Journal of Hydrologic Engineering

The Susquehanna River basin (SRB) is the largest U.S. watershed ($71{,250\mathrm{km}}^2$) draining to the Atlantic Coast. It encompasses portions of New York, Pennsylvania, and Maryland. Given that annual maximum flood events commonly result from either rain-on-snow or hurricanes/tropical storms, determining the potential impacts of climate change on flooding behavior is especially challenging. This paper presents a modeling system that captures these dominant flooding processes, which is well-suited for future research investigating the impacts of regional climate change. For this study, a coupled hydrologic-hydraulic model is developed and used to estimate hourly streamflow for the period from 2000 to 2008, capturing a range annual maximum discharge phenomenon (e.g., rain-on-snow, localized convective events, and hurricanes/tropical storms). The three-layer variable infiltration capacity (VIC-3L) model is used to generate surface runoff and the vertical flux of water through the root zone at a scale of 0.025° (about 2.8 km), which are used as inputs to the Hillslope River Routing (HRR) model that operates on an irregular grid with a mean length scale of 4.7 km to simulate lateral surface and subsurface transport and channel hydraulics. The coupled model is validated using USGS daily streamflow, snow water equivalent (SWE) derived from the advanced microwave scanning radiometer for EOS (AMSR-E) satellite and snow depth from in situ measurements. The coupled model (VIC-HRR) shows good performance for both seasonal baseflow patterns and large flood events (e.g., rain-on-snow and hurricane/tropical storms). Given the SRB is commonly subjected to two types of flood events, the role of snow processes is investigated. Comparing synthetic model scenarios with and without snow processes suggests that if future climate conditions reduce winter snowfall due to warmer temperatures, but maintain total precipitation levels, annual runoff will increase and mean annual peak discharge will decrease.

With constrained Nile water and the cruel drop in per capita freshwater share in Egypt, soil aquifer treatment (SAT) appears an attractive, low-cost, unconventional water resource that is environmentally friendly and relatively easy to operate. Nevertheless, organic micropollutants (OMPs) in treated wastewater pose environmental and health risks if not properly attenuated through vadose zone infiltration. Thus, determination of OMP removal in a SAT system is important for sustainable groundwater management. In this study, a new, simple equation for the prediction of organic micropollutant removal in SAT systems was developed using gene expression programming (GEP). A wide range of 15 OMPs and aquifer conditions were examined along with various real operational aspects of SAT systems, including hydraulic loading rate and dry/wet ratio. The effect of spatial heterogeneity on saturated hydraulic conductivity was considered. Developed GEP models had an average coefficient of determination $R^2$ of 0.92. Monte Carlo simulation (MCS) with 50,000 realizations was used to propagate uncertainty in SAT parameters in order to generate stochastic inputs for the GEP model. It was found that the removal of OMPs in SAT systems is mostly affected by biodegradation rate and soil-saturated hydraulic conductivity, in addition to dry/wet ratio. Finally,the developed GEP models were applied to enhance the criteria for selecting potential sites for SAT systems in Egypt considering OMP. It was shown that a SAT system would perform well at three sites, with a OMP removal efficiency reaching 100%, whereas it would have a removal of only 50% in the other two sites. Uncertainties in predictions were quantified with an average value of 35%. The developed GEP models can serve as basis for preliminary SAT site selection and design, and can substitute for complex commercial modeling software, especially for practitioners and decision makers in feasibility studies. However, for SAT implementation in a selected location, results can be confirmed only through field column tests.

Streambank nutrient loading rates are a growing concern within many watersheds. Only a few studies exist on streambank soil chemistry and phosphorus (P) concentrations, spatial distributions in watersheds, and P loading rates with a consideration of the potential uncertainty associated with the estimates. More so, limited studies compare streambank P loading for streams within similar watersheds and with similar land use and management. The objectives of this research included (1) quantifying the magnitude and spatial distribution of soil pH, electrical conductivity (EC), total P concentration, dissolved P concentration, and the degree of P saturation of streambanks in a watershed; (2) quantifying whether water-soluble phosphorus (WSP) and total phosphorus (TP) loads entering the stream from streambanks are significant based on a combined mass balance and uncertainty analysis approach; and (3) contrasting streambank P concentrations and loadings between two similar streams: Spavinaw Creek (SC) versus Barren Fork Creek (BFC) in eastern Oklahoma. Both SC and BFC flow through the Ozark ecoregion, possess similar geomorphology, and are characterized by similar land uses. Following procedures conducted for BFC, streambank sampling occurred at five sites along SC, and samples were processed for pH, EC, WSP, and TP. Unlike BFC, there were no clear longitudinal trends in WSP, TP, pH, and EC; similar to BFC, no consistent vertical trends were observed. Using estimated sediment loading ($727\times {10}^6\mathrm{kg}$) from aerial images and Monte Carlo analysis, it was estimated from 2003 to 2013 there was $1.5\times {10}^3\mathrm{kg}$ WSP and $1.4\times {10}^5\mathrm{kg}$ TP loaded into SC from streambanks in Oklahoma. Average in-stream estimates were an order of magnitude larger for WSP and comparable for TP. Streambank P contributions and erosion rates along one stream cannot be used to accurately predict P loading along other streams even in similar watersheds with similar hydrology, geomorphology, and land use because of watershed-specific variability in streambank erodibility and streambank P concentrations. Due to the uncertainty associated with critical input parameters, the uncertainty in streambank P loads at the watershed scale can be large and therefore uncertainty analysis approaches should be used in future research.

Long-term water supply data are important for the current practice of water resources management at a target region. However, long-term water outflow data from reservoirs are typically limited at fine time resolution (hourly). In this study, the historical data on water supply from the Shasta Dam were reconstructed by a reservoir operation model with reconstructed reservoir inflow data. Before embarking on this exercise, first the inflow data were reconstructed by means of a watershed environmental hydrology hydroclimate model, with its input provided from historical atmospheric reanalysis data. The reanalysis data used in this study are from the National Center for Atmospheric Research and the National Center for Environmental Prediction (NCAR–NCEP) at spatial resolution of 2.5° (210 km at the modeled region). The NCAR–NCEP reanalysis data were dynamically downscaled to 3-km spatial grid resolution at hourly intervals by means of the regional atmospheric component of the hydroclimate model. The downscaled atmospheric data were then used as input to the hydroclimate model hydrologic module for the simulation of snowmelt and runoff conditions over Shasta Dam watershed. Then the reconstructed runoff from the watershed was used as input to a reservoir operation model for regulating outflow from Shasta Dam. The coupled hydroclimate model and reservoir model were successfully validated at Shasta Dam watershed by means of comparisons of the model simulations against the observations. Hence, the combination of proposed models were able to reconstruct the historical water supply data during a 60-year historical period (1950–2010) from Shasta Dam watershed.

The use of appropriate space and time scales is fundamental to model the water budget in mountainous regions and to give appropriate replies to the initial requests. However, at a daily scale, the determination of precipitation behavior is not an easy task due to its high variability in mountainous areas. Seven years (2001–2008) of accurate precipitation maps (1 km ground resolution) have been produced for the monsoon season over the Koshi River basin (Nepal) to be used for hydrological modeling. Due to field and topographical constraints, the geostatistical method of ordinary cokriging interpolation (OCK) was used to compute precipitation grids over a $57{,800\mathrm{km}}^2$ basin with a rain gauge network made of 47 stations. Using elevation as a covariable, regionalization models were run to produce 976 daily precipitation grids. They describe temporal and spatial variability close to observed data. Comparisons of the OCK results to an Aphrodite’s reference grid (resolution of 25 km) show that the OCK grids are characterized by a higher spatial variability. Both OCK and Aphrodite data sets underestimate observations, with OCK grids showing the best fit to observed data. However, the OCK method, with an eventual adaptation of the cokriging model, appears consistent for situations where the resolution of the precipitation’s spatial distribution is insufficient and an alternate explaining variable such as elevation is available.

This paper describes the potential use of artificial intelligence (AI)-based techniques to predict monthly groundwater level (GWL) and filling missing GWL data. The adaptive neuro-fuzzy inference system (ANFIS) model conjugated to data pre-processing methods was used for GWL modeling of a Tampa Bay plain at three distinct scenarios. In the proposed methodology, the self-organizing map (SOM) clustering method, as spatial preprocessor, and the discrete wavelet transform (DWT), as temporal data preprocessor, were linked to the ANFIS framework. The original time series of precipitation, runoff, and GWL parameters were decomposed into multifrequency subseries by DWT, and then the subseries were imposed as input data into an ANFIS model for each cluster identified by the SOM, for prediction of the GWL value one and multitime step ahead and to fill missing GWL data. The wavelet transform coherence (WTC) technique was also used for selecting dominant input variables of the ANFIS model. The performance of the ANFIS model was compared to the newly proposed hybrid WTC-ANFIS (WANFIS) model under three different scenarios. The obtained results show that the proposed model can predict GWL with reliable accuracy because the SOM-based spatial clustering method could decrease dimensionality of the inputs matrix, and on the other hand, application of DWT and WTC could enhance the performance of the model by exploring the important periods of the process.

Atmospheric variables (being predictors) play a crucial role in statistical downscaling (SD) studies, which subsequently are used to assess the hydrological impacts of climate change on water resources development and management. The spatial extent of these variables has an utmost importance in development of the SD model and for projecting the precipitation series. In this paper, an attempt is made to identify the spatial extent of atmospheric variables in terms of National Center for Environmental Prediction (NCEP) grid points that result in development of an efficient SD model for precipitation. A case study of the Yamuna-Hindon interbasin is considered to illustrate the applicability of the methodology proposed for capturing the effect of spatial domain of atmospheric variables on development of the SD model. The precipitation in the area is likely to be influenced by surrounding climate, topography, atmospheric circulation, thermodynamic processes, etc. Various cases of spatial extent, which covers a combination of NCEP grid points (containing atmospheric data), are constructed and SD models are developed for each case. SD Models are evaluated for performance using statistical measures. The statistical analysis revealed that though the complete coverage of atmospheric variables gives a better result, the atmospheric activities on NCEP grid points located in Himalayan region have a higher influence on precipitation in the Yamuna-Hindon interbasin. The best performing SD model is then practiced with various emission scenarios.

Slow-moving Hurricane Isaac affected the northern gulf coast between August 28th and August 31st, 2012. Previous studies of the event reported on the hydrometeorology of the event across southeast Louisiana and southern Mississippi. This report provides an in-depth examination and analysis of a suspected rainfall extreme in the New Orleans, Louisiana metropolitan area. Event analysis for most natural watersheds involves examination of river discharge data and the modeling of infiltration to infer watershed-average rainfall. New Orleans is unique because its topography does not allow for runoff, and rainfall must be pumped out of the city. A methodology is proposed which uses data from pumping records as a proxy for streamflow out of the New Orleans watershed. A hydrologic model was created to estimate runoff by modeling infiltration using the Green & Ampt method. Modeled runoff was compared to runoff inferred from pumping records to validate rainfall estimates. Modeled runoff was within 1% of the runoff inferred from pumping records; this strongly suggests that a relatively extreme amount of rain—exceeding the 100 year (1% annual chance) event—did occur over parts of New Orleans during Hurricane Isaac.

This study evaluates several error indices of the radar rain rate from the viewpoint of the accuracy of rainfall-runoff analysis. The error indices considered in this study are the mean error (ME), mean absolute error (MAE), normalized standard deviation (NSD), correlation coefficient (CC), bias (BS), and radar rain rate quality criterion (RRQC). These indices were analyzed using the concept of sum of squares (SS) and mean squares (MS) in the analysis of variance (ANOVA), which explains how the bias and random error of the radar rain rate are combined in an error index. Also, these error indices were linked to the errors in the runoff result, such as the total runoff volume, peak flow, and peak time. As an application, two dam basins in Korea were selected, the Chungju Dam basin and the Namgang Dam basin. The radar rain rates from Gwangdeoksan Radar and Gudeoksan Radar were used as input for the rainfall-runoff analysis. Six different rainfall events were applied to secure various rainfall types. As a result, it was found that ME, BS, and RRQC were linearly proportional to the errors in the runoff result. This result indicates that the bias plays a dominant role in the evaluation of the radar rain rate, targeting the rainfall-runoff analysis.