Climate models project stronger warming in the warm season than in the cool season over much of the Western U.S. Across much of this region, the widely-accepted hydrologic signature of climate change is reduced winter snow accumulation and earlier streamflow peaks, which are mainly caused by cool season (October-March) warming. However, the warm season (April-September) warming effect receives far less attention. One of the few studies that investigated the warm season warming using a single land surface model showed that while runoff timing shifts consistently under cool season warming, runoff volume changes are mostly attributable to warm season warming in the western U.S. However, what remains unclear are the mechanisms controlling runoff responses to asymmetrical (warm/cool season) warming, as well as the extent to which the previous results are model-dependent. To answer the questions, we expand the earlier work to include experiments with four land surface models over the four major river basins of the West, with a focus on investigating the reasons causing differences in annual and seasonal streamflow response to asymmetric seasonal warming. Our results show that: (i) the general features of seasonal and annual streamflow responses to asymmetric warming are consistent across models, although the magnitudes vary; (ii) basins with higher ratio of warm to cool season gross incoming water, cooler summer, and colder winters have the strongest relative annual streamflow decreases for warm season warming than cool season warming, as a consequence of strongly influenced evapotranspiration-temperature sensitivity.

This study investigates how radiation and other forcing factors, such as temperature, precipitation, and snow water equivalent (SWE) influence snowmelt runoff in 20 river basins in the Western United States. The forcing values are derived from gridded observations of precipitation and temperature, and other surface variables (such as downward solar and longwave radiation, and humidity). Using daily runoff from U.S. Geological Survey streamflow records, this paper calculated several indexes to describe the timing and magnitude of snowmelt induced spring runoff peaks and rising limb of the hydrograph. Pearson correlation coefficients is used to analyze the relationships between forcings and runoff indexes. The results show that snowmelt runoff tends to be earlier and lower at higher temperature. With higher precipitation or SWE, snowmelt runoff tends to be later and larger. For most sites, with higher net radiation, snowmelt runoff tends to be earlier. But only for some of the sites, with higher net radiation, snowmelt runoff tends to be smaller. For other sites, the magnitude of snowmelt runoff is not correlated with net radiation.

Droughts and floods are among the most catastrophic yet least understood weather and climate threats. Accurate forecasting of droughts and floods is crucial due to their significant financial and human impacts. Concurrently, precise streamflow simulation is critical for effective water management and disaster prevention. The subseasonal drought forecast, vital for water management and disaster mitigation, has been under-studied due to a lack of appropriate meteorological forecast databases until recently. The NOAA's National Water Model (NWM), anchored by its hydrological core Noah Multi-parameterization (Noah-MP), needs a comprehensive comparison with SAC-SMA model-based River Forecast Center (RFC) forecasts to evaluate its flood forecasting efficiency in the Western United States. Accurate daily streamflow predictions are crucial, and a comprehensive, calibrated Land Surface Model (LSM) parameter set is important for reliable streamflow predictions.This dissertation explores evaluations and enhancements of hydrological simulations and forecasts in the Western U.S., with a focus on three key aspects: a) subseasonal forecast accuracy for drought onset and termination using NOAA’s Climate Testbed Subseasonal Experiment (SubX) reforecasts, b) the flood forecasting capabilities of the Noah-MP in comparison to current RFC forecasts, and c) the development of high-resolution calibrated parameters for two notable LSMs, the Variable Infiltration Capacity (VIC) model and Noah-MP. For the first aspect, I employ SubX to drive Noah-MP and produce drought forecasts of different severity and for lead weeks 1-4. I find significant drought termination and onset forecast skill within the initial two weeks and limited skill or no skill at week 4 regardless of drought severity. I find that skill is generally higher for drought termination than for onset for all drought events and that drought prediction skill generally decreases from north to south for all drought events. For the second aspect, I start with selection of appropriate physics options for Noah-MP and calibration of parameters in seven watersheds that form a transect along the U.S. Pacific Coast. I find promising flood prediction capacities of Noah-MP in northern basins but requires refinement for southern basins both in terms of bias and variability. For the third aspect, I develop calibrated and regionalized hydrologic parameters for VIC and Noah-MP at a precision of 1/16� latitude-longitude resolution across 4816 HUC-10 basins in the Western U.S., aiming to enhance the accuracy of hydrological modeling and predictions. In summary, the dissertation provides important contributions to understanding and improving the hydrological simulation and forecast in the Western U.S.

A body of work over the last several decades has demonstrated that most major floods along the U.S. West Coast are attributable to atmospheric rivers (ARs). Recent studies suggest that observed changes in extreme precipitation associated with a general warming have not necessarily lead to corresponding changes in floods, and changes in antecedent hydrological conditions could be a primary causal mechanism. This study examines climate change impacts on AR-related floods and their modulation by antecedent soil moisture (ASM) conditions in three watersheds that form a transect along the U.S. Pacific Coast: the Chehalis River basin in Washington State, the Russian River basin in Northern California, and the Santa Margarita River basin in Southern California. All three basins are rain-dominant and frequented by ARs. I used the Distributed Hydrology-Soil-Vegetation Model (DHSVM), a spatially distributed hydrological model, to reconstruct floods in all three basins.

For historical AR flooding, based on a combination of observed and model-simulated soil moisture in the Russian River basin, my results show that the storm runoff-precipitation ratio during extreme precipitation events is much more strongly related to ASM than to storm total precipitation. If ASM is low, extreme precipitation may not lead to extreme discharge. When I used Global Climate Model (GCM)-projected future atmospheric forcings (primarily precipitation and temperature) in the three basins, my results show that the projected fraction of AR-related extreme discharge events slightly decreases in the Chehalis basin, but increases in the Russian and Santa Margarita River basins. These changes in California are driven by increases in AR-related extreme precipitation events, as well as projected increases in year-to-year volatility of annual precipitation, which increases the likelihood of concurrent occurrence of large storms and wet ASM. I also investigated the subseasonal forecast skill of AR-related flooding using the NOAA/Climate Testbed Subseasonal Experiment (SubX) database, applied to both AR- and non-AR related floods. I found that flood forecast skill drops quickly after week 1. There is some probabilistic forecast skill in week 2, but only a hint of skill in weeks 3-4, especially for annual maximum floods, notwithstanding some probabilistic skill for smaller floods in weeks 3-4.

Climate change brings about several new hydrological changes including changes in intensity and frequency of extreme events such as drought and extreme precipitation. For a more accurate understanding of the impact of these extreme events on water resources, work is being done to move towards methods capable of accounting for nonstationarity. In this work, I examine how the presence of spatial nonstationarity and temporal nonstationarity can affect water resource management in two contexts. First, I consider drought in California linked to warming winter temperatures. The spatial nonstationarity present in California temperature records complicate attempts to create gridded datasets from station observations, and ability of the interpolation method to account for spatial nonstationarity greatly impacts the final products. I examine five gridded datasets to find substantial differences between winter temperature trend magnitudes and spatial patterns largely due to the interpolation method selected as well as to differences in station lists and bias correction methods. The differences in temperature also strongly impact modeled changes in Snow Water Equivalent (SWE) over the last century, which are a vital component of water resources in California. Second, I examine extreme precipitation using Monte Carlo simulations representing events typical in the US to determine the bias and variance in error associated with using a temporally stationary versus temporally nonstationary model. In comparing the performance between a stationary and nonstationary version of the generalized extreme value distribution, I find the classic bias-variance tradeoff in modeling limits the usefulness of using a nonstationary distribution, which should be applied when record length is long, the coefficient of variation of the data is low, and the behavior of temporal nonstationarity is high (large increases in the mean and standard deviation of the data with time). This result complicates the reality of transferring from a stationary method of analysis to a nonstationary, revealing that for many stations in the US, record lengths are too short (fifty years and under) for the nonstationary application to be definitively useful.

The Colorado River is the main source of surface water for the Southwestern U.S. and Mexico. It is heavily regulated by two large reservoirs, and many smaller ones. Although summer precipitation is about the same amount as winter precipitation averaged over the basin, runoff is mainly generated from the melting of snowpack accumulated during the cold season. Therefore, fresh water availability is challenged by warming temperatures that have occurred over the last few decades. A noticeable downward trend in the basin’s naturalized streamflow appears to be related to the increasing temperatures, as precipitation changes have been small. In this dissertation, I evaluate the effects of climatic controlling factors on two major hydrologic components (runoff and snow water equivalent) of the region’s water cycle. The primary tool I use is offline land surface simulations from macro-scale hydrological models. In particular, this dissertation comprises three studies that have or will be published as journal articles. First, I employed the Variable Infiltration Capacity (VIC) model to investigate the causes of the century-long decreasing trend in Colorado River streamflow. I separated the influences of warming temperatures and unevenly (spatially) distributed (and mostly small) precipitation changes by conducting a parallel set of experiments that isolated the various effects. My experiments suggest that more than half of the downward trend in streamflow is attributable to warming temperatures. Compared with an earlier drought period (1953-1968) caused mainly by insufficient precipitation over the UCRB, about half of runoff losses during the Millennium Drought (2000-2014) is attributed to warm temperature. Second, I explored the factors that control snow ablation processes across the West and differences in their representation in a multi-model suite. I selected ten USDA Snow Telemetry (SNOTEL) stations distributed across the mountainous ranges of the Western U.S. Consistent with earlier studies, I find that during the ablation period net radiation generally has stronger effects on melt rates than does air temperature. However, estimates of melt rates vary greatly across the models, in part because of differences in the way they represent the effects of vegetation on the surface energy balance. The canopy effect of each model on snow melting is also evaluated with parallel experiments. Finally, I reconstructed snowpack in the Upper Colorado River Basin (UCRB) over the last 67 years (1949-2015) using the macroscale VIC model implemented at 1/16� latitude-longitude spatial resolution. I then investigated the storms that were associated with accumulation of snow water equivalent (SWE) using 86 SNOTEL stations distributed across the UCRB. In particular, I classified storms associated with SWE accumulation into Atmospheric River (AR) related and non-AR events. The storms I identified (both AR and non-AR) during the study period account for an average of 78.2% of annual peak SWE. On average, 69% of the storms are AR-related; they contribute 56.3% of the annual snowpack maxima. I find no statistically significant basin-wide trends in the number of storms (of either type) or their contributions to SWE. However, in the middle of the basin, there are a number of grid cells with significant upward trends in the storm contributions to snow, which suggests some movement of snow accumulation towards the UCRB mid-zone over the last few decades.

Over 26% of the world's land area and ~8% of its population depend on snowmelt as the primary water source. A typical example is the western U.S. (WUS), where snowmelt has been the center of streamflow research for the past two decades. Under climate warming, snow's contribution to streamflow will decrease, increasing the uncertainty of streamflow fluctuations and challenging streamflow forecasting and management that are based on traditional snow-based methods. Addressing these challenges requires a deep understanding of the linkages among warming, snowmelt, and other runoff-generating processes and how they affect future streamflow. In my dissertation, I take the WUS as an example, using hydrologic modeling and in-situ data to explore the following questions: 1. How does seasonal warming affect annual streamflow in different watersheds, and why? 2. What composes runoff during water-scarce seasons, and how does the composition change under warming? 3. What mechanism dominates annual runoff decline under warming in snow-affected areas, and how do its dynamics affect the evolution of runoff? In 1, I found and explained an asymmetrical pattern that controls annual streamflow response to seasonal warming, which indicates the cooler, inland region streamflow is more sensitive to warm season warming, whereas warmer, coastal region streamflow is more sensitive to cool season warming. I found this asymmetry is explained by the bell-shape variation of evapotranspiration-temperature sensitivity as a function of increasing temperature, filling a long-lasting gap in the sub-annual linkages between water and climate in WUS. In 2, I developed a custom-period water source partition algorithm that uses a handful of output variables from land surface models to quantify the fractional contribution of custom-period rainfall to custom-period streamflow. Using this algorithm, I detected an increasing contribution of seasonal rainfall to summer streamflow under warming, which shed light on the evolution of a new dynamic for summer streamflow generation in the water-short WUS. In 3, I explored the ongoing runoff decline under warming across major WUS snow-covered regions and explored the linkage between the evolving trend dynamics with changing snow. In three major snow-covered basins, I discovered a smaller warming-sensitivity of runoff under a warmer climate with fixed precipitation, which could break the long-lasting static mechanism paradigm that governs hydrologic dynamics and a more rigorous assessment of future streamflow. In summary, the dissertation provides important contributions to understanding ongoing and future runoff evolution in WUS snow-affected regions under a warmer future.

The flood that would result from the greatest depth of precipitation “meteorologically possible”, or Probable Maximum Precipitation (PMP) is used in the design of dam spillways and other high-risk structures. Historically, PMP has been estimated by scaling precipitation depth-area-duration relationships obtained from severe historical storms. Over the last decade, numerical weather prediction models have been leveraged to instead predict precipitation resulting from the addition of moisture at the model boundaries (called relative humidity maximization, or RHM). Despite the major improvement this represents, model-based methods have not yet been applied to produce operational PMP estimates. Several sources of uncertainty still limit their applicability: (i) model uncertainty, (ii) uncertainty regarding whether maximum storm efficiency (moisture conversion to precipitation) was achieved by observed storms and (iii) uncertainty caused by a lack of physically-constrained guidance on moisture amplification. These uncertainties result in differences in model-based PMP estimates among studies that used different model setups, datasets and moisture amplification approaches, which limits confidence in the estimates: could PMP be much larger than a given study indicates? Or are we over-designing for unrealistically large events? Focusing on mountainous watersheds in the coastal western U.S. affected by atmospheric river (AR) storms (the Feather and Willamette River basins among others), I seek to identify the largest sources of uncertainty affecting model-based PMP estimates and ways to reduce these uncertainties. Using the Weather Research and Forecasting (WRF) regional atmospheric model, I performed experiments in which I reconstructed and amplified several severe storms. First (Chapter 2), I produced an ensemble of simulations to characterize model uncertainty (parametrization and propagation of model error). I found that the spread in PMP totals among ensemble members was modest, ranging from +/- 7% of the ensemble mean. Next (Chapter 3), I simulated artificial storms selected from a large sample (~1500 years) produced by a global climate model (the CESM2 large ensemble, or CESM2-LE). I found that the largest PMP estimates they produced was only 8% larger than estimates obtained from observed storms. Finally (Chapter 4), I found that the way moisture maximization is implemented has the largest impact on PMP estimates (with saturation of the atmosphere resulting in up to 20% larger precipitation totals than other approaches). I as a result developed a method that uses climatology to objectively determine the magnitude of moisture amplification. The understanding of the uncertainties in model-based PMP estimates and the tools I developed to characterize and reduce these uncertainties provide key information for the development of more reliable model-based PMP estimation guidance.

Water is a key resource in arid Central Asia (CA) and is heavily affected by climate change and human activities. Temperature across the region has increased drastically especially in the mountain region while precipitation change is less homogeneous. The increased temperature has caused increased melting of glacier and snow which has a large contribution to the runoff in rivers. Human activities such as agriculture irrigation and reservoir management also affect water availability. In the Soviet era, agriculture in CA expanded continuously and large amount of water was extracted from rivers for irrigation. This has caused the catastrophic decline of the Aral Sea. In the post-Soviet era, countries in CA have reorganized their agriculture structure to be self-sufficient. It is important to understand how these changes affect water availability in CA especially under climate change. This dissertation uses lakes as proxy indicators of water availability and assesses how climate and human activities have affected lakes in CA. Seventeen lakes located in three former Soviet republics and western China from seven basins are examined using remote sensing and hydrologic modeling to estimate their changes in area, water level and volume. Agriculture area changes in these basins from seven countries are also examined using remote sensing. It is found that 1) lakes located in the mountains have generally expanded due to the melting glaciers and snow; 2) lakes located in the lowlands have remained relatively stable due to the relative stability of agriculture area; 3) reservoirs exhibit different seasonal patterns due to their major function as power generation reservoirs release water during the winter while irrigation reservoirs release water during the summer; 4) agriculture area in the former Soviet Central Asia republics is highly dependent on precipitation due to the lack of efficient irrigation infrastructure while agriculture in China has continuously expanded due to the adoption of drip irrigation and groundwater extraction. In conclusion, climate is the more dominant factor affecting water availability especially in the mountains causing the lakes to expand while agriculture irrigation has offset some of the surplus in the lowlands causing the lakes to remain relatively stable.