Developing Subseasonal to Seasonal Climate Forecast Products for Hydrology and Water Management

We describe a new effort to enhance climate forecast relevance and usability through the development of a system for evaluating and displaying real‐time subseasonal to seasonal (S2S) climate forecasts on a watershed scale. Water managers may not use climate forecasts to their full potential due to perceived low skill, mismatched spatial and temporal resolutions, or lack of knowledge or tools to ingest data. Most forecasts are disseminated as large‐domain maps or gridded datasets and may be systematically biased relative to watershed climatologies. Forecasts presented on a watershed scale allow water managers to view forecasts for their specific basins, thereby increasing the usability and relevance of climate forecasts. This paper describes the formulation of S2S climate forecast products based on the Climate Forecast System version 2 (CFSv2) and the North American Multi‐Model Ensemble (NMME). Forecast products include bi‐weekly CFSv2 forecasts, and monthly and seasonal NMME forecasts. Precipitation and temperature forecasts are aggregated spatially to a United States Geological Survey (USGS) hydrologic unit code 4 (HUC‐4) watershed scale. Forecast verification reveals appreciable skill in the first two bi‐weekly periods (Weeks 1–2 and 2–3) from CFSv2, and usable skill in NMME Month 1 forecast with varying skills at longer lead times dependent on the season. Application of a bias‐correction technique (quantile mapping) eliminates forecast bias in the CFSv2 reforecasts, without adding significantly to correlation skill.     

Daily Updating of Operational Statistical Seasonal Water Supply Forecasts for the western U.S.1

Abstract: Official seasonal water supply outlooks for the western United States are typically produced once per month from January through June. The Natural Resources Conservation Service has developed a new outlook product that allows the automated production and delivery of this type of forecast year‐round and with a daily update frequency. Daily snow water equivalent and water year‐to‐date precipitation data from multiple SNOTEL stations are combined using a statistical forecasting technique (“Z‐Score Regression”) to predict seasonal streamflow volume. The skill of these forecasts vs. lead‐time is comparable to the official published outlooks. The new product matches the intra‐monthly trends in the official forecasts until the target period is partly in the past, when the official forecasts begin to use information about observed streamflows to date. Geographically, the patterns of skill also match the official outlooks, with highest skill in Idaho and southern Colorado and lowest skill in the Colorado Front Range, eastern New Mexico, and eastern Montana. The direct and frequent delivery of objective guidance to users is a significant new development in the operational hydrologic seasonal forecasting community.     

A Streamflow Forecasting Framework using Multiple Climate and Hydrological Models1

Abstract: Water resources planning and management efficacy is subject to capturing inherent uncertainties stemming from climatic and hydrological inputs and models. Streamflow forecasts, critical in reservoir operation and water allocation decision making, fundamentally contain uncertainties arising from assumed initial conditions, model structure, and modeled processes. Accounting for these propagating uncertainties remains a formidable challenge. Recent enhancements in climate forecasting skill and hydrological modeling serve as an impetus for further pursuing models and model combinations capable of delivering improved streamflow forecasts. However, little consideration has been given to methodologies that include coupling both multiple climate and multiple hydrological models, increasing the pool of streamflow forecast ensemble members and accounting for cumulative sources of uncertainty. The framework presented here proposes integration and offline coupling of global climate models (GCMs), multiple regional climate models, and numerous water balance models to improve streamflow forecasting through generation of ensemble forecasts. For demonstration purposes, the framework is imposed on the Jaguaribe basin in northeastern Brazil for a hindcast of 1974‐1996 monthly streamflow. The ECHAM 4.5 and the NCEP/MRF9 GCMs and regional models, including dynamical and statistical models, are integrated with the ABCD and Soil Moisture Accounting Procedure water balance models. Precipitation hindcasts from the GCMs are downscaled via the regional models and fed into the water balance models, producing streamflow hindcasts. Multi‐model ensemble combination techniques include pooling, linear regression weighting, and a kernel density estimator to evaluate streamflow hindcasts; the latter technique exhibits superior skill compared with any single coupled model ensemble hindcast.     

Using Paleo Reconstructions to Improve Streamflow Forecast Lead Time in the Western United States

In water stressed regions, water managers are exploring new horizons that would help in long‐range streamflow forecasts. Oceanic‐atmospheric oscillations have been shown to influence streamflow variability. In this study, long‐lead time streamflow forecasts are made using a multiclass kernel‐based data‐driven support vector machine (SVM) model. The extended streamflow records based on tree ring reconstructions were used to provide a longer time series data. Reconstructed data were used from 1658 to 1952 and the instrumental record was used from 1953 to 2007. Reconstructions for oceanic‐atmospheric oscillations included the El Niño‐Southern Oscillation, Pacific Decadal Oscillation, Atlantic Multidecadal Oscillation, and North Atlantic Oscillation. Streamflow forecasts using all four oscillations were made with one‐year to five‐year lead times for 21 gages in the western United States. This is the first study that uses both instrumental and reconstructed data of oscillations in SVM model to improve streamflow forecast lead time. SVM model was able to provide “satisfactory” to “very good” forecasts with one‐ to five‐year lead time for the selected gages. The use of all the oscillation indices helped in achieving better predictability compared to using individual oscillations. The SVM modeling results are better when compared with multiple linear regression model forecasts. The findings are statistical in nature and are expected to be useful for long‐term water resources planning and management.     

Exploring Persistence in Streamflow Forecasting

In this study, the authors explore three persistence approaches in streamflow forecasting motivated by the need for forecasting model skill evaluation. The authors use streamflow observations with 15 min resolution from the year 2008 to 2017 at 140 United States Geological Survey streamflow gauges monitoring the streams and rivers over the State of Iowa. The spatial scale of the basins ranges from about 7 to 37,000 km². The study explores three approaches: simple persistence, gradient persistence, and anomaly persistence. The study shows that persistence forecasts skill has strong dependence on basin scales and weaker but non‐negligible dependence on geometric properties of the river network for a given basin. Among the three approaches explored, anomaly persistence shows highest skill especially for small basins, under about 500 km². The anomaly persistence can serve as a benchmark for model evaluations considering the effect of basin scales and geometric properties of river network of the basin. This study further reiterates that persistence forecasts are hard‐to‐beat methods for larger basin scales at short to medium forecast range.     

DOWNSCALED CLIMATE AND STREAMFLOW STUDY OF THE SOUTHWESTERN UNITED STATES1

ABSTRACT: Downscaling coarse resolution climate data to scales that are useful for impact assessment studies is receiving increased attention. Basin-scale hydrologic processes and other local climate impacts related to water resources such as reservoir management, crop and forest productivity, and ecosystem response require climate information at scales that are much finer than current and future GCM resolutions. The Regional Climate System Model (RCSM) is a dynamic downscaling system that has been used since 1994 for short-term precipitation and streamflow predictions and seasonal hindcast analysis with good skill. During the 1997–1998 winter, experimental seasonal forecasts were made in collaboration with the NOAA Climate Prediction Center and UCLA with promising results. Preliminary studies of a control and 2°CO₂ perturbation for the southwestern U.S. have been performed.     

Reducing Streamflow Forecast Uncertainty: Application and Qualitative Assessment of the Upper Klamath River Basin,Oregon1

Abstract: The accuracy of streamflow forecasts depends on the uncertainty associated with future weather and the accuracy of the hydrologic model that is used to produce the forecasts. We present a method for streamflow forecasting where hydrologic model parameters are selected based on the climate state. Parameter sets for a hydrologic model are conditioned on an atmospheric pressure index defined using mean November through February (NDJF) 700‐hectoPascal geopotential heights over northwestern North America [Pressure Index from Geopotential heights (PIG)]. The hydrologic model is applied in the Sprague River basin (SRB), a snowmelt‐dominated basin located in the Upper Klamath basin in Oregon. In the SRB, the majority of streamflow occurs during March through May (MAM). Water years (WYs) 1980‐2004 were divided into three groups based on their respective PIG values (high, medium, and low PIG). Low (high) PIG years tend to have higher (lower) than average MAM streamflow. Four parameter sets were calibrated for the SRB, each using a different set of WYs. The initial set used WYs 1995‐2004 and the remaining three used WYs defined as high‐, medium‐, and low‐PIG years. Two sets of March, April, and May streamflow volume forecasts were made using Ensemble Streamflow Prediction (ESP). The first set of ESP simulations used the initial parameter set. Because the PIG is defined using NDJF pressure heights, forecasts starting in March can be made using the PIG parameter set that corresponds with the year being forecasted. The second set of ESP simulations used the parameter set associated with the given PIG year. Comparison of the ESP sets indicates that more accuracy and less variability in volume forecasts may be possible when the ESP is conditioned using the PIG. This is especially true during the high‐PIG years (low‐flow years).     

A High‐Resolution National‐Scale Hydrologic Forecast System from a Global Ensemble Land Surface Model

Warning systems with the ability to predict floods several days in advance have the potential to benefit tens of millions of people. Accordingly, large‐scale streamflow prediction systems such as the Advanced Hydrologic Prediction Service or the Global Flood Awareness System are limited to coarse resolutions. This article presents a method for routing global runoff ensemble forecasts and global historical runoff generated by the European Centre for Medium‐Range Weather Forecasts model using the Routing Application for Parallel computatIon of Discharge to produce high spatial resolution 15‐day stream forecasts, approximate recurrence intervals, and warning points at locations where streamflow is predicted to exceed the recurrence interval thresholds. The processing method involves distributing the computations using computer clusters to facilitate processing of large watersheds with high‐density stream networks. In addition, the Streamflow Prediction Tool web application was developed for visualizing analyzed results at both the regional level and at the reach level of high‐density stream networks. The application formed part of the base hydrologic forecasting service available to the National Flood Interoperability Experiment and can potentially transform the nation's forecast ability by incorporating ensemble predictions at the nearly 2.7 million reaches of the National Hydrography Plus Version 2 Dataset into the national forecasting system.     

MONTHLY RUNOFF PREDICTIONS BASED ON RAINFALL FORECASTS IN A SMALL OKLAHOMA WATERSHED1

ABSTRACT: Conditions under which monthly rainfall forecasts translate into monthly runoff predictions that could support water resources planning and management activities were investigated on a small watershed in central Oklahoma. Runoff response to rainfall forecasts was simulated using the hydrologic model SWAT. Eighteen scenarios were examined that represented combinations of wet, average, and dry antecedent rainfall conditions, with wet, normal, and dry forecasted rainfall. Results suggest that for the climatic and physiographic conditions under consideration, rainfall forecasts could offer potential application opportunities in surface water resources but only under certain conditions. Pronounced wet and dry antecedent rainfall conditions were shown to have greater impact on runoff than forecasts, particularly in the first month of a forecast period. Large forecast impacts on runoff occurred under wet antecedent conditions, when the fraction of forecasted rainfall contributing to runoff was greatest. Under dry antecedent conditions, most of the forecasted rainfall was absorbed in the soil profile, with little immediate runoff response. Persistent three‐month forecasts produced stronger impacts on runoff than one‐month forecasts due to cumulative effects in the hydrologic system. Runoff response to antecedent conditions and forecasts suggest a highly asymmetric utility function for rainfall forecasts, with greatest decision‐support potential for persistent wet forecasts under wet antecedent conditions when the forecast signal is least dampened by soil‐storage effects. Under average and dry antecedent conditions, rainfall forecasts showed little potential value for practical applications in surface water resources assessments.     

A Comprehensive Python Toolkit for Accessing High‐Throughput Computing to Support Large Hydrologic Modeling Tasks

The National Flood Interoperability Experiment (NFIE) was an undertaking that initiated a transformation in national hydrologic forecasting by providing streamflow forecasts at high spatial resolution over the whole country. This type of large‐scale, high‐resolution hydrologic modeling requires flexible and scalable tools to handle the resulting computational loads. While high‐throughput computing (HTC) and cloud computing provide an ideal resource for large‐scale modeling because they are cost‐effective and highly scalable, nevertheless, using these tools requires specialized training that is not always common for hydrologists and engineers. In an effort to facilitate the use of HTC resources the National Science Foundation (NSF) funded project, CI‐WATER, has developed a set of Python tools that can automate the tasks of provisioning and configuring an HTC environment in the cloud, and creating and submitting jobs to that environment. These tools are packaged into two Python libraries: CondorPy and TethysCluster. Together these libraries provide a comprehensive toolkit for accessing HTC to support hydrologic modeling. Two use cases are described to demonstrate the use of the toolkit, including a web app that was used to support the NFIE national‐scale modeling.     

From Global to Local: Providing Actionable Flood Forecast Information in a Cloud‐Based Computing Environment

Global and continental scale flood forecast provide coarse resolution flood forecast, but from the perspective of emergency management, flood warnings should be detailed and specific to local conditions. The desired refinement can be provided by the use of downscaling global scale models and through the use of distributed hydrologic models to produce a high‐resolution flood forecast. Three major challenges associated with transforming global flood forecasting to a local scale are addressed in this work. The first is using open‐source software tools to provide access to multiple data sources and lowering the barriers for users in management agencies at local level. This can be done through the Tethys Platform that enables web water resources modeling applications. The second is finding a practical solution for the computational requirements associated with running complex models and performing multiple simulations. This is done using Tethys Cluster that manages distributed and cloud computing resources as a companion to the Tethys Platform for web app development. The third challenge is discovering ways to downscale the forecasts from the global extent to the local context. Three modeling strategies have been tested to address this, including downscaling of coarse resolution global runoff models to high‐resolution stream networks and routing with Routing Application for Parallel computatIon of Discharge (RAPID), the use of hierarchical Gridded Surface and Subsurface Hydrologic Analysis (GSSHA) distributed models, and pre‐computed distributed GSSHA models.     

Towards Real‐Time Continental Scale Streamflow Simulation in Continuous and Discrete Space

The National Weather Service (NWS) forecasts floods at approximately 3,600 locations across the United States (U.S.). However, the river network, as defined by the 1:100,000 scale National Hydrography Dataset‐Plus (NHDPlus) dataset, consists of 2.7 million river segments. Through the National Flood Interoperability Experiment, a continental scale streamflow simulation and forecast system was implemented and continuously operated through the summer of 2015. This system leveraged the WRF‐Hydro framework, initialized on a 3‐km grid, the Routing Application for the Parallel Computation of Discharge river routing model, operating on the NHDPlus, and real‐time atmospheric forcing to continuously forecast streamflow. Although this system produced forecasts, this paper presents a study of the three‐month nowcast to demonstrate the capacity to seamlessly predict reach scale streamflow at the continental scale. In addition, this paper evaluates the impact of reservoirs, through a case study in Texas. Validation of the uncalibrated model using observed hourly streamflow at 5,701 U.S. Geological Survey gages shows 26% demonstrate PBias ≤ |25%|, 11% demonstrate Nash‐Sutcliffe Efficiency (NSE) ≥ 0.25, and 6% demonstrate both PBias ≤ |25%| and NSE ≥ 0.25. When evaluating the impact of reservoirs, the analysis shows when reservoirs are included, NSE ≥ 0.25 for 56% of the gages downstream while NSE ≥ 0.25 for 11% when they are not. The results presented here provide a benchmark for the evolving hydrology program within the NWS and supports their efforts to develop a reach scale flood forecasting system for the country.     

The Sensitivity of California Water Resources to Climate Change Scenarios1

Abstract: Using the latest available General Circulation Model (GCM) results we present an assessment of climate change impacts on California hydrology and water resources. The approach considers the output of two GCMs, the PCM and the HadCM3, run under two different greenhouse gas (GHG) emission scenarios: the high emission A1fi and the low emission B1. The GCM output was statistically downscaled and used in the Variable Infiltration Capacity (VIC) macroscale distributed hydrologic model to derive inflows to major reservoirs in the California Central Valley. Historical inflows used as inputs to the water resources model CalSim II were modified to represent the climate change perturbed conditions for water supply deliveries, reliability, reservoir storage and changes to variables of environmental concern. Our results show greater negative impacts to California hydrology and water resources than previous assessments of climate change impacts in the region. These impacts, which translate into smaller streamflows, lower reservoir storage and decreased water supply deliveries and reliability, will be especially pronounced later in the 21st Century and south of the San Francisco bay Delta. The importance of considering how climate change impacts vary for different temporal, spatial, and institutional conditions in addition to the average impacts is also demonstrated.     

Spatiotemporal Evaluation of Simulated Evapotranspiration and Streamflow over Texas Using the WRF‐Hydro‐RAPID Modeling Framework

This study assesses a large‐scale hydrologic modeling framework (WRF‐Hydro‐RAPID) in terms of its high‐resolution simulation of evapotranspiration (ET) and streamflow over Texas (drainage area: 464,135 km²). The reference observations used include eight‐day ET data from MODIS and FLUXNET, and daily river discharge data from 271 U.S. Geological Survey gauges located across a climate gradient. A recursive digital filter is applied to decompose the river discharge into surface runoff and base flow for comparison with the model counterparts. While the routing component of the model is pre‐calibrated, the land component is uncalibrated. Results show the model performance for ET and runoff is aridity‐dependent. ET is better predicted in a wet year than in a dry year. Streamflow is better predicted in wet regions with the highest efficiency ~0.7. In comparison, streamflow is most poorly predicted in dry regions with a large positive bias. Modeled ET bias is more strongly correlated with the base flow bias than surface runoff bias. These results complement previous evaluations by incorporating more spatial details. They also help identify potential processes for future model improvements. Indeed, improving the dry region streamflow simulation would require synergistic enhancements of ET, soil moisture and groundwater parameterizations in the current model configuration. Our assessments are important preliminary steps towards accurate large‐scale hydrologic forecasts.     

Does Including Soil Moisture Observations Improve Operational Streamflow Forecasts in Snow‐Dominated Watersheds?

Changing climate and growing water demand are increasing the need for robust streamflow forecasts. Historically, operational streamflow forecasts made by the Natural Resources Conservation Service have relied on precipitation and snow water equivalent observations from Snow Telemetry (SNOTEL) sites. We investigate whether also including SNOTEL soil moisture observations improve April‐July streamflow volume forecast accuracy at 0, 1, 2, and 3‐month lead times at 12 watersheds in Utah and California. We found statistically significant improvement in 0 and 3‐month lead time accuracy in 8 of 12 watersheds and 10 of 12 watersheds for 1 and 2‐month lead times. Surprisingly, these improvements were insensitive to soil moisture metrics derived from soil physical properties. Forecasts were made with volumetric water content (VWC) averaged from October 1 to the forecast date. By including VWC at the 0‐month lead time the forecasts explained 7.3% more variability and increased the streamflow volume accuracy by 8.4% on average compared to standard forecasts that already explained an average 77% of the variability. At 1 to 3‐month lead times, the inclusion of soil moisture explained 12.3‐26.3% more variability than the standard forecast on average. Our findings indicate including soil moisture observations increased statistical streamflow forecast accuracy and thus, could potentially improve water supply reliability in regions affected by changing snowpacks.     

The Watershed Flow and Allocation Model: An NHDPlus‐Based Watershed Modeling Approach for Multiple Scales and Conditions

The Watershed Flow and Allocation model (WaterFALL^®) provides segment‐specific, daily streamflow at both gaged and ungaged locations to generate the hydrologic foundation for a variety of water resources management applications. The model is designed to apply across the spatially explicit and enhanced National Hydrography Dataset (NHDPlus) stream and catchment network. To facilitate modeling at the NHDPlus catchment scale, we use an intermediate‐level rainfall‐runoff model rather than a complex process‐based model. The hydrologic model within WaterFALL simulates rainfall‐runoff processes for each catchment within a watershed and routes streamflow between catchments, while accounting for withdrawals, discharges, and onstream reservoirs within the network. The model is therefore distributed among each NHDPlus catchment within the larger selected watershed. Input parameters including climate, land use, soils, and water withdrawals and discharges are georeferenced to each catchment. The WaterFALL system includes a centralized database and server‐based environment for storing all model code, input parameters, and results in a single instance for all simulations allowing for rapid comparison between multiple scenarios. We demonstrate and validate WaterFALL within North Carolina at a variety of scales using observed streamflows to inform quantitative and qualitative measures, including hydrologic flow metrics relevant to the study of ecological flow management decisions.     

ANN OUTPUT UPDATING OF LUMPED CONCEPTUAL RAINFALL/RUNOFF FORECASTING MODELS1

ABSTRACT: Artificial neural networks (ANNs) are tested for the output updating of one‐day‐ahead and three‐day‐ahead streamflow forecasts derived from three lumped conceptual rainfall/runoff (R‐R) models: the GR4J, the IHAC, and the TOPMO. ANN output updating proved superior to a parameter updating scheme and to the ‘simple’ output updating scheme, which always replicates the last observed forecast error. In fact, ANN output updating was able to compensate for large differences in the initial performance of the three tested lumped conceptual R‐R models, which the other tested updating approaches were not able to achieve. This is done mainly by incorporating input vectors usually exploited for ANN R‐R modeling such as previous rainfall and streamflow observations, in addition to the previous observed error. For one‐day‐ahead forecasts, the performance of all three lumped conceptual R‐R models, used in conjunction with ANN output updating, was equivalent to that of the ANN R‐R model. For three‐day‐ahead forecasts, the performance of the ANN‐output‐updated conceptual models was even superior to that of the ANN R‐R model, revealing that the conceptual models are probably performing some tasks that the ANN R‐R model cannot map. However, further testing is needed to substantiate the last statement.     

Development and Operational Testing of a Super‐Ensemble Artificial Intelligence Flood‐Forecast Model for a Pacific Northwest River

Coastal catchments in British Columbia, Canada, experience a complex mixture of rainfall‐ and snowmelt‐driven contributions to flood events. Few operational flood‐forecast models are available in the region. Here, we integrated a number of proven technologies in a novel way to produce a super‐ensemble forecast system for the Englishman River, a flood‐prone stream on Vancouver Island. This three‐day‐ahead modeling system utilizes up to 42 numerical weather prediction model outputs from the North American Ensemble Forecast System, combined with six artificial neural network‐based streamflow models representing various slightly different system conceptualizations, all of which were trained exclusively on historical high‐flow data. As such, the system combines relatively low model development times and costs with the generation of fully probabilistic forecasts reflecting uncertainty in the simulation of both atmospheric and terrestrial hydrologic dynamics. Results from operational testing by British Columbia's flood forecasting agency during the 2013‐2014 storm season suggest that the prediction system is operationally useful and robust.     

A New Open‐Access HUC‐8 Based Downscaled CMIP‐5 Climate Model Forecast Dataset for the Conterminous United States

Watershed‐scale hydrologic simulation models generally require climate data inputs including precipitation and temperature. These climate inputs can be derived from downscaled global climate simulations which have the potential to drive runoff forecasts at the scale of local watersheds. While a simulation designed to drive a local watershed model would ideally be constructed at an appropriate scale, global climate simulations are, by definition, arbitrarily determined large rectangular spatial grids. This paper addresses the technical challenge of making climate simulation model results readily available in the form of downscaled datasets that can be used for watershed scale models. Specifically, we present the development and deployment of a new Coupled Model Intercomparison Project phase 5 (CMIP5) based database which has been prepared through a scaling and weighted averaging process for use at the level of U.S. Geological Survey (USGS) Hydrologic Unit Code (HUC)‐8 watersheds. The resulting dataset includes 2,106 virtual observation sites (watershed centroids) each with 698 associated time series datasets representing average monthly temperature and precipitation between 1950 and 2099 based on 234 unique climate model simulations. The new dataset is deployed on a HydroServer and distributed using WaterOneFlow web services in the WaterML format. These methods can be adapted for downscaled General Circulation Model (GCM) results for specific drainage areas smaller than HUC‐8. Two example use cases for the dataset also are presented.     

A Hydraulic MultiModel Ensemble Framework for Visualizing Flood Inundation Uncertainty

While deterministic forecasts provide a single realization of potential inundation, the inherent uncertainty associated with forecasts also needs to be conveyed for improved decision support. The objective of this study was to develop an ensemble framework for the quantification and visualization of uncertainty associated with flood inundation forecast maps. An 11‐member ensemble streamflow forecast at lead times from 0 to 48 hr was used to force two hydraulic models to produce a multimodel ensemble. The hydraulic models used are (1) the International River Interface Cooperative along with Flow and Sediment Transport with Morphological Evolution of Channels solver and (2) the two‐dimensional Hydrologic Engineering Center‐River Analysis System. Uncertainty was quantified and augmented onto flood inundation maps by calculating statistical spread among the ensemble members. For visualization, a series of probability flood maps conveying the uncertainty in forecasted water extent, water depth, and flow velocity was disseminated through a web‐based decision support tool. The results from this study offer a framework for quantifying and visualizing model uncertainty in forecasted flood inundation maps.