Geocartis carries out agricultural and forestry surveys across India: farm and plot mapping, sown area measurement, crop stage and condition recording, localised loss assessment for insurance, acreage verification for contract farming and procurement, and forest boundary, plantation and encroachment survey. Capture is at 5 cm ground sample distance or better with DGPS control, delivered as georeferenced imagery and attributed vector layers. We are based in Ahmedabad and work nationally.
One thing to be clear about at the start. We survey. We do not spray. Agricultural spraying is a separate business with separate aircraft, separate regulation under the Insecticides Act and DGCA’s agricultural operations framework, and separate operator training. If you are looking for spraying services, we are not the right people and we will say so on the first call rather than the third.
What a drone can and cannot tell you about a crop
Agriculture is the sector where drone capability is most exaggerated, so the limits go first.
Measured reliably: the boundary and area of a cultivated plot, whether a plot is sown or fallow, the extent of a crop within a plot, plant or tree counts on row and orchard crops, canopy cover, plant height on taller crops, gaps and missing rows, standing water and waterlogging extent, physically visible damage such as lodging, flooding, hail flattening, fire and grazing, and terrain and drainage across the field.
Indicated, not measured: crop vigour and stress. A vegetation index shows that one part of a field differs from another. It does not tell you whether the cause is water, nitrogen, salinity, pest, disease or soil variation. Ground truthing is what turns an index map into a diagnosis, and a provider who hands you an NDVI map with a diagnosis written on it has skipped that step.
Not produced by a drone at all: yield. Yield estimation from imagery is a model output. It requires calibration against ground observations for that crop, that variety, that region and that season, and the model belongs to whoever built and validated it. We produce the imagery, the indices and the ground-truth locations. We do not put a quintals-per-hectare figure on a report and call it a measurement.
The sensor question matters. True NDVI needs a near-infrared band, which a standard RGB camera does not have. Indices computed from visible bands alone exist and are weaker, more confounded by soil background and lighting, and not interchangeable with NDVI. Any proposal you receive should state which sensor is being flown and which index is being computed from it. If it does not, ask.
Why a drone, when satellites already cover agriculture
Satellite remote sensing does a great deal in Indian agriculture and is the backbone of the national systems. Drone survey is not a replacement for it. It answers a different set of questions, and there are four situations where it is clearly the better tool.
Cloud. The kharif season is when most insurance-critical events happen, and it is also the cloudiest part of the year. Optical satellites cannot see through cloud. A drone flies below it. On a claim that has to be assessed within days of a monsoon event, that difference decides whether there is any usable imagery at all.
Resolution. Satellite imagery commonly used for agricultural monitoring operates at metres per pixel. At 5 centimetres, you can count plants, identify a missing row, see lodging in a corner of a field and measure the boundary of a two-bigha plot. Those are field-level and plot-level questions, and a metre pixel cannot answer them.
Timing. A drone flies on the day you need it. A satellite passes when it passes, and for a localised event the revisit may miss the window entirely.
Localised events. Hailstorm, a burst channel, a fire, wildlife damage, a flood confined to a few fields. These are exactly the losses that are real to the farmer and invisible at the resolution and scale of a wide-area system.
And where satellites are better: whole-district and whole-state coverage, long historical time series, and repeated coverage over a full season at low cost. On a large programme the sensible design uses satellite for the wide baseline and drone for the field-level verification, the disputed cases and the cloud-affected periods. We are happy to be the drone half of that arrangement.
Farm survey
The base service, and the one most other work builds on.
We map the farm as it physically is: plot boundaries and measured area, bunds and field channels, existing structures, wells and bores, farm ponds, access tracks, tree lines, and the terrain across it.
Where this earns its keep:
Land levelling and drainage. A terrain model at centimetre resolution across a field shows where water sits, which is usually the question, and supports levelling design and cut-fill quantities. On a field that waterlogs every monsoon, a survey answers in an afternoon what a farmer has been guessing at for years.
Micro-irrigation and layout design. Accurate plot geometry, levels and existing infrastructure for drip and sprinkler layout, and for pipeline routing across undulating ground.
Orchard and plantation planning and counting. Tree positions and counts on established orchards, gaps for replanting, canopy extent per tree, and layout planning on new plantings.
Farm records and asset documentation. A dated, measured record of what exists, useful for lease negotiation, valuation input, subsidy applications and dispute avoidance between family members over shared land.
Area verification against records. Measured cultivated extent alongside the recorded area on the revenue extract. As with any comparison to a cadastral record, the georeferencing tolerance of the old map is usually larger than the survey accuracy, and we say so rather than presenting the difference as a definitive discrepancy. Our property survey page covers that in more depth.
Crop insurance
The largest institutional market in Indian agriculture for this kind of work, and the one where methodology matters most.
The framework. Pradhan Mantri Fasal Bima Yojana was designed as a technology-enabled scheme and has been steadily moving that way. YES-TECH, the Yield Estimation System based on Technology, was introduced from kharif 2023 for paddy and wheat and extended to soybean in kharif 2024, assigning a minimum thirty percent weightage to technology-based yield assessment alongside crop cutting experiments. By early 2025 it had been adopted by nine states, with Madhya Pradesh moving fully to technology-based yield estimation. CROPIC provides geotagged, time-stamped crop imagery for damage assessment, and WINDS supplies the weather data layer.
The YES-TECH framework itself contemplates drones. Its cost structure treats drones and UAVs as reimbursable alongside satellites and weather stations, and it places responsibility on the state to arrange the approvals needed for field activities including drone flying.
Where we fit. We are not a yield modelling agency and we are not a Technology Implementation Partner. What we provide is the field-level capture that the models and the assessors need.
Sown area and crop verification at insurance unit or plot level, establishing what was actually planted where, which underpins everything downstream.
Localised calamity assessment. Hailstorm, inundation, landslide, cloudburst and fire affecting individual plots rather than a whole unit. These are assessed on a plot basis and they are precisely the cases where wide-area imagery is least useful. A drone survey within days of the event produces dated, georeferenced, high-resolution evidence of the affected extent.
Prevented sowing and mid-season adversity documentation, where the question is whether and how much was planted.
Post-harvest loss documentation, where cut crop lying in the field has been damaged.
Dispute resolution. This is the growing one. As technology-based yield estimation expands, farmers increasingly question assessments they cannot see the basis of. Reporting from Madhya Pradesh has captured that scepticism directly, with farmers arguing that satellite assessment cannot see what a person on the ground can. A drone survey at five centimetres, flown on a stated date over the specific plot in question, is verifiable by the farmer standing next to it. That is a meaningful thing on a dispute.
CCE support. Locating, geotagging and photographing crop cutting experiment plots, and providing the surrounding field context.
What we do not do here. We do not determine yield, we do not compute claim amounts, and we do not assess a percentage loss. Loss determination under PMFBY follows a defined methodology involving designated agencies. We supply measured, dated, georeferenced evidence into that process.
Contract farming
Contract farming runs on trust about acreage, and acreage is measurable.
The recurring problems are familiar to anybody who has run a buy-back programme. Registered acreage does not match planted acreage. Crop is planted on a different plot from the one registered. Contracted crop is sold outside the arrangement, or crop from outside the arrangement is sold into it. Input supplied for a stated area does not match the area actually sown. And when the crop fails, both sides have a different account of what was planted and when.
A drone survey addresses these with measurement rather than affidavit.
Enrolment verification. Measured plot boundary and area at registration, tied to the plot the farmer has actually committed, with a dated image.
Sown area confirmation after planting, confirming the crop, the extent and the plot.
Stage monitoring through the season at agreed intervals, giving the sponsor an independent view of the crop’s progress without a field officer visiting every farm.
Pre-harvest extent for procurement planning, which supports logistics and working capital planning rather than relying on farmer-declared estimates.
Loss documentation where the crop is damaged, protecting both sides.
The economics work at aggregation. Surveying one two-acre plot is expensive per acre. Surveying a contiguous cluster of two hundred plots in a village in a single day is not, and cluster-based contract farming is the normal structure anyway.
A note on consent. Farmer plots are private property and the farmer is a person with rights over data about their land. Any programme we fly for a sponsor requires that participating farmers have been informed and have consented. We ask for confirmation of that, and we do not treat it as a formality.
Agro agencies, input companies and processors
The commercial side of the same capability.
Sugar mills. Cane acreage survey across the command area is an annual exercise that mills already do, usually on foot, and it determines crushing planning, harvesting schedules and advance payments. Aerial survey of a cane command is faster and produces a measured area rather than a reported one. Cane is also a good crop for this: tall, dense, distinctive from above, and grown in blocks.
Seed companies. Isolation distance verification, which is a compliance requirement in seed production and is genuinely awkward to check on the ground. Plot boundary and area verification for production contracts, and rogueing and stand quality assessment.
Input companies and distributors. Crop and area mapping across a territory for demand estimation, trial plot documentation, and demonstration plot recording.
Agri-lenders and FPOs. Measured area and crop verification supporting credit decisions, and portfolio-level monitoring for a lender with exposure across a cluster.
Processors and buyers. Pre-harvest area and condition across a procurement zone, for volume and logistics planning.
Trial and research plots. Precise plot layout capture, plot-level canopy and stand measurement, and consistent dated imagery across a trial season, which is a real improvement on manual scoring for anything measurable from above.
Forest department work
Forestry is a distinct discipline within this and the constraints are different, so it needs stating separately.
What works well from a drone:
Plantation survival and counting. Compensatory afforestation and plantation programmes commit to numbers of saplings across defined areas, and verification is normally sample-based on foot. On young plantations with visible spacing, an aerial count over the whole block is both faster and complete rather than sampled. This is one of the strongest applications in the whole page and it is under-used.
Plantation area verification. Measured extent of a plantation block against the sanctioned area.
Boundary and encroachment. Forest boundary against actual occupation and cultivation, dated and georeferenced, with change measurable between years.
Canopy gap and clear-felling detection. Where cover has been removed, measured as area.
Fire-affected area mapping. Post-fire extent and severity indication, measured rather than estimated.
Nursery, check dam, trench and soil-moisture-conservation work. Location, count and volume measurement on the physical works forest departments build.
Working plan and management support. Base mapping, compartment boundaries, roads, water sources and infrastructure.
What does not work:
Anything under closed canopy. Photogrammetry sees the top of the canopy. Ground level, understorey, regeneration and the terrain beneath dense forest are not visible. Where bare earth is needed under canopy, LiDAR is the sensor, and we will say so rather than delivering an interpolated surface.
Species identification in mixed natural forest. Possible in limited cases with the right sensor and a lot of ground truth. Not something to promise on a general survey.
Timber volume estimation. Height and crown measurement from above feeds an allometric model. The model, the species parameters and the validation are forest science, not survey.
Airspace. Protected areas carry their own restrictions in addition to DGCA’s, and permission from the forest department is required separately. National parks and sanctuaries frequently sit in restricted airspace, and drone operation over wildlife habitat raises disturbance questions that a responsible operator raises before the department has to. In Gujarat that includes Gir, Shoolpaneshwar, the Banni grasslands and the coastal mangrove areas.
How we do it
Step 1: The crop calendar decides the date
This is the point that separates a useful agricultural survey from a wasted one, and it applies to nothing else we do.
A crop is only identifiable, only measurable and only comparable at the right growth stage. Fly a wheat field three weeks too early and you see soil. Fly it three weeks late and you see stubble. The window for a meaningful survey of a given crop in a given region is often a matter of weeks, sometimes less, and it is set by sowing dates, the season and the weather, none of which move for a survey schedule.
So the first conversation is about phenology. What crop, what sowing window, what stage the survey needs to capture, and what the fallback is if weather closes the window. On a multi-plot programme across a cluster, that also means the whole cluster has to be flown within a short period, or plots flown at different stages are not comparable with each other.
Step 2: Scoping, aggregation and consent
Agricultural survey economics are driven by contiguity. A single small plot carries the full cost of mobilisation, control and processing. A contiguous cluster spreads it. We will tell you honestly when a single plot is better served by a ground measurement, and we will structure a programme around clusters where the work is at scale.
Where the survey covers farmers’ land at a sponsor’s or institution’s instruction, we require confirmation that the farmers have been informed and have consented, and that whoever is commissioning has the right to do so.
Step 3: Airspace and local permissions
We check the area against the Digital Sky airspace map before quoting. Green zone to 120 metres above ground level without prior flight permission, yellow zone with permission and a lower ceiling, red zone not at all. Agricultural land near an airfield, and in border districts, is affected more often than people expect.
Beyond DGCA, agricultural survey happens in inhabited rural areas, and turning up unannounced with a drone over somebody’s field is a poor way to start. On institutional programmes we work through the sponsor, the FPO, the panchayat or the department to inform the village beforehand. Forest work needs forest department permission separately.
Our aircraft carry active Unique Identification Numbers and our pilots hold DGCA Remote Pilot Certificates.
Step 4: Control
Ground control on a plot survey goes on stable identifiable features around the area: bund corners, structures, road junctions, permanent marks. Where the deliverable is measured area or a boundary comparison, control is surveyed with DGPS as it would be on any property survey.
Where the deliverable is crop condition mapping rather than measurement, the accuracy requirement is lower and control can be lighter. We do not put a full control regime on a job that does not need one, and we do not skip it on a job that does. Which of the two your job is gets settled at scoping.
Step 5: Flight planning and sensor selection
Ground sample distance is pixel pitch multiplied by flying height divided by focal length. A 20 megapixel one-inch sensor with 5472 pixels across 13.2 millimetres has a pixel pitch of about 2.41 micrometres. With an 8.8 millimetre lens at the 120 metre ceiling that gives roughly 3.3 centimetres GSD, which is ample for plot boundaries and coarse crops.
Plant counting on close-spaced crops needs finer resolution and therefore lower flying, and that trades directly against area covered per day. On a young plantation with small saplings the flying height may be a third of what it would be for a boundary survey. We size this against the deliverable rather than defaulting.
Sensor selection follows the question. RGB for boundaries, area, counting, gaps, physical damage and visual record. Multispectral where a true vegetation index is required and the client has a use for it beyond a colourful map.
Overlap runs high on crop work. A uniform canopy is a repetitive, low-contrast surface and tie point matching degrades on it, particularly on dense mature crops. Eighty percent forward and seventy or more lateral is a floor rather than a target.
Step 6: Capture
Consistent lighting matters more than on any other survey we do, because vegetation indices and even visual comparison are sensitive to illumination. On a repeat programme we fly at the same time of day each visit, and where a multispectral sensor is used we capture calibration panel readings at each flight.
Wind affects crop imagery in a way it does not affect ground: a swaying canopy blurs and moves between overlapping frames, which degrades both the imagery and the matching. Early morning after the light has come up, or late afternoon, is often the compromise between wind and sun angle.
Field boundaries are the practical constraint on access. Take-off points on bunds and tracks, avoiding standing crop, and a crew that does not walk through a field that is about to be harvested.
Step 7: Processing and interpretation
Aerial triangulation, dense matching, and then the agricultural work: plot boundary digitisation, crop extent delineation, counting, gap identification, damage extent mapping, index computation where a suitable sensor was flown.
Crop classification, where it is part of the scope, is interpretation supported by ground truth. We collect or receive ground truth points and state how many were used and what the resulting confidence looks like. A classification delivered without an accuracy statement is a coloured picture.
Step 8: Reporting
Deliverables organised by plot, cluster or block, with the survey date prominent because on agricultural work the date is half the meaning. Measured areas with the basis stated, and a clear separation between what was measured and what was interpreted.
Accuracy and timing
Ground sample distance and accuracy are different quantities. GSD is the ground area one pixel covers. Accuracy is how close a coordinate is to the truth.
For a well-controlled photogrammetric block, horizontal RMSE typically lands between one and two times the GSD and vertical between two and three times. At 5 cm GSD that suggests roughly 5 to 10 centimetres horizontal and 10 to 15 centimetres vertical.
On agricultural work, three qualifications apply.
Measured area on a cultivated plot is only as well defined as the boundary is. A bunded plot has a physical edge and the measurement is clean. A plot boundary running through open cultivation with no bund, wall or marker has no physical edge, and any line drawn there is interpretation. We say which one we are delivering.
Terrain accuracy under a crop is not ground accuracy. Photogrammetry over a standing crop measures the canopy top. Bare earth beneath it is not visible. Terrain surveys of agricultural land should be flown on bare or stubble fields, and we will schedule for that rather than delivering a canopy surface labelled as terrain.
Timing error dominates everything. A survey with excellent accuracy flown at the wrong growth stage is worth less than a rough survey flown at the right one. This is the only sector we work in where that is true, and it is why the calendar conversation comes first.
Deliverables and output file formats
| Deliverable | What it is | Format |
|---|---|---|
| Orthomosaic | Georeferenced image of the plot, cluster or block, with survey date | TIFF (GeoTIFF), JPEG |
| Plot boundary and area layer | Measured plot polygons with area, and boundary type recorded | SHP, KML, DXF, PDF |
| Crop extent layer | Sown area within each plot, with crop where identified | SHP, PDF |
| Sown or fallow status | Plot-level status against the registered list | SHP, XLSX |
| Plant and tree count | Counted positions with totals per plot or block, and gaps marked | SHP, XLSX, PDF |
| Canopy cover and gap layer | Cover percentage and gap polygons | SHP, TIFF, PDF |
| Damage extent layer | Affected area mapped and measured, with damage type recorded | SHP, PDF |
| Vegetation index map | Where a multispectral sensor is flown, with index and sensor stated | TIFF, PDF |
| Digital terrain model | Bare earth for levelling and drainage, flown on bare or stubble fields | TIFF |
| Contours and spot levels | For levelling design and drainage | DXF, SHP |
| Cut and fill report | Land levelling quantities against a target surface | PDF, XLSX |
| Farm asset and feature layer | Wells, bores, farm ponds, structures, channels, tracks, tree lines | SHP, DXF |
| Forest boundary and encroachment layer | Boundary against occupation, dated, with areas computed | SHP, DXF, PDF |
| Ground truth point set | Locations visited, geotagged, with observations | SHP, XLSX, JPEG |
| Dated photo record | Geotagged stills at plot and damage locations | JPEG, PDF |
| Survey report | Method, date, sensor, control, what was measured and what interpreted | |
| Flight log and survey record | Flights, heights, times, drone UIN, pilot licence number |
Raw imagery is handed over with the deliverables.
Where we work
We are based in Ahmedabad and work nationally, with programmes structured around clusters rather than individual plots.
Gujarat. Cotton and groundnut across Saurashtra including Rajkot, Junagadh, Amreli, Jamnagar and Bhavnagar. Castor, cumin and fennel through Mehsana, Patan, Banaskantha and Sabarkantha. Tobacco and banana across the Charotar belt of Anand and Kheda. Sugarcane and horticulture in south Gujarat around Bardoli, Navsari and Valsad. Forest and plantation work in the Dangs, Shoolpaneshwar, Gir and the Kutch mangrove and grassland areas.
Maharashtra and the west. Sugarcane, cotton, soybean, grape and pomegranate belts around Pune, Nashik, Solapur, Ahilyanagar, Nagpur and Chhatrapati Sambhajinagar.
Central India. Soybean, wheat, gram and pulses through Indore, Ujjain, Bhopal, Sehore and Vidisha, and the Chhattisgarh paddy belt around Raipur and Bilaspur. This region matters particularly for insurance work, given how far Madhya Pradesh has gone on technology-based yield assessment.
Northern India. Wheat, paddy, sugarcane, mustard and horticulture across Punjab, Haryana, western Uttar Pradesh and Rajasthan, via Ludhiana, Karnal, Meerut, Lucknow, Kanpur and Jaipur.
Southern India. Paddy, cotton, chilli, maize, sugarcane and plantation crops across Telangana, Andhra Pradesh, Karnataka and Tamil Nadu, via Hyderabad, Warangal, Vijayawada, Guntur, Bengaluru, Ballari, Chennai and Coimbatore.
Eastern India. Paddy, jute, potato and horticulture across West Bengal, Odisha, Bihar and Assam, via Kolkata, Bhubaneswar, Patna and Guwahati.
What we do not do
We do not spray. Not pesticide, not fertiliser, not anything. Different aircraft, different regulation, different business.
We do not estimate yield, determine crop loss percentage or compute insurance claim amounts.
We do not diagnose pest, disease or nutrient deficiency from imagery. We show you where a field differs from itself. An agronomist tells you why.
We do not identify species in mixed natural forest as a general service, and we do not estimate timber volume.
We do not see under canopy. Bare earth beneath forest or a standing crop needs LiDAR or ground survey.
We do not determine land ownership, tenure or legal boundaries.
We do not fly over farmers’ land without confirmation that they have been informed, and we do not fly in protected areas without forest department permission.
Frequently asked questions
Do you provide drone spraying? No. We are a survey company. Spraying is a separate business with separate aircraft, separate regulation and separate operator certification.
Can you tell us the yield? No. Yield estimation from imagery is a model output requiring calibration for the crop, variety, region and season. We provide the imagery, indices and ground truth that feed such a model. We do not put a yield figure on a report.
Can you produce NDVI? True NDVI requires a near-infrared band, so it requires a multispectral sensor. Visible-band indices from an RGB camera exist, are weaker and are not interchangeable with NDVI. We will tell you which sensor is being flown and which index it supports.
Can a drone survey support a PMFBY claim? It can supply dated, georeferenced, high-resolution evidence of sown area and physically visible damage, which is particularly useful for localised calamities and for disputed assessments. Loss determination itself follows the scheme’s defined methodology through designated agencies. We supply evidence into that process rather than replacing it.
Why not just use satellite imagery? For wide-area, repeated, historical coverage, satellite is the right tool and cheaper. Drone wins on cloud, which matters enormously in kharif, on resolution for plot and plant level questions, on flying when you need it rather than when the satellite passes, and on localised events too small to register at satellite scale. A good programme uses both.
How small a plot can you survey? Technically, very small. Economically, small single plots carry the full mobilisation and control cost and are often better served by a ground measurement. Clusters are where drone survey becomes cost-effective, and we will tell you which side of that line your job falls on.
When should the survey be flown? At the crop stage the deliverable requires, which is usually a window of a few weeks. Tell us the crop, the sowing window and what you need to see, and we will identify the window and build the schedule around it rather than around our calendar.
Can you count trees in a plantation? Yes, on plantations with visible spacing, which covers most young compensatory afforestation and orchard planting. We deliver counted positions, totals per block and gaps marked. On closed-canopy mature forest where crowns merge, counting becomes unreliable and we will say so.
Do you need the farmers’ permission? Yes. Where we fly farmers’ land at an institution’s instruction, we require confirmation that the farmers have been informed and have consented.
What file formats do you deliver? Orthomosaic as GeoTIFF and JPEG. Plot, crop, count and damage layers as SHP, KML and DXF. Terrain as GeoTIFF. Contours as DXF and SHP. Index maps as GeoTIFF. Tabular output as XLSX. Reports as PDF.
Get a quote
Send us the area as a KML or shapefile, the crop, the sowing window, and what the survey needs to establish. Tell us whether the work is a single visit or a series across the season, and whether it is one holding or a cluster.
We will identify the survey window against the crop calendar, return the airspace position, and quote the programme with the sensor and resolution stated.