Drone River Survey Services in India

Geocartis surveys rivers across India for mining lessees, district authorities, water resources departments, riverfront and infrastructure projects and their consultants. We cover exposed riverbed and bar volumes, sand mining replenishment surveys on the statutory cycle, bank erosion and channel change, floodplain terrain models, and riverfront and structure survey. Capture is at 5 cm ground sample distance or better, tied to DGPS control on stable ground, with an RMSE report against every dataset. We are based in Ahmedabad and work nationally.

We are a survey and geospatial services company. We do not build drones and we do not sell software.

What an aerial survey can and cannot measure on a river

This needs saying at the top, because it determines everything else.

A camera records reflected light from a surface. On water it records the surface, or a scatter of sun glint and sky reflection, and nothing below. No processing step recovers bed level from an aerial photograph of a flowing river.

So a drone survey of a river gives you, accurately: the exposed bed, the sand bars and islands, the banks, the berms, the floodplain, the vegetation line, the structures, the channel planform at the moment of capture, and the water surface elevation at the edge. It does not give you the submerged bed.

For most Indian river work, that is enough more often than people expect, because the work that matters happens in the lean season when the channel is at its narrowest and the bed is at its most exposed. Sand mining, in particular, happens on the parts of the bed that are dry.

Where the submerged channel is genuinely needed, the answer is bathymetry with an echo sounder from a boat, merged with the aerial survey of everything above water. Bathymetric LiDAR exists and performs poorly in the turbidity typical of Indian rivers, so we do not recommend it as a general solution. Where a project needs full topo-bathy, we say so and we say who should do the wet half, rather than delivering an interpolated bed and letting somebody design against it.

One thing every river survey report should state and most do not. The water level on the day. On a regulated river, releases from an upstream dam or barrage can change the water surface by a metre between one week and the next, which changes how much bed is exposed and therefore how much surface got measured. Two surveys of the same reach at different water levels are not comparable unless the levels are recorded. Every survey we deliver states the water surface elevation at the time of capture and maps the water edge as a polygon, so a later comparison can be done honestly.

Sand mining: the replenishment study cycle

For most commercial river survey work in India, this is the market.

MoEFCC issued the Sustainable Sand Mining Management Guidelines in 2016 and the Enforcement and Monitoring Guidelines for Sand Mining in 2020. Together they turned sand mining monitoring from an inspection exercise into a measured one, and they name the technology. The 2020 guidelines direct states to conduct replenishment studies of riverbeds and to monitor mining with drones, aerial surveys and ground surveys, and recommend drone technology for illegal mining detection, reserve estimation, quantity estimation and land use monitoring.

The replenishment study has a defined survey cycle, and it is the reason a sand mining lease is a recurring survey client rather than a one-off.

Year one requires four surveys.

The first in April, recording the level of the mining lease before the monsoon.

The second at the closing of mines for the monsoon season, which gives the quantity excavated before the monsoon arrives.

The third after the monsoon, which gives the quantum of material deposited and replenished in the lease.

The fourth at the end of March, which gives the quantity excavated during the financial year.

Subsequent years require three surveys. The results across years are what allow the state to establish the replenishment rate for that reach, which is the number the whole permitting regime rests on.

The guidelines also cap riverbed extraction. Mineable material per hectare available for actual mining is not to exceed 60,000 metric tonnes per annum. And the District Survey Report, required under the 2016 guidelines before a lease or letter of intent is granted, identifies areas of aggradation and deposition where mining can be allowed, areas of erosion and proximity to structures where it must be prohibited, and the annual replenishment rate.

Four numbers in that framework come from a survey: pre-monsoon bed level, pre-monsoon excavated quantity, post-monsoon replenished quantity, and annual excavated quantity. All four are volumes, and all four are only as good as the survey that produced them and the control they are tied to.

The control point is the whole game here. A replenishment study is a sequence of differences between surfaces. If each survey sits on its own independently established control, the differences carry the combined error of both surveys plus any datum offset between them, and the replenishment figure becomes noise. We establish permanent control on stable ground outside the active channel at the first survey, mark it properly, and re-occupy it at every subsequent visit for the life of the lease. That is what makes a four-survey year produce three defensible differences rather than three plausible-looking numbers.

On tonnage. The guidelines are written in metric tonnes. A survey measures volume. Converting requires a bulk density for the material in that reach, which varies with grain size, moisture and compaction, and which should come from sampling rather than a standard figure. We deliver volume in cubic metres with the base surface stated, apply your density values where you have them, and show the calculation. Where no density determination exists, we say so rather than inventing precision.

Other river work

Riverfront development. Existing bank geometry, levels, structures, encroachment and the land available for development. Ahmedabad’s Sabarmati riverfront is the best-known project of this type in the country, and similar schemes are now running on rivers through many Indian cities.

Bank erosion and channel change. Repeat surveys registered to the same control show bank retreat, bar migration and planform change. On the Ganga, Brahmaputra and their major tributaries this is a measurement with real consequences for land, settlements and infrastructure. Annual capture builds a record that a single survey cannot.

Floodplain terrain and inundation. Terrain models of the floodplain for flood modelling, inundation visualisation, floodplain zoning support and post-event damage assessment. Rapid capture after a flood gives an accurate picture of extent and damage while the evidence is still on the ground.

River training and protection works. Survey for spurs, revetments, guide bunds, embankments and bank protection, plus progress and as-built measurement during construction.

Structures over and beside the river. Bridges, weirs, barrages, intake works, pump houses and ghats, captured with oblique passes so the vertical faces reconstruct properly. Scour assessment around piers needs bathymetry for the submerged part, and the exposed geometry is measurable from the air.

Dredging and inland waterways support. The above-water half of a dredging survey, and the exposed bed at low water. Channel depth and least available depth work is echo sounder territory.

Encroachment and river land. River land and floodplain boundaries against actual occupation, dated and georeferenced.

How we do it

Step 1: Scoping, the reach and the season

We start with the reach, expressed the way your records express it: chainage along the river, lease boundary, or a defined stretch between two known points.

Then the season, which on river work matters more than on any other survey except canals. The pre-monsoon lean period, broadly April to June across most of India, exposes the maximum bed and is when the statutory April survey falls. The post-monsoon survey has to happen once flows have dropped enough for the bed to be visible and safe. The monsoon itself is not a survey season.

Then the water level question. If a comparison against an earlier survey is part of the scope, we ask for the earlier survey’s water level and control. Where those are unknown, we say at the outset that the comparison will be indicative rather than measured.

Coordinate system is settled before control goes in. Most river work runs on UTM WGS84, and India spans zones 42N to 47N, so a long reach can cross a zone boundary.

Step 2: Control on ground that will still be there next year

River control has a problem no other sector has. The thing you are surveying moves, and it takes your control points with it.

A point established on a sand bar in April may not exist in October. A point on a bank that is actively eroding will be in the water within two seasons. On a braided river the entire channel can shift between epochs.

So control goes on stable ground: high bank above the flood line, road embankments, bridge abutments, permanent structures, land outside the active channel. We accept a longer baseline to reach stable ground rather than placing convenient points that will not survive. On leases where a multi-year replenishment study is planned, the permanent points are chosen for the life of the study and documented with ground-level photographs so a different crew can reoccupy them.

Points are surveyed with calibrated DGPS. Independent checkpoints go in as well, withheld from the bundle adjustment, and they are how we measure accuracy rather than assert it.

Where a reach is wide, control on one bank only produces the same roll error that afflicts any narrow corridor. On a river wide enough to make crossing difficult, that is a logistics problem to solve rather than a corner to cut, and it usually means a longer day.

Step 3: Airspace and permissions

We map the reach against the Digital Sky airspace map before quoting. Green zone up to 120 metres above ground level without prior flight permission, yellow zone with permission and a lower ceiling, red zone not at all. Rivers form state and district boundaries in many places, which means a single reach can involve more than one district administration.

Sand mining leases sit under state minor mineral rules and district administration, and access to a lease usually needs the lessee’s arrangement. On contested reaches, and illegal sand mining is contested in parts of the country, we will not put a crew somewhere the local administration has not been informed. That is a safety position, not a legal formality.

Our aircraft carry active Unique Identification Numbers and our pilots hold DGCA Remote Pilot Certificates.

Step 4: Flight planning

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.

River blocks are wide and shallow rather than narrow and long, which is better geometry than a canal, and they bring their own planning problems.

Sand is the difficulty. A large expanse of clean dry sand is close to the worst case for tie point matching: low texture, uniform colour, and a surface whose appearance changes with viewing angle. Long sand bars can reconstruct poorly or fail to reconstruct at all. We manage that with higher overlap over bar areas, and by placing additional temporary targets across large featureless stretches. On a lease that is mostly open sand, that target placement is a meaningful share of the field day and it is in the quotation.

Water gets masked rather than matched. We map the water edge as a polygon and exclude the wetted area from the terrain model, so the deliverable shows measured ground and a clearly marked boundary rather than a plausible surface that is actually noise.

Structures get oblique passes.

Step 5: Capture

Middle of the day for sun angle, with one river-specific caveat. Sun glint off water at certain angles saturates the sensor and can wash out a whole strip, so the flight direction relative to the sun matters more here than on dry ground.

Flying over water carries a risk that flying over land does not: a drone that comes down on a sand bar is recoverable and one that comes down in the channel usually is not. Flight lines over open water are planned to minimise exposure, and battery reserves are set conservatively.

Access to bars and islands is the practical constraint. Reaching a mid-channel bar to place a target or a control point may need a boat, and that has to be arranged rather than assumed.

The crew keeps a flight log, and on river work it also records water level, weather and flow condition, because those are part of what makes the survey interpretable later.

Step 6: Processing

Aerial triangulation with bundle adjustment against the surveyed control and camera self-calibration. Dense matching, then classification.

River classification has two hard cases. Riparian vegetation, which on many Indian rivers forms dense thickets on bars and banks, sits above ground level and gets classified inconsistently by automatic filters. And the transition from dry bed to wet bed to open water is gradual, so the wetted boundary needs to be set deliberately rather than left to the algorithm.

Where a comparison against a previous epoch is part of the scope, both surfaces are clipped to the common measured area before differencing. Differencing a surface that was dry last time against one that is wet this time produces a volume change that is entirely an artefact of water level, and it is the most common error in river volume reporting.

Step 7: Volumes and reporting

Volumes are computed against the agreed base surface, and the base is stated. For replenishment work the base is normally the previous epoch, which is why control continuity matters so much. For reserve estimation the base may be a defined level from the District Survey Report.

The report states what was measured, what area was excluded and why, the water level at capture, the base surface used, the checkpoint residuals and the software. On a replenishment survey it also states the extent of common measured area between the two epochs, because that is what the difference actually covers.

Step 8: Quality control and delivery

Residuals at withheld checkpoints, RMSE reported horizontally and vertically. Deliverables issued by epoch with projection metadata embedded in every spatial file, structured so the sequence across a study year reads as a series rather than four unrelated jobs.

Accuracy on a river

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. Your figures come from your checkpoints.

Three things matter more on rivers than elsewhere.

Systematic error dominates volume work. Random noise across a large sandy surface largely cancels when integrated into a volume. A systematic vertical bias does not cancel, it multiplies by the area. A 15 centimetre bias across a 20 hectare lease is 30,000 cubic metres of apparent material. On a replenishment study feeding a regulatory determination, that is the number that has to be defensible.

Control continuity beats single-survey accuracy. For a difference between two epochs, what matters is that both sit on the same frame. Two surveys each at 12 centimetre vertical RMSE on shared permanent control give a far better difference than two surveys each at 8 centimetres on independent control.

Sand degrades matching. Featureless bar surfaces produce noisier point clouds than textured ground, and the effect is worst in the middle of large open stretches. Additional targets are the fix, and we plan them rather than discovering the problem in processing.

Vegetation on bars and banks prevents the ground being seen, so the surface there is interpolation and we mark it. Water is excluded rather than guessed.

Deliverables and output file formats

Raw imagery is handed over with the deliverables. On a multi-year replenishment study where a regulator may look back at the series, holding the raw data yourself matters.

Where we work

We are based in Ahmedabad and mobilise nationally.

Gujarat. The Sabarmati through Ahmedabad, including the riverfront reaches, plus the Mahi, Narmada, Tapi, Banas, Saraswati, Vishwamitri, Bhadar and Shetrunji. Sand and minor mineral leases across Sabarkantha, Aravalli, Mahisagar, Panchmahal, Chhota Udepur, Narmada, Bharuch, Surat, Banaskantha and Patan.

Rajasthan and the north-west. The Chambal, Banas, Luni and their tributaries, with access via Jaipur, Kota, Udaipur and Jodhpur.

The Gangetic plain. The Ganga, Yamuna, Ghaghara, Gandak, Sone, Ken and Betwa, across Uttar Pradesh, Bihar and Madhya Pradesh, with access via Lucknow, Kanpur, Prayagraj, Varanasi, Patna, Bhopal and Gwalior. This is the largest sand mining market in the country and the one with the most active enforcement attention.

Punjab, Haryana and the Himalayan foothills. The Sutlej, Beas, Ravi, Ghaggar and Yamuna, and the foothill rivers of Himachal Pradesh and Uttarakhand, via Chandigarh, Ludhiana, Dehradun and Haridwar.

Central and western India. The Narmada, Tapi, Godavari, Wainganga, Mahanadi and their tributaries, via Indore, Jabalpur, Raipur, Nagpur, Nashik and Pune.

Eastern and north-eastern India. The Damodar, Subarnarekha, Mahanadi, Baitarani and Brahmani, and the Brahmaputra and Barak systems where bank erosion measurement is a substantial requirement in its own right, via Kolkata, Bhubaneswar, Ranchi, Jamshedpur and Guwahati.

Southern India. The Krishna, Godavari, Cauvery, Tungabhadra, Penna and Periyar, via Hyderabad, Vijayawada, Bengaluru, Chennai, Tiruchirappalli and Kochi.

What we do not do

We do not measure the submerged bed. No aerial camera does. Echo sounder bathymetry is the method, and it is a separate scope with a separate provider.

We do not conduct surveillance. The 2020 guidelines suggest night-vision drones for watching sand mining sites. That is an enforcement activity, it is not measurement, and it is not what we sell. We produce the dated, georeferenced record that shows what was extracted and where. What an authority does with it is their function.

We do not determine lease boundaries or river land boundaries. We map physical working extent against the boundary you supply.

We do not determine bulk density or convert volume to tonnage on an assumed figure. Density comes from sampling.

We do not see through riparian vegetation. Thickly vegetated bars and banks need LiDAR or ground survey.

We do not survey during the monsoon, and we do not fly in red zones.

Frequently asked questions

Does the law require a drone survey for sand mining? The Enforcement and Monitoring Guidelines for Sand Mining 2020 direct states to conduct replenishment studies of riverbeds and to monitor mining using drones, aerial surveys and ground surveys, and recommend drone technology for illegal mining detection, reserve estimation, quantity estimation and land use monitoring. The specific requirement that applies to your lease comes from your state’s minor mineral rules and your environmental clearance conditions.

How many surveys does a replenishment study need? Four in the first year: April before the monsoon, at closure of mining for the monsoon, after the monsoon, and at the end of March. Three in subsequent years. The series across years is what establishes the replenishment rate for the reach.

What is the extraction limit? The 2020 guidelines state that mineable material per hectare available for actual mining in a riverbed shall not exceed 60,000 metric tonnes per annum. Check your specific clearance conditions, which may be tighter.

Can you measure the bed under water? No. Photogrammetry measures exposed surfaces only. We survey the dry bed, the bars and everything above the water line, record the water level and map the water edge. Submerged bed level requires an echo sounder.

Why does the water level matter so much? Because it determines how much bed was exposed and therefore measured. Two surveys at different water levels cover different areas, and differencing them without accounting for that produces a volume change that is partly or entirely an artefact. We record the level, map the edge, and clip both epochs to the common measured area before differencing.

Can you compare against a survey somebody else did? Only meaningfully if their control and water level were recorded. Where they were not, the datasets can be overlaid but the volume difference cannot be relied on, and we will tell you that rather than producing a number that looks authoritative and is not.

How accurate is it? With proper control, expect horizontal RMSE around one to two times the GSD and vertical around two to three times. At 5 cm GSD that is roughly 5 to 10 centimetres horizontal and 10 to 15 centimetres vertical, measured from withheld checkpoints. For replenishment work, control continuity between epochs matters more than the single-survey figure.

Can you give us tonnage? We give volume in cubic metres with the base surface stated. Tonnage requires a bulk density from sampling of the material in that reach. We will apply your density figures and show the calculation.

What file formats do you deliver? Orthomosaic as GeoTIFF and JPEG. DTM and DSM as GeoTIFF. Contours as DXF and SHP. Cross sections and L-sections as DXF, DWG, PDF and XLSX. Point cloud as LAS and LAZ. Water edge, planform, erosion and lease layers as SHP. Volume and replenishment reports as PDF and XLSX.

Can you survey a reach that crosses a state boundary? Yes, with the airspace and administrative permissions on both sides. Rivers form boundaries in many places and that means more than one district administration, so allow longer for permissions on those reaches.

Get a quote for your reach

Send us the reach or lease boundary as a KML or shapefile, the deliverables you need, and whether this is a one-off survey or part of a replenishment study series. Tell us the season you need it in, and if there is an earlier survey, whether its control and water level were recorded.

We will return the airspace classification, the control plan including permanent points for a multi-year study, a delivery schedule and a fixed price. For a replenishment study we will quote the survey series for the year rather than each visit separately, because that is how it should be budgeted.