A river can change significantly between two conventional field surveys. Sediment accumulates, banks erode, gravel bars move, and the geometry around bridges and embankments gradually shifts. During a major rainfall event, those changes can affect how quickly water moves through the channel and where it starts to spill over.
This article looks at how UAV surveys can help monitor these changes and turn aerial data into practical information for river management. We will cover how drones can be used to map channel geometry, track sediment accumulation and erosion, assess dams, embankments and bridges, and provide data for hydraulic modelling and flood-risk assessment.
Sediment deposition is easy to notice after a flood, especially when new gravel bars or deposits appear along the channel. The more important question is what those deposits do to the hydraulic behavior of the river.
As material accumulates, the shape and elevation of the riverbed can change. Local sections may become shallower, flow can be redirected around deposited material, and hydraulic resistance can vary with the size and distribution of bed particles.
A useful river survey therefore needs more than a visual record of where sediment is present. It needs to describe the geometry of the channel and, where possible, quantify characteristics of the riverbed.
This approach was demonstrated in a river-channel study published in 2026, where researchers combined UAV imagery with deep learning to analyze exposed gravel across the riverbed.
Using YOLOv5, they automatically detected individual gravel particles in UAV-derived orthophotos and converted their image dimensions into physical measurements based on the ground sampling distance (GSD). The resulting particle-size data, including D50, D65, D75 and D90, were then used to estimate Manning’s roughness coefficients for different river sections and feed them into subsequent hydraulic modelling.
A useful river-monitoring dataset can contain several layers of information at once.
Photogrammetry can produce a dense point cloud and DSM describing the terrain surface. From this data, engineers can extract river width, bank elevations, longitudinal profiles and cross-sections, as well as identify local changes in channel form.
This is one of the main strengths of UAV photogrammetry for river monitoring. A series of overlapping images can be processed into georeferenced 3D data, making it possible to compare the same reach at different dates rather than relying only on individual field measurements.
The orthophoto can also provide information about exposed gravel and sediment bars. Depending on image resolution and site conditions, this may include:
This becomes particularly valuable when combined with terrain data. A map showing where the surface has risen can be used alongside an orthophoto showing what material is present.
Two DSMs can be compared to identify areas of elevation gain and loss. Positive changes may indicate deposition, while negative changes can point to erosion or excavation. Repeated observations can also reveal lateral bank movement and shifts in the channel itself.
This turns the survey into a change-detection workflow:
Survey 1 → Survey 2 → elevation difference → sedimentation and erosion map
Over several survey cycles, those observations can become a history of how the river reach is evolving.
The same aerial dataset can cover the riverbanks and structures around the channel. Depending on the application and sensor configuration, monitoring may include embankments, levees, bridge approaches, bridge piers and areas where erosion is threatening infrastructure.
The important part is that infrastructure and channel conditions should not always be analyzed separately. A change in one can influence the other.
A river does not need a dramatic blockage to become more vulnerable to flooding. Gradual changes in channel geometry can also affect how much water the river can carry and how water levels respond during high-flow conditions.
A newly formed sediment bar can reduce the effective flow area, while erosion can change the shape of the banks or weaken an already vulnerable embankment. Structures such as bridge piers can further alter local flow patterns and contribute to higher upstream water levels. During extreme rainfall, these relatively small changes can become much more significant.
This is why detailed terrain data matters in hydraulic modelling. A digital surface model captures the physical shape of the channel and surrounding floodplain, providing the geometry needed for flow simulations. The resolution of that model also affects how much local variation the simulation can represent.
A useful example comes from a 2019 heavy-rainfall event analyzed in the Sanxia River watershed in Taiwan. The modelling used UAV-derived terrain data together with rainfall and riverbed roughness parameters, and tested DSM resolutions ranging from 20 m down to 0.5 m. The 0.5 m model produced the closest match to observed water levels, with a reported accuracy of 95.39%. The accuracy decreased to 94.84% with a 1 m DSM, 92.33% with 5 m, and 87.20% with 20 m data.
There was, however, a clear trade-off between detail and processing time. The 20 m model required 65 seconds to run, compared with 805 seconds for the 0.5 m dataset. At the same time, the overall simulated flood areas were relatively close: compared with the 0.5 m reference, the difference was 1.03% for 1 m, 1.44% for 5 m and 1.94% for 20 m.
UAVs do not predict floods by themselves. Their role is to keep the physical description of the river up to date.
A practical early-warning workflow can look like this:
Periodic UAV survey
↓
Updated orthophoto + point cloud + DSM
↓
Channel and sediment change detection
↓
Updated roughness and terrain parameters
↓
Hydraulic simulation under forecast rainfall or discharge
↓
Flood-risk areas and warning thresholds
This distinction matters. A flood model is only as useful as its representation of the terrain, channel geometry and other relevant hydraulic parameters. When the river changes, the model inputs can become outdated.
The resulting concept is closer to continuous river condition monitoring than to a one-time aerial survey.
The right UAV setup depends on what needs to be measured. For detailed river-channel mapping, LiDAR is useful for building a 3D terrain model, while high-resolution RGB imagery adds the visual detail needed to study sediment bars, exposed gravel and riverbank changes. A practical workflow can combine either of the following platforms with dedicated mapping payloads and point-cloud processing software.
The Matrice 400 is well suited to large-area surveying, while the Zenmuse L3 combines LiDAR and RGB imaging in one payload. This setup can be used to build dense point clouds and terrain models, extract cross-sections and elevation profiles, and compare repeated surveys to identify erosion or sediment accumulation. The RGB data also provides an additional layer for interpreting exposed gravel and channel features.
JOUAV CW-25 + GreenValley LiAir H800
For long river corridors, a VTOL fixed-wing platform such as the JOUAV CW-25 can cover larger areas in a single mission. Paired with the GreenValley LiAir H800, it can provide airborne LiDAR data for terrain and floodplain mapping. This combination is particularly relevant when the monitoring area extends well beyond a single river reach and surveys need to be repeated over a larger watershed.
Raw UAV data needs to be turned into a usable engineering dataset before it can support change detection or hydraulic modelling. DJI Terra can be used with DJI mapping workflows to process photogrammetric and LiDAR data, while LiDAR360 is a strong option for working with UAV point clouds, terrain models and measurements.
For a monitoring program, the important part is not simply generating a 3D model once. The same survey workflow should be repeatable so that datasets from different dates can be compared reliably. This makes it possible to track changes in channel elevation, sediment deposits, bank geometry and other features that may influence flood risk.
The strongest case for UAV river monitoring is not the aerial image itself. It is the range of engineering information that can be extracted from the same survey.
A well-designed UAV workflow can support:
Channel geometry — DSMs, point clouds, cross-sections and longitudinal profiles.
Sediment monitoring — deposition areas, gravel characteristics and changes in riverbed elevation.
Erosion monitoring — bank movement, local scour and areas of terrain loss.
Infrastructure assessment — embankments, levees, bridge surroundings and other critical structures.
Hydraulic modelling — updated terrain and roughness inputs for flood simulations.
Flood-risk management — identification of changing high-risk reaches and more up-to-date model scenarios.
For river managers, this makes UAVs a practical tool for building a consistent record of river conditions. The goal is not simply to document what the channel looks like today, but to understand how it is changing and how those changes may influence sedimentation, infrastructure condition and flood risk.
Drones collect aerial imagery and LiDAR data for mapping river geometry, sediment, erosion and nearby infrastructure.
Yes. Repeat UAV surveys can show changes in riverbed elevation and map new sediment deposits and gravel bars.
Yes. UAV-derived terrain data can be used to update hydraulic models and assess changes that may affect water levels and flood extent.
LiDAR can produce 3D point clouds for channel geometry, cross-sections, riverbanks and elevation-change analysis.
