Soil Organic Carbon Monitoring to Assess Agricultural Climate Change Adaptation Practices in the Arid Regions

Abstract

This research presents an in-depth investigation of Soil Organic Carbon (SOC) dynamics and the concentration of heavy metals, specifically chromium, in the soils of the United Arab Emirates (UAE), an area characterized by its arid climate. The significance of SOC in sustaining agricultural productivity and mitigating climate change, along with the environmental and health risks posed by heavy metals, underscores the urgency of this study. Methodologies employed spanned laboratory-based geochemical analyses, spectral analysis, machine learning modeling, and the utilization of both hyperspectral and multispectral data to predict SOC and DOC levels and chromium content in soils. Field sampling across various agricultural fields in the UAE provided a robust dataset for analysis. The geochemical analysis, including the Walkley-Black method for SOC, combustion for DOC, laser scattering for soil texture, and the use of spectrometers for chromium detection, laid the foundational data. Spectral analysis was then conducted to correlate specific wavelengths with SOC, DOC, texture, and chromium levels, enhancing the understanding of soil characteristics from a remote sensing perspective.

The study introduced machine learning models, particularly Partial Least Squares Regression (PLSR) and Support Vector Regression (SVR), to predict SOC, DOC, texture, and chromium levels from spectral data and obtained high-performing models. This approach not only highlighted the potential of remote sensing technologies in soil analysis but also underscored the challenges in accurately determining SOC in arid environments. Significant correlations were found between SOC levels and spectral reflectance across various wavelengths, indicating the feasibility of using not only proximal remote sensing, but also satellite-based remote sensing for SOC estimation. For chromium, PLSR analysis revealed specific spectral regions associated with chromium concentrations, facilitating the development of predictive models for its quantification in soils. These findings are crucial, given the toxicological significance of chromium and its impact on agricultural and ecological systems. The successful prediction of soil parameters through spectral analysis and machine learning models represents a significant advancement in soil science, offering a non-destructive, cost-effective alternative to traditional soil sampling and analysis methods. This methodology holds promise for widespread application in soil health assessment, precision agriculture, and environmental monitoring.

Results indicated a wide range of SOC levels across the sampled areas, reflecting the diverse soil types and textures and agricultural practices in the UAE. The study's findings emphasize the need for targeted soil management practices to enhance SOC content and mitigate chromium contamination, thereby supporting sustainable agriculture and environmental conservation in arid regions. In conclusion, this research provides a comprehensive analysis of SOC dynamics and chromium concentrations in the soils of the UAE, employing a novel integration of laboratory analysis, spectral analysis, and machine learning models. The findings underscore the potential of remote sensing technologies in soil science, offering valuable insights for sustainable land management and environmental protection in arid regions. Future research should focus on expanding the spectral library for soil properties and exploring advanced machine learning algorithms to enhance the accuracy of SOC and heavy metal predictions, contributing further to global efforts in environmental sustainability and agricultural productivity.

Date of Award	6 May 2024
Original language	American English
Supervisor	Maryam Alshehhi (Supervisor)