Generation of Ensemble Streamflow Forecasts Using an Enhanced Version of the Snowmelt Runoff Model1

Harshburger, Brian J., Von P. Walden, Karen S. Humes, Brandon C. Moore, Troy R. Blandford, and Albert Rango, 2012. Generation of Ensemble Streamflow Forecasts Using an Enhanced Version of the Snowmelt Runoff Model. Journal of the American Water Resources Association (JAWRA) 48(4): 643‐655. DOI: 10.1111/j.1752‐1688.2012.00642.x Abstract: As water demand increases in the western United States, so does the need for accurate streamflow forecasts. We describe a method for generating ensemble streamflow forecasts (1‐15 days) using an enhanced version of the snowmelt runoff model (SRM). Forecasts are produced for three snowmelt‐dominated basins in Idaho. Model inputs are derived from meteorological forecasts, snow cover imagery, and surface observations from Snowpack Telemetry stations. The model performed well at lead times up to 7 days, but has significant predictability out to 15 days. The timing of peak flow and the streamflow volume are captured well by the model, but the peak‐flow value is typically low. The model performance was assessed by computing the coefficient of determination (R²), percentage of volume difference (Dv%), and a skill score that quantifies the usefulness of the forecasts relative to climatology. The average R² value for the mean ensemble is >0.8 for all three basins for lead times up to seven days. The Dv% is fairly unbiased (within ±10%) out to seven days in two of the basins, but the model underpredicts Dv% in the third. The average skill scores for all basins are >0.6 for lead times up to seven days, indicating that the ensemble model outperforms climatology. These results validate the usefulness of the ensemble forecasting approach for basins of this type, suggesting that the ensemble version of SRM might be applied successfully to other basins in the Intermountain West.     

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.     

Observed Changes in Climate and Streamflow in the Upper Rio Grande Basin

Observed streamflow and climate data are used to test the hypothesis that climate change is already affecting Rio Grande streamflow volume derived from snowmelt runoff in ways consistent with model‐based projections of 21st‐Century streamflow. Annual and monthly changes in streamflow volume and surface climate variables on the Upper Rio Grande, near its headwaters in southern Colorado, are assessed for water years 1958–2015. Results indicate winter and spring season temperatures in the basin have increased significantly, April 1 snow water equivalent (SWE) has decreased by approximately 25%, and streamflow has declined slightly in the April–July snowmelt runoff season. Small increases in precipitation have reduced the impact of declining snowpack on trends in streamflow. Changes in the snowpack–runoff relationship are noticeable in hydrographs of mean monthly streamflow, but are most apparent in the changing ratios of precipitation (rain + snow, and SWE) to streamflow and in the declining fraction of runoff attributable to snowpack or winter precipitation. The observed changes provide observational confirmation for model projections of decreasing runoff attributable to snowpack, and demonstrate the decreasing utility of snowpack for predicting subsequent streamflow on a seasonal basis in the Upper Rio Grande Basin.     

Climatic Fluctuations and Forecasting of Streamflow in the Lower Colorado River Basin1

Abstract:  Water‐resource managers need to forecast streamflow in the Lower Colorado River Basin to plan for water‐resource projects and to operate reservoirs for water supply. Statistical forecasts of streamflow based on historical records of streamflow can be useful, but statistical assumptions, such as stationarity of flows, need to be evaluated. This study evaluated the relation between climatic fluctuations and stationarity and developed regression equations to forecast streamflow by using climatic fluctuations as explanatory variables. Climatic fluctuations were represented by the Atlantic Multidecadal Oscillation (AMO), Pacific Decadal Oscillation (PDO), and Southern Oscillation Index (SOI). Historical streamflow within the 25‐ to 30‐year positive or negative phases of AMO or PDO was generally stationary. Monotonic trends in annual mean flows were tested at the 21 sites evaluated in this study; 76% of the sites had no significant trends within phases of AMO and 86% of the sites had no significant trends within phases of PDO. As climatic phases shifted in signs, however, many sites had nonstationary flows; 67% of the sites had significant changes in annual mean flow as AMO shifted in signs. The regression equations developed in this study to forecast streamflow incorporate these shifts in climate and streamflow, thus that source of nonstationarity is accounted for. The R² value of regression equations that forecast individual years of annual flow for the central part of the study area ranged from 0.28 to 0.49 and averaged 0.39. AMO was the most significant variable, and a combination of indices from both the Atlantic and Pacific Oceans explained much more variation in flows than only the Pacific Ocean indices. The average R² value for equations with PDO and SOI was 0.15.     

Evaluation of Ground‐Water and Surface‐Water Exchanges Using Streamflow Difference Analyses1

Abstract: In the karstic lower Flint River Basin, limestone fracturing, jointing, and subsequent dissolution have resulted in the development of extensive secondary permeability and created a system of major conduits that facilitate the exchange of water between the Upper Floridan aquifer and Flint River. Historical streamflow data from U.S. Geological Survey gaging stations located in Albany and Newton, Georgia, were used to quantify ground‐water and surface‐water exchanges within a 55.3 km section of the Flint River. Using data from 2001, we compared estimates of ground‐water flux using a time adjustment method to a water balance equation and found that these independent approaches yielded similar results. The associated error was relatively large during high streamflow when unsteady conditions prevail, but much lower during droughts. Flow reversals were identified by negative streamflow differences and verified with in situ data from temperature sensors placed inside large spring conduits. Long‐term (13 years) analysis showed negative streamflow differentials (i.e., a losing stream condition) coincided with high river stages and indicated that streamflow intrusion into the aquifer could potentially exceed 150 m³/s. Although frequent negative flow differentials were evident, the Flint River was typically a gaining stream and showed a large net increase in flow between the two gages when examined over the period 1989‐2003. Ground‐water contributions to this stream section averaged 2‐42 m³/s with a mean of 13 m³/s. The highest rate of ground‐water discharge to the Flint River occurred during the spring when regional ground‐water levels peaked following heavy winter and spring rains and corresponding rates of evapotranspiration were low. During periods of extreme drought, ground‐water contributions to the Flint River declined.     

MODEL ACCURACY IN SNOWMELT-RUNOFF FORECASTS EXTENDING FROM 1 TO 20 DAYS1

ABSTRACT: This paper examines the performance of snowmelt-runoff models in conditions approximating real-time forecast situations. These tests are one part of an intercomparison of models recently conducted by the World Meteorological Organization (WMO). Daily runoff from the Canadian snowmelt basin Illecille. waet (1155 km², 509–3150 m a.s.l.) was forecast for 1 to 20 days ahead. The performance of models was better than in a previous WMO project, which dealt with runoff simulations from historical data, for the following reasons: (1) conditions for models were more favorable than a real-time forecast situation because measured input data and not meteorological forecast inputs were distributed to the modelers; (2) the selected test basin was relatively easy to handle and familiar from the previous WMO project; and (3) all kinds of updating were allowed so that some models even improved their accuracy towards longer forecast times. Based on this experience, a more realistic follow-up project can be imagined which would include temperature forecasts and quantitative precipitation forecasts instead of measured data.     

AN ORGANIZED SIGNAL IN SNOWMELT RUNOFF OVER THE WESTERN UNITED STATES1

ABSTRACT: Daily‐to‐weekly discharge during the snowmelt season is highly correlated among river basins in the upper elevations of the central and southern Sierra Nevada (Carson, Walker, Tuolumne, Merced, San Joaquin, Kings, and Kern Rivers). In many cases, the upper Sierra Nevada watershed operates in a single mode (with varying catchment amplitudes). In some years, with appropriate lags, this mode extends to distant mountains. A reason for this coherence is the broad scale nature of synoptic features in atmospheric circulation, which provide anomalous insolation and temperature forcing that span a large region, sometimes the entire western U.S. These correlations may fall off dramatically, however, in dry years when the snowpack is spatially patchy.     

Projections of 21st Century Sierra Nevada Local Hydrologic Flow Components Using an Ensemble of General Circulation Models1

Abstract: Sierra Nevada snowmelt and runoff is a key source of water for many of California’s 38 million residents and nearly the entire population of western Nevada. The purpose of this study was to assess the impacts of expected 21st Century climatic changes in the Sierra Nevada at the subwatershed scale, for all hydrologic flow components, and for a suite of 16 General Circulation Models (GCMs) with two emission scenarios. The Soil and Water Assessment Tool (SWAT) was calibrated and validated at 35 unimpaired streamflow sites. Results show that temperatures are projected to increase throughout the Sierra Nevada, whereas precipitation projections vary between GCMs. These climatic changes drive a decrease in average annual streamflow and an advance of snowmelt and runoff by several weeks. The largest streamflow reductions were found in the mid‐range elevations due to less snow accumulation, whereas the higher elevation watersheds were more resilient due to colder temperatures. Simulation results showed that decreases in snowmelt affects not only streamflow, but evapotranspiration, surface, and subsurface flows, such that less water is available in spring and summer, thus potentially affecting aquatic and terrestrial ecosystems. Declining spring and summer flows did not equally affect all subwatersheds in the region, and the subwatershed perspective allowed for identification for the most sensitive basins throughout the Sierra Nevada.     

MODIS Biophysical States and NEXRAD Precipitation in a Statistical Evaluation of Antecedent Moisture Condition and Streamflow1

Abstract: The potential of remotely sensed time series of biophysical states of landscape to characterize soil moisture condition antecedent to radar estimates of precipitation is assessed in a statistical prediction model of streamflow in a 1,420 km² watershed in south‐central Texas, Moderate Resolution Imaging Spectroradiometer (MODIS) time series biophysical products offer significant opportunities to characterize and quantify hydrologic state variables such as land surface temperature (LST) and vegetation state and status. Together with Next Generation Weather Radar (NEXRAD) precipitation estimates for the period 2002 through 2005, 16 raw and deseasoned time series of LST (day and night), vegetation indices, infrared reflectances, and water stress indices were linearly regressed against observed watershed streamflow on an eight‐day aggregated time period. Time offsets of 0 (synchronous with streamflow event), 8, and 16 days (leading streamflow event) were assessed for each of the 16 parameters to evaluate antecedent effects. The model results indicated a reasonable correlation (r² = 0.67) when precipitation, daytime LST advanced 16 days, and a deseasoned moisture stress index were regressed against log‐transformed streamflow. The estimation model was applied to a validation period from January 2006 through March 2007, a period of 12 months of regional drought and base‐flow conditions followed by three months of above normal rainfall and a flood event. The model resulted in a Nash‐Sutcliffe estimation efficiency (E) of 0.45 for flow series (in log‐space) for the full 15‐month period, ?0.03 for the 2006 drought condition period, and 0.87 for the 2007 wet condition period. The overall model had a relative volume error of ?32%. The contribution of parameter uncertainties to model discrepancy was evaluated.     

Streamflow Hydrology Estimate Using Machine Learning (SHEM)

Continuity and accuracy of near real‐time streamflow gauge (streamgage) data are critical for flood forecasting, assessing imminent risk, and implementing flood mitigation activities. Without these data, decision makers and first responders are limited in their ability to effectively allocate resources, implement evacuations to save lives, and reduce property losses. The Streamflow Hydrology Estimate using Machine Learning (SHEM) is a new predictive model for providing accurate and timely proxy streamflow data for inoperative streamgages. SHEM relies on machine learning (“training”) to process and interpret large volumes (“big data”) of historic complex hydrologic information. Continually updated with real‐time streamflow data, the model constructs a virtual dataset index of correlations and groups (clusters) of relationship correlations between selected streamgages in a watershed and under differing flow conditions. Using these datasets, SHEM interpolates estimated discharge and time data for any indexed streamgage that stops transmitting data. These estimates are continuously tested, scored, and revised using multiple regression analysis processes and methodologies. The SHEM model was tested in Idaho and Washington in four diverse watersheds, and the model's estimates were then compared to the actual recorded data for the same time period. Results from all watersheds revealed a high correlation, validating both the degree of accuracy and reliability of the model.     

Reconstructions of Columbia River Streamflow from Tree‐Ring Chronologies in the Pacific Northwest,USA

We developed Columbia River streamflow reconstructions using a network of existing, new, and updated tree‐ring records sensitive to the main climatic factors governing discharge. Reconstruction quality is enhanced by incorporating tree‐ring chronologies where high snowpack limits growth, which better represent the contribution of cool‐season precipitation to flow than chronologies from trees positively sensitive to hydroclimate alone. The best performing reconstruction (back to 1609 CE) explains 59% of the historical variability and the longest reconstruction (back to 1502 CE) explains 52% of the variability. Droughts similar to the high‐intensity, long‐duration low flows observed during the 1920s and 1940s are rare, but occurred in the early 1500s and 1630s‐1640s. The lowest Columbia flow events appear to be reflected in chronologies both positively and negatively related to streamflow, implying low snowpack and possibly low warm‐season precipitation. High flows of magnitudes observed in the instrumental record appear to have been relatively common, and high flows from the 1680s to 1740s exceeded the magnitude and duration of observed wet periods in the late‐19th and 20th Century. Comparisons between the Columbia River reconstructions and future projections of streamflow derived from global climate and hydrologic models show the potential for increased hydrologic variability, which could present challenges for managing water in the face of competing demands.     

Stationarity of Streamflow Records and Their Influence on Bankfull Regional Curves1

Haucke, Jessica and Katherine A. Clancy, 2011. Stationarity of Streamflow Records and Their Influence on Bankfull Regional Curves. Journal of the American Water Resources Association (JAWRA) 47(6):1338–1347. DOI: 10.1111/j.1752‐1688.2011.00590.x Abstract: Bankfull regional curves, which are curves that establish relationships among channel morphology, discharge, drainage area, are used extensively for stream restoration. These curves are developed upon the assumption that streamflows maintain stationarity over the entire record. We examined this assumption in the Driftless Area of southwestern Wisconsin where agricultural soil retention practices have changed, and precipitation has increased since the 1970s. We developed a bankfull regional curve for this area using field surveys of bankfull channel performed during 2008‐2009 and annual series of peak streamflows for 10 rivers with streamflow records ranging from the 1930s to 2009. We found bankfull flows to correlate to a 1.1 return period. To evaluate gage data statistics, we used the sign test to compare our channel morphology to historic 1.5 return period discharge (Q_1.5) for five time periods: 1959‐1972, 1973‐1992, 1993‐2008, 1999‐2008, and the 1959‐2008 period of record. Analysis of the historic gage data indicated that there has been a more than 30% decline in Q_1.5 since 1959. Our research suggests that land conservation practices may have a larger impact on gaging station stationarity than annual precipitation changes do. Additionally, historic peak flow data from gages, which have records that span land conservation changes, may need to be truncated to represent current flow regimes.     

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.     

Changes in Patterns of Streamflow From Unregulated Watersheds in Idaho,Western Wyoming,and Northern Nevada1

Clark, Gregory M., 2010. Changes in Patterns of Streamflow From Unregulated Watersheds in Idaho, Western Wyoming, and Northern Nevada. Journal of the American Water Resources Association (JAWRA) 46(3):486-497. DOI: 10.1111/j.1752-1688.2009.00416.x Abstract: Recent studies have identified a pattern of earlier spring runoff across much of North America. Earlier spring runoff potentially poses numerous problems, including increased risk of flooding and reduced summer water supply for irrigation, power generation, and migratory fish passage. To identify changing runoff patterns in Idaho streams, streamflow records were analyzed for 26 U.S. Geological Survey gaging stations in Idaho, western Wyoming, and northern Nevada, each with a minimum of 41 years of record. The 26 stations are located on 23 unregulated and relatively pristine streams that drain areas ranging from 28 to >35,000 km². Four runoff parameters were trend tested at each station for both the period of historical record and from 1967 through 2007. Parameters tested were annual mean streamflow, annual minimum daily streamflow, and the dates of the 25th and 50th percentiles of the annual total streamflow. Results of a nonparametric Mann-Kendall trend test revealed a trend toward lower annual mean and annual minimum streamflows at a majority of the stations, as well as a trend toward earlier snowmelt runoff. Significant downward trends over the period of historical record were most prevalent for the annual minimum streamflow (12 stations) and the 50th percentile of streamflow (11 stations). At most stations, trends were more pronounced during the period from 1967 through 2007. A regional Kendall test for water years 1967 through 2007 revealed significant regional trends in the percent change in the annual mean and annual minimum streamflows (0.67% less per year and 0.62% less per year, respectively), the 25th percentile of streamflow (12.3 days earlier), and the 50th percentile of streamflow (11.5 days earlier).     

DENSITY CURRENT INTRUSIONS IN AN ICE-COVERED URBAN LAKE1

ABSTRACT: Evidence is presented that snowmelt runoff from an urban watershed can produce density current intrusions (underflows) in a lake. Several episodes of density current intrusions are documented. Water temperatures and salinities measured near the bottom of a 10 m deep Minneapolis lake during the late winter warming periods in 1989, 1990, 1991, and 1995 show significant rapid changes which are correlated with observed higher air temperatures and snowmelt runoff. The snowmelt runoff entering this particular lake (Ryan Lake) has increased electrical conductivity, salinity, and density. The source of the salinity is the salt spread on urban streets in the winter. Heating of littoral waters in spring may also contribute to the occurrence of the sinking flows, but is clearly not the only cause.     

Modeling Hydrology in a Small Rocky Mountain Watershed Serving Large Urban Populations1

Abstract: This article describes the development of a calibrated hydrologic model for the Blue River watershed (867 km²) in Summit County, Colorado. This watershed provides drinking water to over a third of Colorado’s population. However, more research on model calibration and development for small mountain watersheds is needed. This work required integration of subsurface and surface hydrology using GIS data, and included aspects unique to mountain watersheds such as snow hydrology, high ground‐water gradients, and large differences in climate between the headwaters and outlet. Given the importance of this particular watershed as a major urban drinking‐water source, the rapid development occurring in small mountain watersheds, and the importance of Rocky Mountain water in the arid and semiarid West, it is useful to describe calibrated watershed modeling efforts in this watershed. The model used was Soil and Water Assessment Tool (SWAT). An accurate model of the hydrologic cycle required incorporation of mountain hydrology‐specific processes. Snowmelt and snow formation parameters, as well as several ground‐water parameters, were the most important calibration factors. Comparison of simulated and observed streamflow hydrographs at two U.S. Geological Survey gaging stations resulted in good fits to average monthly values (0.71 Nash‐Sutcliffe coefficient). With this capability, future assessments of point‐source and nonpoint‐source pollutant transport are possible.     

Mid‐Range Streamflow Forecasts Based on Climate Modeling – Statistical Correction and Evaluation1

Abstract: Mid‐range streamflow predictions are extremely important for managing water resources. The ability to provide mid‐range (three to six months) streamflow forecasts enables considerable improvements in water resources system operations. The skill and economic value of such forecasts are of great interest. In this research, output from a general circulation model (GCM) is used to generate hydrologic input for mid‐range streamflow forecasts. Statistical procedures including: (1) transformation, (2) correction, (3) observation of ensemble average, (4) improvement of forecast, and (5) forecast skill test are conducted to minimize the error associated with different spatial resolution between the large‐scale GCM and the finer‐scale hydrologic model and to improve forecast skills. The accuracy of a streamflow forecast generated using a hydrologic model forced with GCM output for the basin was evaluated by forecast skill scores associated with the set of streamflow forecast values in a categorical forecast. Despite the generally low forecast skill score exhibited by the climate forecasting approach, precipitation forecast skill clearly improves when a conditional forecast is performed during the East Asia summer monsoon, June through August.     

GENERAL-CIRCULATION-MODEL SIMULATIONS OF FUTURE SNOWPACK IN THE WESTERN UNITED STATES1

ABSTRACT: April 1 snowpack accumulations measured at 311 snow courses in the western United States (U.S.) are grouped using a correlation-based cluster analysis. A conceptual snow accumulation and melt model and monthly temperature and precipitation for each cluster are used to estimate cluster-average April 1 snowpack. The conceptual snow model is subsequently used to estimate future snowpack by using changes in monthly temperature and precipitation simulated by the Canadian Centre for Climate Modeling and Analysis (CCC) and the Hadley Centre for Climate Prediction and Research (HADLEY) general circulation models (GCMs). Results for the CCC model indicate that although winter precipitation is estimated to increase in the future, increases in temperatures will result in large decreases in April 1 snowpack for the entire western U.S. Results for the HADLEY model also indicate large decreases in April 1 snowpack for most of the western US, but the decreases are not as severe as those estimated using the CCC simulations. Although snowpack conditions are estimated to decrease for most areas of the western US, both GCMs estimate a general increase in winter precipitation toward the latter half of the next century. Thus, water quantity may be increased in the western US; however, the timing of runoff will be altered because precipitation will more frequently occur as rain rather than as snow.     

Use of Hydrologic Landscape Classification to Diagnose Streamflow Predictability in Oregon

We implement a spatially lumped hydrologic model to predict daily streamflow at 88 catchments within the state of Oregon and analyze its performance using the Oregon Hydrologic Landscape (OHL) classification. OHL is used to identify the physio‐climatic conditions that favor high (or low) streamflow predictability. High prediction catchments (Nash‐Sutcliffe efficiency of (NS) > 0.75) are mainly classified as rain dominated with very wet climate, low aquifer permeability, and low to medium soil permeability. Most of them are located west of the Cascade Mountain Range. Conversely, most low prediction catchments (NS < 0.6) are classified as snow‐dominated with high aquifer permeability and medium to high soil permeability. They are mainly located in the volcano‐influenced High Cascades region. Using a subset of 36 catchments, we further test if class‐specific model parameters can be developed to predict at ungauged catchments. In most catchments, OHL class‐specific parameters provide predictions that are on par with individually calibrated parameters (NS decline < 10%). However, large NS declines are observed in OHL classes where predictability is not high enough. Results suggest higher uncertainty in rain‐to‐snow transition of precipitation phase and external gains/losses of deep groundwater are major factors for low prediction in Oregon. Moreover, regionalized estimation of model parameters is more useful in regions where conditions favor good streamflow predictability.     

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.