Lifestage Ruleset

Generate Yearling Rulesets

Raw catch from the trap does not consistently differentiate between yearling and young of year (YOY) Chinook. FlowWest presented a proposed methodology at a life history diversity (LHD) ruleset workshop (see lhd shiny for workshop materials) and worked with watershed experts to define a methodology to systematically determine life history for each tributary (described below).

Approach

  1. Set weekly cutoff values: Use visual determination on fork length over time scatter plots to set weekly cutoff values of yearlings vs YOY.
  2. Generate daily cutoff values: Use linear interpolation to extrapolate weekly cutoffs into daily values.
  3. Review & update: Share proposed cutoff values with watershed experts to review. Update as needed.
  4. Apply cutoff to catch data: Use daily cutoff values to generate a yearling column in the catch data.

Set Weekly Cutoff Values & Generate daily cutoff values

FlowWest proposed weekly cutoff values and used these weekly cutoff values to generate daily values using a linear approximation function, approxfun.

generate_cutoff <- approxfun(date, fork_length_cutoff, rule = 2)

The plot below shows the updated cutoff values with linear interpolation of weekly cutoffs for Deer Creek.

Note: You can view all code used to generate plots and tables in this markdown here.

Review & Update

FlowWest shared above plot for each watershed and asked stream experts to review. We incorporated feedback and modified rulesets to better separate yearlings and YOY.

Apply Cutoff to Catch Data

FlowWest took the daily cutoff line (shown in plot above) and used it as a threshold to classify yearling vs YOY in historical catch data. We added an is_yearling column to the catch data and set is_yearling = TRUE for any fish with a fork length that exceeded the yearling cutoff on a given date.

The following code is applied in the weekly_data_summary script.

# Note this is not the final dataset as lifestage is added below
standard_catch_unmarked_w_yearling <- rst_catch |> 
  filter(species %in% c("chinook", "chinook salmon")) |>  # filter for only chinook
  mutate(month = month(date), 
         day = day(date)) |> 
  left_join(daily_yearling_ruleset) |> 
  mutate(is_yearling = case_when((fork_length <= cutoff & !run %in% c("fall","late fall", "winter")) ~ F,
                                 (fork_length > cutoff & !run %in% c("fall","late fall", "winter")) ~ T,
                                 (run %in% c("fall","late fall", "winter")) ~ NA,
                                 T ~ NA))

Fry and Smolt Designations

In addition to differentiating between yearling and YOY it is important for the SR JPE to differentiate between fry and smolt as there will likely be a separate SR JPE for each lifestage. Some monitoring programs assign lifestage based on visual determination or fork length though not all RST data included lifestage data. To ensure lifestage was assigned consistently across streams and was complete, FlowWest developed an approach for differentiating between fry and smolt.

Approach

  1. Create lifestage based on forklength cutoff of 45mm (< 45 - fry, > 45 - smolt)
  2. Determine year specific proportions for fry, smolt, yearling for each stream, site, week and year
  3. Determine general week proportions for fry, smolt, yearling for each stream, week
  4. Apply proportions to data to fill in missing lifestage
  5. Generate rows for when no fish of a particular lifestage are caught

Create lifestage variable

The first step was to apply a lifestage cutoff to catch records that had fork lengths recorded. These cutoffs are fork_length < 45 = fry, fork_length > 45 = smolt, fork_length > yearling_cutoff = yearling.

date stream site subsite site_group count run life_stage adipose_clipped dead fork_length weight julian_week julian_year
1999-01-19 clear creek lcc lcc clear creek 1 fall fry NA NA 40 0 3 1999
1999-01-19 clear creek lcc lcc clear creek 17100 fall NA NA NA NA 0 3 1999
1999-01-20 battle creek lbc lbc battle creek 1 fall fry NA NA 35 0 3 1999
1999-01-20 battle creek lbc lbc battle creek 3 fall fry NA NA 38 0 3 1999
1999-01-20 clear creek lcc lcc clear creek 6 fall fry NA NA 37 0 3 1999
1999-01-20 clear creek lcc lcc clear creek 5 fall fry NA NA 38 0 3 1999
1999-01-20 battle creek lbc lbc battle creek 1 fall fry NA NA 34 0 3 1999
1999-01-20 clear creek lcc lcc clear creek 5 fall fry NA NA 36 0 3 1999
1999-01-20 battle creek lbc lbc battle creek 4 fall fry NA NA 36 0 3 1999
1999-01-20 battle creek lbc lbc battle creek 2 fall fry NA NA 37 0 3 1999

Determine year specific lifestage proportions

There are 47556 entries with missing lifestage due to missing fork length data.

date stream site subsite site_group count run life_stage adipose_clipped dead fork_length weight julian_week julian_year
1999-01-19 clear creek lcc lcc clear creek 17100 fall NA NA NA NA 0 3 1999
1999-01-20 battle creek lbc lbc battle creek 1 fall NA NA NA NA 0 3 1999
1999-01-20 clear creek lcc lcc clear creek 6631 fall NA NA NA NA 0 3 1999
1999-01-20 battle creek lbc lbc battle creek 3147 fall NA NA NA NA 0 3 1999
1999-01-20 battle creek ubc ubc battle creek 1295 fall NA NA NA NA 0 3 1999
1999-01-21 clear creek lcc lcc clear creek 5163 fall NA NA NA NA 0 3 1999

The first step in filling in these missing lifestages was to find the proportion for each lifestage category for a given stream, site, week, and year. This information could then be used to fill in the lifestage for missing rows within a week.

year week stream site percent_fry percent_smolt percent_yearling
1992 42 deer creek deer creek 0 0 1
1992 44 deer creek deer creek 0 0 1
1992 45 deer creek deer creek 0 0 1
1992 46 deer creek deer creek 0 0 1
1992 48 deer creek deer creek 0 0 1
1992 49 deer creek deer creek 0 0 1
1994 40 deer creek deer creek 0 0 1
1994 41 deer creek deer creek 0 0 1
1994 42 deer creek deer creek 0 0 1
1994 43 deer creek deer creek 0 0 1

Determine general weekly lifestage proportions

For weeks that had no fork length data in a given week, we calculated a general lifestage proportion across years. Calculating the proportion for each lifestage category for a given stream, site, and week.

Apply proportions to fill in missing values

We used these proportions to fill in missing lifestage values. See the final lifestage designations below.

date stream site subsite site_group count run life_stage adipose_clipped dead fork_length weight julian_week julian_year week year model_lifestage_method
1999-01-19 clear creek lcc lcc clear creek 1 fall fry NA NA 40 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 battle creek lbc lbc battle creek 1 fall fry NA NA 35 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 battle creek lbc lbc battle creek 3 fall fry NA NA 38 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 clear creek lcc lcc clear creek 6 fall fry NA NA 37 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 clear creek lcc lcc clear creek 5 fall fry NA NA 38 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 battle creek lbc lbc battle creek 1 fall fry NA NA 34 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 clear creek lcc lcc clear creek 5 fall fry NA NA 36 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 battle creek lbc lbc battle creek 4 fall fry NA NA 36 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 battle creek lbc lbc battle creek 2 fall fry NA NA 37 0 3 1999 3 1999 assigned from fl cutoffs
1999-01-20 battle creek ubc ubc battle creek 1 fall fry NA NA 38 0 3 1999 3 1999 assigned from fl cutoffs

Generate rows for when no fish of a lifestage are caught

In order to improve the usability of this dataset (particulary for modeling) we decided to add rows for when a lifestage was not caught. For instance, there may be only fry caught on a particular day and when running the model for smolt that day would not show up in the dataset.

Review lifestage

The following plot shows the general patten in the lifestage field where fry are caught earlier in the year and smolt are caught later in the year.

Battle Creek: 2011