Seepage is the quiet failure mode of dams and levees — water moving where it should not, often invisible until the surface slumps. Because moving groundwater holds a different temperature than the surrounding embankment, thermal drone surveys make seepage paths visible from the air, safely and repeatably. Timing beats sensor money on this mission, by a wide margin.
- 8–14 °CTypical stable groundwater band
- Pre-dawn / eveningSurvey windows
- ≤40 mKRecommended NETD
- SeasonalRepeat interval for change detection
Key takeaways
- Groundwater sits near a stable year-round temperature while embankment surfaces track air and sun. That difference — not the water itself — is what you detect.
- Seepage reads cool in summer and warm in winter. The sign flips with season, which is why untimed surveys confuse analysts.
- Thermal locates candidates; it cannot quantify flow. Every finding is a task for the geotechnical team, not a conclusion.
- Vegetation masks surface temperature completely. Pair every thermal finding with zoom imagery of the same station before tasking follow-up.
On this page
The thermal signature of seepage
Groundwater stays near a stable year-round temperature while embankment surfaces track air and sun. Early morning or evening flights maximise that contrast: wet toes, boils and saturated patches render as coherent anomalies against dry fill. The physics parallels roof moisture survey work — timing beats sensor money.
The consequence that catches new analysts out is that the sign of the anomaly changes with season. In summer, groundwater near 10 °C emerging into a 28 °C embankment reads as a cool patch. In winter, the same water emerging into a 2 °C embankment reads warm. An analyst who has learned “seepage is warm” from a winter dataset will miss every summer finding.
What stays constant is coherence. Seepage produces a spatially organised anomaly — a linear feature along a toe, a rounded boil, a plume spreading downslope — while thermal noise and solar artefacts produce scattered or geometrically implausible patterns.
| Season / condition | Water-vs-embankment ΔT | Anomaly reads | Survey quality |
|---|---|---|---|
| Summer, pre-dawn | 12–18 K | Cool | Excellent |
| Summer, midday | 5–10 K | Cool but noisy | Poor — solar artefacts dominate |
| Winter, pre-dawn | 6–12 K | Warm | Very good |
| Winter, snow cover | Melt pattern | Warm / melt line | Excellent |
| Spring / autumn transition | 2–5 K | Ambiguous | Weak — avoid if possible |
| Evening, 2–4 h after sunset | 8–14 K | Season-dependent | Good |
| After heavy rain | Masked | — | Unusable — surface water dominates |
| High wind (>6 m/s) | Reduced 40–60 % | Weakened | Poor |
Spring and autumn are the worst time to survey. Air and groundwater temperatures converge, the differential collapses, and findings become ambiguous. Schedule surveys for mid-summer and mid-winter, when the contrast is at its seasonal maximum.
Survey design for linear embankments
Fly consistent corridors along crest and toe at fixed altitude, radiometric capture on, repeated seasonally so change detection does the diagnosis. This last point carries most of the value.
A single survey shows anomalies, and embankments have many benign thermal features — buried culverts, historical repairs, differing fill materials, vegetation shadow. Distinguishing those from developing seepage on one dataset is guesswork. Comparing this summer against last summer, at the same station, at the same hour, turns guesswork into a trend.
Vegetation masks surface temperature, so pair thermal findings with zoom imagery of the same station before tasking geotechnical follow-up. A warm patch under scrub may be the scrub.
- Fix the corridor geometry. Crest line and toe line, fixed altitude, fixed gimbal angle. Record them so the next survey reproduces them exactly.
- Choose the seasonal window. Mid-summer or mid-winter, pre-dawn, at least 48 h after significant rain.
- Capture radiometrically. Relative contrast is enough to spot an anomaly but not to compare it against last year’s value.
- Range and geolocate each anomaly, then convert to the station numbering the geotechnical team uses.
- Photograph every finding in the visible channel at the same station, to rule out vegetation and surface features.
- Compare against the previous same-season survey before raising anything. New or growing beats merely present.
From anomaly to geotechnical task
Thermal survey produces candidates. It does not confirm seepage, cannot measure flow rate, and says nothing about internal erosion. Programmes that present thermal findings as conclusions damage their own credibility on the first excavation that finds nothing.
The deliverable that works is a ranked candidate list: station number, anomaly type, size, whether it is new or previously recorded, and a visible-light photograph. The geotechnical team decides what to investigate.
Location accuracy matters because embankments are long and featureless. Laser ranging plus a route overlay converts a coordinate into a station the team can walk to — with a stated tolerance, since coordinate error grows with slant range as described in the geolocation error budget.
Cost and why frequency is the real gain
Traditional levee inspection is a walked survey — slow, weather-dependent, and covering a fraction of the network per season. A drone corridor survey covers tens of kilometres per sortie with a consistent, archived thermal record.
The saving is not primarily labour. It is frequency. Walked surveys are expensive enough that many systems are inspected annually or less; aerial surveys are cheap enough to run each season, and seasonal repetition is precisely what makes change detection possible. A finding that is stable across four surveys is a feature; one that has grown across two is a priority.
The avoided cost on the other side needs no argument. Embankment failure is a catastrophic-consequence event, and the entire monitoring programme is small against a single breach.
Where seepage surveys go wrong
- Surveying in spring or autumn. The differential collapses and findings become ambiguous.
- Surveying after rain. Surface water masks the subsurface signal entirely; allow at least 48 hours.
- Expecting a consistent anomaly sign. Seepage reads cool in summer and warm in winter — analysts trained on one season miss the other.
- Non-radiometric capture. Relative contrast cannot be compared against last year, which removes the main value of the programme.
- Reporting thermal findings as confirmed seepage. They are candidates. Overstating them costs credibility on the first dry excavation.
- Changing corridor geometry between surveys. Different altitude or angle makes change detection unreliable.
Recommended UAVThermal payloads
OP-90P — radiometric 640×512 thermal with zoom for combined anomaly detection and visual confirmation on the same pass.
MV-2P — 130 g 640-class thermal at NETD below 40 mK for routine levee patrol on small aircraft, where the marginal transitional-season cases need the sensitivity.
LX-6R — 30x zoom with laser ranging to geolocate findings precisely on long embankments where station referencing is the hard part.
Related reading
Technology: radiometric temperature measurement and how detector sensitivity is specified.
Field practice: thermal sensitivity (NETD) explained, roof moisture surveys — the same thermal-lag physics — and report deliverables.
Adjacent missions: district heating leak detection and railway earthworks and corridor inspection.
FAQ
When is the best time to fly a seepage survey?
Pre-dawn and evening, when solar loading is minimal and the water-versus-embankment temperature delta is largest. Season matters more than sensor brand: mid-summer and mid-winter give the strongest contrast, while spring and autumn transitions collapse it.
Can thermal quantify seepage flow?
No. Thermal locates where water is emerging or saturating the surface; it says nothing about flow rate or internal erosion. Every finding is a candidate for geotechnical investigation, not a conclusion.
Why does seepage sometimes read cool and sometimes warm?
Because groundwater holds a stable temperature while the embankment tracks air and sun. In summer the water is cooler than the fill and reads cool; in winter it is warmer and reads warm. Analysts trained on one season routinely miss findings in the other.
How soon after rain can we survey?
Allow at least 48 hours. Surface water masks subsurface saturation completely, so a survey flown too soon reads the rain rather than the seepage.