Overview
As part of a class project, I worked with Jaleise Hall, Minnie Ringland to create time series of salmon at Willamette Falls.
Hydroelectric power represents an important renewable and low-emissions energy source, but the construction and development of the water source that the power plants can require sometimes threatens resident fish populations. Willamette Falls, located outside of Portland in northwestern Oregon (see map below), is an important traditional Native American fishing ground. To protect this natural resource and aid the passage of salmon and steelhead runs over the falls, fishways have been constructed and updated over time. Daily fish counts are monitored to ensure that migration of these fish populations continues to be unhindered by the power plant and the falls. This report summarizes findings from studying Willamette Falls monitoring data from 2001 to 2010.
Data were shared by and accessed from Columbia River DART:
Columbia River DART (Data Access in Real Time), Columbia Basin Research, University of Washington
Accessed Feb 1, 2021 at http://www.cbr.washington.edu/dart/query/adult_graph_text.
Show code
world <- ne_countries(scale = "medium",
returnclass = "sf") # pull world data
states <- st_as_sf(map("state",
plot = FALSE,
fill = TRUE)) # pull state data
states <- cbind(states, st_coordinates(st_centroid(states))) # project
falls <- data.frame(longitude = c(-122.61763),
latitude = c(45.35239)) %>% # make point
st_as_sf(coords = c("longitude", "latitude"), # as sf
crs = 4326,
agr = "constant")
ggplot(data = world) + # plot
geom_sf(fill = "antiquewhite") +
geom_sf(data = states,
fill = "peachpuff3") + # add state outlines and fill
geom_sf(data = falls,
size = 4,
shape = 23,
fill = "royalblue3") + # add falls loc
coord_sf(xlim = c(-125, -110),
ylim = c(40, 50),
expand = FALSE) + # set bounding
theme_minimal() + # minimal theme
labs(title = "Willamette Falls location", # add labs
caption = "Bri Baker, 2021")
Figure 1: Map indicating location of Willamette Falls in Oregon.
Show code
willamette_salmon <- read_csv(here("_posts", "2021-02-07-timeseries","data", "willamette_fish_passage.csv")) %>% # read in data
clean_names() %>% # names in tidy format
select(date, coho, jack_coho, steelhead) %>% # select desired species
mutate(date = mdy(date)) %>% # make date class
as_tsibble(key = NULL,
index = date) # convert to tsibble
salmon_longer <- willamette_salmon %>%
replace(is.na(.), 0) %>% # replace na with 0
rename(Coho = coho,
"Jack Coho" = jack_coho,
Steelhead = steelhead) %>% # nicer-looking names for graphing
pivot_longer(Coho:Steelhead, # consolidate species to one column
names_to = "species",
values_to = "counts")
Original time series
Show code
ggplot(salmon_longer, aes(x = date,
y = counts,
color = species)) + # make ggplot
geom_line() + #as a lineplot
labs(x = "Year", # add labs
y = "Count",
title = "Salmon counts at Willamette Falls fish passage",
subtitle = "2001 - 2010",
caption = "Bri Baker, 2021\nSource: Columbia River DART",
color = "Species") +
scale_color_manual(values = c("aquamarine3", "cornflowerblue", "goldenrod1")) + # change colors
scale_x_date(date_breaks = "1 year", # show all years
date_labels = "%Y") +
theme_minimal() + # use theme_minimal
theme(legend.position = c(0.15, 0.75), # move legend
legend.background = element_rect(fill="white",
linetype = "solid",
color = "whitesmoke"), # format legend
axis.text.x = element_text(angle = 30,
vjust = 1,
hjust = 1)) # angle x labels
Figure 2: Timeseries showing Coho (green), Jack Coho (blue), and Steelhead (yellow) salmon counts at Willamette Falls from 2001 to 2010. Seasonal and specific variation is apparent.
Takeaways
- The most apparent pattern in the data is seasonality on an annual basis, with peak sightings occurring mid-year.
- The bulk of Steelhead movement occurs earlier in the year than that of Coho and Jack Coho.
- Coho increased in abundance over the time studied, particularly in 2009 and 2010.
Seasonplots
Show code
salmon_season <- salmon_longer %>%
index_by(yy_mm = ~yearmonth(.)) %>%
group_by(species) %>%
summarize(monthly_mean_counts = mean(counts,
na.rm = TRUE)) # get monthly averages by species
salmon_season %>%
gg_season(monthly_mean_counts) +
scale_color_gradient2(labels = seq(2000, 2010, 1),
breaks = seq(0,10,1),
low = "goldenrod1",
mid = "aquamarine3",
high = "cornflowerblue",
midpoint = 5)+
labs(title = "Mean Seasonal Salmon Counts at Willamette Falls Fish Passage",
y = "Average Counts",
x = "",
subtitle = "2001 - 2010",
caption = "Jaleise Hall, 2021\nSource: Columbia River DART") +
theme_minimal()
Figure 3: This plot shows the monthly averaged seasonal salmon counts for each of the salmon species observed at Willamette Falls Fish. Colored lines represent the different years of observation between 2001 - 2010.
Show code
salmon_season %>%
gg_subseries(monthly_mean_counts) +
labs(title = "Mean Montly Seasonal Salmon Counts at Willamette Falls Fish \nPassage",
y = "",
x = "",
subtitle = "2001 - 2010",
caption = "Jaleise Hall, 2021\nSource: Columbia River DART") +
scale_x_yearmonth(date_labels = "20%y", # %y only add the last two digits of the year.
date_breaks = "3 years") # show every 3 years to look cleaner
Figure 4: The plot takes the monthly averaged seasonal salmon counts and separates them by month to show individual trends occuring for a single month by the corresponding species.
Show code
#NOTE: I did not add theme_minimal() to the subseries plot as it makes the plot look a bit messy in my opinion
Takeaways
- Both the seasonplot and subseries plot make the seasonal difference in peak sightings for each species apparent. As gathered from the original time series, Steelheads are observed year round, but peak in early summer (July). Conversely, Coho and Jack Coho are not frequently observed until their peak in the fall (Aug - Nov).
- There is a decreasing trend in Steelhead abundance from 2001 - 2010. While there is increased abundance of Jack Coho and Coho salmon, the changes in abundance seem random as opposed to following a clear trend.
- Average peak counts of salmon are much higher for Steelhead and Coho than for Jack Coho.
Annual counts by species
Show code
# Using Bri's "pivot_longer-ed" df:
annual <- salmon_longer %>%
group_by_key() %>% # group by species
index_by(yr = ~year(.)) %>% # index by year
summarize(annual_count = sum(counts)) %>% # sum counts by year and species
mutate(yr = as.Date(ISOdate(yr, 1, 1)))
# Using Bri's graph formatting:
ggplot(data = annual, aes(x = yr,
y = annual_count,
color = species)) +
geom_line(size=1.25) +
labs(x = "Year",
y = "Count",
title = "Salmon counts at Willamette Falls fish passage by year",
subtitle = "2001 - 2010",
color = "Species",
caption = "Minnie Ringland, 2021\nSource: Columbia River DART") +
scale_color_manual(values = c("aquamarine3", "cornflowerblue", "goldenrod1")) + # change colors
scale_x_date(date_breaks = "1 year",
date_labels = "%Y") + # show all years in YYYY format
theme_minimal() +
theme(legend.position = c(0.75, 0.75), # move legend
legend.background = element_rect(fill="white",
linetype = "solid",
color = "whitesmoke"), # format legend
axis.text.x = element_text(angle = 45,
vjust = 1,
hjust = 1))# angle x labels
Figure 5: Annual timeseries showing Coho (green), Jack Coho (blue), and Steelhead (yellow) salmon counts at Willamette Falls from 2001 to 2010.
Takeaways
- As seen in the original time series, Steelhead were sighted in greater numbers than Coho and Jack Coho most of the study period, suggesting that the population is more abundant or that these fish are easier to count.
- Once we aggregate the data by year, there do not appear to be any patterns in terms of cyclicality or trends common across species (The monthly variation within a year is hidden).
- Within species, Steelhead populations appear to be trending down, while Coho fish were seen in greater abundance over time, particularly in the last two years of the study period, which could suggest a population increase. Jack Coho fish seem to be present in very low numbers with no change over time, which could be cause for concern.