Crop insurance damage assessment by satellite in Uzbekistan

What is remote crop damage assessment?

In the traditional process the insurer learns about a loss from the farm's claim, and an expert visits the field and writes a report. Remote assessment adds independent observations: regular satellite images of every insured field and a weather record above it.

Its main value is a record of the event that does not depend on what either party says: the date, the outline of the affected area, and the field before and after. This cuts down on blind field visits and on disputes over whether and where the event happened.

The method works for events that change a crop visibly and quickly: water on the field, lodging, a sudden loss of green. Slow losses and anything happening under the leaf canopy are seen poorly or not at all.

How does satellite-based damage assessment work?

The assessment compares the field with itself: what it looked like before the event and what it looks like after. The usual steps are as follows.

• Field outline and crop: without an accurate boundary the share of the affected area cannot be calculated.
• Field history: images from past seasons show how this field normally looks at this growth stage.
• Event date: from weather data and from a sharp change between two images.
• Damage outline: which part of the field changed and by how much.
• Before and after images: a pair of observations that can be attached to the claim file.
• Field inspection: an expert or agronomist establishes the cause and severity of the damage.

What data is used, and which events are easiest to see?

Sentinel-2 optical imagery shows greenness, leaf water and biomass at 10 m pixels. Nominally a field is imaged about every 5 days, but under clouds the real interval is a week or longer. The Sentinel-1 radar emits its own signal and sees crop structure and surface water through clouds and at night. ERA5 weather data with a forecast provides temperature, air humidity, rainfall and wind.

No single source is enough on its own: optical imagery is blind under clouds, radar does not measure chlorophyll, and weather data warns of a threat but does not prove damage. That is why different events are visible in different ways.

• Flooding: radar detects standing water immediately and through clouds, which matters because floods usually come with overcast weather. Optical imagery shows where the crop was harmed 5–15 days after the water recedes.
• Wind lodging: radar sees the change in crop structure immediately. In optical images a lodged crop looks brighter, so NDVI does not help here.
• Hail: the satellite does not see hail itself, only its trace – a sharp drop in greenness between two clear images, roughly 5–12 days after the event.
• Hot dry wind (garmsel): a warning is possible about a day ahead from temperature, air humidity and wind. Imagery confirms the drop in leaf water only several days later.
• Spring frost: the damage happens at night, and the loss of green appears in imagery 4–10 days later.

How is agricultural insurance developing in Uzbekistan?

According to Spot.uz (25 September 2026), by 2027 the target is to insure 45,000 agricultural enterprises and 60% of farm output. The process from claim to payout is moving into the Agrosafe system, and 500 independent experts are planned. The stated priority is less human discretion and a fair assessment of losses.

At the same time the state is bringing satellite monitoring into agriculture: under a decree of 13 March 2026, an Agricultural Monitoring Centre will analyse satellite data, weather and crop vegetation (Gazeta.uz, 16 March 2026).

For insurers and farms this means growing demand for an objective record of the event – the date, the outline and before and after images that can be checked independently of either party.

What can satellites not prove, and what should be checked in the field?

A satellite records the consequences but almost never names the cause. A sudden loss of green after hail looks the same as after strong wind, a fire or early mowing. Without weather radar the hail swath cannot be located either, so the cause is confirmed by inspection and by the farm's own report of the event.

Clouds block optical imagery, often on exactly the days after a storm when the hail trace is clearest. A missing image is a gap in observation, not proof that there was no damage. A 10 m pixel averages everything inside it, so small patches and narrow strips along the field edge can be lost.

Severity cannot be measured directly from imagery either: radar shows that lodging happened but not the stem angle that determines the loss, and shrivelled grain after hot dry wind can only be confirmed by examining the ears.

• The event date and the farm's note of what happened
• Photos of the damage
• The share of damaged leaves, flowers and young fruit – for the loss report
• The share of lodged area and the stem angle
• Depth and duration of standing water and root condition after flooding
• Grain shrivelling at grain fill after hot dry wind

What does it look like in practice? The ClusterOS example

ClusterOS is a field monitoring platform built on open Sentinel-1, Sentinel-2 and Landsat data and ERA5 weather. For insurers it puts together what is needed to verify a claim without sending an expert to every field: the event, its date, the field outline, the field's history since 2019, and images from before and after the event.

The package covers hail, flooding, lodging and hot dry wind. The area and date of flooding are recorded from radar, while a sudden loss of green without a confirmed cause is shown as a loss of green with an unidentified cause, not as hail. If there is no observation because of clouds, the system says so.

Field data is visible only to the owning organisation, and the farm itself opens view-only access for an insurer.

How to get started

Start with the list of insured events and which of them a satellite confirms reliably: radar sees flooding and lodging immediately, the trace of hail and frost appears with a delay, and hot dry wind is anticipated from the weather. For each event, agree in advance what counts as confirmation: an image, a weather record, an inspection or a combination.

• Collect accurate outlines of the insured fields and the crop on each – no sensors or hardware are needed.
• Pull the field history from past seasons so there is a baseline for comparison.
• Agree with farms that they report an event right away, with the date and photos.
• Use imagery to choose which fields and areas to visit, and establish cause and severity by inspection.
• Record gaps in observation caused by clouds so they are not mistaken for an absence of damage.

Related risks

• HailHail from the forecast – a warning before the event, and images of the field before and after it.
• FloodingThe field or part of it is under water. The area and date are recorded – for decisions and insurance.
• Wind lodgingStrong wind flattens crops before harvest. We show the fields where the risk is higher.
• Hot dry wind (garmsel)A hot dry wind dries out crops within days. We warn from the weather forecast.
• Spring frostFrost at emergence or flowering. The warning comes in advance, with time to protect the crop.

Questions