Years to Include

The Spring Run Juvenile Production Estimate uses historical rotary screw trap and adult survey data. In order to improve model performance the modeling team worked with the stream teams to filter out data that is too incomplete to use for modeling. This article describes the process for selecting RST years to include and adult years to include.

Historically this process differed for each tributary and was based on the rules below. This ruleset was complicated and difficult to maintain. In April 2026 we decided to transition to a consistent ruleset to apply across all years and streams. We compared a number of thresholds and determined that a rule where at least 60% of the weeks (45:53, 1:22) were sampled was appropriate - it was not too stringent and did not include too many gaps. In addition, any sites that do not have any mark-recapture efficiency trials are removed.

RST Data

Heat map of all tributaries

The heat map below shows when sampling occurs for all streams. It shows that sampling is more complete across traps from 2004 - 2009 and from mid November to July. Some traps have continued sampling throughout the season. Some years there are gaps within season for specific traps or across multiple traps.

Years to exclude - Historical Ruleset

The modeling team hosted a modeling windows workshop where we discussed approaches to defining the time window that should be included in the juvenile abundance model. At the workshop we decided to use the full sampling season of data for each tributary, site, and year but to exclude years where we were concerned about data completeness. We came up with the following approach to determine excluded years:

  1. First exclude years with lots of missing data from cumulative catch curves. See data-raw/years_to_exclude.csv (exclusion_type = “really low sampling”).
  2. Exclude yearlings from cumulative catch curves.
  3. Use updated cumulative catch curves to determine the critical window (average window over all historical years that captures 75% percent of catch).
  4. Remove additional years where there is no sampling for 4 consecutive weeks within the critical window. See data-raw/years_to_exclude.csv (exclusion_type = “missing four consecutive weeks in critical window”).

Years to exclude - Current Ruleset

Ruleset defined as “at least 19 weeks included within a run year (week 45-53 and 1-22) for a given site” or 60% of weeks sampled. Sites that do not have any (across all years) mark-repcature efficiency trials are also excluded. Additionally, years with less than 3 weeks with catch > 0 are excluded.

Stream Site Run Year Exclusion Type
battle creek lbc 1998 no efficiency data for site across all years; 60% of weeks sampled
battle creek lbc 1999 no efficiency data for site across all years
battle creek lbc 2000 no efficiency data for site across all years
battle creek lbc 2001 no efficiency data for site across all years; 60% of weeks sampled
battle creek lbc 2002 no efficiency data for site across all years
battle creek lbc 2003 no efficiency data for site across all years
battle creek lbc 2004 no efficiency data for site across all years
battle creek lbc 2005 no efficiency data for site across all years
battle creek lbc 2006 no efficiency data for site across all years; 60% of weeks sampled
battle creek lbc 2016 no efficiency data for site across all years; 60% of weeks sampled

… with 58 more rows

Spring run specific ruleset

Although the focus of this effort is on spring run Chinook salmon, the SR JPE modeling was designed to be applicable for modeling all runs as PLAD run designations are available for all runs. The “years to exclude” ruleset is broadly applicable, however, there are some specific cases where there may be site/years beyond those identified in the “years to exclude ruleset” that are not expected for perform well, specifically for spring run. This may happen when the number of weeks sampling is met but there is a large gap in sampling during the peak of spring run outmigration. We developed an additional list to track these cases and apply to modeling.

stream site run_year number_weeks exclusion_type apply_to exclude
sacramento river tisdale 2013 20 spring run specific - 11 consecutive weeks not sampled Dec 16 to Feb 19 spring TRUE
feather river herringer riffle 2008 21 spring run specific - 9 weeks missing Nov 04 to Jan 01 spring TRUE

rst_model_years object

RST modeling years data object can be found at SRJPEdata::rst_model_years

Adult Data

We treated adult data a little differently to account for the two main types of adult data - adult survey data (holding, redd, carcass) and adult passage data (video passage).

Below are the methods we used for excluding years by adult data type:

Survey data

  • Exclude year if survey does not cover core reaches in a year or if less than 50% of reaches are sampled.

    Stream Core Reaches
    Battle Creek R1, R2, R3, R4 (R5, R6)
    Butte Creek A1, B1, C1, D1, E3
    Clear Creek R1, R2, R3, R4, R5 (R5A, R5B, R5C)
    Deer Creek Lower Falls to A line, A line to Wilson Cove, Polk Springs to Murphy Trail, Murphy Trail to Ponderosa Way, Ponderosa Way to Trail 2E17, Trail 2E17 to Dillon Cove, Uper Falls to Potato Patch Camp, Potato Patch Camp to Highway 32 (Red Bridge)
    Feather River 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38
    Mill Creek Mccarthy Place To Savercool Place, Savercool Place To Black Rock, Black Rock To Below Ranch House, Below Ranch House To Above Avery, Above Avery To Pape Place, Pape Place to Buckhorn Gulch, Buckhorn Gulch To Upper Dam, Above Hwy 36,Hwy 36 To Little Hole In Ground, Hole In Ground To Ishi Trail Head, Ishi Trail Head To Big Bend, Big Bend to Canyon Camp, Canyon Camp To Sooner Place
    Yuba River Yuba does not have core reaches with historical survyes, only samples section of river, not appropriate for use in JPE

Video data

  • Exclude year if video is out for more than 4 weeks in the sampling season
  • Exclude year if flows exceed a threshold value on each tributary (overpass weir etc..). Threshold determined by monitoring program.

We used these rules as preliminary way to identify potentially incomplete years and then conducted outreach to stream teams to review our list and highlight any other years that should be excluded or not.

Years to exclude

We utilized strategies to come up with a list of years to exclude from modeling for adult data. See a section of the adult_model_yearstable filtered to show years to exclude below.

Stream Year Data Type Exclusion Type
battle creek 2017 carcass Missing 50% or more of reach coverage
battle creek 2008 carcass Missing 50% or more of reach coverage
battle creek 2019 carcass Missing 50% or more of reach coverage
battle creek 1996 carcass Missing 50% or more of reach coverage
battle creek 1997 carcass Missing 50% or more of reach coverage
battle creek 1998 carcass Missing 50% or more of reach coverage
battle creek 1999 carcass Missing 50% or more of reach coverage
battle creek 2000 carcass Missing 50% or more of reach coverage
battle creek 2001 carcass Missing 50% or more of reach coverage
battle creek 2017 holding Missing reach 2, 3, 5

… with 45 more rows

adult_model_years object

Adult modeling years data object can be found at SRJPEdata::adult_model_years