The implications of the uncertainty associated with suspended sediment load and yield estimates for catchment management decision-making in data and resource-scarce regions

Abstract

The research described here focuses on the uncertainties intrinsic to a citizen science-based approach to directly measuring suspended sediment concentration (SSC) in a remote, data scarce area. The research took place in the rural, largely undeveloped Tsitsa River catchment in the Eastern Cape Province of South Africa and has relevance to other parts of the world where similar biophysical and socio-economic challenges to determining suspended sediment loads (SSLs) and yields (SSYs) are found. The upper parts of the Tsitsa River catchment are characterised by private and commercial land ownership, with little evidence of land degradation. Communal land tenure characterises the lower and middle catchment, where dispersive duplex soils support degraded grasslands typified by water-related erosion processes, evidenced by widespread sheet, rill, and particularly gully erosion. The study was prompted by the Department of Water and Sanitation’s (DWS) 2014 plans to build two large dams in the middle to lower parts of the catchment, and by the Department of Fisheries, Forestry, Environment’s (DFFE) concurrent launch of sustainable land management interventions in the same area (i.e. The Tsitsa Project). DWS and DFFE required measured SSL and SSY estimates to inform dam siltation rates, and to prioritise further interventions, respectively. The candidate designed, established, and oversaw the implementation of an SSC sampling and analysis programme (the focus of her Master’s research). The research described in this thesis used the resulting SSC data and aimed to investigate the uncertainty associated with SSL/SSY estimates, and to consider the implications for catchment management decision-making in this and other remote and resource-scarce regions. The study aim was achieved through four objectives: Objectives 1 and 2 identified and assessed the uncertainty associated with input SSC data and SSL and SSY estimates. The estimated SSLs/SSYs were compared between sub-catchments, and with published measured and modelled SSLs/SSYs. Objectives 3 and 4 critically evaluated the methods used to collect SSC samples, calculate discharge, undertake laboratory analysis, and estimate SSLs/SSYs. The study area was 500 km distant from Rhodes University. Proximity to the site of those undertaking the monitoring was a key factor in responding to variable conditions and thereby constraining epistemic and natural uncertainty. Local residents, here referred to as citizen technicians, collected flood-focused river water samples which were analysed in a laboratory to provide SSC (mg L-1) data. Data analysis covered the period from 1 December 2015 to 31 May 2019, although sampling continued until March 2023. Time-weighted mean concentration was used to estimate daily SSL (t day-1) from instantaneous SSC and discharge. Annual SSL was estimated using a ratio estimator, as weak SSC/discharge relationships precluded the use of sediment rating curves. SSC data was found to be representative of target high discharges for the periods that sites were sampled: The citizen technicians sampled SSC for >80% of the time that their site was active and > 78% of target high flows were typically sampled giving <20%CV variability for high SSC samples. However, sampling from the riverbank may not have been representative of channel cross-section SSC: the associated uncertainty is unknown, and further research is required to confirm assumptions that suspended sediment originates mainly from washload and is distributed homogenously across river channels in South Africa. SSL and SSY estimates both require input of discharge data. Secondary discharge data (m3 s-1) were received from two DWS gauges and, in nine previously ungauged sub-catchments, from stage data collected and calculated by other researchers as part of wider biophysical surveying for the Tsitsa Project. Hydrological variability is a major component of natural uncertainty in SSL/SSY estimation, encompassing the variable hydrological responses of a catchment to long term flood and drought cycles, seasonal rainfall trends, antecedent conditions, and individual rainfall events. Although low flows dominated throughout the project period, the Tsitsa River and its tributaries were found to be flashy, with high seasonal and annual variability contributing to natural uncertainty. Measuring flood discharges is recognised even in established national-level monitoring programmes to be challenging: In this study, limited funding coupled with the remote study area resulted in few, mainly dry-season multi-purpose field visits that allowed few, mainly low-discharge measurements. These in turn resulted in stage/discharge rating curves with a potentially significant but unknown error, particularly for high flows. Hence, the most significant source of input data epistemic uncertainty was found to be associated not with the novel CT-sampling approach to SSC data, but with the measurement and calculation of the discharge data. Annual SSYs ranged from 15 t km-2 yr-1 for Hlankomo (2016) to 1443 t km-2 yr-1 for Inxu (2018), i.e., within the 94 t km-2 yr-1 to 1509 t km-2 yr-1 estimated from reservoir studies for the Eastern Cape region. Mean SSY estimated for the entire ~4900 km2 Tsitsa River catchment (111 t km-2 yr-1) was at the low end of this range, and ~22% of published mean SSY estimated using an uncalibrated model. Mean sub- catchment SSYs from this study were comparable to those estimated from specific reservoir studies, but uncalibrated modelled estimates were typically one to two orders of magnitude higher than rates found here. The error associated with the measured SSLs and SSYs was not quantifiable, but a relative assessment of SSL/SSY uncertainty at each site was made, using criteria such as sampling coverage and record length. Comparisons between sub-catchments showed that mean SSY increased downstream, probably due to the combination of dispersive soils, communal land management practices, and prevalent gullies in the middle and lower Tsitsa River catchment. However there was no clear relationship between SSY and catchment area. It was concluded that project resources were disproportionately allocated towards collecting and analysing water samples with low SSC which comprised the majority of samples but which contributed a negligible amount to estimated SSLs/SSYs. Stratifying daily load data showed that sampling only in the wet season, or only the high flows, would have an insignificant effect on SSL/SSY precision. Decimation of daily SSL showed that sampling could be reduced to three days per week in the larger catchments. These strategies would substantially reduce both sampling and analysis costs, allowing greater temporal and/or spatial coverage. However, socio-economic circumstances in the Tsitsa River catchment mean that the resulting loss of income could lead to citizen technicians seeking other employment, severely disrupting the sampling programme. Instead, it is suggested that citizen technicians should be trained to substitute or supplement low flow samples with turbidity measurements and appropriate site- specific calibration strategies, in order to reduce sample numbers and allow improvements in laboratory analysis. Where practical, SSC sampling sites should be located near functional gauging weirs to reduce the uncertainty associated with measured discharge. In ungauged catchments, funding and equipment should be directed towards more frequently and more accurately measuring high flows to improve the accuracy of stage-discharge rating curves, although the challenge of being present for high flows, and of working safely in medium to high flows and of non-uniform channels with changing bedforms would persist. This study has shown, for the first time, that an innovative, locally-appropriate SSC sampling and analysis strategy can result in precise SSC data, confirming that it can be adapted for use in other resource-scarce regions, with benefits not only for natural resource managers and scientists, but also for people living in the study area. These data allow, again for the first time, the calibration and verification of South African sediment yield models, as well as further research into sub-catchment scale sediment/discharge relationships, for example through events and seasons. They provide a foundation for investigations into key hydrological and geomorphological processes that impact catchment management decision- making in the Tsitsa River and similar catchments. These include amongst others the location and prioritisation of land management initiatives and of water resources infrastructure design and management. Complementary research is required regarding the potential heterogeneity of sediment characteristics throughout channel cross-sections, the volume and rate of movement of the sand-sized particles that form much of the bedload in the Tsitsa River catchment, and the spatial and temporal dynamics of sub- catchment sediment sources.