Descriptor: Archaeological cropmark synthetic signatures (ACSS)

Abstract

With the advent of artificial intelligence and machine learning algorithms, research in the domain of remote sensing archaeology may be greatly assisted by new tools available. Nevertheless, a major barrier relies on the scarcity of existing ground truthing remote sensing data, which eventually hinders training effective models, acquiring general understanding, and performing predictive modeling. In this article, we introduce an extensive dataset of synthetic hyperspectral signatures (ACSS)—of 1 nm interval—simulating cropmarks spectral measurements, which can be used as a proxy indicator for the detection of shallow buried archaeological remains. Simulation generation involved: 1) encoding the observed hyperspectral ground truth signatures, collected over a barley test field, into leaf and canopy bio-physical parameter values, via inversion of the radiative transfer model PROSAIL; and 2) modeling the statistics of the physical parameters by a joint pdf, sampling this distribution, and then running PROSAIL forward to generate two types of simulated signatures, i.e., over buried remains (cropmarks) and over pure soil. The dataset includes 10 000 synthetic signatures from each type in comma-separated value (csv) file format, as well as a Python generator script. To the best of the authors’ knowledge, this is the first time that such an extensive spectral cube has been published, which can be further reused by the scientific community to better understand the formation of cropmarks and optimize archaeological proxies through the development of targeted remote sensing archaeological AI models and architectures. We therefore anticipate that ACSS can be a starting point for the development of targeted data resources in the field of remote sensing archaeology that will facilitate model building toward addressing important questions, which are otherwise hard to answer.