Home » 2021 Yukon River Chinook Run Timing Forecast

2021 Yukon River Chinook Run Timing Forecast

Jun 4, 2021

Yukon River Chinook are predicted to arrive on the delta slightly later than average in 2021. The first significant pulse (15% point) is expected by June 15th and 50% of the run is expected to have arrived by June 23rd.

Despite reports [1] of a warmer-than-average Spring in the Arctic, established indicators of Yukon River Chinook run timing painted a more moderate picture this year. April mean air temperature at the Nome, AK airport (AMATC) was -5.89°C, only slightly warmer than the long-term average of -6.6°C (range -17.1°C to 1.3°C). On the other hand, May mean sea surface temperature (MSSTC), measured along the migratory route between the delta and St. Lawrence Island, was -1.9°C, appreciably cooler than the long-term average of -0.45°C (range -3.8°C to 2.8°C). Similar to what we saw in 2020, spring-time ice cover between the delta and St. Lawrence Island came in just shy of the long-term average of 54% (range 8% to 78%) (Figure 1).

Figure 1: A three-panel plot, each panel showing time series of April mean air temperature (AMATC) at the Nome AK airport, May mean sea surface temperature in the nominal marine staging area (MSSTC), and the proportion of ice cover (PICE) in the nominal marine staging area. Solid vertical line mark indicates each series’ long-term mean and a solid circle mark indicates the 2021 value in each series.

Forecast Detail

The forecast model is based on the relationship previously established in Mundy & Evenson (2011) [2] which identified a set of three environmental variables which correlate with the run timing of Yukon River Chinook: April mean air temperature at the Nome, AK airport (1961–2020; AMATC), May mean sea surface temperature (1961–2020; MSSTC) in the nominal marine staging area just off the Yukon River Delta, and the Spring-time proportion of ice cover (1970–2020; PICE), also measured in the nominal marine staging area. The forecast model is made of three independent sub-models, where each sub-model is a multiple linear regression with a response of either the 15th (FIFDJ), 25th (QDJ), or 50th (MDJ) percentile of cumulative catch-per-unit effort (CPUE) in either commercial or test fisheries located on the delta (1961–2020).

Based upon the historical relationship with CPUE in commercial and test fishery catches on the Yukon River Delta (1961–2020) and the previously mentioned environmental indicators (AMATC, MSSTC, PICE) [2], the predicted dates of three percentiles of CPUE in the Lower Yukon Test Fishery are June 15th (15%), June 18th (25%), and June 23rd (50%) (Table 1).

ModelPercentile of Cumulative CPUEForecasted Date
FIFDJ ~ AMATC + MSSTC + PICE15%June 15th
QDJ ~ AMATC + MSSTC + PICE25%June 18th
MDJ ~ AMATC + MSSTC + PICE50%June 23rd
Table 1: Predicted dates (June 15th, 18th, 23rd) of three percentiles (15%, 25%, 50%) of cumulative catch-per-unit-effort (CPUE) and the associated model formula for each.

The forecasted date for the 50th percentile of June 23rd is slightly later than the long-term average of June 21st. The main contributor to this shift away from average were the relatively cool sea surface temperatures (MSSTC) seen during the month of May which historically correlates with later runs (Figure 2).

Figure 2: A three-panel plot showing the historical (1961–2020) relationship between the 50th percentile of cumulative catch-per-unit-effort (MDJ) and the three environmental variables AMATC, MSSTC, and PICE. A vertical line indicates the 2021 value for each environmental data series.

Forecast Performance

We quantified the accuracy of the forecast sub-models using hindcasting with the most recent fifteen years of runs and calculated a set of prediction-oriented evaluation criteria for each (Table 2). All three sub-models (FIFDJ, QDJ, MDJ) perform similarly, with mean absolute prediction errors (MAPE) 2.13–2.33 days. We also calculated prediction intervals by multiplying each model’s standard error by 2σ to get interval widths of 7.71–8.35 days, which contained the observed value 80–93% of the time. Last, we calculated the maximum absolute residual of 6–9 days and prediction bias of -1.53 to -1.93 days which means the forecast tends to be slightly earlier than observed run timing. The forecast’s bias is an area for future forecast improvements.

ModelMAPE (SD) (days)Bias (days)Max. Residual (days)
FIFDJ ~ AMATC + MSSTC + PICE2.33 (2.26)-1.538
QDJ ~ AMATC + MSSTC + PICE2.13 (2.13)-1.609
MDJ ~ AMATC + MSSTC + PICE2.20 (1.93)-1.936
Table 2: Model performance metrics for the three sub-models that make up the forecast. MAPE: Mean Absolute Prediction Error and its standard deviation. Bias is the mean residual. Max. Residual is the maximum absolute residual.
Figure 3: A three-panel plot showing the predicted against the observed values for three percentiles of cumulative catch-per-unit-effort (FIFDJ, QDJ, MDJ). A diagonal line marks the values corresponding to a perfect forecast.

See the project website as the season progresses for up to date information about how catches at LYTF match the forecast and historical comparisons of catch data. For more information about how the forecast was produced, see https://github.com/yukon-forecasting/2021-forecast. For more information on the forecast model and its underlying biological hypothesis, see Mundy & Evenson (2011) [2].


Prepared and reviewed by Bryce Mecum (brycemecum@gmail.com) and Phil Mundy (proymundy@gmail.com). Data management and web page support by Holly Kent (kent@aoos.org) and Will Koeppen (will@axiomalaska.com). Yukon Chinook in-season salmon data and management agency coordination by Fred West (fred.west@alaska.gov). Financial and material support provided by the Alaska Ocean Observing System, NOAA National Marine Fisheries Service, and the Alaska Department of Fish & Game.


[1] https://nsidc.org/arcticseaicenews/2021/05/a-step-in-our-spring/

[2] Phillip R. Mundy, Danielle F. Evenson, Environmental controls of phenology of high-latitude Chinook salmon populations of the Yukon River, North America, with application to fishery management, ICES Journal of Marine Science, Volume 68, Issue 6, July 2011, Pages 1155–1164, https://doi.org/10.1093/icesjms/fsr080