Gold Prospectivity Mapping in Tanzania: A Remote Sensing Case Study by Farmonaut

Introduction to Gold Exploration in Tanzania

Tanzania, a country renowned for its significant gold endowments, presents a promising yet challenging landscape for mineral exploration. Traditional methods, reliant on extensive fieldwork, are often time-consuming and costly, particularly in remote and rugged terrains. To address these challenges, Farmonaut, a leader in satellite-based remote sensing solutions, conducted a groundbreaking project to map gold prospectivity across a 10-square-kilometer Area of Interest (AOI) in Tanzania. This case study explores how Farmonaut leveraged advanced remote sensing techniques — integrating optical spectral analysis, thermal remote sensing, synthetic aperture radar (SAR), geomorphic modelling, and magnetic geophysics — to identify and prioritize high-potential gold exploration targets, delivering cost-effective and data-driven insights to guide field investigations.

Keywords: Gold prospectivity mapping, remote sensing, Tanzania gold exploration, satellite imagery, mineral exploration, Farmonaut.

Project Objectives

The primary goal of this project was to conduct a comprehensive remote sensing analysis to systematically identify, characterise, and rank high-potential gold exploration targets within the AOI. By utilising non-invasive satellite data from multiple sources, Farmonaut aimed to:

Process and analyse a multi-year time-series of satellite imagery (2018–2025) across optical, thermal, radar, and geophysical domains to map mineral alteration zones.
Identify key hydrothermal alteration assemblages — including phyllic, argillic, propylitic, and iron oxides — associated with gold deposits, using both Landsat/Sentinel-2 band ratios and ASTER Crósta ratio indices.
Develop a robust Gold Potential Index (GPI) integrating spectral, thermal, SAR, geomorphic, and magnetic geophysical signals with seasonal and temporal weighting for enhanced reliability.
Detect artisanal small-scale mining (ASM) workings and assess placer gold potential from geomorphic and hydrological indicators.
Deliver a prioritised list of exploration targets, ranked by a quantitative scoring system incorporating depth estimates, proxy grade, and structural analysis, to optimise subsequent field efforts.

This approach not only reduces exploration costs but also minimises environmental impact by focusing fieldwork on the most promising areas.

Keywords: Gold Potential Index, hydrothermal alteration, satellite-based mineral exploration, Tanzania gold deposits, artisanal mining detection.

Study Area: Tanzania’s Gold-Rich Region

The AOI, spanning 1,000 hectares in a prospective region of Tanzania, was selected for its geological potential and favourable conditions for remote sensing. The region’s arid to semi-arid climate and sparse vegetation cover provide ideal conditions for analysing surface geology and mineralogy using satellite imagery. Tanzania’s rich gold mining history, coupled with its complex geological structures, makes it a prime candidate for advanced exploration techniques.

The specific location of the AOI was strategically chosen to align with known gold-bearing formations, with regional-scale faults and shear zones suspected as conduits for mineralising fluids. Farmonaut’s analysis leveraged these geological insights to enhance the accuracy of its prospectivity mapping.

Keywords: Tanzania gold mining, geological structures, remote sensing for mineral exploration, arid climate analysis.

Data Sources and Technology

Farmonaut’s analysis integrates seven categories of satellite and ancillary datasets, ensuring a comprehensive multi-sensor view of surface mineralogy, terrain, thermal properties, and structural geology. Where a preferred dataset is unavailable or has insufficient coverage, fallback datasets are applied automatically.

2.1 Optical Imagery

Primary: Landsat 8 OLI / Landsat 9 OLI-2, Collection 2 Tier 1 Level-2 Surface Reflectance (30 m), radiometrically calibrated and atmospherically corrected.
Supplement: Sentinel-2 SR Harmonised (10–20 m, resampled to 30 m) — applied where Landsat coverage is insufficient.
Long observation window: 2018 – present; Recent window: 2023 – present.
Cloud filter: ≤ 30% scene cloud cover to ensure spectral quality.

2.2 Elevation & Terrain

Primary DEM: Copernicus GLO-30 (30 m global digital elevation model).
Fallback DEM: USGS SRTM v3 (30 m).
Derived products: Slope, aspect, multi-directional hillshade (4 azimuths), topographic roughness index (TRI), and residual topography — used to refine anomalies and exclude topographically unsuitable areas.

2.3 Thermal Remote Sensing

ASTER GED (AG100 v3, 100 m): NASA ASTER Global Emissivity Dataset — long-term land-surface temperature used to identify thermal anomalies indicative of hydrothermal activity.
MODIS LST (MOD11A1 v6.1, 1 km): Daily day/night land-surface temperature used to compute the diurnal temperature range anomaly — an indicator of subsurface moisture or lithological variation in bare terrain.

2.4 Synthetic Aperture Radar (SAR)

Sentinel-1 C-band: IW mode, VV + VH polarisations, ascending and descending orbits (5 × 20 m GRD) — used for structural lineament detection and texture analysis.
ALOS PALSAR L-band: JAXA annual mosaic HH + HV (25 m), 2015–2021 — L-band penetrates vegetation more effectively than C-band, providing additional structural heterogeneity information.

2.5 ASTER L1T SWIR

Dataset: ASTER/AST_L1T_003 (15 m VNIR, 30 m SWIR), acquisition window 2000–2008 (pre-detector failure).
Purpose: Crósta ratio indices for argillic, Al-OH/sericite, carbonate/chlorite, and ferric iron alteration zones — providing finer-resolution mineralogical discrimination than Landsat alone.

2.6 Land Cover & Geomorphology

Dynamic World v1 (Google / WRI, 10 m): Near-real-time land cover — bare probability and hard-label persistence used to distinguish geological bare ground from agricultural bare ground.
ESA WorldCover v2 (10 m, 2021): Annual land cover classification — tree, shrub, grassland, cropland, built-up, and bare classes used for surface context assignment.
MERIT Hydro v1: Flow accumulation (km²), flow direction, and stream network — used to derive placer geomorphology indicators including knickpoints, meander traps, and confluences.
Open Buildings v3 (Google building footprints): Used for artisanal-mining exclusion from settlement areas.
LARSE GEDI v2 (25 m): Canopy height and biomass density — proxy for canopy obscuration of spectral and SAR signals.

2.7 Magnetic Geophysics

EMAG2 v3: Earth Magnetic Anomaly Grid (2 arc-minute resolution, ~3.4 km), upward-continued to 4 km altitude — used to detect magnetic anomalies correlated with mineralised zones. When this dataset is unavailable, a topographic-lithology proxy derived from residual topography is substituted.

Keywords: Landsat 8 OLI, Sentinel-2, ASTER SWIR, Sentinel-1 SAR, ALOS PALSAR, SRTM DEM, Copernicus GLO-30, MODIS LST, EMAG2 magnetic anomaly, temporal analysis.

Methodology: A Multi-Stage Integrated Approach

Farmonaut employed a rigorous, multi-stage methodology to map gold prospectivity, fusing optical, thermal, radar, geomorphic, and geophysical data streams into a single composite Gold Prospectivity Index (GPI).

1. Data Pre-Processing

Landsat 8/9 imagery was masked using the QA_PIXEL band to remove cloud, cloud shadow, and cirrus-affected pixels, with surface reflectance and thermal scaling applied per the USGS Collection 2 specification. Sentinel-2 imagery was masked using the QA60 bitmask and harmonised to Landsat through per-band linear correction. Images were labelled dry-season (October–April) or wet-season (May–September) and separate median composites computed for each season across the full observation window. Dry-season composites served as the primary spectral input, as vegetation stress and bare-ground exposure are maximised during this period. SAR pre-processing included separation of Sentinel-1 data by polarisation and orbit direction, calculation of temporal mean and standard deviation mosaics, and conversion of ALOS PALSAR digital numbers to backscatter in dB. Grey-level co-occurrence matrix (GLCM) texture was also computed for both SAR datasets at 30 m resolution.

2. Spectral Analysis

Fourteen spectral band-ratio and normalised-difference indices were computed from the dry-season optical composite to target hydrothermal alteration minerals associated with gold deposits. Key indices included:

Phyllic Alteration: Indicative of sericite and quartz, common in gold systems.
Argillic Alteration: Associated with clay minerals like kaolinite (SWIR1/SWIR2 ratio).
Propylitic Alteration: Linked to epidote and chlorite (SWIR1/Red ratio).
Iron Oxides: Surface weathering products of sulfide minerals (SWIR1/NIR ratio).
Silicification: Indicative of silica-rich zones (SWIR1/SWIR2).
Gossan Index: Targeting gossanous and oxidised zones ((Red + SWIR1) / (NIR + SWIR2)).
Ferric & Ferrous Iron: Separate indices distinguishing oxidised from chloritic iron assemblages.
Hydrothermal Alteration Composite: Broad hydrothermal index ((SWIR1 + SWIR2) / NIR).

Where pre-2008 ASTER L1T SWIR scenes were available, four additional Crósta ratio indices were computed — targeting argillic, Al-OH/sericite, carbonate/chlorite, and ferric iron assemblages — and combined into a weighted ASTER composite (argillic 0.35, Al-OH 0.25, carbonate 0.20, ferric iron 0.20). Vegetation (NDVI) and moisture (NDWI) indices served as masking layers to suppress non-geological signals.

3. Auxiliary Layer Generation

Beyond spectral indices, six auxiliary signal layers were generated to enrich the GPI:

Thermal Anomaly: ASTER GED land-surface temperature and MODIS LST diurnal range were each normalised and averaged into a combined thermal score.
Bare-Ground Analysis: Dynamic World bare-probability layers were decomposed into seasonal variability, geometric irregularity (GLCM contrast), isolation index, and slope position — distinguishing persistent geological exposures from agricultural bare ground.
Geobotanical Anomaly: NDVI z-scores relative to a 500 m Gaussian neighbourhood identified vegetation stress potentially induced by metal toxicity or altered soil chemistry above mineralisation.
Canopy & Laterite Penalty: LARSE GEDI canopy density was used to penalise Sentinel-1 SAR signals and to compute a laterite index (combining Red/Blue ratio, inverted clay index, slope flatness, and PALSAR HV backscatter) that suppresses GPI values where deep weathering may obscure primary mineralisation signatures.
Placer Geomorphology: Three indicators — knickpoints, meander traps, and stream confluences — were derived from the MERIT Hydro dataset and combined into a composite placer score (knickpoints 0.4, meander traps 0.3, confluences 0.3).
Magnetic Geophysics: EMAG2 v3 magnetic anomaly departures were normalised and combined with residual topography at a 0.70/0.30 weighting to produce a structural-geophysical signal layer.

4. Gold Potential Index (GPI) Modelling

Farmonaut developed a pixel-wise weighted GPI by normalising each input signal to [0, 1] and combining them via weighted overlay. Seasonal GPIs were generated with weights of 60% dry season and 40% wet season to account for climatic variations. A temporal stability metric (Coefficient of Variation) prioritised persistent signals over transient anomalies, enhancing target reliability. A surface context classification — seven mutually exclusive classes including bedrock outcrop, regolith, cleared cropland, settlement, pasture, dense vegetation, and artisanal workings — determined per-class GPI floor thresholds, suppressing low-confidence signals in challenging land-cover conditions. A detectability index derived from NDVI, GEDI canopy density, and the laterite penalty was also computed to communicate the confidence level of each GPI pixel explicitly.

5. Target Identification & Artisanal Mining Detection

Using the GPI map, Farmonaut flagged anomalies exceeding per-context GPI floor thresholds. Terrain and land cover masks (NDVI, NDWI, slope, elevation) filtered out false positives caused by vegetation, water bodies, or steep slopes, ensuring only geologically relevant anomalies were retained. Features smaller than approximately 7,200 m² were discarded to eliminate single-pixel noise. In parallel, artisanal small-scale mining (ASM) workings were detected using a conjunctive six-criteria classifier requiring: bare persistence ≥ 0.40, geometric irregularity ≥ 0.50, isolation index ≥ 0.40, slope position ≥ 0.15, building distance ≥ 80 m, and GEDI canopy density ≤ 0.40. Features meeting all six criteria were classified as artisanal mining signals, with confidence graded as High (≥ 5 criteria), Moderate (4), or Low (3).

6. Anomaly Characterisation & Economic Analytics

Raster anomalies were converted into polygons and each polygon analysed individually. A minimum-area oriented bounding box (tested at 18 rotation angles) was fitted to derive geological strike, vein width, and a recommended drill azimuth perpendicular to strike. Lineament density was computed from three sources — topographic gradient, Sentinel-1 VV standard deviation, and Sentinel-1 orbit-direction difference — to quantify structural complexity within a 500 m buffer. A coherence score integrating polygon elongation and smoothness further characterised the likely continuity of each vein or shear zone.

Anomalies were ranked using a custom scoring system that considered:

Size: Larger anomalies indicated broader mineralised zones.
Seasonal Contrast: Stronger signals in both seasons suggested robust targets.
Average GPI: Higher scores reflected intense alteration.
Temporal Stability: Persistent signals increased confidence.
Proxy Grade & Depth: Estimated g/t Au and depth-to-ore incorporating alteration intensity, thermal score, structural lineament density, geobotanical anomaly, and bare persistence — with a 15% mining dilution factor and 0.3 g/t cut-off grade applied.
Placer Classification: Anomalies assessed against flow accumulation, stream proximity, and composite placer geomorphology score, assigned one of seven placer type classifications.

This repeatable methodology delivered a prioritised list of high-confidence exploration targets complete with structural, economic, and placer attributes.

Keywords: Spectral analysis, ASTER Crósta ratios, SAR lineament detection, Gold Potential Index, temporal stability, anomaly ranking, artisanal mining detection, placer geomorphology, remote sensing methodology.

Results: Data-Driven Exploration Targets

Farmonaut’s analytical workflow yielded clear, actionable results, presented through prospectivity maps and a ranked table of anomalies.

Prospectivity Maps

Map 4.1: Final Gold Anomaly Map
This primary deliverable displayed prioritised anomaly polygons — filtered through spectral, thermal, SAR, geomorphic, and terrain criteria — overlaid on a true-colour satellite composite of the AOI. Each polygon is attributed with structural strike, proxy grade, depth estimate, coherence score, artisanal mining signal, and placer classification.
Map 4.2: Combined GPI Heatmap
The GPI heatmap illustrated prospectivity scores across the AOI, with “hot” colours (red, orange) indicating high potential and “cool” colours (blue, green) denoting lower potential. The heatmap integrates all seven data source categories and provides regional context for the final anomaly polygons.

Coordinates of Major Anomalies

These coordinates, accompanied by structural orientation, recommended drill azimuth, proxy grade, and detectability scores, guided precise field targeting, ensuring efficient resource allocation.

Keywords: Gold anomaly map, GPI heatmap, exploration target prioritisation, Tanzania mineral mapping, drill targeting.

Discussion: Insights and Implications

The results provided a robust foundation for advancing gold exploration in the AOI, with significant implications for future efforts.

Interpretation of Results

The non-random distribution of anomalies suggested structural controls on mineralisation, likely along regional faults or shear zones. These geological features, common in gold systems, served as conduits for hydrothermal fluids, enhancing the prospectivity of identified targets. The alignment of anomalies with SAR-derived lineaments and magnetic geophysical anomalies further validated Farmonaut’s multi-sensor methodology. Artisanal mining signals co-located with high-GPI anomalies additionally confirmed the empirical gold endowment of several target areas.

Strengths of the Methodology

Farmonaut’s approach offered several key innovations:

Multi-Sensor Data Fusion: By integrating optical (Landsat/Sentinel-2), ASTER SWIR, thermal (ASTER GED + MODIS LST), SAR (Sentinel-1 + ALOS PALSAR), geomorphic (MERIT Hydro, Dynamic World, ESA WorldCover, GEDI), and magnetic (EMAG2) datasets, the methodology captured a far richer picture of mineralisation potential than any single data source could provide.
Temporal Analysis: By analysing multi-year data and comparing seasonal composites, Farmonaut filtered out transient features like vegetation flush or soil moisture changes. The temporal stability metric (Coefficient of Variation) ensured only persistent, geologically significant signals were prioritised, reducing false positives.
Enhanced Prioritisation Score: The multi-criteria scoring system integrated alteration intensity, spatial extent, temporal stability, seasonal contrast, structural coherence, thermal anomaly, geobotanical stress, and geophysical response. This nuanced ranking allowed exploration teams to focus on targets with optimal characteristics, maximising efficiency.
Detectability Transparency: A dedicated detectability index explicitly communicated the confidence level of each GPI pixel, ensuring users understood where canopy cover, laterite, or vegetation attenuated the signal — a critical safeguard against misinterpretation in densely vegetated or deeply weathered terrain.

These strengths dramatically increased target confidence and minimised the risk of wasted field resources.

Keywords: Structural controls, multi-sensor fusion, temporal analysis, prioritisation score, detectability index, gold exploration efficiency.

Benefits of Remote Sensing for Gold Exploration

Farmonaut’s remote sensing approach offers several advantages over traditional exploration methods:

Cost-Effectiveness: Non-invasive multi-sensor analysis reduces the need for extensive fieldwork prior to drill targeting.
Time Efficiency: Rapid processing of large, multi-source datasets accelerates target identification across large AOIs.
Environmental Impact: Minimising fieldwork reduces ecological disturbance, particularly in sensitive or remote terrain.
Scalability: The methodology can be applied to other regions or adapted for other mineral systems.
Reconnaissance Screening: The approach is specifically designed for desk-top screening where ground access is limited, enabling informed go/no-go decisions before committing to field programmes.

By leveraging satellite data, Farmonaut empowers exploration companies to make informed decisions with minimal risk.

Keywords: Remote sensing benefits, cost-effective exploration, environmental sustainability, scalable mineral mapping, reconnaissance screening.

Why Choose Farmonaut for Mineral Exploration?

Farmonaut’s expertise in satellite-based solutions sets it apart as a trusted partner for mineral exploration. Key differentiators include:

Advanced Technology: Fusion of Landsat 8/9, Sentinel-2, ASTER SWIR, Sentinel-1, ALOS PALSAR, Copernicus GLO-30, MODIS LST, ASTER GED, Dynamic World, ESA WorldCover, MERIT Hydro, GEDI, and EMAG2 for comprehensive multi-sensor analysis.
Innovative Methodology: Multi-stage approach combining spectral, thermal, SAR, geomorphic, geobotanical, and geophysical innovations in a single integrated GPI framework.
Actionable Insights: Prioritised target lists with structural attributes, proxy grade, depth estimates, drill orientations, artisanal mining signals, and placer classifications for comprehensive field planning.
Proven Results: Successful application in Tanzania’s challenging terrain, with outputs validated against known geological structures and artisanal workings.

Farmonaut’s commitment to data-driven excellence ensures clients achieve exploration goals efficiently and sustainably.

Keywords: Farmonaut mineral exploration, satellite-based solutions, innovative methodology, actionable insights.

Conclusion

Farmonaut’s gold prospectivity mapping project in Tanzania demonstrates the transformative potential of integrated multi-sensor remote sensing in mineral exploration. By fusing advanced optical imagery (Landsat 8/9, Sentinel-2, ASTER SWIR), thermal datasets (ASTER GED, MODIS LST), synthetic aperture radar (Sentinel-1, ALOS PALSAR), geomorphic and land-cover layers (Copernicus GLO-30, Dynamic World, ESA WorldCover, MERIT Hydro, GEDI), and magnetic geophysics (EMAG2), Farmonaut delivered a prioritised list of high-confidence exploration targets within a 10-square-kilometer AOI. The methodology’s temporal stability, multi-criteria prioritisation, artisanal mining detection, placer geomorphology assessment, and explicit detectability reporting ensured reliable, field-ready results that guided efficient investigations while minimising costs and environmental impact.

This case study underscores Farmonaut’s leadership in satellite-based exploration solutions, offering a scalable and sustainable approach for gold and other mineral exploration projects worldwide. For companies seeking to optimise their exploration efforts, Farmonaut provides the technology, expertise, and insights needed to succeed.

Keywords: Gold prospectivity mapping, multi-sensor remote sensing, Tanzania exploration, Farmonaut expertise.

Call to Action

Ready to revolutionise your mineral exploration strategy? Contact Farmonaut today to learn how our satellite-based solutions can deliver data-driven insights for your next project. Visit farmonaut.com for more information.