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.

RST Data - Years to Include in Model

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”).

Cumulative catch curves

The cumulative catch curve below shows cumulative catch over time for Battle Creek. This plot shows that in 2007 there was only sampling through mid January on Battle Creek. We used similar plots for each tributary to exclude years where there is very limited sampling.

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

We utilized the above cumulative catch curves and heatmaps to come up with a list of years to exclude from modeling. See a section of the “years to exclude” table below.

For ongoing data collection and more recent seasons, we applied an automatic check as a first pass to determine if a year should be excluded. Any automatic exclusion records will be reviewed manually before being excluded from data. We use the following criteria to asses if a recent RST season should be excluded from analysis:

  • If 25% of the weeks are missing (this test is used in place of the original criteria of “exclude years with lots of missing data from cumulative catch curves”)

We check this list against our original method annually after it is run to confirm that it is making the correct exclusion decisions.

Years to Exclude Table

Stream Site Run Year Exclusion Type Notes
battle creek lbc 1998 automatic less than 75% of weeks sampled
battle creek lbc 2006 automatic less than 75% of weeks sampled
battle creek lbc 2016 automatic less than 75% of weeks sampled
battle creek ubc 2007 really low sampling only 12 week of data
battle creek ubc 2003 really low sampling Stream Team noted this as year that is incomparable and also efficiency data exists where there is no catch data
battle creek ubc 2015 missing four consecutive weeks in critical window NA
battle creek ubc 1998 low sampling NA
battle creek ubc 2001 low sampling NA
battle creek lbc 2001 low sampling NA
butte creek adams dam 1997 automatic less than 75% of weeks sampled

… with 55 more rows

Applying to Modeling Datasets

In order to apply the years to exclude information to the modeling datasets we did some additional processing to create a table describing the stream, site, year that should be included in the SR JPE modeling.

The table below shows a section of this table:

stream site run_year
feather river eye riffle 1998
yuba river hallwood 2023
butte creek okie dam 2018
yuba river hallwood 2024
butte creek okie dam 2021
butte creek okie dam 2022
butte creek okie dam 2023
butte creek okie dam 2020
butte creek okie dam 2025
yuba river hallwood 2025

… with 219 more rows

Adult Data - Years to Include in Model

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 above cumulative catch curves and heatmaps to come up with a list of years to exclude from modeling for adult data. See a section of the “years to exclude” table 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 redd Missing reaches 2, 3, 5, and 6

… with 47 more rows

Applying to Modeling Datasets

In order to apply the years to exclude information to the modeling datasets we did some additional processing to create a table describing the stream, year by data type that should be included in the SR JPE modeling.

The table below shows a section of this table:

year stream count data_type lower_bound_estimate upper_bound_estimate confidence_level
2001 battle creek 30 redd NA NA NA
2002 battle creek 71 redd NA NA NA
2003 battle creek 156 redd NA NA NA
2004 battle creek 35 redd NA NA NA
2005 battle creek 46 redd NA NA NA
2006 battle creek 121 redd NA NA NA
2007 battle creek 132 redd NA NA NA
2008 battle creek 38 redd NA NA NA
2009 battle creek 85 redd NA NA NA
2010 battle creek 89 redd NA NA NA

… with 253 more rows