Date of Award
Master of Science
Over the years there is an increasing evidence of climate change on the available water resources. The interaction of hydrological cycle with climate variability and change may provide information related with several water management issues. The current study analyzes streamflow variability of the United States due to large-scale ocean-atmospheric climate variability. In addition, forecast lead-time is also improved by coupling climate information in a data driven modeling framework. The spatial-temporal correlation between streamflow and oceanic-atmospheric variability represented by sea surface temperature (SST), 500-mbar geopotential height (Z500), 500-mbar specific humidity (SH500), and 500-mbar east-west wind (U500) of the Pacific and the Atlantic Ocean is obtained through singular value decomposition (SVD). For forecasting of streamflow, SVD significant regions are weighted using a non-parametric method and utilized as input in a support vector machine (SVM) framework. The Upper Rio Grande River Basin (URGRB) is selected to test the applicability of the proposed forecasting model for the period of 1965-2014. The April-August streamflow volume is forecasted using previous year climate variability, creating a lagged relationship of 1-13 months. To understand the effect of predefined indices such as El Nino Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO), and Atlantic Multidecadal Oscillation (AMO) on the regional streamflow, a wavelet analysis is also performed for regions developed by from 2014 National Climate Assessment (NCA). Moreover, different SVD approach is performed for streamflow of each of the six NCA regions named as Great Plains, Midwest, Northeast, Northwest, Southeast, and Southwest. In regional case, SVD is applied initially with streamflow and SST; and that spatial-temporal correlation is later correlated with Z500, SH500, and U500 separately to evaluate the interconnections between climate variables. SVD result showed that the streamflow variability of the URGRB was better explained by SST and U500 as compared to Z500 and SH500. The SVM model showed satisfactory forecasting ability as the observed and forecasted streamflow volume for different selected sites were well correlated. The best results were achieved using a 1-month lead to forecast the following 4-month period. Overall, the SVM results showed excellent predictive ability with average linear correlation coefficient of 0.89 and Nash-Sutcliffe efficiency of 0.79. Whereas regional SVD analysis showed that streamflow variability in the Great Plains, Midwest, and Southwest region is strongly associated with SST of ENSO-like region. However, for Northeast and Southeast region, U500 and SH500 were strongly correlated with streamflow as compared to the SST of the Pacific Ocean. The continuous wavelet analysis of ENSO/PDO/AMO and the regional streamflow patterns revealed different significant timescale bands that affected their variation over the study period. Identification of several teleconnected regions of the climate variables and the association with the streamflow can be helpful to improve long-term prediction of streamflow resulting in better management of water resources in the regional scale.
This thesis is Open Access and may be downloaded by anyone.