Water is allocated on paper. We measure what the fields actually take.
AgriTech Hub Kazakhstan measures irrigation water use, crop type and land productivity field by field, across whole regions of Kazakhstan and Central Asia. All of it from satellites. Nothing installed in the ground, nothing to maintain.
2 654 real fields in the Talas valley, from our 2024 Zhambyl project. On paper, hover a field to light up its whole irrigation zone: one figure for the lot. From space, hover the same field to get its own numbers.
Swipe the map sideways to see the whole command area.
Baizak and Zhambyl districts, a 52 by 20 km window: 2,654 fields and 22,373 hectares, out of 43,354 analysed across the region in 2024. The recorded figure is water delivered to a whole irrigation zone, split across its hectares. The measured figure is what each field actually consumed, from every source including rain. Different quantities, and that is the point. One is a plan. The other is what happened.
The last mile in water accounting
Water is counted carefully all the way down the system. Then it reaches the field, and the counting mostly stops.
Rivers and reservoirs
Flow meters
Canals
Gauges
Irrigated fields
Metered on a few fields, unknown on the rest
Dark markers are fields with their own meter. Everything grey is unaccounted for.
All three panels are our own Zhambyl mapping: real rivers, real canals and command areas, real field boundaries.
Watersheds and canals are well instrumented almost everywhere. The field is where it thins out. Pivot and drip systems often carry their own meters, and a few gauged fields sit alongside them, but they are the exception. For most of an irrigated region, nobody knows what any single field consumed.
Satellites cover the rest. Every field gets a number, metered or not, measured the same way. A whole region becomes comparable, instead of a handful of instrumented farms.
The Aral Sea did not fail overnight. It failed invisibly.
Water was diverted on plans and assumptions. Never on measured use.
The loss built for years before anyone could see it. That pattern is still running. Across Central Asia water is allocated, logged and reported, and almost never measured where it is actually spent.
The danger is not only that there is less water. It is that nobody sees the shortfall until it has been paid for, in yield, in income, in political trust.
of the world's fresh water goes to agriculture
of irrigation water in the region is lost before it reaches a crop
projected fall in Central Asian surface water by 2050, against demand rising up to 51%
the year the region is on course to cross the chronic water-scarcity line of 1,700 m³ per person
Three layers. One picture of the ground.
Three services, one dataset. Pick one to see what it delivers, and what it looked like on a real season.
A living water balance, not a declared one
Consumption measured at every field in a command area, through the season, from evapotranspiration rather than volumes declared upstream. The gap between the two is usually where the story is.
How a season gets measured
Four stages over a season of imagery. Machine learning does the work. Our own specialists check it on the ground.
Classify the crops
Every field mapped and identified by crop, with boundaries drawn from imagery rather than cadastre.
Map the infrastructure
Canals, drains, reservoirs and command areas digitised, so consumption can be traced to the system that delivered it.
Model the water
Vegetation and land-surface temperature give evapotranspiration, and from that the real consumption and productivity of each field.
Validate and deliver
Field teams check the model against the ground. Results go into a dashboard, a policy brief, or a ministry's own systems.
Sample project: Zhambyl region, 2024
One project, opened up. Both charts come from the same 2 654 fields on the map above. Every figure is the region’s own, not an industry average.
Water applied, month by month
Consumption is not one annual number. It peaks in July and stays high into September. That is when an allocation decision actually bites.
What each crop returns per cubic metre
The same water buys very different results. Sugar beet returns twenty-one times what wheat does per cubic metre, on land drawing from the same canals.
Water productivity is total yield divided by total water applied across every field of that crop inside the mapped window. Crops on less than 800 hectares, fallow ground and non-vegetated parcels are excluded.
From one canal to a national system
It started on one irrigation canal. The same method now runs across borders.
First pilot, Turkestan region
Productivity monitoring along the Arys-Turkestan canal. The first test of field-level measurement against how a canal was actually run.
Work with international development institutions
Projects delivered with the Asian Development Bank, the World Bank and IFC, alongside USDA and Michigan State University research programmes.
Scaled to two regions
Water consumption measured across Turkestan and Kyzylorda, the two regions under the most pressure.
Decision-support system, three southern regions
The measurement became a system officials could plan with, across the three southern oblasts.
Across the border, and deeper at home
Eight regions of a client country brought under one monitoring system, delivered under confidentiality. The same season, 43,000 fields in Zhambyl were mapped for crop, yield and water productivity. That is the dataset behind the map at the top of this page.
Country-scale crop classification
Agricultural monitoring across the entire territory of a client country. Every field in a nation, mapped and classified. The client is not named here, under a confidentiality agreement.
A shared data layer for the region
Expansion across Central Asia and an international platform for sharing satellite data on transboundary water. Talks underway in India, Pakistan, Mexico and the United States.
Why the same numbers matter to everyone
Water does not respect borders, and neither should the data. When everyone upstream and downstream reads the same figure, measured the same way, a fight over a shrinking resource becomes something you can negotiate. Neutral measurement is not only an efficiency tool. It is a basis for trust.
Collaborators and partners
Past and present, across research, finance and delivery.
About the Hub
AgriTech Hub Kazakhstan is a corporate fund, a non-commercial organisation working from Astana. We bring the technology the country’s agrarian sector needs into the hands of the people who run it.
We find, adapt and transfer the technology the sector is missing, and close the gaps that hold industry and agriculture back.
We do not model from a desk. Government-scale delivery, satellite analytics and field validation sit in one team. That is what turns space data into something a ministry can run on.
The maps on this site were not bought in. Every boundary, crop class and water figure is ours, produced season after season. The 2024 Zhambyl analysis is one of them.
The team
A satellite calculation stays a presentation all too easily. Diana built a team that turns it into a working instrument: from framing the question and checking it in the field to a system that runs every season. She started with drone survey of fields and agronomic mapping, trained Kazakhstan’s first UAV operators, then became the youngest deputy chair of the board at Kazakhstan Gharysh Sapary, the national space company. Later an advisor to the Minister of Digital Development. Biotechnology, agrochemistry and soil science, with geospatial study at Michigan State University.
A satellite image holds neither crop type nor water use. Both have to be computed, and that is Ruslan’s part: crop classification, digitising the irrigation network, and calculating water consumption and water productivity for every field. Every map on this page comes out of his processing chains in Google Earth Engine, ArcGIS, QGIS and ESA SNAP, checked against ground surveys he runs himself. Six seasons at AgriTech Hub. M2 in aerospace and autonomous systems from Paris-Saclay; a master’s in automation and control from Auezov University.
A method tuned on one farm behaves differently across millions of hectares. Bakdaulet keeps the accuracy standing at that scale: he trains and validates the classifiers, runs the pipelines that turn raw imagery into structured spatial databases, and builds those into finished analytical systems. Six seasons at AgriTech Hub. Python, GDAL, Google Earth Engine, the ArcGIS stack and Power BI, with a master’s in project management from Al-Farabi Kazakh National University.
Common questions
What governments, banks and water authorities usually ask before a first project.
How is satellite water monitoring different from meters and sensors?
Meters count water at the headgate. Sensors cover the few points where someone installed one. Satellites measure consumption at every field at once, with no hardware in the ground and nothing to maintain. Across millions of fragmented fields, that is the only approach that pays at national scale.
What satellite data do you use?
Public satellite archives going back decades, plus regional climate data, elevation models and soil layers. That gives 10 to 30 metre resolution, so results come per field, not per district. Where a client supplies high-resolution commercial imagery we use that too, which takes the detail down to sub-field patterns. Our specialists validate the models on the ground every season.
Can you cover a whole region or a whole country?
Yes. Water consumption analysis across Turkestan, Kyzylorda and Zhambyl. Crop classification and agricultural monitoring across the whole territory of a client country, under a confidentiality agreement. National coverage is the normal unit of work, not an upgrade.
How accurate is satellite crop classification?
Above 94 per cent for the major crops. Crop type comes from the season's imagery and is checked against ground surveys our own teams run. The 2024 Zhambyl analysis on this page separated eleven classes across 43,354 fields. We report accuracy per class with every delivery, so a ministry can see where the model is strong and where it needs more ground truth.
Who uses this kind of analysis?
Ministries and water authorities responsible for irrigation and allocation. Development banks and donors financing water and agriculture programmes. Anyone who needs numbers that hold up in an audit, a dispute or a policy decision. We have delivered with the Asian Development Bank, the World Bank, IFC, USDA and Michigan State University.
Do you work outside Kazakhstan?
Yes. National-scale work delivered for a client in a neighbouring country, expansion across Central Asia underway, and talks in India, Pakistan, Mexico and the United States. Water here crosses borders. The measurement has to as well.
How can satellite data help manage water resources in Kazakhstan?
Kazakhstan allocates water through plans and reports, but consumption at the field is rarely measured. Satellite remote sensing closes that gap: it gives a water figure for every irrigated field in a region, on one method, so an authority can compare command areas, find losses and defend an allocation decision.
What does remote sensing give agriculture in Kazakhstan?
Crop type and field boundaries for a whole region without a cadastre, yield forecasts before harvest, water productivity per crop, and salinisation and degradation maps. On the 2024 Zhambyl season that covered 43,354 fields in a single analysis.
Let's measure something
Run an irrigation system, finance one, or answer for the numbers a water policy rests on? We will show you your own territory measured rather than declared.
Prefer to write directly?