Geothermal resource surface heat anomaly mapping for power siting
Thermal infrared satellites can flag surface heat anomalies linked to geothermal activity before a single borehole is drilled. This page explains what Landsat 8/9 TIRS, ASTER, and ECOSTRESS actually resolve, and where the physics stops.
Sensors
- Landsat 8/9 TIRS: Two thermal bands centred near 10.9 µm and 12.0 µm, 100 m native resolution (resampled to 30 m in products), 16-day revisit per satellite, 8-day combined. Suitable for landscape-scale anomaly mapping and multi-decadal time series back to 1982 via earlier Landsat TM/ETM+ thermal bands.
- ASTER TIR: Five thermal bands from 8.125 to 11.65 µm at 90 m resolution. Superior emissivity separation via the Temperature-Emissivity Separation (TES) algorithm. On-demand tasking rather than systematic coverage; archive extends to 1999. Particularly useful over hydrothermally altered ground where emissivity varies strongly with mineralogy.
- ECOSTRESS: ISS-mounted thermal radiometer, five TIR bands at approximately 70 m resolution, irregular revisit (roughly 1–5 days at mid-latitudes depending on ISS orbit precession). Captures diurnal temperature cycles missed by sun-synchronous sensors, which matters for separating thermal inertia signals from surface heat flux.
- SRTM DEM: 30 m global digital elevation model used to model adiabatic lapse-rate cooling (approximately 6.5 °C per 1,000 m) so that elevation-driven temperature gradients are removed before anomaly detection. Without this step, ridge-top cold signatures and valley warm-pooling contaminate the geothermal signal.
What a thermal pixel is actually measuring
Satellite thermal sensors record upwelling radiance from the top few millimetres of the land surface. That radiance is a function of both temperature and emissivity, and the two are entangled. Retrieving land-surface temperature (LST) requires either an assumed emissivity from a land-cover map or a multi-band retrieval such as the split-window algorithm used with Landsat TIRS or the TES algorithm applied to ASTER's five TIR bands. Errors in emissivity of 0.01 translate to LST errors of roughly 0.5–1 °C, which is material when geothermal anomalies of interest may be only 2–5 °C above background.
The honest implication: satellite thermal data is a surface expression sensor. It cannot measure reservoir temperature, fluid chemistry, or permeability. A bright thermal anomaly in a pixel could reflect a fumarolic field, a shallow hydrothermal system, a sun-facing bare rock slope, or a recently burned patch of ground. Distinguishing these requires ancillary data and careful baseline construction, not just a single image.
Building a defensible anomaly baseline
A single night-time Landsat scene is not an anomaly map. Genuine geothermal heat flux is a persistent signal; solar heating, seasonal vegetation change, and soil moisture variation are transient. The standard approach is to build a multi-year seasonal baseline from the full Landsat archive, then flag pixels that consistently exceed the local percentile threshold across seasons. Night-time acquisitions are strongly preferred because they suppress solar-loading artefacts, though Landsat's fixed 10:00 local overpass time means daytime scenes are the norm and night-time coverage is sparse outside high-latitude winter.
ECOSTRESS partially fills that gap. Its non-sun-synchronous orbit captures pre-dawn passes over the same site multiple times per month, allowing diurnal thermal inertia analysis. High thermal inertia materials (wet soil, water-saturated hydrothermal ground) cool slowly overnight; low inertia materials (dry rock, ash) cool fast. A pixel that is anomalously warm at 03:00 local time is a much stronger candidate than one that is warm only at midday.
SRTM-derived lapse-rate correction is applied before anomaly scoring. The procedure is straightforward: regress LST against elevation across a reference zone with no known geothermal activity, subtract the fitted lapse rate, and work with the residual. Residuals persistently above roughly 2–3 °C across multiple seasons warrant field investigation. Below that threshold, emissivity uncertainty alone can explain the signal.
Hydrothermal alteration as a corroborating signal
Heat is not the only surface expression of a geothermal system. Hydrothermal fluids alter host rock mineralogy, producing clay minerals (kaolinite, alunite, smectite) and iron oxides that have distinctive spectral signatures in shortwave infrared. ASTER's SWIR bands (six bands from 1.6 to 2.43 µm, 30 m resolution) and Landsat 8/9 OLI bands 6 and 7 can map these alteration assemblages. Overlaying a TIR-derived thermal anomaly with a SWIR-derived alteration map significantly reduces false positives: a pixel that is both thermally anomalous and spectrally altered is a much more compelling prospect than either signal alone.
This cross-sensor approach is well established in the published literature on geothermal exploration in Iceland, the East African Rift, and the Taupo Volcanic Zone. It does not require proprietary data; the method runs entirely on open Landsat and ASTER archives.
Where the method fails, and what to do about it
Dense vegetation is the most common masking problem. A closed forest canopy with leaf area index above roughly 3 effectively blocks the thermal signal from the ground beneath. LST in that case reflects canopy temperature, not surface heat flux. Partial workarounds include targeting naturally unvegetated ground (fumarolic kill zones, bare alteration patches) or using SAR coherence to identify areas of sparse canopy where thermal retrieval is less compromised.
Cloud cover is a hard limit. Thermal infrared cannot penetrate cloud. In persistently cloudy tropical geothermal regions (Indonesia, parts of Central America, East Africa), a single cloud-free scene may take months to acquire. Compositing over a full year of Landsat acquisitions is often necessary, which means the anomaly map reflects an annual mean rather than any instantaneous state.
Emissivity uncertainty over hydrothermally altered ground is higher than over vegetated or water surfaces because alteration mineralogy is spatially heterogeneous and poorly captured by standard land-cover emissivity libraries. ASTER TES reduces this uncertainty relative to Landsat split-window, but even TES carries residual errors of 0.5–1.5% in emissivity over complex surfaces. Analysts should report LST anomaly confidence intervals, not point estimates.
Finally, satellite thermal data says nothing about reservoir depth, temperature at depth, or producible enthalpy. A strong surface anomaly over a thin, low-enthalpy system is commercially worthless. The satellite screen narrows the field for ground surveys; it does not replace magnetotelluric profiling, gravity surveys, or exploratory drilling.
From pixel to prospect: a practical screening workflow
A first-pass screen typically proceeds in four stages. First, compile all available cloud-free Landsat 8/9 night-time and ECOSTRESS pre-dawn scenes over the target area. Second, apply split-window LST retrieval, SRTM lapse-rate correction, and seasonal baseline subtraction to produce a persistent anomaly score per pixel. Third, overlay ASTER or Landsat SWIR-derived alteration maps to identify pixels with both thermal and mineralogical signatures. Fourth, rank candidate zones by anomaly magnitude, spatial coherence (isolated single pixels are suspect; clusters of tens to hundreds of pixels are more credible), and proximity to mapped faults or volcanic structures from published geological surveys.
The output is a prioritised prospect list with associated uncertainty flags, not a resource estimate. It is the kind of deliverable that justifies a targeted airborne survey or a handful of shallow temperature-gradient boreholes, rather than a full drilling campaign into the unknown. That is the appropriate claim for satellite thermal screening, and it is a genuinely useful one: a 50 km² prospect list costs orders of magnitude less to assemble from satellite data than from ground surveys alone.
Satellize runs this workflow on open Landsat, ASTER, and ECOSTRESS archives, adding commercial high-resolution tasking where a client holds the appropriate licence. The analytics stack is the same one underpinning the Tonga crop-estimation programme: open-data ingestion, rigorous baseline construction, and outputs formatted for GIS integration rather than internal consumption.
What satellite data cannot tell a developer
It is worth being direct about the ceiling. No satellite sensor in current operation can measure subsurface temperature, fluid flow rate, dissolved gas content, or the structural permeability that determines whether a geothermal resource is commercially exploitable. The International Energy Agency and national geological surveys are consistent on this point: remote sensing is a reconnaissance tool, not a resource characterisation tool.
The practical value is in reducing the search space before committing to expensive ground campaigns. A well-constructed satellite thermal anomaly screen can eliminate large areas of obviously cold, unaltered ground and focus attention on a small number of coherent, persistent anomalies worth investigating further. That is a real saving, particularly in remote or politically complex terrain where mobilising ground teams is itself costly. The satellite data earns its place in the workflow by being cheap, repeatable, and archive-backed, not by pretending to see deeper than physics allows.
Typical figures
| Spatial resolution (thermal) | 70 m (ECOSTRESS), 90 m (ASTER TIR), 100 m native / 30 m resampled (Landsat 8/9 TIRS) |
| Revisit cadence | 8 days combined Landsat 8+9; 1–5 days ECOSTRESS (ISS-dependent); ASTER on-demand tasking only |
| Thermal bands | Landsat TIRS: 10.6–11.2 µm and 11.5–12.5 µm; ASTER TIR: 5 bands 8.125–11.65 µm; ECOSTRESS: 5 bands 8.28–12.13 µm |
| LST retrieval accuracy (typical) | ±1–2 °C over homogeneous surfaces with known emissivity; ±2–4 °C over hydrothermally altered or mixed-cover terrain |
| Minimum detectable anomaly (practical) | Approximately 2–3 °C above seasonal baseline after lapse-rate correction; sub-2 °C anomalies are within emissivity uncertainty |
| Elevation correction data | SRTM 1-arc-second (approximately 30 m) global DEM; lapse rate typically 5.5–6.5 °C per 1,000 m applied as linear regression residual |
| Archive depth | Landsat thermal back to 1984 (TM Band 6); ASTER from 1999; ECOSTRESS from 2018 |
| Cloud penetration | None. TIR is blocked by cloud; compositing over 12+ months required in persistently cloudy regions |
| Alteration mapping bands | ASTER SWIR 6 bands 1.6–2.43 µm at 30 m; Landsat OLI Bands 6–7 at 30 m for clay and iron-oxide indices |
| Deliverable formats | GeoTIFF LST anomaly grids, GeoPackage prospect polygons, PDF ranked prospect report |
Analytics Satellize can run
| Land-surface temperature anomaly map | Split-window LST retrieval (Landsat TIRS) or Temperature-Emissivity Separation (ASTER TIR); multi-year seasonal baseline subtraction; SRTM lapse-rate correction | GeoTIFF raster of persistent thermal anomaly score (°C above baseline) at 30–100 m resolution, with per-pixel uncertainty band |
| Ranked geothermal prospect polygons | Spatial clustering of pixels exceeding 2 °C anomaly threshold across at least 60% of cloud-free acquisitions; scored by anomaly magnitude, cluster area, and fault proximity from published geological maps | GeoPackage vector layer with prospect scores and uncertainty flags, ready for GIS overlay with ground survey data |
| Hydrothermal alteration index map | Band-ratio and spectral mixture analysis on ASTER SWIR or Landsat OLI Bands 6–7 to map kaolinite, alunite, and iron-oxide assemblages consistent with acid-sulphate or propylitic alteration | GeoTIFF alteration probability layer co-registered to thermal anomaly map; joint thermal-alteration prospect ranking table |
| Diurnal thermal inertia analysis | Comparison of ECOSTRESS pre-dawn and afternoon LST retrievals; apparent thermal inertia (ATI) computed from temperature amplitude and albedo; high-ATI zones flagged as consistent with fluid-saturated hydrothermal ground | ATI raster and summary statistics per prospect polygon; PDF interpretation note |
| Multi-decadal anomaly trend report | Mann-Kendall trend test on annual maximum LST anomaly per prospect polygon across Landsat archive (1984 to present); flags intensifying or declining surface expressions | Time-series plots per prospect and tabular trend statistics; supports environmental baseline documentation for permitting |
| First-pass prospect prioritisation report | Composite scoring of thermal anomaly persistence, alteration index, topographic setting, and published fault/volcanic proximity; calibrated against publicly documented geothermal fields (e.g. Olkaria, Wairakei) used as spectral-thermal reference sites | Ranked PDF report with prospect maps, uncertainty discussion, and recommended ground-survey priorities |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.