Overview

In situ water quality monitoring instruments take continuous measurements of various chemical and physical parameters. The longer these instruments stay in the field, the further sensor readings drift from their true values. The purpose of this package is to correct water quality monitoring data sets for instrumental drift in a reliable, reproducible method.

Getting Started with R

If you are new to R, welcome! You will need to download R. We also recommend downloading RStudio. Once you have those installed, you can install driftR:

If you want, install the devtools package to install the development version of driftR:

macOS users should also download and install XQuartz.

Packages

In addition to driftR, you will also want to install and load at least two other packages:

  • dplyr - a set of common verbs for cleaning and wrangling data
  • readr - tools for reading and writing plain-text data files

Both of these packages can be installed with the install.packages() function. Since both are part of the tidyverse, you can also install both at once by using install.packages("tidyverse").

Equations

The drift correction equations that are implemented in driftR were originally published as part of Dr. Elizabeth Hasenmueller’s dissertation research (see page 32).

Correction Factor

The correction factor equation (implemented in dr_factor()) is as follows:

\(\textrm{Let:}\)

  • \({f}_{t} = \textrm{correction factor}\)
  • \(t = \textrm{time interval}\)
  • \(\sum{t} = \textrm{total time}\)

\[ {f}_{t} = \left( \frac { t }{ \sum { t } } \right) \]

One-Point Calibration

The one-point calibration equation (implemented in dr_correctOne()) is as follows:

\(\textrm{Let:}\)

  • \(C = \textrm{drift corrected parameter}\)
  • \(m = \textrm{uncorrected in-situ measurement parameter}\)
  • \({f}_{t} = \textrm{correction factor}\)
  • \({s}_{i} = \textrm{calibration standard value}\)
  • \({s}_{f} = \textrm{calibration standard value after drift has occurred}\)

\[ C = m + {f}_{t} \cdot \left( { s }_{ i }-{ s }_{ f } \right) \]

Two-Point Calibration

The two-point calibration equation (implemented in dr_correctTwo()) is as follows:

\(\textrm{Let:}\)

  • \(C = \textrm{drift corrected parameter}\)
  • \(m = \textrm{uncorrected in-situ measurement parameter}\)
  • \({f}_{t} = \textrm{correction factor}\)
  • \({a}_{t} = \textrm{drift correction using lower concentration calibration standard}\)
  • \({b}_{t} = \textrm{drift correction using higher concentration calibration standard}\)
  • \({a}_{i} = \textrm{lower calibration standard value}\)
  • \({b}_{i} = \textrm{higher calibration standard value}\)
  • \({a}_{f} = \textrm{lower calibration standard value after drift has occurred}\)
  • \({b}_{f} = \textrm{higher calibration standard value after drift has occurred}\)

\[ { a }_{ t } = { a }_{ i } + {f}_{t} \cdot \left( { a }_{ i }-{ a }_{ f } \right) \] \[ { b }_{ t } = { b }_{ i } - {f}_{t} \cdot \left( { b }_{ i }-{ b }_{ f } \right) \] \[ C=\left( \frac { m-a_{ t } }{ { b }_{ t }-{ a }_{ t } } \right) \cdot \left( { b }_{ i }-{ a }_{ i } \right) + { a }_{ i } \]

Using driftR

There are six main functions in the driftR which each serve their own purpose in data cleaning process:

  1. dr_read() - import water quality data
  2. dr_factor() - apply correction factors
  3. dr_correctOne() - one point calibration
  4. dr_correctTwo() - two point calibration
  5. dr_drop() - drop observations 1) at the start and finish of a data set, 2) over a specific date range, or 3) expressionally.
  6. dr_replace() - replace observations with NA 1) for specific date ranges or 2) expressionally.

Importing Data

To import a data set, the dr_read() function is utilized. The argument file tells the function where the data is located.The argument instrument tells the function what instrument collected the data for formatting purposes. Accepted instruments are currently YSI Multiparameter V2 Sonde (instrument = "Sonde"), YSI EXO2 (instrument = "EXO"), and Onset U24 Conductivity Logger (instrument = "HOBO"). The argument defineVar is a logical statement where if FALSE, the data will be imported with no modifications, and if TRUE, the “units” observation will be removed and the data will be stored as numeric variables. The argument cleanVar is a logical statement that, if TRUE, will remove all special characters, numbers, and spaces from variable names in order to increase the ease of working with the data in R. For example, for the YSI Sonde 6600, turbidity is exported from the instrument as Turbidity+. This makes it hard to call the variable in R because the + is seen as an operation. The argument case is a string that tells dr_read how you want the clean variable names formatted. The default is “snake”, but there are lots of options to choose from. The options and their respective outputs can be found in the clean_names function of the janitor package.

If you want to use the package and do not have your own data, you can load the sample data included in the package using the following syntax:

waterTibble <- dr_read(file = system.file("extdata", "rawData.csv", package = "driftR"),
                       instrument = "Sonde", defineVar = TRUE, cleanVar= TRUE, case = "snake")

If your data are from another instrument model or brand, please refer to our article on importing data from other sources.

Creating Correction Factors

The next step in the data cleaning process is creating correction factors using dr_factor(). The correction factors that are generated are used to determine how much drift is experienced by each observation in the data set. The argument .data is the working data frame for the correction. corrFactor is the name of the variable that will contain the correction factors. The argument dateVar is the date variable for the data set and timeVar is the time variable. The argument keepDateTime is a logical term that, if TRUE, will keep an intermediate dateTime variable and export it with the correction factors.

waterTibble <- dr_factor(waterTibble, corrFactor = corfac, dateVar= Date,
                         timeVar = Time, keepDateTime = TRUE)

Correcting the Data

After creating the correction factors, you can correcting the data for drift. In order to correct the data, there needs to be some measurement taken that tells the user what the instrument should be reading compared to what the instrument is actually reading. This step can be done with either one or two standard measurements. If one standard measurement is taken, then the function dr_correctOne() is used. The argument .data is the working data frame for the correction. sourceVar is the variable name that you want to correct and cleanVar is the name of the variable that will contain the corrected data. The arguments calVal and calStd are what the instrument was reading and the standard value (i.e. what it should have been reading), respectively. The factorVar argument is the result generated from the dr_factor() function.

waterTibble <- dr_correctOne(waterTibble, sourceVar = SpCond, cleanVar = SpCond_corr,
                           calVal = 1.07, calStd = 1, factorVar = corfac)

If two standard measurements are taken, then the function dr_correctTwo() is utilized. The calValLow and calValHigh arguments are what the instrument was reading for the low and high concentration standard measurements respectively and the calStdLow and calStdHigh arguments are what the instrument should have been reading for the low and high concentration standard measurements respectively.

waterTibble <- dr_correctTwo(waterTibble, sourceVar = pH, cleanVar = pH_corr, calValLow = 7.01,
                           calStdLow = 7, calValHigh = 11.8, calStdHigh = 10, factorVar = corfac)

Dropping Observations

After the data has been corrected, some of the data will likely need to be removed (i.e. dropped) in order to account for the equilibration of the sensors as well as time out of the water during preparation for calibration. It is important to do this step last because dropping data before using dr_correctOne() or dr_correctTwo() will result in the corrections being inconsistent. To do this, the dr_drop() function is used. The argument .data is the working data frame for the correction. The arguments head and tail are the number of observations to be dropped from the beginning and end of the data set respectively. Additionally, sometimes, an instrument may malfunction for a short period, so a date range can be specified to drop data. The arguments for this are dateVar, which is the name of the date variable, timeVar, which is the name of the time variable, from, which is the starting date, and from, which is the ending date. Additionally, there is a tz argument, which stands for time zone. The default for tz is the computer’s time zone, but if data was collected in one time zone and then the data correction is implamented in a different time zone, tz will need to be specified as the time zone where the data was taken. If only the from argument is specified, then to is assumed to be the end of the dataset and if only to is specified, from is assumed to be the beginning of the dataset. Lastly, if there are intermitten times where the instrument is taking inaccurate data such as when a sensor port is leaking and shorting the instrument, an expression can be used to drop all of the bad data. The argument exp is used in this case (e.g., SpCond > 9000).

# drop all data from the begining and end
waterTibble <- dr_drop(waterTibble, head = 6, tail = 6)

# drop all data over the date range
waterTibble <- dr_drop(waterTibble, dateVar = Date, timeVar = Time, 
                       from = "2018-01-03", to = "2018-01-06")
waterTibble <- dr_drop(waterTibble, dateVar = Date, timeVar = Time, to = "2018-01-06")
waterTibble <- dr_drop(waterTibble, dateVar = Date, timeVar = Time, from = "2018-01-03")

#drop all data for observations that match the expression
waterTibble <- dr_drop(waterTibble, exp = SpCond > 9000)
waterTibble <- dr_drop(waterTibble, exp = turbidity < 0)
waterTibble <- dr_drop(waterTibble, exp = pH >= 9)

Replacing incorrect data with NA

Sometimes, individual sensors go bad or give inaccurate data, rather than the entire instrument. The function dr_drop cannot be used in these instances because all of the data will be dropped instead of just the data from the bad sensor. The function dr_replace can be used in these instances and will replace the selected data with NA, which R reads as blank or missing. The argument .data is the working dataframe. The argument sourceVar is the variable you want to remove data from. The argument cleanVar is an optional argument that must only be specified if overwrite = FALSE. cleanVar is the name of a new variable that will be generated with the removed data so that there is a copy of the original data still stored within the dataframe. The argument overwrite is a logical statement that if TRUE will remove the data from the specified variable and store that within the same variable name. However, if FALSE, a new variable will be created with the specified data missing. There are two methods to replace data, by a date range and by an expression. Like dr_drop, the arguments for the date range are dateVar, timeVar, from, to, and tz. The argument for the expression is exp.

Tidying Data

Renaming Variables

Outputs sometimes contains special characters in variable names. For example, “Turbidity+” is a variable name created by the YSI Multiparameter V2 Sonde, which is imported into R as Turbidity.. Neither the original name or R’s attempt at clarifying it follow good variable naming practices. You can rename Turbidity. using the rename() function from dplyr to accomplish this:

Note how the back tick symbols are used to surround non-standard variable names.

Reordering Variables

Variables will need to be re-ordered after using driftR to get similar variables next to each other in your data frame. You can use the select() function from dplyr to accomplish this:

Like all other dplyr functions, select() can be also included in a pipe.

Removing Unnecessary Variables

If there are unnecessary variables left in your data set at the end of the post-processing stage, you can also use the select() function from dplyr to remove them. The function accepts the data frame name followed by a comma and negative sign in front of the variable to be removed. For example, the NitrateN variable does not contain any non-zero observations in our example, so it can be removed.

If there are multiple variables to be removed, a list of the variables can be provided inside -c(varlist):

Exporting Data

Finally, data can be exported to csv (our recommended file format because it is plain-text and non-proprietary) using the readr package’s write_csv() function:

A Full Session

Below is an example of a full data set correction:

# load needed packages
library(driftR)
library(dplyr)
library(readr)

# import data exported from a Sonde
# example file located in the package
waterTibble <- dr_read(file = system.file("extdata", "rawData.csv", package="driftR"),
                            instrument = "Sonde", define = TRUE)

# calculate correction factors
# results stored in new vector corrFac
waterTibble <- dr_factor(waterTibble, corrFactor = corrFac, dateVar = Date,
                         timeVar = Time, keepDateTime = TRUE)

# apply one-point calibration to SpCond;
# results stored in new vector SpConde_Corr
waterTibble <- dr_correctOne(waterTibble, sourceVar = SpCond, cleanVar = SpCond_Corr,
                             calVal = 1.07, calStd = 1, factorVar = corrFac)

# apply one-point calibration to Turbidity.;
# results stored in new vector Turbidity_Corr
waterTibble <- rename(waterTibble, Turbidity = `Turbidity.`)
waterTibble <- dr_correctOne(waterTibble, sourceVar = Turbidity, cleanVar = Turbidity_Corr,
                    calVal = 1.3, calStd = 0, factorVar = corrFac)

# apply one-point calibration to DO;
# results stored in new vector DO_Corr
waterTibble <- dr_correctOne(waterTibble, sourceVar = DO, cleanVar = DO_Corr,
                             calVal = 97.6, calStd = 99, factorVar = corrFac)

# apply two-point calibration to pH;
# results stored in new vector ph_Corr
waterTibble <- dr_correctTwo(waterTibble, sourceVar = pH, cleanVar = pH_Corr,
                             calValLow = 7.01, calStdLow = 7, calValHigh = 11.8,
                             calStdHigh =  10, factorVar = corrFac)

# apply two-point calibration to Chloride;
# results stored in new vector Chloride_Corr
waterTibble <- dr_correctTwo(waterTibble, sourceVar = Chloride, cleanVar = Chloride_Corr,
                             calValLow = 11.6, calStdLow = 10, calValHigh = 1411,
                             calStdHigh =  1000, factorVar = corrFac)

# drop observations to account for instrument equilibration
waterTibble <- dr_drop(waterTibble, head=6, tail=6)

# replace the pH data in the specified date range with NA
waterTibble <- dr_replace(waterTibble, sourceVar = pH, overwite = TRUE, dateVar = Date,
                          timeVar = Time, from = "2018-02-05", to = "2018-02-09")

# reorder variables
waterTibble <- select(waterTibble, Date, Time, dateTime, SpCond, SpCond_Corr, pH, pH_Corr, pHmV,
                      Chloride, Chloride_Corr, AmmoniumN, NitrateN, Turbidity, Turbidity_Corr,
                      DO, DO_Corr, corrFac)

# export cleaned data
write_csv(waterTibble, path = "waterData.csv", na = "NA")

Our vignette on tidy evaluation in driftR includes an example session using magrittr pipe operators (%>%).