Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Computational and Data Sciences

First Advisor

Hesham M. El-Askary

Second Advisor

Mohamed Allali

Third Advisor

Thomas C. Piechota

Fourth Advisor

Daniele C. Struppa


This dissertation is the accumulation of the application of adaptive, empirical learning-based methods in the study and characterization of the El Niño Southern Oscillation. In specific, it focuses on ENSO’s effects on rainfall and drought conditions in two major regions shown to be linked through the strength of the dependence of their climate on ENSO: 1) the southern Pacific Coast of the United States and 2) the Nile River Basin. In these cases, drought and rainfall are tied to deep economic and social factors within the region. The principal aim of this dissertation is to establish, with scientific rigor, an epistemological and foundational justification of adaptive learning models and their utility in the both the modeling and understanding of a wide-reaching climate phenomenon such as ENSO. This dissertation explores a scientific justification for their proven accuracy in prediction and utility as an aide in deriving a deeper understanding of climate phenomenon. In the application of drought forecasting for Southern California, adaptive learning methods were able to forecast the drought severity of the 2015-2016 winter with greater accuracy than established models. Expanding this analysis yields novel ways to analyze and understand the underlying processes driving California drought. The pursuit of adaptive learning as a guiding tool would also lead to the discovery of a significant extractable components of ENSO strength variation, which are used with in the analysis of Nile River Basin precipitation and flow of the Nile River, and in the prediction of Nile River yield to p=0.038. In this dissertation, the duality of modeling and understanding is explored, as well as a discussion on why adaptive learning methods are uniquely suited to the study of climate phenomenon like ENSO in the way that traditional methods lack. The main methods explored are 1) differentiable Programming, as a means of construction of novel self-learning models through which the meaningfulness of parameters arises from emergent phenomenon and 2) empirical decompositions, which are driven by an adaptive rather than rigid component extraction principle, are explored further as both a predictive tool and as a tool for gaining insight and the construction of models.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.