Advances in numerical weather forecasts have brought forward considerable societal benefits and raised expectations for higher resolution, more accurate, and longer predictions. Despite the consistent progresses achieved, the prediction of precipitation remains a less satisfyingly tackled task, with skills falling far behind those of other atmospheric variables. This dissertation serves as an inspection of prediction capacity and an exploration of predictability for the precipitation process, with a particular focus on the region of West Coast United States.

The sources of predictability, accuracy requirements, and optimal model configurations are distinct regarding the considered forecasting scales and ranges. To identify the successes and deficiencies in predictions and benchmark further advances, a seamless assessment of precipitation prediction skill for short range up to subseasonal scale range is conducted. The evaluation is based on the Subseasonal-to-Seasonal Prediction Project retrospective forecast database. The prediction skill–lead time relationship is evaluated, using multiple models, and measured by both deterministic and probabilistic skill scores. Results show advantageous deterministic skills for the evaluated models at Week-1. The best-performing models achieved r ≈ 0.6 for Week-2 predictions.

The potential sources of predictability at extended range from some of the key climate variations are investigated based on a composite of statistical evidences and numerical predictions. Results show that periods of heavy precipitation associated with ENSO are more pre- dictable at the extended range period. The excessive precipitation and improved extended- range prediction skill during ENSO periods are attributed to the meridional shift of baroclinic systems as modulated by ENSO. Through examining precipitation anomalies conditioned on the MJO, I verified that active MJO events systematically modulate the area’s precipita- tion distribution. Most of the evaluated models are still struggling to represent the MJO or its associated teleconnections, especially at phases 3–4. However, some models do exhibit enhanced extended-range prediction skills under active MJO conditions.

The advantageous precipitation prediction skill for short to medium range originates from a steady accumulation of scientific achievements in (1) inferring atmospheric initial states, (2) resolving atmospheric fluid dynamics, and (3) approximating unresolved atmospheric processes. Evaluation results suggest that we have not fully realized the potentials of these advances in fostering a corresponding improvement in precipitation prediction. Here, the old art of forecasting by reading weather chart and advances in deep learning for image recognition are combined to shed light on the precipitation prediction task from a top-down, data-driven viewpoint. A deep convolutional neural network (CNN) model is trained to learn precipitation-related dynamical features from the surrounding dynamical and moisture fields by optimizing a hierarchical set of spatial convolution kernels. The model applies an “end-to- end” learning strategy to automatically search, synthesize, and extract salient spatial features from the resolved high-dimensional atmospheric field for accurate precipitation estimation at daily scale. Experiments for different regions across the contiguous United States show that, provided with enough data, precipitation estimates from the CNN model outperform the reanalysis precipitation products, as well as the statistical downscaling products using linear regression, nearest neighbor, random forest, or fully-connected deep neural network.

The idea of “end-to-end” learning for inferring unresolved precipitation process based on resolved atmospheric field is further explored for hourly scale quantitative precipitation fore- cast. Hourly precipitation observations from various sources are collected, quality controlled, and concatenated to compose a unique long-term (1980/1/1- 2018/12/31) high temporal resolution precipitation observation dataset. A general framework for statistically modeling of spatiotemporal data and making use of inconsistently available observations is developed.

Hourly precipitation predictions using the deep neural network model give r ≈ 0.8 at 2◦ ×2.5◦ spatial scale, while the baseline numerical model achieved r ≈ 0.5. The best performance at hourly, gauge-point scale reaches the order of r ≈ 0.6 for some gauges. However, there is high skill variance in estimating precipitation at such a stringent spatiotemporal resolution. To further test the proposed model in practical forecasts, dynamical retrospective forecast experiments for two atmospheric river land-falling events are carried out using the Weather Research and Forecasting (WRF) model. The WRF dynamical simulations are used to force the trained neural network model for alternative precipitation process predictions. Simulation results verified the consistency and robustness of the proposed approach. It should be noted that the methods here are not intended to replace precipitation-related parameterization schemes using a “black box” model, rather, the target is to set a benchmark for precipitation prediction from a data-driven perspective, and offer directions for improving precipitation related parameterizations.

Overall, this work conducted a systematical evaluation of precipitation prediction skills across a spectrum of critical scales and ranges. Sources of predictability at subseasonal scale are explored based on a composite of statistical analysis and numerical prediction. The potential of deep learning for seeking evidences in improving precipitation prediction is explored by combining high quality observation data with numerical dynamical predictions.

Streamflow simulation and forecasting is an important approach for water resources management and flood mitigation, and both physically-based and data-driven models have been used to predict the streamflow at daily and hourly scale. Although physically-based models can help understand the underlying mechanisms of the hydrological process based on mathematical and physical equations, the performance of the models largely depends on the availability and quality of the spatial information. The computation cost also limits its application in the fast, reliable and accurate streamflow simulation and forecasting. In contrast, data-driven models, especially deep learning models, are capable of learning the nonlinear relationships between the inputs and outputs without explicitly referring to the physical process. In addition, deep learning techniques, powered by graphics processing units (GPUs) can allow researchers to train models 10 or 20 times faster, which is more promising for inputs-outputs simulation.

Currently, the deep learning models rely heavily on prior streamflow information to make predictions, while the hydrological observations are limited in the real world applications. In this Thesis, the effectiveness of deep learning models on streamflow simulation without using the previous streamflow information was explored. In specific, the three deep learning models the ANN, RNN and LSTM models were used to simulate the streamflow at multiple lead time and different sites using the precipitation and soil moisture. In addition, the effectiveness of deep learning models on learning from a physically-based model to generate pixel-to-pixel streamflow mapping has also been tested.

All three deep learning models can achieve reliable and accurate streamflow simulations at the lead time of 1-5 h at the Ninnekah site with Pearson correlation coefficient (r) over 0.9 for all three models. Based on the prediction of the peak and low flow on different sites, the LSTM model shows more consistent and reliable streamflow simulation results and can work better in different basin scales. For the U-net model, it can learn the patterns of ParFlow-CLM model well and achieves average r of 0.770 and 0.820 in the whole domain for validation and calibration, respectively. However, the models have limitations in predicting the peak due to the feature of the imbalanced data.

Precipitation is a crucial input variable for hydrological and climate studies. Rain gauges can provide reliable precipitation measurements at a point of observations. However, the uncertainty of rain measurements increases when a rain gauge network is sparse. Satellite-based precipitation estimations “SPEs” appear to be an alternative source of measurements for regions with limited rain gauges. However, the systematic bias from satellite precipitation estimation should be estimated and adjusted. In this dissertation, a method of removing the bias from the precipitation estimation from remotely sensed information using artificial neural networks cloud classification system (PERSIANN-CCS) over a region where the rain gauge is sparse is investigated, and the method consists of three major parts.

The first part investigates the calculation of monthly empirical quantile mapping of gauge and satellite measurements over several climate zones as well as inverse weighted distance for the interpolation of gauge measurements. Seven years (2010–2016) of daily precipitation estimation from PERSIANN-CCS was used to test and adjust the bias of estimate over Saudi Arabia. The first six years (2010–2015) are used for calibration, while one year (2016) is used for validation. The results show that the mean yearly bias is reduced by 90%, and the annual root mean square error is reduced by 68% during the validation year. The experimental results confirm that the proposed method can effectively adjust the bias of satellite-based precipitation estimations.

The second part is to provide a merged product by using the interpolated gauge estimates and the bias-corrected PERSIANN-CCS (0.04°×0.04°) estimates. Interpolated gauge estimates are calculated by the inverse weighted distance to the existing gauge points. The merged product is calculated based on the uncertainty of SPE and gauge estimation. Results show both the annual mean bias and root mean square error are reduced by 98% and 83% respectively. Statistical analysis of the merged product over different temporal scales and climate regions are discussed.

The third part of this dissertation is to examine the effectiveness of the bias-corrected approach that is based on quantile mapping and climatic zones in hydrological modeling. Three watersheds, which are Bisha, Byash, and Hail watersheds are chosen over Saudi Arabia. The evaluation is carried between 2011 and 2016 where the monthly and yearly runoff volumes are available. To model, the monthly and annual runoff volumes, Soil Conservation Service (SCS) model is implemented. Three evaluation statistics include correlation coefficient, mean bias, and root mean square error. The results show that the proposed method is effective for removing the bias from the SEPs, and for improving hydrological modeling.

It has become more important in recent years to have a general estimate on how much water storage there will be during the spring season to support urban, industrial, and agricultural usage and forecasting drought. In this thesis, ground-based data and climate indices were applied to Principal Component Regression to build a prediction model to estimate seasonal streamflow in spring with winter temperature, precipitation and other climate indices. With data gathered from PRISM, NOAA, CEDC, and USGS and then testing all possible variables and combinations, it was found that using precipitation and temperature with climate indices such as SOI and PNA can provide useful information to forcaste streamflow in spring season.

Precipitation as an essential component of the hydrologic cycle has a great importance to be measured accurately due to various applications such as hydrologic modeling, extreme weather analysis, and water resources management. Among different methods, meteorological satellites are one of the instruments that are widely used for precipitation estimation in fine spatial and temporal resolution. Precipitation Estimation from Remotely Sensed Imagery using Artificial Neural Network Cloud Classification System (PERSIANN-CCS) uses infrared (IR) data from Geostationary Earth Orbit (GEO) satellites to retrieve precipitation based on relationship between clout top temperature (Tb) and rainfall rate (RR) using a neural network technique. The complexity of Tb-RR relationship for estimating precipitation causes uncertainty in PERSIANN-CCS rainfall product. This research is focused on improving PERSIANN-CCS rainfall retrieval using several approaches:

1) Bias adjustment of PERSIANN-CCS rainfall estimates using PMW satellite rainfall data: Using multi satellite data can enhance the quality of rainfall estimation considerably; in this research we have combined the rainfall data from PERSIANN-CCS and PMW rainfall to enhance the bias of PERSIANN-CCS precipitation estimates. The results showed improvement of rainfall estimation during summer and winter time.

2) Increasing the rainfall detection by including warm clouds rainfall: PERSIANN-CCS currently cannot detect rainfall from clouds with temperature warmer than 253 K. This study explores the impacts of increasing the temperature threshold on precipitation estimation. The results show that increasing the threshold level can improve the PERSIANN-CCS rainfall detection.

3) Generating a probabilistic framework for precipitation retrieval: The current version of PERSIANN-CCS retrieves precipitation based on the exponential function fitted to Tb-RR. The major assumption behind this relationship is that the heavier rainfalls are associated with colder clouds which cause underestimation of warmer clouds and overestimation of colder clouds rainfall. The probabilistic approach uses the corresponding sample relationship between cloud temperature and rainfall rate. The model is evaluated during a full summer season which showed improvement in both detection and estimation of rainfall in compare with the current PERSIANN-CCS algorithm.

A combination of accelerated population growth and severe droughts has created pressure on food security and driven the development of irrigation schemes across sub-Saharan Africa. Irrigation has been associated with an increase in malaria risk but the underlying mechanism is not well understood. While useful in investigating transmission dynamics, malaria models are limited in irrigated areas as they typically infer mosquito abundance from rainfall. Fundamentally, this assumes rainfall as a proxy for larval habitats and can lead to contradictory transmission results. The availability of water for breeding is governed by hydrological processes which are highly non-linear and spatially variable. The incorporation of hydrologic modeling is therefore essential to understand the effect of irrigation on malaria.

The overarching goal of this dissertation is to study the impact of irrigation on the spatiotemporal distribution of malaria larval habitats and transmission using a hydrology-based malaria model. First, a three-dimensional, distributed hydrologic model was applied to simulate malaria larval habitats at a sugarcane plantation site in Arjo, Ethiopia. The scarcity of field data was overcome by integrating remotely sensed data with the model. Results suggest that at least half of the irrigated farms had a high probability of larval habitat occurrence. Irrigation dampened and prolonged the seasonality of the larval habitats, with a significant shift from semi-permanent to permanent habitats.

Land surface depressions highly influence the transformation of rainfall to ponding for aquatic larval habitat formation. Hence, I sought to improve the representation of physically meaningful surface depressions in the digital elevation model (DEM) used in the hydrologic model. A new topographic conditioning workflow, Depression-Preserved DEM Processing (D2P), was developed and evaluated through a case study in a pond-dominated watershed in the United States. D2P successfully resolved 86% of the ponds at 10 m DEM resolution. A hydrologic simulation was performed using the D2P processed DEM and demonstrated a more robust characterization of surface water dynamics.

Lastly, I extended the Arjo hydrologic modeling framework with D2P to enhance the larval habitat estimation in an agent-based malaria model and examine transmission dynamics. The inclusion of hydrologic processes increased the variability of larval habitats which resulted in significantly lower transmission. The application of irrigation enabled the development of mosquitoes in dry seasons while stabilizing growth in rainy seasons. The model also revealed that malaria transmission was sustained all-year round and intensified during the main transmission season, with the peak shifted forward by around one month. Finally, I showed how habitat heterogeneity could affect the spatiotemporal dynamics of malaria transmission. Understanding changes to malaria transmission dynamics by irrigation is important for the development of mitigation strategies. The framework presented in this dissertation particularly helps larval source management as a supplementary vector control by identifying malaria hotspots and prioritizing resources for operational planning.

Soil moisture is a vital component of the hydrologic cycle. It modulates the partitioning of infiltration and runoff as well as controls surface moisture and energy fluxes. Until recent years, it has also been one of the most overlooked components because of the difficulty to effectively measure and model it at scales most relevant to the hydrologic community. With a rise of new observation technologies and a growth in modeling efforts, this dissertation aims to investigate and establish the most useful ways to incorporate observed soil moisture information into a conceptually-based distributed hydrologic model developed for operational streamflow prediction by the U.S. National Weather Service. A common platform of soil saturation ratios for comparison between observations and model simulations of soil moisture is established using past observations mindfully combined with calculations of soil hydraulic properties based on readily available characteristics from soil surveys. Data assimilation using an Ensemble Kalman Filter is used to update conceptual model storages with observed volumetric soil moisture. The assimilation is modified to address underdispersiveness that occurs following a precipitation event and after prolonged dry periods. A second filtering step is developed to spread innovation at pixels collocated with observations to model pixels that are not “observed.” The use of soil moisture observations as a tool for calibration is also investigated, and a new two-step hybrid calibration scheme that utilizes both soil moisture observations and observed streamflow is introduced. The effects of soil moisture data assimilation and soil moisture-based calibration on simulated streamflow are also analyzed.

Development of soil saturation ratio comparison efforts and data assimilation techniques are conducted in the Russian River Basin in northern California. Both in situ soil moisture probes and Soil Moisture Ocean Salinity (SMOS) surface soil moisture estimates are tested in the data assimilation experiments. The soil moisture-based calibration development takes place in the Turkey River Basin in northeastern Iowa, and is further tested in the Russian River Basin. Results show that while both uses of soil moisture are effective in improving estimates of the soil moisture state, the calibration scheme is more stable and necessary to improve simulated streamflow.

The simulation and optimization of reservoir systems has attracted a great deal of attention in the field of hydrology and water resources management. Although many advances have been made, the gap between theoretical and real-world operation of reservoir systems still exists. Here, multiple tools and algorithms are proposed to bridge the existing gap to some extent. These tools are developed to aid decision makers, engineers, and scientists to understand, simulate, and improve the operational rules of reservoir systems.

In this dissertation, I propose an optimization framework, titled shuffled complex-self adaptive hybrid evolution (SC-SAHEL), to optimize the controlled discharge from reservoirs. This new optimization tool can solve a wide range of optimization problems, using a self-adaptive search mechanism. The algorithm employs multiple search methods from different optimization tools and selects the most suitable method for the problem space. This process reveals the potential of each search mechanism during the course of the search and enhances the efficiency and effectiveness of the search. The SC-SAHEL framework is tested on multiple benchmark problems and showed superior performance in comparison to single search methods. The framework is applied to the Folsom reservoir to maximize hydropower generation by tuning the controlled discharge. The results showed that the SC-SAHEL algorithm is superior to other single search method algorithms for finding near optimum solutions. The results also demonstrated the robustness of the SC-SAHEL framework for solving a wide range of reservoir optimization problems.

In addition to the optimization framework, a new data-mining algorithm, title generalized model tree (GMT), is proposed for simulating rule-based hydrologic systems. The new framework is developed based on the decision tree models and employs linear regression for prediction and model induction. The newly developed framework is tested on several benchmark datasets to compare its performance with other popular decision tree models. The framework can generate simple models to replicate different rule-based hydrologic systems. The simple structure of the inducted models makes them easy to implement and use in any programming languages and modeling framework. The framework is employed to simulate controlled discharge from multiple reservoirs across the Contiguous United States (CONUS). The results revealed the potential of the GMT framework for generating reservoir routing models. The models inducted by the GMT framework can be employed as a reservoir module within the hydrologic models, especially in large scale hydrologic modelling. In addition, the GMT framework model structure can reveal useful information about the underlying structure of the system. Hence, it can reveal information about importance of decision variables in the real operation of the reservoir systems.

The proposed tools benefit stakeholders to understand and improve their management practices of reservoir systems. The simulation method reveals the hydrologic response of natural systems to historical reservoir operation through reservoir routing, while shed light on the importance of decision variable in real operation of the system. The optimization algorithm finds the optimum operational rules for the reservoir systems for the specified goal and objective. The outcome of these two algorithms can be used for evaluating reservoir management practices, while considering the hydrologic behavior of the watersheds.