ASSESSING THE EFFECT OF RATING CURVE UNCERTAINTY IN STREAMFLOW SIMULATION USING MACHINE LEARNING MODELS (CASE STUDY OF KULFO WATERSHED, SOUTHERN ETHIOPIA

Abstract

Accurate streamflow simulation and comprehending its associated uncertainty are crucial for effective water resource management. However, the uncertainty of rating curves from which streamflow data is derived remains poorly understood. This study aims to simulate streamflow at the outlet of the Kulfo watershed under rating curve uncertainty conditions. To achieve this, the bootstrap resampling technique (BSRT) was employed to establish the rating curve and estimate associated uncertainty. Further, it integrated with Recurrent neural network (RNN) models, specifically long short-term memory (LSTM), bidirectional long short-term memory (BiLSTM), and gated recurrent units (GRU) to assess the effect of this uncertainty on streamflow simulation accuracy and reliability. Different sets of streamflow data, derived from the fitted rating curve and its lower and upper uncertainty bands were utilized to train RNN models independently, and used for scenario analysis. Different Lag times of daily and monthly areal rainfall and discharge are used as model inputs, and executed in Python programming language. By comparing the results from different rating curve uncertainty scenarios, the study evaluates the effect of this uncertainty on streamflow simulation. The findings reveal the excellent potential of RNN models for streamflow simulation, with the BiLSTM model outperforming LSTM and GRU at both temporal scales. Moreover, the study highlights that rating curve uncertainty propagates significantly onto streamflow simulations, particularly in high-flow regions. Consequently, the rating curve uncertainty in Kulfo River led to an uncertainty of the streamflow of about 17.8 m3 s -1 and 9.8 m3 s -1 , representing about 25% and 22% at peak discharge for daily and monthly simulations respectively. These findings underscore the importance of considering rating curve uncertainty in streamflow simulation to ensure accurate results. Thus, the simulated flow should not be directly used for decision-making in the watershed. Instead, streamflow should be treated as an uncertain variable and managed by incorporating rating curve uncertainty in decision-making