Probing Venus Surface Iron Contents With Six-Band Visible Near-Infrared Spectroscopy From Orbit

Machine learning models enable interpretation of orbital spectral measurements of Venus using laboratory calibration data collected at Venus surface temperatures. Partial least squares models show that total iron content can be accurately predicted using data from the six bands (two in the 1.02 mu m window). Prediction errors on total wt% FeO are 0.50 for common subalkaline volcanic rocks. Accuracy is 0.42 for wt% FeO in alkaline rocks, and 2.47 for all 18 igneous samples studied to date. These robust capabilities will allow discrimination of basalt versus rhyolite/granite and elucidate the rock type of the enigmatic tessera terrain on Venus. Plain Language Summary Because the surface of Venus is robed by a thick CO2-rich atmosphere and dense clouds, it has long been thought that orbital spectroscopic measurements of its surface were unfeasible. However, a small range of the CO2 spectrum near 1 mu m lacks absorption features, providing windows that enable surface analyses. This paper uses a machine learning method known as partial least squares regression to quantify how igneous rock types such as basalt and rhyolite can be distinguished using 440 degrees C laboratory data for those six bands. This advance will make it possible to unequivocally distinguish between basalt and granite/rhyolite rock types on the surface of Venus and resolve questions of its geological history. Key Points Accuracies of wt% FeOtotal are +/- 0.50 for subalkaline rocks, +/- 0.42 for alkali rocks, and +/- 2.47 for 440 degrees C lab spectra of both Six-band remote sensed spectra can easily distinguish basalt ( x=9.80 wt% FeO) from granitic/rhyolitic ( x=2.19 wt% FeO) rocks Partial least squares models show that total iron content can be accurately predicted from high-temperature lab spectra