Cropland management practices have traditionally focused on maximising the production of food, feed and fibre. However, croplands also provide valuable regulating ecosystem services, including carbon (C) storage in soil and biomass. Consequently, management impacts the extents to which croplands act as sources or sinks of atmospheric carbon dioxide (CO2). And so, reliable information on cropland ecosystem C fluxes and yields are essential for policy-makers concerned with climate change mitigation and food security. Eddy-covariance (EC) flux towers can provide observations of net ecosystem exchanges (NEE) of CO2 within croplands, however the tower sites are temporally and spatially sparse. Process-based crop models simulate the key biophysical mechanisms within cropland ecosystems, including the management impacts, crop cultivar, soil and climate on crop C dynamics. The models are therefore a powerful tool for diagnosing and forecasting C fluxes and yield. However, crop model spatial upscaling is often limited by input data (including meteorological drivers and management), parameter uncertainty and model complexity. Earth observation (EO) sensors can provide regular estimates of crop condition over large extents. Therefore, EO data can be used within data assimilation (DA) schemes to parameterise and constrain models. Research presented in this thesis explores the key challenges associated with crop model upscaling. First, fine-scale (20-50 m) EO-derived data, from optical and radar sensors, is assimilated into the Soil-Plant-Atmosphere crop (SPAc) model. Assimilating all EO data enhanced the simulation of daily C exchanges at multiple European crop sites. However, the individually assimilation of radar EO data (as opposed to combined with optical data) resulted in larger improvements in the C fluxes simulation. Second, the impacts of reduced model complexity and driver resolution on crop photosynthesis estimates are investigated. The simplified Aggregated Canopy Model (ACM) – estimating daily photosynthesis using coarse-scale (daily) drivers – was calibrated using the detailed SPAc model, which simulates leaf to canopy processes at half-hourly time-steps. The calibrated ACM photosynthesis had a high agreement with SPAc and local EC estimates. Third, a model-data fusion framework was evaluated for multi-annual and regional-scale estimation of UK wheat yields. Aggregated model yield estimates were negatively biased when compared to official statistics. Coarse-scale (1 km) EO data was also used to constrain the model simulation of canopy development, which was successful in reducing the biases in the yield estimates. And fourth, EO spatial and temporal resolution requirements for crop growth monitoring at UK field-scales was investigated. Errors due to spatial resolution are quantified by sampling aggregated fine scale EO data on a per-field basis; whereas temporal resolution error analysis involved re-sampling model estimates to mimic the observational frequencies of current EO sensors and likely cloud cover. A minimum EO spatial resolution of around 165 m is required to resolve the field-scale detail. Monitoring crop growth using EO sensors with a 26-day temporal resolution results in a mean error of 5%; however, accounting for likely cloud cover increases this error to 63%.

Carbon dioxide is taken up by crops during production and released back to the atmosphere at different geographical locations through respiration of consumed crop commodities. In this study, spatially distributed county-level US cropland net primary productivity, harvested biomass, changes in soil carbon, and human and livestock consumption data were integrated into the prior terrestrial biosphere flux generated by the Boreal Ecosystem Productivity Simulator (BEPS). A global time-dependent Bayesian synthesis inversion with a nested focus on North America was carried out based on CO2 observations at 210 stations. Overall, the inverted annual North American CO2 sink weakened by 6.5% over the period from 2002 to 2007 compared to simulations disregarding US crop statistical data. The US Midwest is found to be the major sink of 0.36±0.13 PgC yr-1 whereas the large sink in the US Southeast forests weakened to 0.16±0.12 PgC yr-1 partly due to local CO2 sources from crop consumption.

Carbon dioxide is taken up by crops during production and released back to the atmosphere at different geographical locations through respiration of consumed crop commodities. In this study, spatially distributed county-level US cropland net primary productivity, harvested biomass, changes in soil carbon, and human and livestock consumption data were integrated into the prior terrestrial biosphere flux generated by the Boreal Ecosystem Productivity Simulator (BEPS). A global time-dependent Bayesian synthesis inversion with a nested focus on North America was carried out based on CO2 observations at 210 stations. Overall, the inverted annual North American CO2 sink weakened by 6.5% over the period from 2002 to 2007 compared to simulations disregarding US crop statistical data. The US Midwest is found to be the major sink of 0.36±0.13 PgC yr-1 whereas the large sink in the US Southeast forests weakened to 0.16±0.12 PgC yr-1 partly due to local CO2 sources from crop consumption.