Inseason Forecasts Methods and Information

Smolt Passage Inseason Forecasts

The Smolt Passage Inseason Forecasts predict the arrival distributions of stocks of outmigrating juvenile salmon at several monitoring sites along the Snake and Columbia Rivers. The tool uses "real time" information about the current status of the runs along with current hydrographic information to predict the future progress of the migrating fish.

For the real time analysis, we use PIT tag, Smolt Index and hydrosystem data. In addition, stock-specific estimates of arrival distributions at the lower dams use flow forecasts provided by the Bonneville Power Administration and projected system operations (spill plans, transport operations, etc.) - taking into account the fraction of fish removed from the system for transport.

Current transport operations require that all fish collected at Snake River transport projects (Lower Granite, Little Goose, and Lower Monumental dams) be transported, except for a small fraction returned to the river at Lower Granite for a transport survival study. Any changes of these strategies will be incorporated into forecasts.

Throughout the migration season, the predictions are generated daily with daily updated information. This tool is intended to provide information to the public and to help managers in decisions about mitigation efforts such as flow augmentation, spill scheduling and fish transportation. By providing current information on the status of particular salmon runs, the managers will be able to operate the Columbia River system in a manner that will maximize salmon survival while minimizing the costs of mitigation procedures.

The Smolt Passage Inseason Forecasts utilizes two separate programs to generate downstream passage distributions: Program RealTime and juvenile passage model, COMPASS (CRiSP1 prior to 2007).

RealTime RealTime uses an empirical pattern matching routine to predict arrival distributions at the first detection point in the migratory route (usually Lower Granite Dam on the Snake River) for a wide variety of wild salmon stocks.
COMPASS (implemented 2007) The Comprehensive Passage (COMPASS) model takes the predictions from RealTime and uses hydrological, fish behavioral and dam geometry information to simulate the movement and survival of juvenile salmonids through the remainder of the Columbia River system. The downstream passage component of COMPASS represents downstream migration and survival of salmon populations through the Snake and Columbia rivers. It is written in the C programming language and was derived from CRiSP, a previous salmon passage model. COMPASS computes release-specific daily fish passage at all river segments and dams. The model is composed of four submodels: dam passage, reservoir survival, travel time, and hydrological processes. Parameter values and stock calibration provided by NOAA Fisheries.
CRiSP1 (used through 2006) The Columbia River Salmon Passage model takes the predictions from RealTime and uses hydrological, fish behavioral and dam geometry information to simulate the movement and survival of juvenile salmonids through the remainder of the Columbia River system.

Smolt Passage PIT Tag Forecasts

Program RealTime uses spill-adjusted detections of PIT-tagged fish to make predictions of run-timing and passage distribution of wild Snake River runs-at-large at Lower Granite Dam. The COMPASS model uses the predicted passage distribution at Lower Granite to predict the passage at Little Goose, Lower Monumental, Ice Harbor, and McNary dams.

In addition, Program RealTime uses spill-adjusted detections of PIT-tagged fish to make predictions of run-timing and passage distribution at McNary Dam of specific ESU populations in the Columbia and Snake basins. The COMPASS model uses the predicted passage distribution at McNary to predict the passage at John Day, The Dalles, and Bonneville dams.

2008-2012 Joint RealTime/COMPASS inseason predictions
Species Stock Name Forecast Location NOAA Fisheries Calibration for COMPASS
RealTime COMPASS
Yearling Chinook Selected Composite Wild LWG LGS to BON Wild Snake River Yearling Chinook, Feb. 2008
Yearling Chinook PIT-Tagged Wild Run-At-Large LWG LGS to BON Wild Snake River Yearling Chinook, Feb. 2008
Steelhead Snake River Wild LWG LGS to BON Wild Snake River Steelhead, Feb. 2008
Yearling Chinook Snake River ESU Spring/Summer MCN JDA to BON Wild Snake River Yearling Chinook, Feb. 2008
Steelhead Snake River ESU Steelhead MCN JDA to BON Wild Snake River Steelhead, Feb. 2008
Steelhead Upper Columbia River ESU Steelhead MCN JDA to BON Wild Snake River Steelhead, Feb. 2008

Yearling Chinook Composite Stocks

Composites are the combined data of the streams they include, treated as a single stock.

  • Select Composite Wild (2008 - present) - determined annually based on wild spring/summer stocks released above Lower Granite that have at least 3 historical years with 100 detections and a sufficient release size in the current year to expect 100 detections based on historical detection percentages for the stock.
  • PIT-Tagged Wild Run-At-Large - all wild spring/summer stocks released above Lower Granite.

For previous compositions of the composite stocks, please refer to the Inseason Publications web page for the post season analyses.

ESU Populations

ESU (Evolutionarily Significant Unit) is the distinct populations listed for protection under the Endangered Species Act (ESA). DPS (Distinct Population Segment) is the distinct steelhead population listed for protection under the Endangered Species Act (ESA). See ESA Listing Pages, Protected Resources, NMFS, NOAA, for detailed information on ESA listings and ESU/DPS status and descriptions of specific stocks.

  • The Snake River Spring/Summer Chinook ESU Wild Only Subpopulation includes all naturally spawned populations of spring/summer-run Chinook salmon in the mainstem Snake River and the Tucannon River, Grande Ronde River, Imnaha River, and Salmon River subbasins. The ESU Wild Only Subpopulation does not include the 15 artificial propagation programs that are part of the Snake River Spring/Summer Chinook ESU. Definition implemented by Columbia River DART: March 2010.

  • The Snake River Fall Chinook ESU Wild Only Subpopulation includes all naturally spawned populations of fall-run Chinook salmon in the mainstem Snake River below Hells Canyon Dam, and in the Tucannon River, Grande Ronde River, Imnaha River, Salmon River, and Clearwater River. The ESU Wild Only Subpopulation does not include the 4 artificial propagation programs that are part of the Snake River Fall Chinook ESU.

  • The Snake River Sockeye ESU includes all anadromous and residual sockeye salmon from the Snake River Basin, Idaho, as well as artificially propagated sockeye salmon from the Redfish Lake captive propagation program. Definition implemented by Columbia River DART: March 2010.

  • The Snake River Steelhead DPS Wild Only Subpopulation includes all naturally spawned anadromous O. mykiss (steelhead) populations below natural and manmade impassable barriers in streams in the Snake River Basin of southeast Washington, northeast Oregon, and Idaho. The ESU Wild Only Subpopulation does not include the 6 artificial propagation programs that are part of the Snake River Steelhead DPS. Definition implemented by Columbia River DART: March 2010.

  • The Upper Columbia River Steelhead DPS Wild Only Subpopulation includes all naturally spawned anadromous O. mykiss (steelhead) populations below natural and man-made impassable barriers in streams in the Columbia River Basin upstream from the Yakima River, Washington, to the U.S.-Canada border. The DPS Wild Only Subpopulation does not include the 6 artificial propagation programs that are part of the Upper Columbia River Steelhead DPS. Definition implemented by Columbia River DART: March 2010.

  • Chinook Fall, Upper Columbia River consists of PIT tagged chinook where the run is designated as Unknown or Fall, the rearing type is designated as Unknown or Wild, the release site is on the Columbia River above the Snake Confluence, and the McNary detection date occurs after June 20th of the same year the fish was released.

Subyearling Chinook

Since 2007, Subyearling Chinook predictions are generated by Program RealTime only. The predictions are based on detections of PIT-tagged wild subyearling fall chinook tagged and released from April to July, from 1993 to the present, by William P. Connor (U.S. Fish and Wildlife Service, Dworshak Fisheries Complex), with releases made along the Snake River between RK 268 and RK 224. Detections of these subyearling chinook at Lower Granite constitute a fairly good representation of the subyearling fall chinook run at large, particularly during the first and middle portions of the run.

Sockeye

Predicted by Program RealTime only. For the year 2002, there were no PIT-tagged hatchery sockeye releases from Alturas Lake (ALTURL) or Redfish Lake (REDFL). Passage predictions are based on a composite of PIT-tagged hatchery sockeye releases from Alturas Lake Creek (ALTULC), Redfish Lake Trap (RLCTRP) and Sawtooth Trap (SAWTRP).

Smolt Combined Passage Indices at Rock Island, Lower Granite, McNary, John Day, and Bonneville dams

Program RealTime predictions are based upon the historical outmigration pattern detected at a specific point, and its estimation accuracy therefore depends on regularity in these patterns. The annual patterns of fish passage over McNary Dam are influenced by a number of elements, such as the transportion of fish by barge around upriver dams (Lower Granite, Lower Monumental, and Little Goose Dams), the release of hatchery fish at various times in the season, etc. This is true of Rock Island Dam as well, where irregularities in the patterns of fish passage over Rock Island Dam occur largely due to hatchery releases upriver from the dam. These unpredictable events increase the variability of outmigration patterns from year to year, making accurate predictions more difficult and resulting in wider confidence intervals for the RealTime predictions.

Notes on 2000 Smolt Passage Inseason Forecasts

Fish Passage Center (FPC) passage indices of wild fish were once used by the RealTime forecasting project to make run-timing predictions of runs-at-large of ESA-listed fish to Lower Granite Dam. Beginning in 1999 with subyearling fall chinook salmon, and continuing this year with yearling chinook salmon and steelhead trout, hatcheries have released unmarked fish into the rivers, rendering obsolete the FPC's categorization by rearing-type. Currently, the RealTime forecasting project is using spill-adjusted detections of PIT-tagged fish to make predictions of run-timing and passage distribution of wild Snake River runs-at-large.

Notes on 1999 Smolt Passage Inseason Forecasts

On June 3 and 5, 1999, Captain John Landing and Big Canyon hatcheries released 670,000 unmarked subyearling chinook above Lower Granite Dam. Without marks, the Fish Passage Center (FPC) is unable to distinguish these hatchery-reared fish from wild runs. Our methodology until that time depended on the accurate daily counts of wild fish from the FPC to reliably forecast the wild subyearling outmigration status. Forecasts for previous years based on Smolt Indices can be viewed from the Archive page.

After June 6, the 1999 forecasts were based on detections of PIT-tagged wild subyearling fall chinook at Lower Granite Dam. Subyearling chinook were tagged and released at regular intervals from April into July, or until water temperatures neared 20°C or catches neared zero, from 1993 to the present, by William P. Connor (U.S. Fish and Wildlife Service, Dworshak Fisheries Complex) with releases made along the Snake River between RK 268 and RK 224. Detections of these subyearling chinook at Lower Granite constitute a fairly good representation of the subyearling fall chinook run at large, particularly during the first and middle portions of the run.

Critical Parameters Used by COMPASS / CRiSP1

Migration Seasons 2008-Present, COMPASS

Fish Guidance Efficiency

For dams with bypass systems, FGE is a function of flow through the powerhouse (FPH) and day of the season.

logit(FGE) = beta0 + beta1 * FPH + beta2 * day

Spill Passage Efficiency

SPE models the proportion of fish passing via the spillway. SPE is a function of proportion of flow through the passage route (FSPILL) and the total river flow (FTOTAL).

logit(PSPILL) = beta0 + beta1 * logit(FSPILL) + beta2 * FTOTAL

For the full discussion of FGE and SPE, please refer to the Comprehensive Passage (COMPASS) Model Review DRAFT manual, 29 February 2008.

2008-Present Critical Parameters Used by COMPASS
Project RSW Spill Cap Yearling Chinook Steelhead
Mortality Mortality
RSW Spill Sluiceway Bypass Turbine RSW Spill Sluiceway Bypass Turbine
Lower Granite a  
Little Goose b 8.0 0.02 0.028 0.036 0.077 0.02 0.028 0.050 0.070
Lower Monumental b 8.0 0.02 0.039 0.078 0.119 0.02 0.039 0.078 0.119
Ice Harbor b 7.9 0.030 0.035 0.003 0.129 0.015 0.010 0.000 0.129
McNary b 18.7 0.021 0.038 0.087 0.097 0.021 0.041 0.048 0.110
John Day b 19.2 0.021 0.036 0.035 0.201 0.021 0.027 0.118 0.201
The Dalles   0.076 0.006 0.182 0.076 0.006 0.182
Bonneville   0.031 0.072 0.052 0.033 0.041 0.047
Bonneville II   0.020 0.052 0.047 0.122
a. For Inseason Forecasts, Lower Granite Dam is not modeled in COMPASS. Fish passage distribution and timing input into the COMPASS model at Lower Granite Dam is produced by Program RealTime based on historic and current PIT Tag detections.
b. RSW parameters implemented for 2009 migration season.

Migration Seasons 1996-2006, CRiSP1

There are two forms of spill efficiency equations used for the 1996-2006 migration seasons.

[NL] Non-linear Spill Efficiency equation
Y = (aX/100 + b(X/100)² + C(X/100)³) * 100 + e
[L] Linear Spill Efficiency equation
Y = a +bX +e, where b is value reported in table

During 2003-2006, both non-linear and linear equations were used. During 1996-2002, only the linear form was used.

2003-2006 Critical Parameters Used by CRiSP
Project Yearling Chinook Steelhead Subyearling Chinook
FGE Spill Eff. Mortality FGE Spill Eff. Mortality FGE Spill Eff. Mortality
Turbine Bypass Spill Turbine Bypass Spill Turbine Bypass Spill
Lower Granite 75% NL 7% 2% 2% 81% NL 7% 2% 2% 53% NL 10% 2% 2%
Little Goose 78% NL 8% 1% 0% 81% NL 8% 5% 0% 53% NL 10% 2% 2%
Lower Monumental 49% NL 8% 5% 3% 82% NL 7% 7% 3% 49% NL 10% 2% 2%
Ice Harbor 54% 1.0 L 10% 2% 2% 93% 1.0 L 1% 2% 2% 54% 1.0 L 10% 2% 5%
McNary 83% 1.0 L 10% 2% 2% 89% 1.0 L 10% 2% 2% 62% 1.0 L 10% 3% 2%
John Day 73% 1.6 L 10% 2% 2% 85% 1.6 L 10% 2% 2% 32% 2.0 L 10% 2% 2%
The Dalles 12% 2.0 L 19% 4% 5% 3% 2.0 L 19% 5% 5% 10% 1.8 L 16% 7% 6%
Bonneville 39% 1.0 L 10% 10% 2% 41% 1.0 L 10% 10% 2% 9% 1.0 L 10% 18% 2%
Bonneville II 48% - 10% 2% - 48% - 10% 2% - 28% - 6% 2% -

2000-2002 Critical Parameters Used by CRiSP
Project Yearling Chinook Steelhead Subyearling Chinook
FGE Spill Eff. Mortality FGE Spill Eff. Mortality FGE Spill Eff. Mortality
Turbine Bypass Spill Turbine Bypass Spill Turbine Bypass Spill
Lower Granite 78% 1.0 7% 2% 2% 81% 1.0 7% 2% 2% 53% 1.0 7% 2% 2%
Little Goose 82% 1.0 7% 2% 2% 81% 1.0 7% 2% 2% 53% 1.0 7% 2% 2%
Lower Monumental 61% 1.2 7% 2% 2% 82% 1.2 7% 2% 2% 49% 1.2 7% 2% 2%
Ice Harbor 71% 1.0 7% 2% 2% 93% 1.0 7% 2% 2% 46% 1.0 7% 2% 2%
McNary 95% 1.0 7% 2% 2% 89% 1.0 7% 2% 2% 62% 1.0 7% 2% 2%
John Day 64% 1.0 7% 2% 2% 85% 1.2 7% 2% 2% 34% 1.2 7% 2% 2%
The Dalles 46% 2.0 7% 2% 2% 40% 2.0 7% 2% 2% 46% 2.0 7% 2% 2%
Bonneville 38% 1.0 7% 2% 2% 41% 1.0 7% 2% 2% 16% 1.0 7% 2% 2%
Bonneville II 44% - 7% 2% 0% 48% - 7% 2% 0% 18% - 7% 2% 0%

1996-1999 Critical Parameters Used by CRiSP
Project Yearling Chinook Steelhead Subyearling Chinook
FGE Spill Eff. Mortality FGE Spill Eff. Mortality FGE Spill Eff. Mortality
Turbine Bypass Spill Turbine Bypass Spill Turbine Bypass Spill
Lower Granite 56% 1.0 7% 2% 2% 76% 1.0 7% 2% 2% 35% 1.0 7% 2% 2%
Little Goose 60% 1.0 7% 2% 2% 81% 1.0 7% 2% 2% 30% 1.0 7% 2% 2%
Lower Monumental 55% 1.2 7% 2% 2% 63% 1.2 7% 2% 2% 31% 1.2 7% 2% 2%
Ice Harbor 54% 1.0 7% 2% 2% 77% 1.0 7% 2% 2% 31% 1.0 7% 2% 2%
McNary 72% 1.0 7% 2% 2% 62% 1.0 7% 2% 2% 40% 1.0 7% 2% 2%
John Day 58% 1.0 7% 2% 2% 72% 1.2 7% 2% 2% 26% 1.2 7% 2% 2%
The Dalles 34% 2.0 7% 2% 2% 36% 2.0 7% 2% 2% 43% 2.0 7% 2% 2%
Bonneville 30% 1.0 7% 2% 2% 65% 1.0 7% 2% 2% 15% 1.0 7% 2% 2%
Bonneville II 54% - 7% 2% 0% 52% - 7% 2% 0% 24% - 7% 2% 0%

Adult Passage Inseason Forecasts

Originally developed in 2002, the Adult Passage Inseason Forecasts predict the arrival distributions of adult chinook migrating upstream at several projects along the Snake and Columbia rivers. The tool uses "real time" information about the current visual counts at Bonneville Dam along with current hydrographic information to predict the future progress of the migrating fish. For the real time analysis, we use adult passage counts and hydrosystem data provided by the U.S. Army Corps of Engineers and flow forecasts provided by the Bonneville Power Administration.

The Adult Passage Inseason Forecasts utilizes separate programs to generate upstream passage distributions: Escapement Forecaster, Adult Upstream Model, and Adult Peak Predictor. Also integral to the process are preseason run size predictions generated using multiple methods.

The Escapement Forecaster predicts the arrival timing and run-size of adult spring Chinook and fall Chinook salmon at Bonneville Dam and the Adult Upstream Model predicts the passage timing of the fish at dams above Bonneville Dam. The Adult Peak Predictor generates predictions for spring Chinook at Bonneville Dam that begin with a preseason prediction of run timing from a Genetics and Environment Timing model and a preseason run-size prediction based on linear regression last year's adults, jacks, and environmental sensitivity of jacks. Each day during the spring and fall chinook runs, the predictions are updated based on daily updated information. Starting in 2009, the Adult Peak Predictor run timing update is used as an input to the stock separation component of the AUM model. Starting in 2011, the Adult Peak Predictor run timing update of the peak day is used as an input to the Escapement Projector for determining timing of switching from the startup method into the pattern matching method.

This suite of tools is intended to provide information to the public and to fish managers.

Escapement Forecaster

The Escapement Forecaster is pattern matching algorithm that compares the current year's data to historical years of adult passage counts to predict arrival distributions and run size at Bonneville Dam. The pattern matching algorithm is similar to the smolt Program RealTime prediction algorithms (Townsend RL, Skalski JR, Yasuda D. Evaluation of the 1996 Predictions of the Run-Timing of Wild Migrant Subyearling Chinook in the Snake River Basin using Program RealTime. Monitoring and Evaluation of Smolt Migration in the Columbia Basin [Internet]. 1997;II. Available from: http://pisces.bpa.gov/release/documents/documentviewer.aspx?doc=91572-2). Using pattern matching algorithms, it compares the inseason adult visual counts at Bonneville Dam to the historical counts (data courtesy of the U.S. Army Corps of Engineers, NWD) to predict the percent of the run that has passed Bonneville Dam on the current day and to project the remaining timing of the run. The Escapement Forecaster also makes run-size forecasts for Bonneville Dam. During the season, we predict run-size with 90% confidence intervals. There is a blending routine based on Julian date which smoothly switches from the preseason prediction to the pattern match prediction. There is also a three-day weighted smoothing routine to minimize the effects of daily count fluctuations.

Through the 2005 Adult Spring Chinook run, we used a Bayesian algorithm with lognormal predictive density and with the pattern match forecast as a prior distribution to produce confidence intervals. Starting with the 2005 Adult Fall Chinook Run Size forecasts, a new method was employed to calculate the 90% confidence intervals for the next day cumulative run forecast and the final day total run size forecast. Mean squared error (MSE) of the total run forecast was calculated using deviations of actual runs from run predictions made for each day in all historical years since 1982. Run predictions were generated using the same methods employed for the current year's forecast. MSE is used to estimate the standard error and 90% confidence intervals (depicted by the gray lines) of the run forecast. These same methods are applied for the next day cumulative run size forecast.

Adult Upstream Model

The Adult Upstream Model takes the projected escapement at Bonneville as input and predicts the arrival timing at dams above Bonneville Dam. The model contains a temperature and flow based submodel for reservoir passage and submodels for dam passage, fallback, and straying. In addition, the model uses a standard bioenergetic model to predict the fish energy consumption against the current during migration. River flow and temperature are modeled with the smolt passage model (see Water Quality Inseason Forecasts below).

The Adult Upstream Model has been calibrated to radio tag data (Bjornn, T.C., M.L. Keefer, C.A. Peery, K.R. Tolotti, R.R. Ringe, P.J. Keniry, and L.C. Stuehrenberg. 2000. Migration of adult spring and summer chinook salmon past Columbia and Snake River Dams, through reservoirs and distribution into tributaries, 1996. 2000. Technical Report 2000-5. U.S. Geological Survey, Idaho Cooperative Fish and Wildlife Research Unit, University of Idaho, Moscow, ID.) as well as all adult PIT-tag returns detected at Bonneville since 1998 and all available upstream detection sites--McNary, Ice Harbor, Lower Granite, Priest Rapids, and Wells.

Adult Peak Predictor

The Adult Peak Predictor generates predictions for spring Chinook at Bonneville Dam that begin with a preseason prediction of run timing from a Genetics and Environment Timing model and a preseason run-size prediction based on linear regression last year's adults, jacks, and environmental sensitivity of jacks. The daily, inseason methods to simultaneously estimate distribution parameters from the observations-to-date include use of up-to-date environmental conditions, historical bounds on parameters, and the mathematical properties of the gaussian distribution. The complete run of Chinook in the Columbia River is the sum of three sub-runs—spring, summer and fall—each well characterized by a gaussian distribution. (Anderson and Beer. 2009. Oceanic, riverine, and genetic influences on spring Chinook salmon migration timing. Ecological Applications. 19(8):1989-2003. Available from: http://dx.doi.org/10.1890/08-0477.1.) For more information, please refer to Run timing of adult Chinook salmon passing Bonneville dam on the Columbia River.

Starting in 2009, the run timing update is used as an input to the stock separation component of the AUM model. Starting in 2011, the run timing update of the peak day is used as an input to the Escapement Projector for determining timing of switching from the startup method into the pattern matching method.

Preseason Adult Run Size Prediction at Bonneville Dam

Preseason Run Size Prediction at Bonneville, Adult Spring Chinook and Adult Fall Chinook, Date-based Method, with Final Observed

Year Species Run Dates Preseason Method1 Preseason Forecast
at Bonneville
Final Observed Run2
at Bonneville
Preseason/Final Daily Forecast Record
2013 Spring Chinook 3/15-6/15 NOAA preseason forecast 218000 View
2012 Spring Chinook 3/15-6/15 linear regression of last year's adults, jacks, and environmental sensitivity of jacks 599000 186434 3.21293 View
2011 Spring Chinook 3/15-6/15 TAC preseason forecast 209000 205382 1.01762 View
2010 Spring Chinook 3/15-6/15 TAC preseason forecast 424400 277350 1.53020 View
2009 Spring Chinook 3/15-6/15 linear regression last year's jacks 294318 147470 1.99578 View
2008 Spring Chinook 3/15-6/15 linear regression last year's jacks 307037 151853 2.02194 View
2007 Spring Chinook 3/15-6/15 linear regression last year's jacks 83064 80807 1.02793 View
2006 Spring Chinook 3/15-6/15 linear regression last year's jacks 106683 126156 0.84564 View
2005 Spring Chinook 3/15-6/15 linear regression last year's jacks 212365 97382 2.18074 View
2004 Spring Chinook 3/15-5/31 linear regression last year's jacks 245395 170152 1.44221 View
2003 Spring Chinook 3/15-5/31 linear regression last year's jacks 121388 192010 0.63220 View
2002 Spring Chinook 3/15-5/31 linear regression last year's jacks 244000 268813 0.90769 View
2013 Fall Chinook 8/1-11/15  
2012 Fall Chinook 8/1-11/15 linear regression of last year's adults, jacks, and environmental sensitivity of jacks 399374 350047 1.14092 View
2011 Fall Chinook 8/1-11/15 linear regression last year's jacks 272444 401576 0.67844 View
2010 Fall Chinook 8/1-11/15 linear regression last year's jacks 290153 467524 0.62062 View
2009 Fall Chinook 8/1-11/15 linear regression last year's jacks 249419 283691 0.87919 View
2008 Fall Chinook 8/1-11/15 linear regression last year's jacks 253817 314995 0.80578 View
2007 Fall Chinook 8/1-11/15 linear regression last year's jacks 231184 157784 1.46519 View
2006 Fall Chinook 8/1-11/15 linear regression last year's jacks 228547 299161 0.76396 View
2005 Fall Chinook 8/1-11/15 linear regression last year's jacks 239015 415684 0.57499 View
2004 Fall Chinook 8/1-11/15 linear regression last year's jacks 237237 583269 0.40674 View
2003 Fall Chinook 8/1-11/15 linear regression last year's jacks 218149 610075 0.35758 View
2002 Fall Chinook 8/1-11/15 linear regression last year's jacks 305402 473786 0.64460 View

1. Preseason forecast based on linear regression of the previous year's jacks to the adult passage counts used dataset 1982 to the present.
2. Adult passage visual counts provided as a courtesy by U.S. Army Corps of Engineers.

Preseason Run Size Prediction at Bonneville, Adult Spring Chinook, Area Under Curve Method, with Final Observed

Year Species Preseason Forecast1
at Bonneville
Final Observed Run2
at Bonneville
Preseason/Final Daily Forecast Record Preseason Methods
2013 Spring Chinook 147000     View NOAA Forecast of Adult Returns for coho and Chinook Salmon
2012 Spring Chinook 299600 961063 3.11739 View 2012 Preseason Peak Arrival and Run Size
2011 Spring Chinook 143600 146554 0.97984 View 2011 Preseason Peak Arrival and Run Size
2010 Spring Chinook 242700 221923 1.09362 View 2010 Preseason Peak Arrival and Run Size
2009 Spring Chinook 186890 80873 2.31091 View 2009 Preseason Peak Arrival and Run Size
2008 Spring Chinook 197500 114032 1.73197 View 2008 Preseason Peak Arrival and Run Size

1. Preseason forecast based on linear regression last year's adults, jacks, and environmental sensitivity of jacks and using "Area Under Curve" counting.
2. Adult passage visual counts provided as a courtesy by U.S. Army Corps of Engineers.
3. For a full explanation of the 2012 final observed result, please read the "Evaluation of the 2012 Predictions of Run-size and Passage Distributions of Adult Chinook Salmon (Oncorhynchus tschawytscha) Returning to the Columbia and Snake Rivers" to be released in 2013.

Stock Specific Migration Timing

Since 2005, the Daily Run Size & Passage Prediction (Date-based) is for the period March 15 - June 15 to best match the Columbia River Fisheries (CRM) spring management period. The U.S. v Oregon Technical Advisory Committee 2014 Upriver Spring Chinook Forecast is 227,000 fish to the river mouth, as of January 22, 2014. For 2014, we are using NOAA preseason forecasts. For information on the preseason forecasts and methods, please refer to Inseason Forecasts Methods and Information.

Starting with the 2005 migration season, the Passage Predictions are available for four stocks: Snake River, Upper Columbia River, Lower Columbia River, and Hanford Reach/Yakima. Percent of the run arriving at Bonneville Dam is based on historic run timing and observed visual count data (courtesy of U.S. Army Corps of Engineers, NWD). Stock separation of the run at Bonneville is based on stock composition, reach conversion rates, and run-timing of selected PIT-tagged stocks for all available years of adult detections at PIT Tag observation sites.

  • Snake River: Predicted percent passage of Adult Chinook through the projects on the Columbia mainstem and the lower Snake River. Arrivals to the upstream dams (The Dalles, John Day, McNary, Ice Harbor, Lower Monumental, Little Goose, and Lower Granite) are forecast with the Adult Upstream Model.
  • Upper Columbia River: Predicted percent passage of Adult Chinook through the projects on the Columbia River with destination above Priest Rapids Dam. Arrivals to the upstream dams (The Dalles, John Day, McNary, Priest Rapids, Wanapum, Rock Island, Rocky Reach, Wells) are forecast with the Adult Upstream Model.
  • Lower Columbia River: Predicted percent passage of Adult Chinook through the projects on the Lower Columbia River with destination above Bonneville but not above McNary Dam. Arrivals to the upstream dams (The Dalles, John Day) are forecast with the Adult Upstream Model.
  • Hanford Reach/Yakima: Predicted percent passage of Adult Chinook through the projects on the Columbia River with destination above McNary Dam but not above Priest Rapids Dam, these include Yakima and Hanford Reach stocks. Arrivals to the upstream dams (The Dalles, John Day, McNary) are forecast with the Adult Upstream Model.

Water Quality Inseason Forecasts

Temperature Forecasts

The "real time" Temperature Algorithm was originally developed in 1996 to predict the current year's water temperature values. The Temperature Algorithm is a multi-method algorithm that uses historical mean data, year-to-date observed data, and a forecast of flow to predict temperature values at multiple locations in the Columbia and Snake Rivers. Initially, Lower Granite, Priest Rapids, and The Dalles pools were selected as representative temperatures of the Snake, Mid-Columbia, and mainstem Columbia, respectively. Rock Island was added in 2001 due to the inconsistent reporting of temperature at Priest Rapids Dam. In 2004, we added all possible pools in the Columbia River Basin.

A general trend of negative correlation between flow and water temperature can be seen in data from the Snake and Columbia rivers. By looking at yearly averages of water temperature and flow, one can see that years with higher than average flows have lower than average water temperatures and similarly years with lower than average flow have higher than average water temperatures.

Using a flow forecast file for the current year, a prediction of temperature for can be made using the above relationship. Water temperature is very noisy data being influenced by several variables including: air temperature and other weather conditions, water volume and reservoir geometry, snowpack, upstream water releases. Consequently, the flow/temperature relationship only explains a small amount of the variation of water temperature within a year and between years. As a result, averaged historical data plays a large part in the predictions made, with the above relationship only predicting a small amount of variation about the mean.

Each year before the fish migration season, the mean temperature and mean flow profiles are updated by recomputing the daily historical mean values using all years where data is available. The historical and year-to-date observed temperature and flow data are obtained from the Columbia River DART database.

The temperature algorithm initially sets the value for each day in the year to the historical mean temperature for that day. Then, the historical mean temperature values are replaced with observed values where available. Next, the last 3 days of available temperatures are averaged to predict today's temperature. Averaging over the last three days is an attempt to smooth out some of the day to day variation and to provide a safeguard against bad data giving the algorithm a faulty starting point. Given the averaged starting point, the next 4 weeks of temperatures are calculated by taking the previous day's temperature and adding to it the average daily temperature increment for that day (the increment for day j is the historical mean value on day j minus the historical mean value on day j-1).

Over time, the current trend of temperature becomes less and less useful and eventually uncorrelated with future temperatures. After four weeks, the short-term predictor is phased out of the calculation. For the next three weeks, an intermediate predictor is used. It is a simple linear regression against predicted flow which is used to adjust the historical mean temperature to predict temperature values. After this period, the historical mean value is used as the prediction.

The algorithm developed for temperature has many desirable features. It concurs with the most up-to-date data, it is consistent with historical seasonal patterns in temperature, and it uses predicted flows to make moderate adjustments.

Temperature predictions are run throughout the migration season, April to September, as flow and spill forecasts are provided. The dotted black line on the forecast graphs indicates the date on which the forecast was produced. The observed data is updated daily which allows for the comparison of early forecasts to the actual observed values. Output from the temperature forecasts are used as input to the Total Dissolved Gas forecasts as well as the Smolt Passage and Adult Passage forecasts.

Total Dissolved Gas Forecasts

Total Dissolved Gas forecasts are produced by the COMPASS model (CRiSP.1 prior to 2007) using gas production equations originally developed as a part of the U.S. Army Corps of Engineers "Gas Abatement Study." Gas production equations implemented in the juvenile passage models included equations developed by Waterways Experiment Station (WES) and Columbia Basin Research. The equations predict tailwater gas production as a function of spill. Temperature forecasts, flow and spill forecasts, and year-to-date observed data are inputs to the gas modeling forecasts.

Notes on 1998 TDG Inseason Forecasts

1998 was the second year of predictions for the total dissolved gas forecaster. Because of changes in operations in recent years and data availability at many tailrace sites, the forecaster was fit to data from 1995-1998. 1998 was a more moderate flow year with a few newly modified structures and has thus yielded some important information about forebay and tailrace gas distributions. New flow deflectors were added to John Day and Ice Harbor dams since the 1997 season. The flow deflectors were added to reduce gas production and hence new gas equations were added for the 1998 season. Also new to 1998 was the tailrace water quality monitor added to Wells Dam.

A few changes were made mid-season due to reflect this new information and to significantly improve the TDG model in CRiSP.

Powerhouse Flow Entrainment

Because gas levels were significantly lower in 1998 than in the last two years, a larger variety of forebay conditions were available to fit the forebay model. A new feature was added to the CRiSP model that allowed not only the generation of TDG at a tailrace, but also allowed partial gassing of the powerhouse flow due to entrainment by the spillway flow. Previously TDG from the forebay was passed through the powerhouse, and the tailrace flows from the spillway and powerhouse were either instantaneously mixed or mixed at a slower rate as the river moved downstream. There seemed to be noticeable entrainment below Wanapum, Lower Granite and Little Goose Dam resulting in more TDG in the downstream forebay than if only the spillway flow had been gassed. Entrainment coefficients were fit to 1995-1998 forebay data. Entrainment coefficients seem to improve the overall fit, though a detailed analysis still needs to be done. The change in the model is reflected in the forecasts dated May 18,1998 or later.

Ice Harbor Dam

Ice Harbor's new TDG production equation was based on water quality and spill data from March 1998. This was enough to predict the 1998 season. At the end of the year, a new equation will be fit using the whole year's data.

John Day Dam

John Day Dam's new TDG production equation was based on early spill data and was provided by Army Corps of Engineers. This equation significantly underestimated the amount of TDG produced at JDA. A new equation was fit to the year-to-date water quality data through June 8, 1998. All forecasts prior to this date had the original estimate for gas production, and all forecasts after this date use the new production equation.

The new equation improved the underestimation, however there still seems to be an underestimate of TDG at John Day. This could be the result of CRiSP currently using daily average values for spill. There is significantly more spill at JDA during the night. This is causing an underestimation of gas at night. The day equation is a "flatter" equation, and so the underestimation of the night TDG was enough to throw off the daily average. This could be improved if spill was imported into the model on a hourly or every 6 hour average to capture the day/night differences in spill.

Lower Monumental Dam

Lower Monumental Dam had a poor fit in 1998. Previous years also tended to underestimate TDG. A new equation was fit for both night and day dynamics, as well as a new entrainment coefficient (see above), in order to improve TDG predictions at this dam. Predictions after May 18th reflect the new model coefficients for LMN.

Wells Dam

The new water quality monitor, WELW, added to Wells Tailrace has provided information for the first time on TDG levels below Wells Dam. A new equation was fit to WELW first on May 11, and then on June 11; each resulting in a better fit. CRiSP had previously been significantly overestimating the TDG levels produced by Wells. The old equation was based on forebay data at Rocky Reach from the high gas years 1996-1997 and proved to be inadequate. This equation will be refit once the full year's data is in.

Further Information

For further information on methods and seasonal performance, please refer to the year specific post season analysis reports listed on Inseason Publications.