Canary Islands Underground Water Leak Detection Using Aerial Thermography

Underground water leak detection presents a fundamentally different challenge to traditional building or electrical thermography. Unlike exposed systems, buried potable water infrastructure cannot be directly observed and does not inherently generate strong thermal contrast at the surface. In this case, the site was located within a paved urban environment underlain by volcanic substrates typical of the Canary Islands. These conditions introduce several complicating factors:

Without careful control, these variables can easily produce misleading thermal artefacts. As a result, the project was approached with a strong emphasis on governance, protocol development, and cautious interpretation.

Thermography does not detect water directly. Instead, it measures surface temperature distribution, which may be influenced by subsurface conditions such as moisture content, altered thermal inertia, or disturbed ground associated with leakage. When water escapes from a buried pipe, it can modify how surrounding materials absorb, store, and release heat. Moisture-laden materials typically warm and cool more slowly than dry equivalents. In paved environments, this can manifest as subtle, spatially coherent thermal patterns rather than dramatic hot or cold spots.

Crucially, these effects are indirect and comparative. Thermography cannot confirm pipe integrity, leak rate, flow direction, or volume. Its value lies in supporting prioritisation and guiding further investigation when applied within known constraints.

For this project, the objective was therefore defined as follows:

Phase 1 was deliberately conducted at a known leak location rather than as a blind search. This decision was fundamental to the success of the project.

By working at a site with confirmed surface water emergence and repair activity, observed thermal behaviour could be compared directly against reality. This allowed the methodology itself to be tested, refined, and validated without the pressure to “find something” where nothing may exist. This approach also enabled disciplined learning. Features that appeared anomalous at first could be re-evaluated, downgraded, or excluded as evidence accumulated, reinforcing correct professional behaviour rather than confirmation bias.

A structured data collection protocol was developed before any analysis commenced. Given the qualitative nature of thermography and the subtle signals expected, ad-hoc or opportunistic capture was considered inappropriate.

Key elements of the protocol included:

The protocol was applied consistently across all flights, enabling meaningful comparison between datasets and reducing interpretive uncertainty.

Two survey windows were selected to capture different stages of the surface thermal cycle.

Dusk surveys were conducted after daytime solar loading to observe cooling behaviour and potential evaporative effects. While visually striking, these datasets proved sensitive to recent surface activity, shading, and transient conditions.

Dawn surveys, conducted before significant solar loading, produced the most interpretable results. Under near-equilibrium conditions, subtle thermal differences associated with retained heat in moisture-affected sub-base materials became clearer and more stable.

This finding was one of the most significant outcomes of Phase 1 and directly informs best practice moving forward.

Surveys were conducted at three altitudes to assess the impact of ground sample distance on interpretability. High-altitude data provided useful context but lacked the spatial resolution required to resolve small-scale surface behaviour associated with leakage. Mid-altitude data offered improved clarity but remained limited for diagnostic purposes. Low-altitude surveys, achieving a ground sample distance of approximately 3 cm, were essential. At this resolution, localised anomalies, plume behaviour, and subtle gradients could be evaluated with confidence.

This confirmed that spatial resolution is a critical limiting factor in aerial thermographic leak assessment.

All thermal data were processed using a controlled radiometric workflow. Environmental parameters recorded during capture were applied during post-processing to ensure consistency. Thermal spans were selected separately for dawn and dusk datasets to reflect different surface conditions. A single colour palette was used consistently within each dataset to avoid visual bias. RGB imagery was reviewed alongside thermal data at every stage. This proved essential for identifying and excluding confounding factors such as surface reinstatement patches, kerb geometry, staining, shading, and street furniture. Selected datasets were processed into thermal and RGB orthomosaics using WebODM. This allowed spatial comparison across the site and provided an accessible visual tool for stakeholder review.

At the confirmed surface water breach, thermography revealed a repeatable localised warm anomaly during dawn surveys. This anomaly was spatially coherent, distinct from surrounding asphalt, and aligned precisely with visible surface water emergence and staining. Adjacent to the breach, a diffuse plume-like thermal feature was observed. Line profile analysis demonstrated a stepped temperature drop followed by shallow thermal decay, consistent with retained heat in moisture-affected near-surface materials rather than transient surface heating. These findings were consistent across multiple datasets and supported by RGB evidence, providing a high degree of interpretive confidence at this location.

Elsewhere on site, several linear thermal features were identified crossing the road surface. Initial inspection suggested potential subsurface anomalies.

However, correlation with RGB imagery revealed surface indentations and manhole covers aligned with these features. The thermal signatures were stable, narrow, and lacked plume behaviour or surface water expression. These characteristics were consistent with service ducts or trench backfill zones rather than active leakage. Reclassifying these features as baseline infrastructure signatures was a critical step in avoiding false positives and reinforced the importance of contextual interpretation.

Adjacent open land and vegetated areas were reviewed as part of the Phase 1 assessment. Thermal behaviour in these zones was dominated by vegetation structure, shading, and heterogeneous ground materials.

No repeatable patterns consistent with underground water leakage were identified. As a result, these areas were classified as non-diagnostic for this application. This finding reinforces that paved service corridors offer the most reliable medium for thermographic leak assessment in environments of this type.

Phase 1 confirmed that aerial thermography can contribute meaningful decision-support information at known or suspected underground water leak locations when applied correctly. It also clearly demonstrated the limitations of the technique. Thermography does not confirm leaks in isolation and cannot replace established detection methods. Its value lies in screening, prioritisation, and guiding targeted investigation when integrated with infrastructure knowledge and engineering judgement.