Surface Biology and Geology Optimal Design using Intrinsic Dimensionality

The information content that can be derived from spectroscopic imagery increases with finer ground sampling distance, finer spectral resolution, more frequent revisit, and higher signal-to-noise ratios (SNRs). However, these parameters are not independent, and it is thus impossible to design a mission to maximize all of them simultaneously. We present an instrument model and simulation environment that enable us to find the optimal combination of these four mission design parameters, using intrinsic dimensionality (ID) as the metric. ID is the size of the signal subspace – the maximum degrees of freedom when noise can be disregarded – and is a metric that is independent of any one particular algorithm or application area. This study is important for upcoming missions such as NASA’s Earth System Observatory mission to study the Earth’s Surface Biology and Geology (SBG), which will comprise a visible to shortwave infrared spectrometer in addition to a multi-channel thermal radiometer on a separate platform. When evaluating a desert site and a dry topical forested site, we find that spectral resolution drives information content, with a significant drop in normalized ID (15-45% decrease) when simulating 15 nm spectral bandwidth as opposed to 10 nm bandwidth. Ground sampling distances in the range 30-50 m provided the optimal balance between spatial resolution and SNR, although more frequent revisit, potentially by combining data from multiple missions, would maximize total information content.