Search

Climate models play a vital role in predicting how climate change will impact various systems, particularly water resources, by enabling hydrologic simulations. The accuracy and reliability of simulations heavily depend on the quality of input data, including climate model outputs. Despite its improvement in replicating observations, the latest generation of Coupled Model Intercomparison Project 6 (CMIP6) still exhibits systematic biases that need to be addressed. Through simplistic assumptions, existing bias correction models have evolved without much consideration of biases in the simulated trends and frequencies which are critical for water resources management. However, considering these attributes increases the complexity of a bias correction model, potentially leading to instability in the corrected future projections. This thesis presents two time-frequency bias correction alternatives. The first utilizes discrete wavelet transform to correct the underlying trend in climate model simulations, referred to as Discrete Wavelet-based Bias Correction (DWBC). Although the time-varying trend is adequately bias-corrected, certain information in the non-dyadic spectrum is left uncorrected. To address this issue, the second alternative, Continuous Wavelet-based Bias Correction (CWBC), is proposed by utilizing continuous wavelet transform. Through the use of CWBC, current climate model simulations exhibit nearly perfect correction in terms of power spectrum and spectral density. The effectiveness of CWBC is demonstrated by correcting bias in climate model simulations that exhibit extreme changes. The final part of the thesis highlights the curse of dimensionality as a common challenge in statistical hydrology and multi-dimensional bias correction models, resulting from the significant parameterization needed to correct dependence biases across multiple variables and locations. To assess its independence from the number of variables and/or locations, CWBC is applied to correct biases in a large and complex spatially distributed multi-variables field, and it is found to perform consistently well. CWBC produces more realistic future climate model simulations by addressing biases in each climate variable independently. Given the potential of CWBC to improve the accuracy of hydrologic simulations by correcting systematic bias in climate model outputs, its use could have important implications for water resources management and hydrologic design at global scales.