Monitoring farm soil health using an integrated field and earth observation based approach

Utilised agricultural land covers more than 60% of the terrestrial area in the UK, usually sustaining a wide range of crop and livestock enterprises among different agricultural use change included in the land covers. Past soil nutrient loss and degradation have been attributed to intensive farming practices, and have resulted in substantial release of soil carbon to the atmosphere. Existing and popular carbon accounting tools currently did not account agricultural land types of arable crops, temporary ley, permanent grass and woodland fields for monitoring real-time soil and plant properties. A better measurement of soil organic carbon (SOC) level is required in order to provide quality and useful SOC benchmarks and predictions for timely agricultural accounting and management. This thesis examines both SOC and plant properties across different land types on farms in Gloucestershire. Field investigations of SOC content and properties and plant biomass cover were undertaken in Summer and Spring seasons using real-time near-infrared spectroscopy (NIRS), laboratory methods of loss-on-ignition (LOI) and Dumas, aerial imagery (UAV and satellite) and spatial interpolations of inverse distance weighting (IDW) and ordinary kriging (OK). The current study first determined SOC contents among the NIRS, LOI and Dumas methods, where key results showed that strong and positive correlations were among all the three methods, with permanent grass fields having the highest correlation between the Dumas and NIRS method compared to arable and temporary ley fields. This study then used the NIRS method to further measure and assess both SOC and plant changes at the same sampling point across the four land types in the studied farms. Key results indicated that arable and temporary ley fields across the land types could act as potential fields for carbon sinks based on high proportion of the fields falling into the moderate and good SOC/ clay ratio threshold levels (ranging from 22% to 46%). In addition, the study used aerial imagery from UAV and satellite to predict field vegetation coverage and SOC estimates with key vegetation indices of normalized green-blue difference index (NGBDI), green leaf index (GLI) and brightness index (BI), particularly for the land types of permanent grass fields and woodland with denser biomass distribution. Key SOC and plant properties, such as carbon stock (in both g/kg and t/ha), SOC/ plant organic carbon (POC) ratio and SOC/ clay ratio can also be predicted by the key indices from the novel combined techniques which are suggested to serve as vegetation indices-SOC predictors for improving carbon accounting on-farm. The study finally performed SOC predictions on field sampling points by both the NIRS and the two interpolation methods, and reflected very high correlations among the NIRS and IDW method (r > 0.9), followed by the OK method (r > 0.7). Further regional and even national studies on SOC estimates among different soil types, land uses and farms for quality SOC benchmarks with seasonal patterns is needed to allow landowners to make timely, practical and appropriate land practice decisions and management on their farms.