EPSCoR SCTC Aquatic Ecology
Juvenile salmon size and growth data
Analysis in progress
This document plots juvenile Chinook and Coho salmon length and weight data by species, drainage, year, and age.
The intended use for this data is to provide size input data for bioenergetic simulations.
Questions to consider:
- Are statistically significant growth trends observed using the seasonally-shifting-modes technique for:
- Weight
- Fork Length
- Chinook vs. Coho
- 2015 and/or 2016
Create mean length & weight values for fish grouped at level of:
- river
- year
- species
- season (“sample event”)
- age
# create table of mean lengths/weights by season for each year/spp/river
dat1 <- dat %>%
group_by(river,year,spp,sample.event.num,age) %>%
summarise(avg_wt = mean(wt_g),
avg_len = mean(len_mm),
n_wt = validn(wt_g), # number of fish with weight data
n_len = validn(len_mm), # number of fish with length data
n_diet = validn(sample.id)) %>% # number of fish diet sampled
mutate(season = fct_recode(sample.event.num, #code sampling events as seasons
"Early Summer"= "1",
"Mid-Summer" = "2",
"Late Summer" = "3",
"Fall" = "4")) %>%
mutate(river_spp = paste(river,"\n",spp)) %>%
mutate(count_label = paste0("n=",n_len)) %>%
ungroup()
# filter(n_wt > 1) # not sure why I had this in original codePlot size trends
Create dataframes for each species
Size Trends
1.) Coho
2.) Chinook
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><> <><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Coho Lengths
Coho Weights
Coho Catch Summary
## river year season wt_g
## Beaver Creek :1410 2015:1226 Early Summer: 485 Min. : 0.200
## Russian River :1064 2016:1879 Mid-Summer : 941 1st Qu.: 1.700
## Ptarmigan Creek: 365 Late Summer :1165 Median : 3.400
## Kenai River : 266 Fall : 514 Mean : 4.437
## 3rd Qu.: 6.500
## Max. :24.200
## len_mm
## Min. : 30.00
## 1st Qu.: 54.00
## Median : 67.00
## Mean : 68.76
## 3rd Qu.: 83.00
## Max. :132.00NOTE: Probably not useful to examine growth trends for Age 2 Coho because they are most likely all smolting out and departing.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><> <><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Chinook Lengths
Chinook Weights
Chinook Catch Summary
## river year season wt_g
## Beaver Creek :210 2015: 142 Early Summer:180 Min. :0.500
## Russian River : 92 2016:1002 Mid-Summer :414 1st Qu.:1.300
## Ptarmigan Creek: 39 Late Summer :349 Median :1.900
## Kenai River :803 Fall :201 Mean :2.196
## 3rd Qu.:2.700
## Max. :8.800
## len_mm
## Min. :38.00
## 1st Qu.:49.00
## Median :55.00
## Mean :56.45
## 3rd Qu.:62.00
## Max. :90.00Question: could growth inputs for bioenergetics somehow be generated though a von bertellanfy relationship???!?!?!
Adapt this code to markdown
<><><><><><><><><><><><><><><><><><><><><><><><><><>
To what temporal scale can groupings be reduced?
When did sampling occur?
a.) Test for significant difference among seasons for each spp/year. (Kruskall-wallis, nonparametric).
b.) If not consistent, test for differences between years for each spp/river. If not sigfnificant, consider combining years. (Question: Would this mean re-doing Age-Length Keys?)
QUESTION
Using the ALK-assigned ages, can the case be made for combining the datasets by year?
# prepare dataset
all_coho <- all_coho %>%
transform(year = as.character(year),
age = as.numeric(age),
lcat5 = as.numeric(lcat5))
# For each species/river, does the overall relationship between length and age
# differ by YEAR ??
# All Coho (NO, p = 0.38)
mod1 <- multinom(age~lcat5,data=all_coho,maxit=500)## # weights: 9 (4 variable)
## initial value 3411.191156
## iter 10 value 928.471817
## final value 920.788889
## convergedmod2 <- multinom(age~lcat5*year,data=all_coho,maxit=500)## # weights: 15 (8 variable)
## initial value 3411.191156
## iter 10 value 1050.628746
## iter 20 value 912.894106
## iter 30 value 911.794034
## final value 911.793999
## convergedanova(mod1,mod2)## Model Resid. df Resid. Dev Test Df LR stat. Pr(Chi)
## 1 lcat5 6206 1841.578 NA NA NA
## 2 lcat5 * year 6202 1823.588 1 vs 2 4 17.98978 0.001239786# Beaver Creek Coho (YES, p = 0.18e-05)
all_coho_bc <- all_coho %>%
filter(river == "Beaver Creek")
mod1 <- multinom(age~lcat5,data=all_coho_bc,maxit=500)## # weights: 9 (4 variable)
## initial value 1549.043327
## iter 10 value 486.346880
## final value 484.844741
## convergedmod2 <- multinom(age~lcat5*year,data=all_coho_bc,maxit=500)## # weights: 15 (8 variable)
## initial value 1549.043327
## iter 10 value 499.857349
## iter 20 value 471.810597
## iter 30 value 470.039665
## final value 469.890056
## convergedanova(mod1,mod2)## Model Resid. df Resid. Dev Test Df LR stat. Pr(Chi)
## 1 lcat5 2816 969.6895 NA NA NA
## 2 lcat5 * year 2812 939.7801 1 vs 2 4 29.90937 5.106827e-06# Russian River Coho (YES, p = 0.001)
all_coho_rr <- all_coho %>%
filter(river == "Russian River")
mod1 <- multinom(age~lcat5,data=all_coho_rr,maxit=500)## # weights: 9 (4 variable)
## initial value 1168.923475
## iter 10 value 272.776274
## final value 264.491376
## convergedmod2 <- multinom(age~lcat5*year,data=all_coho_rr,maxit=500)## # weights: 15 (8 variable)
## initial value 1168.923475
## iter 10 value 271.834760
## iter 20 value 247.102050
## iter 30 value 245.368602
## final value 245.366756
## convergedanova(mod1,mod2)## Model Resid. df Resid. Dev Test Df LR stat. Pr(Chi)
## 1 lcat5 2124 528.9828 NA NA NA
## 2 lcat5 * year 2120 490.7335 1 vs 2 4 38.24924 9.954295e-08# Ptarmigan Creek Coho (YES, p = 0.04)
all_coho_pc <- all_coho %>%
filter(river == "Ptarmigan Creek")
mod1 <- multinom(age~lcat5,data=all_coho_pc,maxit=500)## # weights: 9 (4 variable)
## initial value 400.993485
## iter 10 value 142.599507
## final value 142.243360
## convergedmod2 <- multinom(age~lcat5*year,data=all_coho_pc,maxit=500)## # weights: 15 (8 variable)
## initial value 400.993485
## iter 10 value 159.003677
## iter 20 value 140.125902
## iter 30 value 139.013568
## iter 40 value 138.270323
## iter 50 value 138.224318
## final value 138.223089
## convergedanova(mod1,mod2)## Model Resid. df Resid. Dev Test Df LR stat. Pr(Chi)
## 1 lcat5 726 284.4867 NA NA NA
## 2 lcat5 * year 722 276.4462 1 vs 2 4 8.040542 0.09010432# Evidence generally supports that there is no significant difference between years at the
# overall level, but evidence suggests that a difference between years exists when
# examining at the scale of individual watershed.
# What does it look like when data is combined among years ????
# Create new df for mean values & labels;
# values not grouped by year
# create table of mean lengths/weights by season for each spp/river
dat2 <- dat %>%
group_by(river,spp,sample.event.num,age) %>%
summarise(avg_wt = mean(wt_g),
avg_len = mean(len_mm),
n_wt = validn(wt_g), # number of fish with weight data
n_len = validn(len_mm), # number of fish with length data
n_diet = validn(sample.id)) %>% # number of fish diet sampled
mutate(season = fct_recode(sample.event.num, #code sampling events as seasons
"Early Summer"= "1",
"Mid-Summer" = "2",
"Late Summer" = "3",
"Fall" = "4")) %>%
mutate(river_spp = paste(river,"\n",spp)) %>%
mutate(count_label = paste0("n=",n_len)) %>%
ungroup()