In the thickly forested Hokuroku kuroko district of northern Japan, Hede, Koike, Kashiwaya, Sakurai, Yamada, and Singer (2017) mapped vegetation-anomaly zones from the temporal variation of a Landsat ETM+ vegetation index (VIGS) and showed that these zones trace interpreted ring structures, contain most major deposits located in dense forest, and correspond well to a geologically derived ore-deposit exploration index. The method's own authors, however, left its central scientific question unresolved: no ground-truth soil-geochemical or plant-spectroscopic measurements were made to confirm that the mapped anomalies are caused by mineralization-related soil conditions rather than by soil acidity alone, forest-type differences, topography, drought, or other external stressors; several large known deposits outside forest areas fell into the background class; and anomaly-class composition differed between deciduous and evergreen forest, with deciduous senescence suspected of masking metal-induced stress. This problem asks what ground-truthing the causal validation in the Hokuroku setting requires — the field programme of soil geochemistry and plant reflectance spectra at anomaly and background sites — and whether such a study has since been carried out, together with the characterization of how detection performance depends on forest type, seasonal phenology, and land-cover context, so that satellite-derived vegetation anomalies can be accepted or rejected as a reliable prospecting signal beneath thick vegetation.

Background

Optical satellite remote sensing is a standard tool of early-stage mineral exploration, but the established workflows — mapping rock, alteration-mineral, and gossan spectral signatures — work mainly in arid and semiarid regions where the ground is exposed. Under thick vegetation the canopy hides the soil and rock signal, so conventional lithological mapping fails. Geobotanical remote sensing attacks this problem from the opposite direction: instead of removing the vegetation, it exploits the vegetation as a sensor. The idea rests on geobotany, the observation that plants growing over mineralized ground can take up metals in toxic quantities and show physiological stress — chlorosis (yellowing from chlorophyll loss), reduced growth, and altered water balance — which changes leaf and canopy reflectance in the visible, near-infrared, and shortwave-infrared (SWIR, roughly 1.3–2.5 micrometre) wavelength ranges.

A vegetation index (VI) is a band combination that condenses reflectance into one number per pixel tracking plant physiological activity; the most common is the normalized difference vegetation index (NDVI), built from red and near-infrared bands. Hede, Kashiwaya, Koike, and Sakurai (2015) first developed the Vegetation Index considering Greenness and Shortwave infrared (VIGS), which adds green and SWIR terms so that it responds to both chlorophyll-related stress and water stress; laboratory growth experiments on Japanese mustard spinach in metal-doped soils, and a tropical-forest case study in Sumatra validated against soil geochemistry, indicated that VIGS detects metal-induced stress better than NDVI.

Hede et al. (2017) then carried the approach to the Hokuroku district of Akita Prefecture, northern Japan — one of the world's richest volcanogenic massive sulfide (VMS) districts. A VMS deposit forms on or beneath the seafloor from hydrothermal fluids driven by submarine volcanism; the Miocene 'kuroko' ('black ore') deposits of Hokuroku, rich in Cu, Pb, Zn, Ba, Au, and Ag, are the type example, and the district also hosts vein-type epithermal Au–Ag deposits. The area is mountainous and more than 84% covered by dense forest, roughly half deciduous and half evergreen. From five cloud-free Landsat Enhanced Thematic Mapper Plus (ETM+) scenes acquired between late July and mid-October of 2000 and 2002 (three scenes from August–September 2000 and two from July and October 2002), the authors computed VIGS per pixel, took the ratio of the temporal mean to the temporal standard deviation m/σm/\sigma as an anomaly intensity (the premise being that stressed vegetation fluctuates more between summer and autumn), and then applied a chain of image-enhancement steps: forced invariance (a scene-statistics correction that suppresses systematic dependence of the ratio on elevation), smoothing by ordinary kriging (a geostatistical interpolation method, here with 300 m cells), and residual analysis (subtracting an 8.5 km moving-average regional trend) to isolate local anomalies. Classifying residuals into background, low-anomaly, and high-anomaly classes — with thresholds set separately for deciduous and evergreen forest using a land-cover map — the anomaly zones traced interpreted ring structures associated with the Miocene submarine caldera system, enclosed most major deposits situated in dense forest, and corresponded well to a published ore-deposit exploration index map built by multivariate regression of seven geological and geophysical factors.

The validation, however, was entirely remote: the anomaly map was checked against independent geological and geophysical compilations, never against ground measurements at the anomaly sites. The authors' own discussion makes the gaps explicit. First, the causal chain is unconfirmed: the sulfide-rich deposits should produce strongly acidic, metal-bearing topsoil by oxidation, and the authors nominate soil acidity as 'the most plausible factor' for the vegetation stress — but this is a hypothesis, and the anomalies could equally arise from natural acidity, forest-type composition, topography, drought, or other non-mineralization effects. Second, detectability depends on land-cover context: several large kuroko deposits (Matsuki, Matsumine, Shakanai, Kosaka, and Uchinotai) fell into the background class because they lie outside the forest areas where the method presupposes vegetation. Third, anomaly classes behaved differently by forest type — the high-anomaly class was 63% evergreen forest, and the authors argue that senescence (seasonal aging and leaf decline) in deciduous trees, which dominate at higher elevations, can conceal metal-induced stress, plausibly explaining why two mines (Nurukawa and Towada-ginzan) were missed. The authors' dissertation summary of this program states the required next step plainly: ground-truthing 'may include extensive chemistry and spectroscopy analyses of rocks, soil, and leaf samples'. No such ground-truth study in Hokuroku existed at the time, and the task below is to determine what such a ground-truthing programme must comprise and whether it has since been performed.

Problem Statement

Specify, for the Hokuroku district of Akita Prefecture, northern Japan, the ground-truth investigation that would establish whether the vegetation-anomaly zones mapped from multi-temporal Landsat ETM+ VIGS analysis by Hede et al. (2017) are causally linked to mineralization-related soil conditions — elevated metal contents (e.g., Cu, Pb, Zn) and the associated soil acidity expected from oxidizing sulfide mineralization — rather than to alternative, non-mineralization causes (natural soil acidity alone, forest-type composition, topographic or edaphic gradients, drought or other seasonal stressors), and determine whether such an investigation has already been carried out. Concretely, this requires georeferenced ground-truth measurements at sites spanning the published high-anomaly, low-anomaly, and background classes — soil geochemical analyses (metal contents and pH, with laboratory quality control) and plant reflectance spectra (foliar and/or canopy, with recorded acquisition dates), together with a statistical analysis that discriminates the mineralization-stress explanation from the candidate confounders — and a documented determination, from the literature, of whether any such study of the Hokuroku anomaly sites has been performed since Hede et al. (2017) and, if not, of what remains open. It further requires characterizing the domain of validity of the detection: how anomaly performance depends on forest type (deciduous versus evergreen), on seasonal phenology including deciduous senescence, and on land-cover context — in particular whether the failures at known deposits located outside forest areas (e.g., Matsuki, Matsumine, Shakanai, Kosaka, Uchinotai) are an intrinsic boundary of the geobotanical approach or a correctable artifact. The single overarching objective is to determine whether, and under what conditions, satellite-derived vegetation anomalies in thickly vegetated terrain can be trusted as a mineral-prospecting signal; the causal ground-truthing and the land-cover/phenology robustness analysis jointly serve that objective.

References

  1. Hede, A. N. H., Koike, K., Kashiwaya, K., Sakurai, S., Yamada, R., & Singer, D. A. (2017). How can satellite imagery be used for mineral exploration in thick vegetation areas? Geochemistry, Geophysics, Geosystems, 18(2), 584–596. DOI: 10.1002/2016GC006501. https://doi.org/10.1002/2016GC006501
  2. Hede, A. N. H. (2016). Development and application of geobotanical remote sensing methods for mineral exploration in thick vegetation areas. Doctoral thesis, Kyoto University. DOI: 10.14989/doctor.k19411. http://hdl.handle.net/2433/204588
  3. Hede, A. N. H., Kashiwaya, K., Koike, K., & Sakurai, S. (2015). A new vegetation index for detecting vegetation anomalies due to mineral deposits with application to a tropical forest area. Remote Sensing of Environment, 171, 83–97. DOI: 10.1016/j.rse.2015.10.006. https://doi.org/10.1016/j.rse.2015.10.006
  4. Zhang, C., Ren, H., Liang, Y., Liu, S., Qin, Q., & Ersoy, O. (2017). Advancing the PROSPECT-5 Model to Simulate the Spectral Reflectance of Copper-Stressed Leaves. Remote Sensing, 9(11), 1191. DOI: 10.3390/rs9111191. https://doi.org/10.3390/rs9111191
  5. Liu, L., Ji, M., & Buchroithner, M. (2018). A Case Study of the Forced Invariance Approach for Soil Salinity Estimation in Vegetation-Covered Terrain Using Airborne Hyperspectral Imagery. ISPRS International Journal of Geo-Information, 7(2), 48. DOI: 10.3390/ijgi7020048. https://doi.org/10.3390/ijgi7020048
  6. Saepuloh, A., Ratnanta, I. R., Hede, A. N. H., Susanto, V., & Sucipta, I. G. B. E. (2023). Radioactive remote signatures derived from Sentinel-2 images and field verification in West Sulawesi, Indonesia. Environmental Monitoring and Assessment, 195(10), 1243. DOI: 10.1007/s10661-023-11868-5. https://doi.org/10.1007/s10661-023-11868-5
  7. Kubo, T., Gonnokami, H., Hede, A. N. H., & Koike, K. (2025). Combining vegetation index with mineral identification for detection of high-geothermal-potential zones using hyperspectral satellite data. Geothermics, 125, 103194. DOI: 10.1016/j.geothermics.2024.103194. https://doi.org/10.1016/j.geothermics.2024.103194
  8. Huang, X., & Hui, Z. (2026). Analysis of Vegetation Ecological Anomaly Response in the Xiangshan Uranium Mining Area Based on Multi-Source Remote Sensing Data Fusion. Forests, 17(6), 661. DOI: 10.3390/f17060661. https://doi.org/10.3390/f17060661
  9. Árvai, V., & Albert, G. (2026). From Canopy Phenology to Lithological Signals: Evaluating Biophysical Traits with Machine Learning in the Hațeg Basin. Remote Sensing, 18(16), 2783. DOI: 10.3390/rs18162783. https://doi.org/10.3390/rs18162783