years to 2026, and the Almond Board of Australia puts them at about 7.5% of all plantings, up from 2% in 2023.
LessonA variety that saves on bees can still sell for less: Lightle reported that buyers preferred the older Nonpareil and the price gap widened.
Credit
Zaiger Genetics; University of California Cooperative Extension
2023
Japan cuts off Chinese pollen
Japanese pears do not set fruit with pollen of their own variety, and many growers pollinate the flowers by hand with pollen they collect or buy. A 2016 survey by Japan's Central Fruit Association estimated that about a third of pear growers bought pollen, most of it imported from China. After fire blight, a bacterial disease of pears and apples that Japan has never had, was confirmed in China, Japan stopped imports of fire blight host plants from China, pollen included, on 30 August 2023. Prefectures told growers not to use Chinese pollen and to collect more of their own, and a pear grower in Inagi, Tokyo, said a newly grafted pollen-source tree takes three years to give usable pollen.
LessonAn orchard that depends on imported pollen can lose it overnight to a plant-health ban.
Credit
a failure of supply; import halt by Japan's Ministry of Agriculture, Forestry and Fisheries
Watch for
Machines pollinate avocado orchards
BloomX, an Israeli company, says its Yahav 2400 machine, launched commercially in 2026, gathers avocado pollen from male flowers with an electric field onto a fabric meant to mimic a bee's body and carries it to female flowers. It says growers using its avocado and blueberry machines in Peru, South Africa and Israel recorded average yield increases of about 20%, with returns of up to 10.5 times the cost, and that one operator covers up to 3.5 ha (about 9 acres) a day.
Achieved whena university or grower trial not run by BloomX publishes fruit set and yield for replicated machine-pollinated and bee-only avocado blocks over a full season.
Claimed by
BloomX, October 2026, reported by Ynet Global, which notes that no independent assessment was provided
Open
01
Map bee visits per tree
A network of 30 to 50 small solar-powered camera traps set on flowering branches across one almond or apple block, counting bee visits per flower per hour tree by tree through bloom, and a map that sets those counts beside each tree's distance from the hives and the fruit set on tagged branches of the same trees.
The briefCredit · unclaimed02
Hives matched to bloom
A method that counts open flowers per tree from drone or tractor-mounted camera images a few days before full bloom and sets the number of hives for each block from that count, tested against the usual fixed number of hives per acre.
The briefCredit · unclaimed
Open 01 · the brief
Map bee visits per tree
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A network of 30 to 50 small solar-powered camera traps set on flowering branches across one almond or apple block, counting bee visits per flower per hour tree by tree through bloom, and a map that sets those counts beside each tree's distance from the hives and the fruit set on tagged branches of the same trees.
Why it is still open
Growers judge pollination by the number of hives brought in and by fruit set counted by hand weeks later. Open camera traps for flower visitors exist: Insect Detect tracks insects landing on an artificial flower platform (Sittinger and colleagues, PLOS ONE, 2024), and a Raspberry Pi system running YOLOv5 was checked against direct observation in natural plant communities (Serra-Marin and colleagues, Methods in Ecology and Evolution, 2025). Reilly and colleagues found pollinator limitation in five of seven US crops at 131 sites using observers and hand pollination (Proceedings of the Royal Society B, 2020). Researchers at the James Hutton Institute detect bumblebees on cherry flowers in timelapse images with a recall of 0.959 and a precision of 0.991 (BeeSAM2, a 2025 preprint), a detector rather than a block map tied to fruit set. Searches of Europe PMC and Crossref found no camera network that maps visits across a commercial orchard and ties them to fruit set.
Built with what exists
The open-source Insect Detect camera trap design, or a Raspberry Pi with a camera module; Ultralytics YOLO models trained on labeled images of bees on almond or apple flowers; CVAT for labeling; QGIS for the block map.
First step
Build three camera traps this month, label 1,000 images of honeybees on fruit-tree flowers, and agree with one almond grower on a block to watch at the February 2027 bloom.
Done looks like
Forty cameras running through one full bloom in a commercial block, with camera counts checked against a person watching the same flowers for at least 50 ten-minute periods and agreeing within 15%, and a published map of visits per tree beside fruit set on four tagged branches per camera tree.
Open 02 · the brief
Hives matched to bloom
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A method that counts open flowers per tree from drone or tractor-mounted camera images a few days before full bloom and sets the number of hives for each block from that count, tested against the usual fixed number of hives per acre.
Why it is still open
Almond growers rent hives by the acre; a University of California farm advisor gave two to two and a half hives per acre (about 5 to 6 per ha) as the usual rate (Farm Progress, 2017). Blossom density has been mapped from drone images in apple orchards to guide thinning (Yuan and colleagues, Horticulturae, 2023), and Orchard AI, bought by John Deere in November 2026, counts bloom per tree from vehicle cameras to guide where to spray. BeeHero says its pollination planner sets the frames of bees for each almond block from tree age, planting density, the mix of varieties and the weather, an estimate of the flowers rather than a count of them. No trial was found that sets hive numbers from a measured flower count and compares the harvest.
farmprogress.comapi.crossref.orggrowingproduce.combeehero.io
Built with what exists
An off-the-shelf camera drone flown only over the cooperating grower's land and within local drone rules; OpenDroneMap to stitch the images; an Ultralytics YOLO flower-detection model; a spreadsheet rule that turns flowers per hectare into hives per block.
First step
Ask a grower with two similar almond blocks to have both flown at about 10% bloom in February 2027, and record the beekeeper's hive count and placement for each block.
Done looks like
Two seasons of paired blocks of the same variety in at least three orchards, one stocked by flower count and one at the usual rate, with hives used, nut set and yield per hectare published for every pair.