John Ndiritu

A variable-length block bootstrap method for multi-site synthetic streamflow generation

Abstract The complexity of stochastic streamflow generators limits their practical use, highlighting the need for effective but simpler approaches. An attempt to meet this objective is presented using variable-length bootstrapping (VLB) of annual flows, and a weighted method of fragments for disaggregation and perturbing the bootstrapped annual flows. The perturbations enable generation of annual flows different from those present in the historical record, thereby overcoming one of the main limitations of the classical bootstrap method. The VLB replicates adequately nine historical annual statistics of a five-site problem, and reproduces the annual serial and cross-correlations better than STOMSA—a state-of-the-art parametric generator. The VLB achieves reasonable validation using the sum of minimum flows and the reservoir storage size test. Because of the modification of the monthly flow distribution caused by the weighted averaging of fragments, the VLB cannot be safely used for within-year analysis, but is a potentially robust annual streamflow generator. Citation Ndiritu, J. (2011) A variable length block bootstrap for multi-site synthetic streamflow generation. Hydrol. Sci. J. 56(3), 362–379.

Yield–reliability analysis for rural domestic water supply from combined rainwater harvesting and run-of-river abstraction

Abstract In many rural areas, reticulated water supply from large water resources schemes does not exist and many households obtain water from groundwater (GW), rainwater harvesting (RWH), run-of-river (ROR) flow, or combinations of these. While comprehensive yield–reliability analysis for large water resource systems is the norm, typical rural water supply analysis unrealistically aggregates data into monthly or annual time steps and does not incorporate reliability. A daily time-step simulation of household supply and frequency analysis of the number of days supplied in each year is used herein to: (a) demonstrate a realistic approach of yield–reliability analysis for RWH, ROR and combined RWH and ROR supply; and (b) show how combined utilization of the two can lead to improved supply. Integration of RWH and ROR supply has been found to improve yield and reliability substantially. The limitations of using the simpler mass curve analysis and the effect of applying a monthly time step are demonstrated. Citation Ndiritu, J., Odiyo, J. O., Makungo, R., Ntuli, C. & Mwaka, B. (2011) Yield–reliability analysis for rural domestic water supply from combined rainwater harvesting and run-of-river abstraction. Hydrol. Sci. J. 56(2), 238–248.

Using data-derived perturbations to incorporate uncertainty in generating stochastic areal rainfall from point rainfall

Abstract Raingauge measurements are commonly used to estimate daily areal rainfall for catchment modelling. The variation of rainfall between the gauges is usually inadequately captured and areal rainfall estimates are therefore very uncertain. A method of quantifying these uncertainties and incorporating them into ensembles of areal rainfall is demonstrated and tested. The uncertainties are imposed as perturbations based on the differences in areal rainfall that result when half of the raingauges are alternately omitted. Also included is a method of: (a) estimating the proportion rainfall that falls on areas where no gauges are located that are consequently computed as having zero rain, and (b) replacing them with plausible non-zero rainfalls. The model is tested using daily rainfall from two South African catchments and is found to exhibit the expected behaviour. One of the two parameters of the model is obtained from the rainfall data, while the other has direct physical interpretation. Editor D. Koutsoyiannis; Associate editor C. Onof Citation Ndiritu, J., 2013. Using data-derived perturbations to incorporate uncertainty in generating stochastic areal rainfall from point rainfall. Hydrological Sciences Journal, 58 (8), 1704–1717.