Quantifying Riverscape Connectivity with Graph Theory

Fluvial catchments convey fluxes of water, sediment, nutrients and aquatic biota. At continental scales, crustal topography defines the overall path of channels whilst at local scales depositional and/or erosional features generally determine the exact path of a channel. Furthermore, constructions such as dams, for either water abstraction or hydropower, often have a significant impact on channel networks.The concept of ';connectivity' is commonly invoked when conceptualising the structure of a river network.This concept is easy to grasp but there have been uneven efforts across the environmental sciences to actually quantify connectivity. Currently there have only been a few studies reporting quantitative indices of connectivity in river sciences, notably, in the study of avulsion processes. However, the majority of current work describing some form of environmental connectivity in a quantitative manner is in the field of landscape ecology. Driven by the need to quantify habitat fragmentation, landscape ecologists have returned to graph theory. Within this formal setting, landscape ecologists have successfully developed a range of indices which can model connectivity loss. Such formal connectivity metrics are currently needed for a range of applications in fluvial sciences. One of the most urgent needs relates to dam construction. In the developed world, hydropower development has generally slowed and in many countries, dams are actually being removed. However, this is not the case in the developing world where hydropower is seen as a key element to low-emissions power-security. For example, several dam projects are envisaged in Himalayan catchments in the next 2 decades. This region is already under severe pressure from climate change and urbanisation, and a better understanding of the network fragmentation which can be expected in this system is urgently needed. In this paper, we apply and adapt connectivity metrics from landscape ecology. We then examine the connectivity structure of the Gangetic riverscape with fluvial remote sensing. Our study reach extends from the heavily dammed headwaters of the Bhagirathi, Mandakini and Alaknanda rivers which form the source of the Ganga to Allahabad ~900 km downstream on the main stem. We use Landsat-8 imagery as the baseline dataset. Channel width along the Ganga (i.e. Ganges) is often several kilometres. Therefore, the pan-sharpened 15m pixels of Landsat-8 are in fact capable of resolving inner channel features for over 80% of the channel length thus allowing a riverscape approach to be adopted. We examine the following connectivity metrics: size distribution of connected components, betweeness centrality and the integrated index of connectivity. A geographic perspective is added by mapping local (25 km-scale) values for these metrics in order to examine spatial patterns of connectivity. This approach allows us to map impacts of dam construction and has the potential to inform policy decisions in the area as well as open-up new avenues of investigation.