A blind test of a methodology for the detection of mineralization through the remote sensing of vegetation

In 1980 and 1981 a field experiment was conducted at a series of paired mineralized and background sites near the towns of Mineral, in Louisa County Virginia and New Store, in Buckingham County, Virginia. At each site spectral measurements (using a three-band hand-held radiometer) of excised leaves were taken systematically throughout the growing season and compared to soil samples analyzed for concentrations of micronutrients and trace metals. This work showed that the reflectances in thermatic mapper bands 3 and 5 (TM3 and TM5) from mineralized vegetation were higher than from nonmineralized vegetation during late summer to early fall and the strengths of the relationships between the concentrations of trace metals in the soil and the reflectances were significant and strongest in the fall.For the 1982 field season, an experiment was set up to test the results of the previous two years. This experiment was performed without knowledge of the specific location of any soil anomaly. To find them a 420-tree grid was laid out to cover an area of 1.2 by 0.5 km located near Arvonia, Virginia. The only a priori knowledge available was: (1) the presence of a soil anomaly somewhere in the area, (2) the previous results, which suggested that variation in leaf reflectance related to mineralization was most evident in the early fall just prior to senescence, and (3) publicly available geologic information--strike, dip, and broad lithologic classes. Using only reflectance information from hand-held radiometers, collected on four dates in late summer to early fall, and some relatively simple data analysis techniques, six areas representing approximately 20 percent of the test site were identified. One of the six areas persisted throughout the entire analysis procedure and was selected as the most promising candidate. When the soil geochemistry was made available by the cooperating mineral exploration company, the selected area was found to lie directly on the location of maximum copper mineralization.By examining the decision error structure as a function of spectral band, chemical species, and threshold values for both the trace metal concentrations and percent reflectance, we were able to make formal statements about the quality or "goodness" of results and of the technique used.