Repository logo

Evaluating the potential of using satellite earth observation data to quantify the contribution of riparian total evaporation to streamflow transmission losses.

Thumbnail Image



Journal Title

Journal ISSN

Volume Title



Numerous perennial rivers which flow through arid and semi-arid environments in South Africa, have become severely constrained as water resources abstractions are close to exceeding, or have exceeded the available supply and ecosystem resilience. This is a common phenomenon, as river basins are increasingly developed and often over allocated, in order to maximize socio-economic benefits through consumptive water use, often at the expense of the environment. Thus, managing and maintaining environmental water requirement (EWR) flow allocations in these circumstances becomes increasingly important but all the more challenging, especially during periods of water scarcity. The Letaba River situated in the semi-arid north-eastern region of South Africa is a typical example of a river system in which water governance challenges and infrastructural development have resulted in flows within the river no longer resembling the natural flow regime. This situation has improved to some extent after the establishment of river operating rules and an adaptive operational water resources management system. However, one of the major challenges with successfully implementing and managing EWR flows to date has been the uncertainty regarding the magnitude and influence of streamflow transmission losses (TL’s) on flows within the river system. TL’s along the Letaba are thought to be a significant proportion of streamflow during dry periods and this therefore constrains the ability to meet target EWR flows, as it is often the case that specified EWR releases from the Tzaneen dam are not adequately met further downstream at EWR target gauges. To ensure that water provisions and in particular EWR flows can be managed more effectively and efficiently in the future, it is imperative that the hydrological processes contributing to TL’s are quantified at various spatial and temporal scales. Considering this statement as a point of departure, the overall objective of this thesis was to reduce the uncertainty associated with TL’s by attempting to acquire an improved hydrological process understanding of the natural drivers of loss in this system, so that TL’s along the Letaba River can be more accurately quantified. This research involved, conducting detailed characterizations of hydrological processes along a 14 km reach of the Groot Letaba River which has similar land use activities and hydrological characteristics to the broader river system. Particular emphasis was placed upon establishing the influence of riparian total evaporation (inclusive of open water evaporation) on TL’s, as this process is a major contributing factor to the water balance of arid and semi-arid environments, yet has seldom been incorporated or adequately represented into TL’s estimation procedures. These investigations were centred on evaluating the potential of using a satellite-based approach to acquire spatially explicit estimates of evapotranspiration (ET) during the low flow period in this river system (May to October), which typically represents a critical period with regards to water shortages. For this purpose, the satellite-based surface energy balance (SEBS) model and satellite earth observation data acquired from Landsat and Moderate-resolution imaging spectroradiometer (MODIS) were used to estimate ET. However, the trade-off between the spatial and temporal resolution associated with these data sets can limit the reliability of satellite-based ET modelling (except where occasionally correct). Consequently, the SEBS ET estimates from these data sets were used as inputs to two relatively simplistic approaches (actual crop coefficient or Kcact and output downscaling with linear regression or ODLR) to quantify ET at a moderate spatial resolution (30 m) on a daily time step. These ET estimates were compared against in-situ ET estimates using a one sensor Eddy Covariance system to quantify any uncertainties associated with the satellite-derived estimates. To further investigate spatial and seasonal variations in source contributions to plant water uptake during the investigation period, stable isotope analysis (of 18O and 2H) and a Bayesian mixing model were coupled with the satellite derived ET estimates. The insights acquired from these investigations, were then used to derive baseline estimates of TL’s. This involved using the satellite-derived daily ET time series in conjunction with data obtained from a parallel investigation focusing on quantifying the rapport between surface and sub-surface water storage processes. Initial comparisons of ET estimates acquired using the Kcact and ODLR approaches against ECET were fairly poor yielding RMSE values of; 1.88 and 2.57 mm d-1 and 1.10 and 2.39 mm d-1 (for two replicate transects), respectively. The poor performance of these techniques was largely attributed to the SEBS ET estimates used as inputs to these techniques, as SEBS may overestimate evapotranspiration during conditions of water stress. This limitation was overcome using an evaporative calibration factor (termed the environmental stress factor or ESF) into the original SEBS formulation (SEBS0), to correct for the overestimation of the latent heat flux (LE) and the evaporative fraction (EF). The ESF calibration factor was empirically derived and then integrated into SEBS0, so as to better represent the influence of water stress on the EF and consequently LE. The implementation of the modified version of SEBS (SEBSESF) was shown to significantly improve the estimation of energy fluxes, which in turn resulted in an improved correlation and an increase in the percentage of modelled ET estimates within an acceptable accuracy range (± 15 to 30 %) when compared against in-situ observations. Through the application of this modified version of SEBS (SEBSESF), the ability of the ODLR and Kcact approaches to develop a time-series of daily moderate spatial resolution ET estimates could now be demonstrated. The use of SEBSESF ET estimates as inputs to the Kcact approach was shown to compare most favourably to ECET, yielding correlation coefficient and Nash-Sutcliffe efficiency values of 0.79 and 0.60, respectively. With the ability of this satellite-based approach to adequately represent ET within this environment now confirmed. Stable isotope analysis (of 18O and 2H) and a Bayesian mixing model were coupled with the Kcact derived ET estimates, to further investigate spatial and seasonal variations in plant water uptake dynamics. The results of these investigations showed that soil water was the main contributing source to ET. While stream and groundwater use during transpiration was also prevalent within the study area and increased with aridity, the magnitude of the contribution of these sources to transpiration was fairly minimal and not as significant as generally reported in literature. The insights gained from these investigations, as well as those obtained from the quantification of surface and sub-surface water storage processes, assisted in deriving baseline estimates of TL’s along the length of river reach studied. In general, it was found that during the latter stages of the dry season (August to October) TL’s accounted for approximately 5 to 15 % of the flow in the river system, with riparian total evaporation and in particular transpiration the dominant contributing processes to this loss. Through linkages with the recent gazetting of the Letaba Management Class (resource objective setting) and the mandatory implementation of EWR flows, it was shown that flows within the river system were unable to meet low flow targets and are required to be increased in order to fulfil this requirement, whilst simultaneously accounting for TL’s. It should be noted that while the various investigations undertaken in this study enabled the estimation of TL’s and the contribution of processes viz. riparian ET to TL’s, the estimates provided could not be verified due to the lack of reliable upstream (inflow) flow gauge data. Although the investigations and observations detailed in this study provide an understanding of the system for a limited period in time, they would substantially benefit from longer-term monitoring, so that the assumptions and related uncertainties that had to be factored into the analysis could be reduced. Overall the study has detailed key hydrological processes influencing TL’s along the Groot Letaba River, providing invaluable insights on existing knowledge gaps and contributing new knowledge to this research area. It is envisaged that this will enable the establishment of an improved conceptual understanding of the system, which may prove to be beneficial for future hydrological modelling applications in this region.


Doctor of Philosophy in Hydrology. University of KwaZulu-Natal. Pietermaritzburg, 2017.