Land use, land cover and land management change threatens the sustainability of ecosystem services upon which individuals and communities depend. However, quantifying the effects of large-scale environmental change on regional water resources and climate is challenging because of a lack of appropriate data as well as fundamental limitations of environmental models. This thesis focuses on the development of integrated modelling systems for representing feedback mechanisms between human activities and the environment at regional scales. India is selected as a case study because of the unprecedented scale of environmental change in this country over recent decades.
Land use change modelling is identified as a viable method for reconstructing historical land use/land cover at regional scales. This is facilitated through the development of a new modelling framework which allows users to perform the entire modelling workflow in the same environment and provides a consistent interface to different spatial allocation models. Hence, the modelling framework enables model intercomparison and ensemble experiments. It furthermore promotes reproducible science because it allows applications to be expressed programmatically.
An adapted version of the Change in Land Use and its Effects (CLUE) land use change model is used to reconstruct historical land use/land cover in India between 1956–2010. The model algorithm explicitly accounts for competition between land use/land cover categories as a result of dynamic socio-economic and biophysical conditions. A further dataset showing the irrigated area of various crops is developed by spatially disaggregating historical agricultural inventory data based on maps of cropland extent and biophysical suitability. Land use/land cover maps are supplied to an offline historical simulation of the Joint UK Land and Environment Simulator (JULES), a process-based land surface model, to generate soil moisture values across the Gangetic plain. Simulated soil moisture values are modified to account for the effects of irrigation. The procedure exploits the characteristics of the irrigated area dataset in order to account for the growing season of individual crops.
Existing tools for coordinating complex workflows in the hydrological sciences are strongly coupled to underlying modelling frameworks. As a result, they lack flexibility and often necessitate refactoring of the source code of model components. Exploring these issues further, an experiment is devised in which the data processing language R is set up as a workflow orchestration tool for hydrological data analysis and modelling. A new software package implements a set of classes for representing multi-dimensional hydrological data and to provide a common interface to hydrological models. The experimental set-up is demonstrated through two example applications drawn from hydrology and the emerging discipline of socio-hydrology. These serve to highlight the flexibility of the R system for workflow orchestration and model coupling but also draw attention to several areas for future development.