Evaluation of streamflow estimates for the Rovuma River

AbstractReliable estimates of historic streamflow are important when estimating future flows and water resources availability based on factors such as climate change, population growth, and changes in land use or land cover. Many regions across the globe have limited streamflow observations. Additional information about streamflow in these basins is critical to water resources planning and economic development strategies. In southeastern Africa, the remote Rovuma River lies on the border between Mozambique and Tanzania. There are limited historic measurements in the main tributary, the Lugenda River, and no publicly available observations from recent years. Improved knowledge of the water resources availability and seasonal and annual variability of this river will enhance transboundary river basin management discussions. A combination of methods, including index-gauge methods and a macro-scale hydrological model are used to estimate historic streamflow conditions in the Rovuma River. These methods incorporate data from remote sensing, gridded global soil data, a composite runoff dataset, and in situ observations. The hydrological model was tested in a nearby gauged basin yielding a Nash–Sutcliffe efficiency ratio of 0.8, an efficiency ratio based on mean historical streamflow by month of 0.6, an efficiency ratio based on inverse flows (sensitive to low flows) of 0.9, and a coefficient of determination equal to 0.99. In the Rovuma River, the mean and standard deviation of the index gauge-estimated mean monthly flows agree with streamflow estimates using the Variable Infiltration Capacity (VIC) hydrologic model with a 0.25 decimal degree spatial resolution. A closer look at precipitation records suggests that the model results provide a more accurate historic flow record than the index gauge methods due to small-scale precipitation events. Model inputs and results are evaluated by leveraging available regional in situ data in comparison to remote sensing data input data. Uncertainties in the streamflow estimates are high, however, additional in situ measurements can reduce these uncertainties. This combination of methods could prove useful for estimating flows in other rivers in southern Africa and other regions with intermittent or sparse streamflow observations