Deep transfer learning of global spectra for local soil carbon monitoring

There is global interest in spectroscopy and the development of large diverse soil spectral libraries (SSL) to model organic carbon (SOC) monitor, report, verify (MRV) its changes. The reason that increasing SOC can improve food production mitigate climate change. However, ‘global’ modelling with such hyperdimensional SSLs do not generalise well locally, e.g. at a field scale. To address this challenge, we propose deep transfer learning (DTL) leverage useful information from large-scale assist local modelling. We used one global, three country-specific data sites DTL localise estimates individual fields or farms each country. With DTL, transferred instances SSLs, representations one-dimensional convolutional neural networks (1D-CNNs) trained on both Transferring effectively SSL most accurately estimate site, reducing root mean square error (RMSE) by 25.8% average compared Our results highlight effectiveness value diverse, for accurate predictions. Applying could anywhere world more accurately, rapidly, cost-effectively, enabling MRV protocols monitor