In many rural areas in arid and semi-arid regions, balancing agricultural and environmental water demands is a key challenge facing resource managers, which is complicated by the interconnection of groundwater and surface water resources. In California, water management decisions are increasingly supported by hydrologic models, but these can often provide confusing or overwhelming amounts of information. The Scott River watershed (HUC8 18010208), a 2,109 square km undammed rural basin in northern California, was used as a case study to develop a suite of tools, informed by a groundwater-surface water model, for water managers, including: 1) a hydrologic proxy for the ecological success of a key aquatic species; 2) applying a quantitative multi-benefit framework (i.e., incorporating ecological, agricultural, and cost objectives) to an existing groundwater planning process; and 3) two locally-tailored, seasonal, quantitative predictions of fall-season watershed behavior as a complement to historically-used water year type categories. Chapter 1 aims to quantify hydrologic conditions that support persistence of the Scott Valley coho salmon (Oncorhynchus kisutch) run. We applied the functional flows framework to characterize the hydrology of each water year measured at a key long-term stream gauge. Taking advantage of a nearly two-decade ecological monitoring dataset, we built linear models to predict coho salmon reproductive success using combinations of one and two hydrologic metric predictors. We used an ensemble of the three best linear models to formulate a Hydrologic Benefit function, summarizing the ecological services provided by the hydrology in different seasons into a single index value per water year.

In Chapter 2, we apply a multi-benefit framework to a portfolio of management actions, which were proposed for the Scott Valley jurisdiction during the recent development of a long-term Groundwater Sustainability Plan (GSP). We developed a summary statistic or proxy for each of the three primary policy objectives in the GSP (environmental, agricultural, and project cost). We then used them to summarize the results of 40 management scenarios developed for the Groundwater Sustainability Plan (GSP) in Scott Valley in Northern California, which were simulated using an existing integrated surface- and groundwater model. We found that a trade-off in benefits for fish and farms was evident in every category of infrastructure investment (a proxy for project cost), though greater infrastructure investment can achieve some reductions in this trade-off. Additionally, although the GSP management priorities emphasized infrastructure investments, both infrastructure-based and regulatory approaches fell within the Pareto-optimal set of management options under a strict application of these objective functions. Finally, regarding regulatory actions, management interventions targeted at low-flow periods produced a more efficient gain in environmental flow value per cost in agricultural productivity.

In Chapter 3, we propose methods to predict, approximately five months in advance, two key hydrologic metrics in the Scott River watershed. Both metrics are intended to quantify the transition from the dry to the wet season, to characterize the severity of a dry year and support seasonal adaptive management. The first metric is the minimum 30-day dry season baseflow volume, V_min, which occurs at the end of the dry season (September-October) in this Mediterranean climate. The second metric is the cumulative precipitation, starting Sept. 1st, necessary to bring the watershed to a "full" or "spilling" condition (i.e. initiate the onset of wet season storm- or baseflows) after the end of the dry season, referred to here as P_spill. As potential predictors of these two values, we assess maximum snowpack, cumulative precipitation, the timing of the snowpack and precipitation, spring groundwater levels, spring river flows, reference ET, and a subset of these metrics from the previous water year. We find that, though many of these predictors are correlated with the two metrics of interest, of the predictors considered here, the best prediction for both metrics is a linear combination of the maximum snowpack water content and total October-April precipitation. These two linear models could reproduce historic values of V_min and P_spill with an RMSE of 1.4 Mm3 / 30 days (19.4 cfs) and 20.7 mm (0.8 inches), respectively.

The tools developed for this case study could be of value for other local jurisdictions with similar features, including a Mediterranean and/or intermontane (snow-fed) climate, an undammed watershed, and challenges balancing agricultural and environmental water needs. The method for empirically deriving the highest-priority hydrologic functions for a threatened species could be used in other watersheds (if sufficient ecological data records are available) to evaluate trade-offs and support water management decisions in human-altered novel ecosystems. The development of basin-specific objective functions, especially ones using output from existing hydrologic models, could help quantify management decision trade-offs and improve stakeholder communication in ongoing water planning efforts throughout the region. And finally, although careful consideration of baseline conditions used as a basis for prediction is necessary, seasonal predictive indices could be used by governance entities to support adaptive management in an uncertain future climate.

Agriculture is responsible for 90% of nitrate leaching to groundwater in the Central Valley. This nitrate comes from excess fertilizer in the soil which is not taken up by crops. One method to reduce excess fertilizer, and therefore nitrate leaching, is by matching the crop’s nitrogen demand through high frequency, low concentration (HFLC) fertilizer applications. This method has been implemented at the study site, a 56-ha commercial almond orchard in Modesto, CA, since 2017. Over five years of data have been collected from a vadose zone and groundwater monitoring well network at the orchard to study how this nutrient management practice impacts nitrate leaching rates as well as the nitrate concentrations in shallow groundwater. This project develops numerical models of water flow and nitrate transport used to 1) quantify the resulting reduction in nitrate loading rates from the orchard, and 2) predict when improvements in shallow groundwater quality may be measured. The models integrate flow and transport in the unsaturated zone and shallow groundwater at the field-scale. A key challenge investigated is the role that spatial variability observed in both the geologic media and in nitrate concentrations plays in the vadose zone and in groundwater. The models demonstrate that spatial variability observed in groundwater nitrate concentrations at the sub-field scale primarily results from sub-field scale spatial variability in nitrate leaching from the vadose zone. HFLC is predicted to reduce nitrate leaching from an already well-managed orchard by 40%. Low recharge rates (7 cm/year on average) caused by dry climate and efficient irrigation make for long response times (up to 50 years or more) in field-scale improvements in shallow groundwater quality. Additionally, nitrate concentrations in groundwater recharge under HFLC are predicted to remain above the public health criteria of 10 mg/L NO3-N. These findings imply that under dry current climate conditions, reliance on nutrient efficiency goals alone may not resolve groundwater nitrate issues in the Central Valley.

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.