UC Davis

This dissertation presents three interconnected studies that leverage advanced computational techniques, including Natural Language Processing (NLP), Computer Vision, Machine Learning, and Big Data analytics to gain insights into various aspects of hydrologic sciences and drought research.

In the first study, we applied NLP to assess topic diversity in approximately 75,000 research articles from eighteen water science and hydrology journals published between 1991 and 2019. We found that individual water science and hydrology research articles are becoming increasingly diverse in the sense that, on average, the number of topics represented in individual articles is increasing, which may be a sign of increasing interdisciplinarity. This is true even though the body of water science and hydrology literature as a whole is not becoming more topically diverse. Topics with the largest increases in popularity were Climate Change Impacts, Water Policy & Planning, and Pollutant Removal. Topics with the largest decreases in popularity were Stochastic Models and Numerical Models. At a journal level, Water Resources Research, Journal of Hydrology, and Hydrological Processes are the three most topically diverse journals among the corpus that we studied.

The second study focused on understanding the relationship between droughts and drought awareness, which is crucial for decision-making, policy development, and socioeconomic outcomes related to water management and conservation strategies. We used computer vision (UNet models) to analyze nonlinear, lagged correlations between Standardized Precipitation Evapotranspiration Index (SPEI) and Google Trends Search Interest within the Continental United States (CONUS). We also used Twitter data to asses people's sentiments about droughts. The most important drivers of this relationship are the variability and ranges of drought trends and severity, as well as climatic extremes. This relationship was the strongest for Western states, followed by Northeastern, Southeastern, and Central regions. Search interest tends to lag droughts by a period of ~1-3 months. We also found evidence that reductionist linear approaches, such as Principal Component Analysis, might not be as effective as UNet models in capturing the nuanced relationship between droughts and drought awareness at various dimensions and scales. We subsequently applied sentiment analysis on a set of 2.5 million georeferenced tweets related to droughts and found that people's sentiments towards drought have become increasingly positive with decreasing neutral sentiments since 2014 within the United States.

In the third study, we propose a novel approach for global drought prediction using the Vision Transformer (ViT) model, leveraging its ability to contextually learn spatial and temporal patterns from high-dimensional climate data. Using a sliding window approach, we trained the ViT model on a global dataset spanning from January 1970 to December 2004, using Sea Surface Temperature (SST), 2-meter Air Temperature (T2M), and Total Precipitation (TP) as input variables, and the Standardized Precipitation Evapotranspiration Index (SPEI) (looking ahead 0, 1, and 2 months) as the target variable. The model's performance is evaluated on a test dataset from January 2005 to December 2020 using accuracy, precision, recall, and F1 score metrics. Our results demonstrate the ViT model's effectiveness in predicting drought occurrences, with high accuracy scores ranging from 0.9456 to 0.9475 and precision scores from 0.8747 to 0.8781 for a three-month prediction horizon. The model's relatively lower recall scores (0.6285 to 0.6465) indicate room for improvement in capturing all drought occurrences, particularly in regions with complex or sporadic drought patterns. The findings of this study indicate substantial potential of the ViT model in predicting increasingly complex meteorological drought occurrences on a global scale.

Collectively, these studies contribute to the advancement of hydrologic sciences by providing operational tools and insights for researchers, policymakers, and all stakeholders in the field of water resources science and management.