Dube, Y2022-07-222022-07-222021https://hdl.handle.net/10539/33056A research report submitted to the School of Animal, Plant and Environmental Sciences, Faculty of Science, University of the Witwatersrand, in fulfilment of the requirements for the degree of Master of Science, 2021South Africa is developing a research infrastructure called the Enhanced Freshwater and Terrestrial Ecological Observation Network (EFTEON). It is focused on a set of landscapes, where observations of climate, terrestrial ecology (land cover, phenology, and land-atmosphere fluxes), biodiversity (communities and populations), social-ecological circumstances (land use, livelihoods) and hydrology (soil water, river flows, aquifers and biogeochemistry) will be co-located and integrated. Automated instruments for river flow monitoring are well-established, but automated, continuous measurements of water chemistry for a comprehensive set of constituents is not yet fully operational anywhere in the world. A great deal of experimentation with different sensor systems is taking place. The pilot experiment in automated monitoring of water salinity at landscape scale described here is a technology testbed, to see if the sensors are sufficiently robust for long-term deployment, and if they can accurately quantify an important aspect of the stream chemistry. Deployment under field conditions for 14 months allowed robustness to be determined, both in terms of fraction of useful data collected and sensor drift. Accuracy was assessed by calibrations before, during and after deployment. Whether the achievable accuracy was sufficient for the landscape-scale biogeochemical questions posed by the network was assessed by seeing if it is possible to achieve a salt budget closure between several tributaries and the main stem of the river within + 10% error, based on data from a network of five sensors. The successfully designed and implemented observation network consisted of five stations equipped with a Decagon Conductivity-Temperature-Depth device that integrates water depth over a weir (to get flow), electrical conductivity (to get salt content), and water temperature sensors connected to a Em50 Logger. Field robustness was at the lower limits of acceptability (65% of potential data flows were achieved) across a varying and challenging real-world conditions at the Sabi-Sand Catchment. The biggest component of data loss was theft of equipment, followed by unexplained logger failure. Nevertheless, the quantity of continuous data records obtained far exceeds the quantity of historical manual grab sample data records and has demonstrated the competency of automated water quality observation systems to capture hydrochemical events which are not presented in traditional manual water quality monitoring systems (for example, episodic events captured at observation station X3H015). The sensor arrangement was fit to capture riverine salt fluxes across time and space, but salt budget closure to within 10% was not possible. The achieved closure was within 30% error. The computation of catchment salt fluxes with existing monthly estimates derived from manual sampling is not adequate to represent the short-duration hydrological events that occur within a catchment. The implemented observation network has characterised the water quality status of clear mountain headwaters to slow moving rivers with varying land use changes influencing stream chemistry. This range is representative of many river systems in South Africa. Thus the methods developed in this study can be modified and used for the development of continuous water quality monitoring systems elsewhere in the countryenThesis