In hdugan/NTLlakeloads: Tools for the analysis of long-term NTL-LTER data

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  message = FALSE
)

Why?

A better way of interpolating nutrient data in the lakes. Both for mass calculations and for data visualization.

We often linearly interpolate water quality observations between observation depths. This is typically fine if a lake is well mixed, but if it's stratified it introduces a lot of error around the thermocline. Here, I provide functions to interpolate water quality observations leveraging information on lake stratification from temperature data. Often there is more temperature data than water quality data, which results in better interpolation.

The functions in this package allow you to:

• Load NTL-LTER data from EDI
• Interpolate water quality data to a weekly, 1-m interval, dataframe
• Plot data
• Calculate weekly mass and mean annual mass

Why is it called lakeloads? Mass cacluations can be used to infer loading.

Installation

You can install NTLlakeloads from github using devtools:

install.packages("devtools")
devtools::install_github("hdugan/NTLlakeloads")
library(NTLlakeloads)

library(NTLlakeloads)

Get LTER data

# Load NTL datasets
LTERtemp = loadLTERtemp() # Download NTL LTER data from EDI
LTERnutrients = loadLTERnutrients() # Download NTL LTER data from EDI
LTERions = loadLTERions() # Download NTL LTER data from EDI
LTERsecchi = loadLTERsecchi() # Download NTL LTER data from EDI

Additionally, these datasets can be viewed at: https://github.com/hdugan/NTLviewer

Variables available for plotting

# Available variables
availableVars()

# Available variables that are not depth-discrete. Used with weeklyInterpolate.1D.
availableVars.1D()

Simple heat map for Mendota and Sparkling

library(metR)
# # Create interpolated dataframe. Uses simple linear temperature interpolation 
df.ME.temp = weeklyTempInterpolate(lakeAbr = 'ME', maxdepth = 24, dataset = LTERtemp)

plotTimeseries.year(df.interpolated = df.ME.temp$weeklyInterpolated,   
                    var = 'wtemp', chooseYear = 2018, binsize = 1, 
                    legend.title = 'Temp (°C)') +
  metR::scale_fill_divergent_discretised(high = 'red4', mid = '#e8e3a7', low = 'lightblue4', midpoint = 15) +
  labs(title = 'Lake Mendota, water temperature 2018')
  # theme(legend.position = 'none') # removes legend if you want 

# For Sparkling also add sampled points
df.SP.temp = weeklyTempInterpolate(lakeAbr = 'SP', maxdepth = 19, dataset = LTERtemp)

plotTimeseries.year(df.interpolated = df.SP.temp$weeklyInterpolated,   
                    var = 'wtemp', chooseYear = 2018, binsize = 1, 
                    legend.title = 'Temp (°C)') +
  geom_point(data = df.SP.temp$observations, aes(x = sampledate, y = depth), size = 0.2) + # add sampling points
  # scale_fill_viridis_d(option = 'H') +
  metR::scale_fill_divergent_discretised(high = 'red4', mid = '#e8e3a7', low = 'lightblue4', midpoint = 15) +
  labs(title = 'Sparkling Lake, water temperature 2018')

Interpolate weekly total phosphorus data for Lake Mendota

If LTERtemp = loadLTERtemp() has not been run, this function downloads the NTL temperature dataset. This function works by joining the water quality dataset (e.g. nutrients) with the temperature dataset. It uses the temperature profile to fit a gam, which is used to vertically interpolate data to a 1-m depth increment. The idea here is that this method will pick up large changes around the thermocline better than linear interpolation.

# printFigs = TRUE to output series of interpolated profiles (but slower)
# See help file for argument descriptions
df.ME = weeklyInterpolate(lakeAbr = 'ME', var = 'totpuf_sloh', dataset = LTERnutrients, maxdepth = 24, 
                          constrainMethod = 'zero', setThreshold = 0.1, printFigs = F)

Plotting entire timeseries

# See help file for available arguments
plotTimeseries(df.interpolated = df.ME$weeklyInterpolated, var = 'totpuf_sloh')

Plot specific year with observed data

# With observations
plotTimeseries.year(df.interpolated = df.ME$weeklyInterpolated, observations = df.ME$observations,  
                    var = 'totpuf_sloh', chooseYear = 2008)

# Without observations, but adding legend title and decreasing binsize 
plotTimeseries.year(df.interpolated = df.ME$weeklyInterpolated, var = 'totpuf_sloh', 
                    binsize = 0.06, chooseYear = 2016, legend.title = 'TP (mg/L)')

Customize plots

Since these functions return a ggplot, they can be additionally customized with ggplot options. For example:

library(MetBrewer)
library(ggplot2)

plotTimeseries.year(df.interpolated = df.ME$weeklyInterpolated,
                    var = 'totpuf_sloh', chooseYear = 2015, binsize = 0.1, 
                    legend.title = 'TP (mg/L)') +
  scale_fill_manual(values=met.brewer("Hokusai2", 7)) + # override color default
  geom_point(data = df.ME$observations, aes(x = sampledate, y = depth), size = 1, 
             fill = 'gold', shape = 22, stroke = 0.2) + # sampling observations
  labs(title = 'Lake Mendota, total phosphorus 2015', x = 'Date') +
  theme_minimal(base_size = 10)

Calculate mass at annual or weekly timescales

# Conversion from g to kg
df.mass.annual = calcMass(df.ME$weeklyInterpolated,lakeAbr = 'ME', 
                          time.res = 'annual', conversion = 1e3)

Example of plotting annual mass

library(ggplot2)

ggplot(df.mass.annual, aes(x = year, y = mass)) +
  geom_path() +
  geom_point() +
  ylab('TP (kg)') +
  labs(title = 'Lake Mendota mean annual TP mass', caption = 'Calculated from NTLlakeloads') +  
  theme_bw(base_size = 10) +
  theme(axis.title.x = element_blank())

Decompose weekly mass timeseries to analyse trends and seasonality

df.mass = calcMass(df.ME$weeklyInterpolated,lakeAbr = 'ME', 
                   time.res = 'weekly', conversion = 1e3)

decomposeTS(df.mass, lakeAbr = 'ME', var = 'totpuf_sloh')

hdugan/NTLlakeloads documentation built on June 16, 2024, 6:59 p.m.