Predicting soil carbon sequestration potential of Irish soils from spectral data

Soil carbon sequestration has emerged as a critical natural climate solution for mitigating climate change while simultaneously improving soil health and agricultural productivity. However, accurate assessment of soil carbon stock and its sequestration potential faces significant scientific and methodological challenges. Employing mid-infrared (MIR) spectroscopy, three interconnected challenges in soil carbon sequestration potential assessment were systematically addressed in this thesis. First, the frequent absence of soil bulk density data from legacy databases was tackled through developing spectral models using MIR spectroscopy and machine learning algorithms. The developed spectral model achieved promising performance compared to traditional pedotransfer functions, with consistent accuracy across different soil horizons and depth layers, providing a robust solution for converting soil carbon concentrations to area-based stocks. Second, a novel local quantile regression framework was developed to optimise the estimation of mineral-associated organic carbon (MAOC) saturation capacity when soil mineralogy information is unavailable. This approach used MIR spectral dissimilarity of soil samples to define local neighbourhoods for quantile regression analysis, yielding more conservative but realistic estimates of MAOC sequestration potential compared to global quantile regression approaches. The framework revealed that 53.04 Mt C could be sequestered as MAOC in the 5-20 cm layer across mineral soils in grassland in the Northern half of Ireland, demonstrating substantial regional carbon storage potential. Third, an integrated spatial assessment framework was developed that combines digital soil mapping with multiple carbon sequestration assessment approaches to create practical decision-support tools. By integrating different assessing soil carbon sequestration potential approaches, a four-class soil carbon sequestration potential classification framework that provides targeted management strategies for different soil conditions. The classification framework identified four distinct areas: dual saturation areas (34.87%) requiring protection strategies, dual high potential areas (32.93%) ideal for carbon farming projects, and intermediate areas with specific limitations in either MAOC (20.24%) or total SOC (11.95%) accumulation. This framework successfully translates complex soil carbon science into actionable guidance for sustainable land management and climate policy development. This thesis demonstrated the exciting potential of integrating advanced analytical techniques with soil spectroscopy for large-scale carbon assessment. Moreover, it provides not only a comprehensive assessment of soil carbon sequestration potential in Ireland, but also transferable frameworks applicable to other regions, contributing to effective implementation of climate mitigation strategies through targeted land management interventions. Meanwhile, there are several limitations must be acknowledged. For example, the carbon assessment was constrained to the 5-20 cm depth interval, excluding the top 5 cm layer and deeper soil profiles. Besides that, the applicability of developed spectral models across diverse soil types, climatic conditions, management regimes, and different MIR instruments requires further local validation.