Network Analysis of Simulated and Real Indigenous Irrigation System

Abstract

Small-scale Indigenous Irrigation Systems (IIS) are water-sharing societies which have been observed to persist for long periods of time finding a dynamic equilibrium with the environment. This persistence is thought to be mostly due to the institutions and system structures which evolve to maintain stability despite internal and external changes. They have been described as the most ancient and ubiquitous example of public infrastructure system, however, the way in which they grow, evolve and maintain stability is a contested and also controversial topic. The study of IIS is interdisciplinary and generally classified under the umbrella term Social-Ecological Systems.

Advances in computing performance and software have enabled simulation modelling to be quicker, cheaper and more accessible. Alongside this, recent scientific understanding of complex systems and network theory has led to new interdisciplinary theories on universal scaling and preferential attachment based growth in systems of many interacting components. This research aims to harness this potential by building an abstract simulated IIS in a generative agent-based model environment.

The model assumes that the IIS network grows through preferential attachment to optimise space, which is the mechanism of growth found in previous models and a common assumption in biological systems as it increases efficiency. Different growth strategies relating to local or global information and stochastic processes are tested and find a range of space optimal configurations.

The results find that all models form rooted planar tree networks. Stochastic processes are important for the model to search for different model configurations and finding almost optimal configuration. The optimal solution is found from collecting global information of the network and selecting growth globally, however this is computationally expensive, and inefficient so unlikely to be found in real systems. The networks formed through random asynchronous processes follow scale-free laws which coincide with similar scaling exponent as other sub-critical networks such as rivers.

Two vastly different real-world IIS networks are then analysed and compared with the simulated models. The real-world networks show differences comparable to the differences found in the simulated models. The reasons for these differences are speculated to be due to a number of factors including geomorphology and managerial arrangements, however given the limited data collected no firm conclusion can be made. It is recommended that further data is collected and analysed in order to confirm this.

More information

Identifiers

ORCID for John Dearing: orcid.org/0000-0002-1466-9640

ORCID for Craig Hutton: orcid.org/0000-0002-5896-756X

Catalogue record

Export record