In agroimpacts/longisland: Bouy Data Change in Long Island

knitr::opts_chunk$set(echo = TRUE)

R code should follow the style we have been using all along.

1. Introduction (SA)

The Long Island Sound Study in Connecticut, New York, and Road Island shores serves as a new robust and localized case study to learn from and possibly improve. It's a "bi-state partnership" that includes networks with federal and state agencies, user goups, organizations, and individuals that are concerned with the health of the Sound (About the Long Island Sound Study). The study uses Ecosystem Indicators measuements to assess the health of the Lond Island Sound ecosystem among climate change impacts like sea level rise, temperature change, and ocean acidification (What You Should Know). The main indicators fall into categories on waters and watersheds, habitats and wildlife, and community sustainability and resilience. Because of the broad effects of climate change to the region, quantitative indicators relate to a range of topics, those including physical, biological, chemical, and socioeconomic (LISS Ecosystem Targets and Supporting Indicators). Physical measurements like temperature and salinity of water can be used to track water mass movements. These indicators and the level of dissolved oxygen can indicate how suitable the water is for marine life. Other chemical and biological analyses on animal tissue can indicate what chemicals and biological material is present in the water, including contaminents. Minitoring of water and sampling of species can occure on a continuous, temportary, seaosnal, or emergency basis. Many buoy locations where monitoring occurs happens at different intervals and time rnages, which affects this analysis (Water Quality Monitoring).

Water quality indicators used here include surface temperature, dissolved oxygen, and salinity. We focus on these metrics because we understand them better and because of time and scope contraints. The measurments come from the Water Quality Monitoring Program from the Connecticut Department of Energy and Environmental Protection on behalf of the study (Department of Environmental Protection). A helpful map to get a sense of the study area can be found here (Long Island Sound Monitoring Stations...). This map shows where the four main regions of the study. From west to east these regions are The Narrows, Western Sound, Central Sound, and Eastern Sound. Notice that The Narrows is losets to New York City, the Western Sound is closest to Bridgeport and other populated areas, and the Central and Eastern Sound are located between less-populated regions (Long Island Sound Report Card).

(KP) The data that is presented in this project is meant to provide us with valuable information about certain variables that may impactthe water quality of the Long Island Sound as mentioned before. To help look at and examine these qualities we will examine the different buoy points and the amount of change they experienced in terms of density, conductivity, temperature, salinity, and dissolved oxygen. There are 29 buoy points located along the Sound that was collected and we will examine for this project.

2. Methods

We first load all the libraries we need, and even a little extra just in case we need them. (SA)

We assess in detail the dissolved oxygen, temperature, and salinity indicators at four selected buoy locations and reiterate their significance, and create graphical ouptuts to accompany the study's online graphics and information. We include some statstical analysis: descriptive statistics, correlation, variance, and regression to explain the chosen indicators at the four buoys. We also include water quality index metrics, shellfish harvest data, and fish abundance data to illistrate the health of the sound. We also elaborate on consider data and time limitations and ways in which this project could improve in the future.

library(devtools)
library(geospaar)
library(dplyr)
library(tidyr)
library(tidyverse)
library(ggplot2)
library(fs)
library(mapview)
library(rgeos)
library(rgdal)
library(rmarkdown)
library(sf)
library(sp)
library(tibble)
library(stats)
library(utils)
library(knitr)
library(gstat)
library(raster)

We develop a "mastersheet" with coordinate points at 29 buoy locations, and water quality change for the temperature, dissolved oxygen, and salinity indicators calculated from measurements at two dates spanning at least 10 years and at the same time of year, during the summers. We import this sheet as a csv file from our working directory, and add coordinates as an sf oject. This data should be downloaded into the given local directory from the code above, but this is the code we use to upload data locally. We also delete NA values and change the names of the data. (SA)

library(longisland)

mastersheet <- readr::read_csv(
  system.file("extdata/Mastersheet.csv", package = "longisland")) %>%
  st_as_sf(., coords = c("long", "lat"), crs = 4326)
mastersheet_na.omit <- na.omit(mastersheet)
mast <- mastersheet_na.omit

To map, we first use the mapview function from the mapview package to view the data on an interactive display with a proper map of the Long Island Sound as a background extent. As an x and y coordinate is provided for each buoy point, the exact location is presented in the mapview. For each different mapview there is a different column is taken from Mastersheet.csv. As each column provides data for a certain variable, each mapview represents buoy location data associated with a variable (eg, Dissolved Oxygen Change). When making these mapviews we add a new sequence and a legend, and keep the colors as the defualt as shown below. These maps will allow us to see where the buoys are located along the sound and the distribution of the variables. (KP)

library(mapview)
mapview_study_regions <- mapview(mast, zcol = "Region", legend = TRUE)
mapview_study_regions

mapview_ox <- mapview(mast, zcol = "Dissolved Oxygen Change (mg/L)", at = seq(-3, 4, 1), legend = TRUE)
mapview_ox

mapview_sal <- mapview(mast, zcol = "Salinity Change (PSU)", at = seq(-1.5, 1.5, 0.5), legend = TRUE)
mapview_sal

mapview_temp <- mapview(mast, zcol = "Temperature Change (Fahrenheit)", at = seq(-1, 3, 0.5), legend = TRUE)
mapview_temp

mapview_con <- mapview(mast, zcol = "Conductivity Change (S/m)", at = seq(-0.3, 0.3, 0.1), legend = TRUE)
mapview_con

mapview_den <- mapview(mast, zcol = "Density Change (kg/m^3)", at = seq(-1.8, 0.9, 0.3), legend = TRUE)
mapview_den

We now create an aerial interpolation to better have a better visual of what areas of the Long Island Sound may have higher or lower values of each change that is being examined at all 29 buoy points. To do this we are using the Mastersheet data as the extent of our raster data. Mastersheet_New.csv is the same file as Mastersheet.csv, the only difference is on Mastersheet_New.csv I had to change the column name "Conductivity Change" to "Con_Change", and all other variables to "abbrevation_change" because aerial interpolation would not run if there was a space in the column name. For each interpolation I used data of the US outline from getData and then performed an inverse mask on it to allow the interpolation to only cover areas not on US land. In this case it was the Long Island Sound. Using this feature allows us to examine the general pattern of change throughout the Long Island Sound for each variable, and have a better idea where lower or higher values of change are occurring. (KP).

mastersheet_new <- readr::read_csv(
  system.file("extdata/Mastersheet_new.csv", package = "longisland")) %>% 
  st_as_sf(., coords = c("long", "lat"), crs = 4326)
r <- raster(extent(mastersheet_new), res = 0.03)
US_Data <- getData('GADM', country='USA', level=0) %>% st_as_sf
# US_Data %>% st_as_sf %>% 
#   filter(NAME_1 %in% c("New York", "Connecticut", "New Jersey")) -> US_Data 
# US_Data %>% st_geometry %>% plot(col = "grey")
US_Data <- st_crop(US_Data, st_bbox(mastersheet_new)) %>% as_Spatial

system.file("extdata/Mastersheet_new.csv", package = "longisland")
dat <- read_csv(system.file("extdata/Mastersheet_new.csv", package = "longisland"))
mastersheet_new <- readr::read_csv(system.file("extdata/Mastersheet_new.csv", package = "longisland")) %>% st_as_sf(., coords = c("long", "lat"), crs = 4326)
dat <- mastersheet_new %>% as_tibble %>% 
  select(-geometry) %>% 
  bind_cols(., mastersheet_new %>% st_coordinates %>% as_tibble) %>% 
  rename(x = X, y = Y)
invdistdo <- gstat(formula = Do_change ~ 1, locations = ~x + y, data = dat)
invdistrdo <- interpolate(object = r, model = invdistdo)
invdistrmsk_do <- mask(x = invdistrdo, mask = US_data, inverse = TRUE)
title_do <- c("Dissolved Oxygen Change (kg/m^3)")
plot(invdistrmsk_do, main = title_do)

invdistsal <- gstat(formula = Sal_change ~ 1, locations = ~x + y, data = dat)
invdistrsal <- interpolate(object = r, model = invdistsal)
invdistrmsk_sal <- mask(x = invdistrsal, mask = US_data, inverse = TRUE)
title_sal <- c("Salinity Change (PSU)")
plot(invdistrmsk_sal, main = title_sal)

invdisttemp <- gstat(formula = Temp_change ~ 1, locations = ~x + y, data = dat)
invdistrtemp <- interpolate(object = r, model = invdisttemp)
invdistrmsk_temp <- mask(x = invdistrtemp, mask = US_data, inverse = TRUE)
title_temp <- c("Temperature Change (Fahrenheit)")
plot(invdistrmsk_temp, main = title_temp)

invdistcon <- gstat(formula = Con_change ~ 1, locations = ~x + y, data = dat)
invdistrcon <- interpolate(object = r, model = invdistcon)
invdistrmsk_con <- mask(x = invdistrcon, mask = US_data, inverse = TRUE)
title_con <- c("Conductivity Change (S/m)")
plot(invdistrmsk_con, main = title_con)

invdistden <- gstat(formula = Den_change ~ 1, locations = ~x + y, data = dat)
invdistrden <- interpolate(object = r, model = invdistden)
invdistrmsk_den <- mask(x = invdistrden, mask = US_data, inverse = TRUE)
title_den <- c("Density Change (kg/m^3)")
plot(invdistrmsk_den, main = title_den)

We create a sf plot in addition, but it does not explain the extent as well as the mapview and is not interactive. However it is interesting to see and easy to interpret. We index the mastersheet by column first to split the dataframe, or index by column. We change the astletics and make quantile breaks with unique break incriments for the lengend for the dissolved oxygen, temperautre, and salinity seperatly. For example: (SA)

mast_sal_p <- plot(mast[8], axes = TRUE, cex = 1.75, pch = 10, main = "Salinity Change at Buoys", key.pos = 1, breaks = "quantile", key.width = lcm(1.25), key.length = 1.0) + scale_y_continuous(breaks = c(-2, -1.75, -1.50, -1.25, -1, -.75, -.50, -.25, 0, .25, .50, .75, 1, 1.25)) + theme(plot.title=element_text(face="bold", size=14))

Data on water quality change is extracted at individual buoy locations in the analysis from The Narrows, the Western Sound, the Central Sound, and the Eastern Sound (Long Island Sound Coastal Observatory...). We use 13 individual buoy locations. Some points from the mastersheet do not have complete data between time periods. (SA)

BI2 <- readr::read_csv(
  system.file("extdata/I2.csv", package = "longisland")) 
BI2_na.omit <- na.omit(BI2)
bI2 <- BI2_na.omit
bI2

To look at the change in the indicator values, we first create a gpplot of the bC1 buoy at The Narrows to show the dissolved oxygen, temperature, and salinity change over time. This buoy time period spans between December 05, 1994 and November 06, 2017, and has 434 points. We create asthtics for the plot, but keep it relatively simple. Here is an example of a code chunk: (SA)

bC1 <- readr::read_csv(
  system.file("extdata/C1.csv", package = "longisland")) 
bC1_oxygen <- ggplot(bC1, aes(newdate, oxygen, fill = group)) +
geom_point(shape= 21, fill= "#33CCCC", colour= "black", size=2) + xlab("Date") + ylab("Dissolved Oxygen") + ggtitle( "Dissolved Oxygen at bC1") + theme_light() +labs(subtitle = "Oxygen in mg/L", caption = "Enought dissolved oxygen content is an important indicator of good water quality") + theme(plot.title=element_text(face="bold", size=14))
bC1_oxygen

We do the same process for b16 at the Western Basin, b28 at the Central Basin, and bJ2 at the Eastern Basin, and the chunk of code to plot dosn't change much except for the input data and output names, as well as title and caption information. (SA)

We produce complimentary statistical analysis for the four buoy points we graphed earlier in ggplot. We use the summary function to find the minimum, 1st quartile, median, mean, 3rd quartile, and maximum values from the dissolved oxygen, temperautre, and salinity columns. We take the values produced to create a new dataframe with three columns for summary type (minimum, Q1, etc), water quality (dissolved oxygen, temperautre, salinity) and value (summary value). here is an example chunk of code for that: (SA)

summarybC1df <- data.frame(SummType = rep(c("Min", "Q1", "Med", "Mean", "Q3", "Max"), each = 3), WaterQual = rep(c("Oxygen", "Temp", "Sal"), 6), Value = c(5.235, -0.585, 23.11, 8.193, 5.485, 25.66, 9.470, 13.556, 26.61, 9.615, 13.233, 26.59, 10.995, 21.071, 27.49, 14.386, 25.762, 29.05))

summarybC1df

We then make sure that these columns are of the factor type, using the as.factor function for the SummType column: (SA).

summarybC1df$SummType <- as.factor(summarybC1df$SummType)

We then reorder these factors to force a conceptually resonable order. We do not rename the dataframe: (SA)

summarybC1df$SummType <- factor(summarybC1df$SummType, levels=c("Min", "Q1", "Med", "Mean", "Q3", "Max"))

summarybC1df$SummType

We then create a bar graph with asthetics (SA).

summary_bC1_bar <- ggplot(data = summarybC1df, aes(x = WaterQual, y = Value, fill = SummType)) + geom_bar(stat = "identity", position = position_dodge()) + ggtitle("Summary of Descriptive Stats at bC1") + theme(plot.title = element_text(face="bold", size=14)) + scale_fill_manual(values=c("#33CCCC", "#33CC00", "#996633", "#33FFFF", "#336600", "#CC9933"))

summary_bC1_bar

In addition to creating a bar graph outlining summary statistics for the four buoy points, we take a look at correlation and variance between these indicators by using simple code from the cor and var functions. We first create a new dataframe to use for correlation and remove unnecesary columns: (SA)

bC1_cor<- dplyr::select(bC1, -ID, -date, -newdate, -depthcode, -biosil, -den)

bC1_cor

We then find the correlation (pearson) and variance between the three indicators. We add three variance functions with different index values or names (depending on the data type): (SA)

cor(bC1_cor, use = "complete.obs", method = "pearson")

var(bC1_cor[1], y = bC1_cor[2], na.rm = TRUE, use = "complete.obs")

We create a new dataframe, order it, and plot it as a bar graph to explain visually and be easier to interpret. The framework is the same for creating bargraphs for the summary statistics section. (SA)

bC1_cordf <- data.frame(corType = rep(c("Variance", "Correlation"), each = 3), Comparison = rep(c("OxTemp", "TempSal", "OxSal"), 2), Value = c(-.655, -10.776, .002, .016, .118, .016))

bC1_cordf

bC1_cordf$corType <- factor(bC1_cordf$corType, levels=c("Correlation", "Variance"))

bC1_cor_bar <- ggplot(data = bC1_cordf, aes(x = Comparison, y = Value, fill = corType)) + geom_bar(stat = "identity", position = position_dodge()) + ggtitle("Correlation at bC1") +
    theme(plot.title = element_text(face="bold", size=14)) + scale_fill_manual(values=c("#33CCCC", "#33CC00"))

bC1_cor_bar

We then take a quick look as some functions that show more advanced metircs on the indicators, but we do not really need to investigate this for the analysis. The lm function conveniently plots four outputs that take a little time to interpret, but are essy to produce. Though we do not go into these plots in depth, they do expres this kind of inconsistency and lack of correlation between variables that we see in the bar graphs above. The scatter.smooth function create a scatterplot of two independent variables, which I create for one buoyfor the dissolved oxygen indicator ~ temperature just to display: (SA)

b16 <- readr::read_csv(
  system.file("extdata/b16.csv", package = "longisland")) 
b16
scatter.smooth(x = b16$oxygen, y = b16$temp, main = "Oxygen ~ Temperature", pch = 19, cex = .75, col = "#33CCCC") +  theme(plot.title=element_text(face="bold", size=14))

The loess.smooth function smooths "volatile" time sies. A least squares regression is performed on on "localized subsets", so it shows the smoothed version of the regression from above (Prabhakaran, 2017). We are unable to add titles and axis lables, so one should note the order in which they apply these commands. (SA)

loess.smooth(x = b16$oxygen, y = b16$temp, span = 1/3, degree = 1,
    family = c("symmetric", "gaussian"), evaluation = 50) %>% plot

loess.smooth(x = b16$oxygen, y = b16$temp, span = 2/3, degree = 1,
    family = c("symmetric", "gaussian"), evaluation = 50) %>% plot

loess.smooth(x = b16$oxygen, y = b16$temp, span = 3/3, degree = 1,
    family = c("symmetric", "gaussian"), evaluation = 50) %>% plot

Some data use use expresses preexisting non-buoy data on the Long Island Sound Study website that we transform here and analyze. We use water quality index categorizations for each Sound region between 2002-2011 that are either "good", "fair", or "poor". Using ggplot, we create a bar graph of this simple indicator to show the proportions against each other in 2011 for each region. I first create a dataframe similar to above bar graphs, order the data by index category similar to above bar graphs, then graph and add asthetics. Note, that the order of the regions is not correct going from west to east. The code looks like this: (SA)

allbasinsdf2 <- data.frame(WaterQual = rep(c("good", "fair", "poor"), each = 4), Basin = rep(c("The Narrows", "Western", "Central", "Eastern"), 3), Percent = c(19.3, 22.22, 46.81, 53.80, 63.2, 65.74, 48.94, 46.20, 17.50, 12.04, 4.26, 0))

allbasinsdf2$WaterQual <- factor(allbasinsdf2$WaterQual, levels=c("good", "fair", "poor"))

allbasinsbar2 <- ggplot(data = allbasinsdf2, aes(x = Basin, y = Percent, fill = WaterQual)) + geom_bar(stat="identity", position=position_dodge()) + scale_fill_manual(values=c("#33CC00", "#33CCCC", "#996633")) + ggtitle("Water Quality Index at Long Island Sound Basins") +
    theme(plot.title=element_text(face="bold", size=14))

allbasinsbar2

To explain the affect water quality change can have on the Sound, we upload harvest data for oysters, clams, and lobsters for the years 1990-2017 and 1981-2017 for Connecticut and New York. I create seperate dataframes to index with by shellfish type, and delete rows with missing infromation. We create line graphs with ggplot to see how the amount of shellfish harvested changes over time (x axis is year, y axis is bushels for oysters, bushels for clams, and landings for lobsters). We only use New York harvest data. We use ggplot and create asthetics. Unfortunately, we have a hard time plotting the lines over eachother: (SA)

shellny <- readr::read_csv(
  system.file("extdata/shellfishny.csv", package = "longisland")) 
shellny_oy <- shellny 
shellny_clams <- shellny

shellny_lob = shellny_lob[-36,]

shellny_lob_p <- ggplot(shellny_lob, aes(year, lobsters, linetype = lty)) +
geom_line(shape= 21, lwd = 1, lty = 1,  colour= "red", size=2) + xlab("Year") + ylab("Lobster Landings") + ggtitle("Lobster Landings") + theme_light() + labs(subtitle = "NY Landings between 1981 and 2015", caption = "LL reflect change of fish. mang. practices, socioeconomic conditions & species abundance")
shellny_lob_p

The warm and cold water fish is also made into a line graph using ggplot that groups each sample by the fall and spring between 1984-2014. Fin fish at the sound are seperated into two categories of prefering temperature below 15C or prefer temperautre ranging between 11-22C, or warmer temperatures. This is the average number of species in each group captured in the spring and fall collection dates (Howell & Gottschall, 2017). The plot is formatted as above, and asthetics are added, but the same issue of not being able to plot lines over each other seperates this index into four graphs when they should be two. Here is the fish index graph the Fall for warm water fish. (SA)

warmcoldfall <- readr::read_csv(system.file("extdata/warmcoldfall.csv", package = "longisland"))
wcfall <- warmcoldfall
wcspring <- warmcoldspring
wcfall_w_p <- ggplot(wcfall[4], aes("year", warmwater)) + 
geom_line(shape= 21, lwd = 1, colour= "red", size=2) +
xlab("Year") + ylab("Warm Water Fish Counts") + 
ggtitle("Warm Water Fish Counts During Fall") + theme_light()
wcfall_w_p

wcfall_c_p <- ggplot(wcfall[3], aes("year", coldwater)) + 
geom_line(shape= 21, lwd = 1, colour= "blue", size=2) +
xlab("Year") + ylab("Cold Water Fish Counts") + 
ggtitle("Cold Water Fish Counts During Fall") + theme_light()
wcfall_c_p

wcspring_w_p <- ggplot(warmcoldspring, aes(year, warm)) + 
geom_line(shape= 21, lwd = 1, colour= "red", size=2) +
xlab("Year") + ylab("Warm Water Fish Counts") + 
ggtitle("Warm Water Fish Counts During Spring") + theme_light() 
wcspring_w_p

wcspring_c_p <- ggplot(warmcoldspring, aes(year, cold)) + 
geom_line(shape= 21, lwd = 1, colour= "blue", size=2) +
xlab("Year") + ylab("Cold Water Fish Counts") + 
ggtitle("Cold Water Fish Counts During Spring") + theme_light()
wcspring_c_p

3. RESULTS (SA and KP)

Dissolved Oxygen:

Dissolved oxygen enters water at different rates from natural processes. One, diffusion from water turbidity directly adds oxygen from the surrounding air. Two, photosynthesis of plant life in water creates oxygen as a by-product (McCaffrey, n.d.). The mapview output for this indicator shows one point with over a -3 decrease, two points with a -3 to -2 mg/l decrease, four points with a -2 to -1 mg/l decrease, 7 points with a small -1 to 0 decrease, and 12 points with an overall increase between 0 and 3. The two largest decreases are located at The Narrows, however are relativly close to the two largest increases bordering The Narrow and the Western Sound. This does not show a strong visual trend of overall dissolved oxygen decrease. The sf plots show the same infomation as the mapview, just in a different format. For example, this is the direct comparison of the mapview output for dissolved oxygen and the sf output. In addition while the points do not show much of a pattern, the areal interpolation shows as what appears to be a greater decrease in dissolbed oxygen values in the Narrows (closer to New York City) with no change or slight increases across most of the rest of the Sound.

#mapview

mapview_ox

#areal interpolation
plot(invdistrmsk_do, main = title_do)
#sf plot

mast_ox_p <- plot(mast[6], axes = TRUE, cex = 1.75, pch = 10, main = "Dissolved Oxygen Change at Buoys", key.pos = 1, breaks = "quantile", key.width = lcm(1.25), key.length = 1.0) + scale_y_continuous(breaks = c(-1, -.50, 0, .50, 1, 1.50, 2, 2.50, 3, 3.50)) + theme(plot.title=element_text(face="bold", size=14)) #we need to run this whole chunk because the mast_ox_p by itself produces NULL

The color differenciation between the two is different, so there is no dirct comparison of change in mg/l of dissolved oxygen, though the sf plot is not necearily harder to interpret. Though there is no background extent, the colors of the buoys are more distinct, so it is easier to recognise values according to the legend. We think in the future we might want to recognise the ease at which we can interpret point information from the sf using the defualt colors, unless we can find a way to change the legend colors in the mapview.

Temperautre:

The study's website notes that the Long Island Sound water temperature is increaseing at a rate of about 1.8 degrees celcius per century. The trend notes in a graph on the website shows the increaseing trend. However, water temperatures in the region stay relativly consistent from day to day (What You Should Know). The mapview output for water temperature shows most buoy points experienced fairly large jumps in temperature with increases from around 0 to 3 degrees Fahrenheit. There is no significant visual pattern of decrease and increase in points. Here is a comparison between the mapview temperature change output, the areal interpolation, and the sf plotting tmeperature change output. It is still a little easier to interpret the sf plot just by way of having more hues, however the mapview is interactive and shows all point values from the mastersheet when clicked. The areal interpolation though shows the areas where general temperature increase is occurring which looks like closer to New York City in the Narrows Region and a little section of the eastern part of the sound, although that is likely influenced by one point. In addition, the sf plot necesitates the complete code to map in this case, and the mapview is allisgned a name and takes up less space.

#mapview

mapview_temp

#areal interpolation
plot(invdistrmsk_temp, main = title_temp)

#sf plot

mast_temp_p <- plot(mast[7], axes = TRUE, cex = 1.75, pch = 10, main = "Temperature Change at Buoys", key.pos = 1, breaks = "quantile", key.width = lcm(1.25), key.length = 1.0) + scale_y_continuous(breaks = c(-1, -.50, 0, .50, 1, 1.50, 2, 2.50, 3, 3.50)) + theme(plot.title=element_text(face="bold", size=14))

Salinity

Salinity is a measurement of dissolved salts in water. It is usually higher during periods of low water flows. Therefore, sea level rise over time can artificially decrease salinity level, decreasing its capacity to improve water conductivtiy and provide nutrients for marine life. Salinity at the study is measured in PSUs (Practical Salinity Unit), which is similar to the TDS (Total Dissolved Solids) measurement (Howell & Gottschall, 2017). The mapview output for salinity shows a variety of change, as expected. Overall though only five points experienced an increase in salinity. The areal interpolation shows that most of the middle of the sound is experiecning a decrease with no real pattern showing different levels of change occurring in different parts of the Sound. There is an overall decrease in salinity, which can support the overall sea level rise trend.

#mapview

mapview_sal

#areal interpolation
plot(invdistrmsk_sal, main = title_sal)

#sf plot

mast_sal_p <- plot(mast[8], axes = TRUE, cex = 1.75, pch = 10, main = "Salinity Change at Buoys", key.pos = 1, breaks = "quantile", key.width = lcm(1.25), key.length = 1.0) + scale_y_continuous(breaks = c(-2, -1.75, -1.50, -1.25, -1, -.75, -.50, -.25, 0, .25, .50, .75, 1, 1.25)) + theme(plot.title=element_text(face="bold", size=14))

Conductivity

By looking at the mapview and areal interpolation of Conductivity change it can be seen that there is a mixture between increases and decreased in Conductivity, but overall more buoy points expereinced a decrease in Conductivity with 18 of the 29 points doing so. Also through the mapview it is seen that there were four buoy points that had a decreased in conductivity between .2 and .3 (S/m), while only two points had an increase of the same magnitude. Most of the areal interpolation indicates either consistency or a negative change of Conductivity with some positive values along the Narrows and a couple isolated ones elsewhere as shown in both maps. As most of the buoy temperatures experienced an increase in temperature it would be expected for Conductivity to do the same, since increasing the temperature of a solution will decrease viscosity, and increase the mobility of ions in the water thus decreasing conductivity (Barron and Ahston). But it is important to acknowledge that there are other factors that impact Conductivity.

#mapview

mapview_con

#areal interpolation
plot(invdistrmsk_con, main = title_con)

Density

The mapview and areal interpolation show a general decrease in water density change. There are a few isolated spots where an increase occurs but overall a decrease trends is mostly indicated through the whole map as shown in the areal interpolation. On the areal interpolation it may indicate that there is more consistency or increase as some of the buoy points are more isolated than others giving them greater weight across the area of the Long Island Sound. But overall this is what would have been expected with increasing water temperatures at most buoy points since the density of water tends to increase as temperatures warm.

#mapview

mapview_den

#areal interpolation
plot(invdistrmsk_den, main = title_den)

Individual Buoy Assessemnt

We found that we do not have enough space to elaborate on all four buoys mentioned in the methods, so we focus on the bC1 point at The Narrows. The dates at buoy C1 span between 12/05/1994 and 11/30/2017 (about 23 years) and contains 434 valid points (after using rm.na). Points are taken at about equal intervals once a month (except for missing rows). The dissolved oxygen content at buoy C1 at The Narrows shows a downward trend over time and some variability. There is a lump in the middle that looks like a seasonal trend of increased oxygen, maybe at a time of greater water turbidity and amount of photosynthesis in the water. However we later note that there is also a noticable six-month-period gap between 6/6/2006 and 12/11/2006 that might have something to do with the downward tail in dissolved oxygen at the end, where there should be another leverage point downward, and then head up again before decreasing at the end. We are not sure how to explain the lack of evindent seasonal trends considering how many years there is in this dataframe, and consider it just hard to interpret visually without a trendline. There is fairly large variability between about 5 mg/L and 13 mg/l over time, though most of the points seeem to fall under 10 mg/l.

In fact, the bar graph underneath called summary_bC1_bar2 shows the mean for dissolved oxygen at just under 10 mg/l. The maximum value is at about 13 mg/l, and the minimum value is at about 5 mg/l. Because of the same range between intervals of the minimum to Q1, Q1 to the mean, the mean to Q3, and Q3 to the max. So, the total distirbution of points in this 23-year timespan seems to be normally distributed, which makes sense because of the large number of sample points.

#dissolved oxygen

bC1_oxygen

#summary statitics as bC1 

summary_bC1_bar2

The temperature output is interesting. It shows a seasonal trend in the span of about 23 years. Between each year there is about a 10-degree gap between the lowest temperature of the year at about January and the highest temperature of the year at about August. If there was a polynomial trendline added we would see it quiver between each year. It's difficult to find specific indications about general water temperature increase at bC1 because it is only one buoy and a small timespan, however I do notice that the maximum August temperature between 2015 and 2017 at greater than those from the early 2000s. This could indicate decreasing water quality. There is a gap in time between six month gap between 6/6/2006 at 12.0588 degrees celcius and 12/11/2006 at 13.3754 degrees celcius. If this gap was filled there would be another lump upward showing temperature increase about 20 degrees celcius. There are a few more smaller gaps, including a 3-month gap between 2/27/1998 and Jun 01 1998 at the beginning that is noticable in the graph.

The summary statistics of the temperautre at bC1 can be found again with summary_bC1_bar2. It explains the greater range of temperature values than dissolved oxygen values, showing the maximum at a lot higher than the minimum. However, the mean and the medium are about the same, suggesting that collected together, the data follows a general normal distirbution.

#temperature

bC1_temp

#summary statitics as bC1 

summary_bC1_bar2

For salinity, there seems to be a back-and-forth between very high and very low values that could folow weather patterns. Usually, when there is a dry period without rain for a few days, it rains enough to make up for the difference in the span of only a few hours which can rapidly decrease salinity. Some spans without rain are more extreme than others depending on regional climatic pattenerns. There are stronger storms at fairly regular intervals, which provide more rain than normal and decrease salinity further. Increase in salinity is due to a decrease in water level and an increase in human industrial and activity discharge, which changes more steadily and at a slower pace than rainfall per week. This chart would probably look different in a region like Seattle where it rains more regularily. There, the graph would likely have a smaller range of values, as there is less opportunity for water to evaporate before the next rainfall.

The summary statistic bar graph below shows less overall range between values than the other graphs, though the values themselves are higher overall becaue of the different PSU units. The data alltogether looks fairly normally-distributed.

#salinity

bC1_sal

#summary statitics as bC1 

summary_bC1_bar2

In addition, I check the correlation and variance bar graph that I created earlier in the code below. It shows that there is a large negative variance between the temperature and salinity variables. This means that there is a large differenc beetween the sqaurd inputs values to the expected values, and reminds us of the differences between the temperature and salinity distributions. The temperature variable has a smaller range of values and a cleaner, seaosnal trend, while the salinity variable has less trend and correlation between it and date, creating higher variability when compared.

#correlation and variance

bC1_cor_bar

Other Indicators

This is a good summary bar graph showing the pre-assigned water quality categories from the study. It shows that The Narrows has the highest proportion of poor water quality, which could be related to its close proximity with New York City and other populated areas. The Western Basin has the next highest proportion of poor water quality, which could be related to its proximity to densly-populated areas south and noth of the New York areas, including Bridgeport. This graph gives the impression that there is overall O.K. water quality in the sound and there is no emmidiate crisis. Ecosystem protection management practices and water qualty monitoring has helped the Sound become a healthier place since the 1980s. However, at The Narrows the larger proportion of 20% poor water quality should be addressed.

allbasinsbar2

Next, we will quickly review results for shellfish harvest and abundance measurments.

#Lobster harvest in New York

shellny_lob_p

#Oyster harvest in New York
shellny_oy_p

#Clam harvest in New York
shellny_clams_p

The fish abundance data shows an interesting trend. We see that in the fall and spring warm water fish abundance is generally increasing between 1984-2014, and cold water fish are generally decreasing in the fall and spring. Here it would have been helpful to plot the blue cold water fish and red warm water fish over eachother for both seasons to make two plots instead of four, but we were unable to do that.

#fall

wcfall_w_p #warm water fish index
wcfall_c_p #cold water fish index

#spring

wcspring_w_p #warm water fish index
wcspring_c_p #cold water fish index

4. Discussion (SA, KP)

Dissolved Oxygen

Widespread decrease in dissolved oxygen in natural water can have a detrimental affect on marine life, which filter oxygen through gills stay alive. The study notes that 5 mg/l can have a mildly detrimental affect on marine life, and levels below 1.5 mg/l can cause severe affects. Decreased levels of dissolved oxygen can reduce the abundance and diversity of adult fish, reduce their growth rate, reduce the abundance of especially slow-miving fish and shellfish like lobsters, and reduce their resistance to disease (Hypoxia, The Problem). As mapview and interpolation show there is more of a genearl decrease in dissoled oxygen levels, which is particularly prominent at points within the Narrows Region, with some areas of less DO change or positive DO change. However, the results from bC1 in the bC1_oxygen plot do not show widespread decreases in dissolved oxygen below 5 mg/l, which is good. However, it would be helpful if we had plotted averages in dissolved oxygen over the regions of interest for maximum time spans to smooth the data and chunk it into larger extents. We did not think about doing that at the beginning of the project.

Temperature

The lower the temperature of the water, the easier it is for oxygen molecules to diffuse with water. Water at 0 decrees celcius will hold 14.6 of oxygen per litre, but water at 30 degrees celcius will hold only 7.6 mg/L. Water temperaure change also affects the rate of photosynthesis in plants, reiliency to disease, and overall health and devleopment in aquatic life (McCaffrey, n.d.).

The Long Island Sound study notes that the overall mean suface water temperautre in 1976 through 2015 was 3.9 degres elcius in the winter, 11.22 degrees celcius in the spring, 20.07 degrees celcius in the summer, and 12.24 degrees celcius in the fall (O'Donnell, n.d.). The general mapview and areal interpolation showed that there were mostly areas of increased temperature across the Sound with some buoy points having almost consistent temps, or a slight decrease in temperatures. This may indicate that temperatures are warming across the Sound, particularly areaas closer to the Narrows, and some other isolated points throughout the Sound. However, most August months at bC1 was at over 20 degrees celcius, an indication that it is reaching past the upper threshold and expanding the range of temperatures, as well as increasing the overall temperatre. Though the trend in the region is understood, the over-simplicated methodology of temperature differencing from two points does not explain it well. Again, it would be better if we had used an aggregation method with more points and more times, taking the average temperautre over a span of time from the later time dates and subtracting it from that of the earlier time dates. It would also be helpful if the buoy data did not include large gaps in time, so in a later run we could only use datasets that are complete in this measure, and also use the same time scales and incriments. It would also be helpful if we had the ability to add trendlines to see a clearer and more concrete picture of the trend instead of simply conceptualizing it.

Salinity

An appropriate concentration of salinity is important for aquatic plants and aniamls in any watershed, and salinity beyond the normal range for any species can be highly detrimental to their health. Salinity also affects the avaliability of nutrients to enter plant roots (McCaffrey, n.d.). The eastern end of the sound should have 35 parts per thousand in salinity and the western should have 25 (Long Island Sound By the Numbers). The salinity at bC1 ranges between 23.11 and 29.05 PSU, which overall is a little high. However, the measurments were all taken during the summer, when there is a tendency for the Sound water to evaporate more. Again, we cannot realistically link changes in climatic condiditons in this region to the time differencing values here, so we cannot link changes in precipitation and water turbidity to peaks in salinity to conclude with. Although in the mapview and areal interpolations there are increases in salinity overall with most buoy points experiecning an increase, with some isolated points having a small decrease. But the different can contribute to the times of year this data points were collected as evaportation can play a role in the amount of salinity. In the future, it would be great to compare these results to similar ones from another region with differnt climatic conditions. Maybe, a region with more variables rainfall would give us a better picture of salinity change of a region with different climatic conditions.

Conductivity

This variable was looked at through the maps. But as indicated in the results section there was a general decrease in Conductivity across the Sound. As the data did show a temperature increase overtime at most buoy points it would be expected for overall Conductivity to increase as well (Barron and Ashton). It would be valuable for future reference to continue looking at where Conductivity is increasing and decreasing. By looking at mapview and areal interpolation it was seen that more of the increases occurred in the Narrows Region of the Long Island Sound, while it appeared that more of the decreasing values were at buoys out in the middle of the Sound. This could also be indicative of the amount of pollution that is entering the Long Island Sound, as the increase of dissolved solids in water will increase conductivity (EPA). If pollution in the Sound was signficant enough to be impacting conductivity on the Sound, than that would mean the Narrows Region would be most impacted by pollution as it is experiencing the greatest increases of Conducitivity. This would be something interesting to examine in the future. Also changes in Conductivity showed relatively the same patterns in certain areas that temperature experienced when looking at the areal interpolation, meaning temperature may have impacted how much Conductivity increased or decreased to an extent. It would be interesting to look at temperatures role in Conductivity as well as other impacts such as pollution to see if warming of the Earth or pollution is making a bigger impact on Conductivity levels on the Sound.

Density

Overall there was shown to be a decrease in water density for the most part across the Long Island Sound which is expected with temperature increase. The areal interpolation shows that there is a general decrease across the Sound, with few areas of density consistency or increase. The mapview does a better job showing that most points have negative density changes with a few isolated spots of positive change occurring with no real pattern. With as many points that experienced temperature increase it is expected to see water density decrease due to water expanding as a result of warmer temperatures (Temperature Effects on Density). As temepratures are expected to generally rise due in the Long Island Sound to warming of the Earth (What you Should Know), it would be interesting to see if water desnity will continute to show a general decrease.

Other Indexes

The warm and cold-water fish index graphs in wcfall_w_p, wcfall_c_p, wcspring_w_p, and wcspring_c_p are the most expressive of water quality change. It is clear that warmer fish are finding the region more suitable, and colder fish are finding the region less suitable. The study's website suggests that there have been shifts in migratory patterns in fish as surface water temperature warm, with both death and steady decline in abundance of cold-water fish like lobster and winter flounder, and steady migration and prosperation of warm-water fish like summer flounder (What you Should Know). Though we do not include economic data in this research, it would be interesting in the future to follow economci trends of harvest for both winter and summer flounder, or another compatable warm/cold pairing of fish. This can have some affect on the local fishing economy and link to overall economic health of the region, one of the main indicators of changing Long Island Sound health noted online.

5. References (SA, KP)

About the Long Island Sound Study. (n.d.). Retrieved May 4, 2019, from http://longislandsoundstudy.net/about/about-the-study/

Barron, J. J., & Ashton, C. (n.d.). The Effect of Temperature on Conductivity Measurement. Retrieved from https://www.camlab.co.uk/originalimages/sitefiles/Tech_papers/TempCondMeas.pdf.

Department of Environmental Protection, Long Island Sound Water Quality Monitoring. (2014, August 21). Retrieved May 4, 2019, from https://www.ct.gov/deep/cwp/view.asp?a=2719&q=325534&deepNav_GID=1654

EPA. (2016, August 16). Retrieved from https://www.epa.gov/national-aquatic-resource-surveys/indicators-conductivity

Howell, P., & Gottschall, K. (2017). Status and Trends: LISS Environmental Indicators. Retrieved May 4, 2019, from http://longislandsoundstudy.net/indicator/warm-watercold-water-fish-index/

Hypoxia, The Problem. (n.d.). Retrieved May 4, 2019, from http://longislandsoundstudy.net/about/our-mission/management-plan/hypoxia/

Long Island Sound By the Numbers. (n.d.). Retrieved May 4, 2019, from http://longislandsoundstudy.net/about-the-sound/by-the-numbers/

Long Island Sound Coastal Observatory Historical Data Access. (n.d.). Retrieved May 4, 2019, from http://lisicos.uconn.edu/data_stn.php

Long Island Sound Monitoring Stations About Basemap Gallery Layer List. (n.d.). Retrieved May 4, 2019, from http://ctdeep.maps.arcgis.com/apps/webappviewer/index.html?id=f3fb1bb862fc4781ab2561f348bc3d0c

Long Island Sound Study. (n.d.). Retrieved May 4, 2019, from http://longislandsoundstudy.net/research-monitoring/liss-ecosystem-targets-and-supporting-indicators/

Long Island Sound Report Card. (2019). Retrieved May 4, 2019, from https://www.ctenvironment.org/report-card

McCaffrey, S. (n.d.). WATER QUALITY PARAMETERS & INDICATORS. Retrieved May 4, 2019, from https://sswm.info/sites/default/files/reference_attachments/MCCAFFREY ny Water Quality Parameters & Indicators.pdf

O'Donnell, J. (n.d.). Water Temperature. Retrieved May 4, 2019, from http://longislandsoundstudy.net/ecosystem-target-indicators/water-temperature/

Prabhakaran, S. (2017). R-statistics.co. Retrieved May 4, 2019, from http://r-statistics.co/Loess-Regression-With-R.html

Temperature Effects on Density. (n.d.). Retrieved from http://butane.chem.uiuc.edu/pshapley/GenChem1/L21/2.html

Water Quality Monitoring. (n.d.). Retrieved May 4, 2019, from http://longislandsoundstudy.net/research-monitoring/water-quality-monitoring/

What You Should Know. (n.d.). Retrieved May 4, 2019, from http://lissclimatechange.net/what-you-should-know/#collapse6

agroimpacts/longisland documentation built on May 18, 2019, 12:27 p.m.