CAN REMOTE SENSING MAP MERCURY POLLUTION IN VEGETATED AREAS? THE CASE OF TARKWA, GHANA, AFRICA

Mercury pollution is a serious threat to ecosystem and human health. A significant source of environmental mercury is artisanal—commonly illegal—gold mining. The remote and unregulated nature of small-scale artisanal gold mining makes monitoring mercury contamination difficult. Previous work showed that vegetation on mercury contaminated soils reflects less in IR and green wavelengths. We use Landsat data from 1986-2018 to test whether it is possible to detect and quantify mercury contamination associated with artisanal gold mines near Tarkwa, Ghana. This site provides a useful test case because mercury contamination in soils surrounding the mines has been measured.

We used the Tasseled Cap Transformation to convert RGB channels into brightness, greenness, and wetness, and experimented with correlating various indices derived from these values with measured mercury levels (n = 13). Greenness alone yielded a correlation coefficient of 0.58 (p = 0.036); an index that summed all three parameters produced a correlation of 0.62 (p = 0.025). To determine whether we could distinguish contaminated from uncontaminated vegetation using these metrics, we sampled 300 points in each of three vegetation types: (a) undisturbed primary forest, (b) secondary forest far from mining sites (uncontaminated), and (c) secondary forest within 500 m of mining sites that ground-sampling showed to be contaminated with high levels of mercury. Both greenness and the combined index distinguished contaminated from uncontaminated secondary forest at α < 0.001.

Using published values of soil mercury for calibration, we developed regression models to reconstruct changes in contamination in 2006 and 2015, years for which clear images were available. In 2006, with 95% confidence, secondary forest beyond 500 m of an active mine showed no contamination using either index; inside the 500 m radius, the contamination ranged from 74-91 mg/kg (greenness) or 0-13mg/kg (combined index). In 2015, far from the mine, greenness shows no contamination and combined index 0-8 mg/kg. Within the 500 m of the mine, 95% confidence intervals produced from greenness suggested 8-24 mg/kg of mercury contamination and 0-17mg/kg (combined index). Next, we apply the same approach to agricultural vegetation surrounding Tongguan, China.