Flow alteration by diversion hydropower in tributaries to the Salween river: a comparative analysis of two streamflow prediction methodologies

A multi-model approach was applied to reconstruct long-term flow records in 32 ungauged rivers developed with small diversion hydropower stations. Potential hydrologic alteration was assessed for flow records simulated by a catchment similarity model and the multi-criteria Streamflow Prediction under Extreme Data-scarcity (SPED) framework. Model validation based on limited observed data suggests that the SPED flow predictions are substantially more accurate than those generated by the catchment similarity model (NSE of 0.74 and 0.22, respectively and Correlation Coefficient of 0.87 and 0.72, respectively). Both flow prediction techniques indicated that flow signatures were altered substantially by diversion hydropower. Mean annual flows decreased by a mean of 76–86% across the 32 rivers and flow became more predictable in most rivers (47–94% mean increase in predictability). Frequency and duration of high flows decreased and duration of low flow events increased substantially. Slopes of rising hydrograph limbs and recession limbs increased respectively by a mean of 123–161% and 254–720%. While direction of detected flow alteration was similar regardless of model choice, severity of alteration was consistently greater based on the analysis of flows simulated by the multi-objective SPED model. Overall, the agreement of the multi-model analysis indicates that the signal of flow alteration by diversion hydropower in the study rivers supersedes uncertainty associated with flow prediction. While both models may be appropriate for applications such as change detection analysis, prescriptive management actions, such as establishing flow targets for environmental flow regimes, should be based on flow records generated by models adept at simulating rainfall-runoff processes targeted to individual basins, such as SPED.