Smart Municipality Eye: What Dubai's AI Drones Are Actually Looking For

Dubai Municipality has an aerial monitoring programme for its green spaces called Smart Municipality Eye. Drones fly the parks, verges, medians and landscaped corridors, and artificial intelligence reads what the cameras bring back: plant health, irrigation faults, early signs of pest activity. The stated coverage is 52 million square metres, which is 52 square kilometres of planting spread across an emirate.

It was announced in November 2025, during Urban Future Week at the Museum of the Future, alongside a memorandum of understanding with the Emirati manufacturer Eanan Al Samma for Remote-Controlled Aircraft Manufacturing. The framing from Dubai Municipality was blunt about the intent. "We are not merely planting trees, we are designing and preserving Dubai's environmental and visual identity with precision," said Director General Eng. Marwan Ahmed bin Ghalita.

Strip out the launch language and there is a real engineering proposition underneath, and it is one that applies to any organisation responsible for a large landscape. Here is what the system is actually doing.

The Announcement, in Numbers

Two things the announcement did not give: a fleet size, and a rollout timeline. The MoU covers exchanging technical expertise, developing solutions jointly and piloting operational models, which is the language of a programme still being built rather than one already flying every district on a fixed schedule.

Why the Area Figure Is the Whole Argument

Fifty two square kilometres cannot be inspected by walking it. Not weekly, not monthly.

So conventional landscape maintenance runs on exception reporting. Something looks wrong, someone raises it, a crew goes out. By the time a tree is visibly dying, or a bed has gone brown, the cause has usually been running for weeks. A blocked dripper starves a row of shrubs long before the shrubs show it. Red palm weevil hollows out a trunk from the inside, and the crown collapses at the end of that process, not the start.

Aerial mapping changes the unit of work. Instead of inspecting the sites someone complained about, you capture all of them on a repeating cycle and let the analysis rank where to send people. The number of horticulturists does not change. What changes is that they stop guessing which of two hundred locations needs them on Tuesday.

What the Sensors Are Actually Reading

Dubai Municipality has not published a payload list, so treat what follows as the physics of the problem rather than a spec sheet. There are four layers that do this job, and most vegetation programmes use two or three of them together.

• RGB, plain visible imagery. What a person would see, except from above, geotagged, and framed identically every time you fly it. Good for canopy gaps, missing plants, storm damage, encroachment and counting stock. It is also the layer anyone can interpret without training.
• Multispectral, including near infrared. Healthy foliage reflects near infrared strongly. Stressed foliage reflects less of it, and that drop happens before the leaf changes colour to the human eye. Indices such as NDVI turn the difference into a map you can threshold. This is the layer that buys you time.
• Thermal. Transpiration cools a leaf, so a well watered canopy runs cooler than a water stressed one under the same sun. On the ground side, a leak or a burst line shows as an unexpectedly cool wet patch, and the dead zone downstream of a blockage shows as a hot one. Irrigation faults are a thermal problem more than a visual one.
• Hyperspectral. Dozens or hundreds of narrow bands instead of four or five broad ones, which lets a specific disease signature be separated from general stress. Eanan markets a platform called OptoSky built for exactly this, aimed at catching palm disease early.

Combine the layers and you get something a ground inspection cannot produce: a stress map of an entire district, on one date, comparable against the same district last month.

Who Is Building the Aircraft

Eanan Al Samma is not a reseller. The company runs what it describes as the first commercial drone manufacturing facility in the UAE, based in the Mohammed Bin Rashid Aerospace Hub at Dubai South, with around 85 percent of its drone components produced locally and the balance, autopilots included, still imported.

Its existing work is agricultural and environmental rather than cinematic. Aerial seeding of mountain terrain around Khor Fakkan with the Ghaith seeder platform. Feed delivery for aquaculture. And the palm health case, which matters here: the UAE has roughly four million palm trees, and detecting blight in one of them from the air, early, is worth more than treating a hundred of them late.

That is a sensible partner for a green space programme. It is also a signal about where UAE drone procurement is heading, which is towards locally manufactured, task specific airframes rather than general purpose camera drones.

The Honest Caveat: One Flight Is Not Predictive

Predictive maintenance is the phrase in every release. It is achievable. It is not automatic, and it is worth being precise about why.

A single survey produces a snapshot. It tells you that block C is reflecting less near infrared than block D. It does not tell you whether block C has always been like that, whether it is drying out, or whether it is simply a different species with a different signature. Prediction needs a baseline and a trend, which means repeat flights at a fixed interval, with consistent altitude, overlap and radiometric calibration, so that this month is genuinely comparable to last month.

The AI carries the same dependency. A model that flags plant stress reliably has been trained on labelled local examples: these species, this soil, this irrigation regime, this glare, at these times of year. Models trained on temperate agriculture do not transfer cleanly to Gulf landscape planting. That is what a pilot phase is for, and it is the part that takes the longest.

Every flag still needs ground truthing at first, too. The value of the system is not that it replaces the horticulturist. It is that it tells the horticulturist which fifteen locations out of two hundred to look at this week.

If You Own the Landscape, Not the City

The municipality has 52 square kilometres. A master planned community has a few hundred hectares. A hotel estate, a campus, a golf course, a logistics park with a landscaped buffer, a developer sitting inside a defects liability period: all of them have the same problem at a smaller scale, and none of them get a government drone programme.

The method transfers directly. Fly a fixed grid, capture RGB plus multispectral and thermal, process to an orthomosaic and index maps, repeat monthly or quarterly, and compare. What comes out is a ranked list of locations with a reason attached, and a dated record you can hand to a landscaping contractor or hold them to.

Landscape irrigation is one of the larger recurring operating costs on a Gulf property, and most of the waste in it is invisible from a car window. A leak under a bed and an over irrigated zone both cost money every day they run.

The Flying Is Still Regulated

None of this is a hobby flight. In Dubai, every mission needs a no objection certificate from the Dubai Civil Aviation Authority, issued for that specific location, altitude, date and time, on top of operator authorisation and aircraft registration. Law No. 4 of 2020 governs unmanned aircraft in the emirate, and flying outside the approvals is a prosecutable offence, not a fine on the way out.

Add Ministry of Defence approval whenever the mission includes photography, and Dubai Air Navigation Services approval for controlled airspace or anything above 400 feet. Survey work over a residential community also sits close to uninvolved people, which pushes the risk classification up. Build the paperwork into the programme calendar, not into the week before the flight.

So, Is This the Future of Maintaining Dubai's Parks?

The direction is not really in question. Once a city can see the condition of every planted square metre on a repeating cycle, going back to waiting for complaints stops being a defensible position. The open questions are pace and proof: how quickly the pilot becomes routine operations, and whether the AI earns enough trust on local species that anyone acts on its output without sending someone to check first.

Getting the Same Picture of Your Own Green Space

COENGR EYE flies vegetation and site condition surveys across the UAE under GCAA authorisation, with insured crews and permit coordination handled in house. RGB, multispectral and thermal capture, processed into orthomosaics and index maps you can hand straight to a maintenance contractor, on whatever repeat cycle the site needs.

Tell us the site area and how often you want it flown, and we will scope the survey and the approvals together. Reach us at marketing@coengr.com.