Chinook Salmon Escapement in the Chena and Salcha Rivers and Coho Salmon Escapement in the Delta Clearwater River

Salmon enumeration projects in the Tanana River drainage will be conducted by the Alaska Department of Fish and Game (ADF&G) on the Chena, Salcha, and Delta Clearwater rivers. The primary purpose of these projects is to estimate spawning escapement abundance and determine whether or not the established escapement goals for these 3 rivers are met. Chinook salmon Oncorhynchus tshawytscha escapement for the Chena and Salcha rivers will be estimated using tower-counting and sonar techniques and coho salmon O. kisutch escapement in the Delta Clearwater River will be estimated by a visual boat survey at peak escapement.

Operational Plan

A recent operational plan (2019-2023) is available here. Field and analysis methods have not undergone any meaningful changes since this document, as of 2025.

Overview

All code and data archived here pertain to estimating the escapement of Chinook and chum salmon in the Chena and Salcha Rivers, which are tributaries to the Tanana River. The two species ran concurrently, with the Chinook run arriving and peaking a few days or weeks earlier than the chum run, but with considerable overlap in timing. The runs for each species were consistently similar in timing between the two rivers. All field activity was timed in order to capture the entirety of the Chinook run in each season, and the majority of the chum run.

Daily passage has been estimated (2014-2025) using a suite of sequence of enumeration methods, depending on data availability and data quality. These were:

• Expanded visual (tower) counts: During periods of adequate water clarity, technicians counted fish of both species from counting towers during 20-minute intervals in each hour. Counts were then mathematically expanded using formulae for point estimates and standard errors for each day.

• Sonar apportionment: Sonars were also operational at the same sites as the counting towers, in order to provide an alternate means of enumeration when water clarity was inadequate for visual counts. The lengths of all salmon in the sonar record were digitally measured, and a Bayesian mixture model was employed to estimate the species of each sonar target (i.e. Chinook or chum salmon). In addition to the sonar data, the mixture model relied on two additional inputs: the distribution of lengths associated with each species / sex / river (in terms of mean, SE, and SD), and the respective timing of the runs for each species.

• Running average interpolation: Small data gaps were filled using simple running average interpolation methods.

• Hierarchical run-timing model: Larger data gaps were filled using a hierarchical Bayesian model that fit the functional form of the run-timing curve for all years, with curve hyperparameters estimated hierarchically.

Folder Contents

Folders are included for analysis years 2024 and 2025. Each year contains the following subfolders (or similar):

R

• 1_ChenaSalchaYEAR.R: This script reads and processes the sonar data, and estimates values for the length and differential run-timing priors to be incorporated in the sonar mixture model. Respective sets of priors (length and run-timing) are each estimated using hierarchical Bayesian models, which are each run in this script. All results of this script are saved as .Rdata files, which are loaded as appropriate in subsequent scripts.

• 2_CSmixmodelYEAR.R: This script defines and runs the sonar mixture model, and saves the simplified result as a .Rdata file.

• 3_CSvisualdataYEAR.R: This script reads the hourly visual (tower) counts, and compares the totals to sonar. Much of the script is devoted to a comparison of the time periods with complete visual counts and sonar records ("apples to apples"" comparison), which also compares both species combined (visual) to the total number of sonar targets. Theoretically, these should match!

• 4_CSestimatesYEAR.R: Finally, this script puts it all together. It re-reads the visual counts and computes the visual expansion formulae for both visual and sonar estimates as necessary, checks the visual expansion results against the spreadsheet, then conducts the running average and hierarchical run-timing interpolations as needed. It then compiles the set of "final" daily estimates and writes these to external .csv files.

raw_data

A master spreadsheet (.xlsx) for each river (Chena and Salcha) is typically compiled by the project biologist. This contains all hourly visual counts and expansion formulae, as well as historical runs, etc.

These live on local computers and are .gitignore'd, and all data sheets that are relevant to R code are extracted as .csv files.

flat_data

• AllCarcassData_uptoYEAR.csv: All carcass samples for both rivers, which is used for length priors in the sonar mixture model. This is manually appended every year.

• RIVER_historic_YEAR.csv: This should be the NON-EXPANDED daily counts for both species, which is used in the differential run-timing model that creates inputs for the sonar mixture model. This is manually appended every year.

• RIVERSPECIES_expansionsheet.csv: This is a copy of the visual expansion sheet from the master .xlsx file for each river / species, and is imported as a check of the visual expansion results.

• RIVERSPECIES_historic_runtiming.csv: This is a table of EXPANDED daily estimates for all years, and is used as an input to the hierarchical run-timing model. This is manually appended every year.

• RIVERSPECIES_vis20min.csv: This is a table of raw 20-minute visual counts, and is redundant with the first columns of RIVERSPECIES_expansionsheet.csv. Oops.

• RIVERClarity_vis20min.csv: This is a table of the associated water clarity values, and is structured the same as RIVERSPECIES_vis20min.csv.

• RELEVANT SONAR FILES are likely to have different names each year, but will be obvious. These have been batch-processed to compile all the relevant .txt files into one (or multiple) master .csv file(s).

Rdata

• CSpriorsYEAR.Rdata: This file contains the length and differential run-timing priors that are used in the sonar mixture model.

• hier_runtimingYEAR.Rdata: This file contains output from the hierarchical run-timing model used for final interpolations.

• mixmodelYEAR.Rdata: This file contains output from the sonar mixture model.

• sonardataYEAR.Rdata: This file contains processed sonar data, which is used as an input for the sonar mixture model.

• vis_YEAR.Rdata: This file contains processed visual count data, which is used for final estimates.

output

• ChenaSalcha_SummaryTotals_YEAR.csv: A summary table of all daily estimates & SE's, as well as what percentage of the total run (in terms of numbers of fish, not days) estimated to have been enumerated by each method.

• ChenaSalcha_RIVERSPECIES_YEAR.csv: A summary table of daily estimates & SE's and enumeration method used, as well as estimates from all methods.

• ChenaSalcha_RIVERSPECIES_vcov_YEAR.csv: A variance-covariance matrix of daily estimates. Since a nontrivial amount of covariance is present between daily estimates from the sonar mixture model and particularly the hierarchical run-timing models, the variances of the full-season totals will not be the simple sum of the daily variances.