In ISU-STRIPS/STRIPSyield: STRIPS yield data

\hypersetup{ colorlinks = true, % Colours links instead of ugly boxes urlcolor = blue, % Colour for external hyperlinks linkcolor = blue, % Colour of internal links citecolor = red % Colour of citations }

knitr::opts_chunk$set(
  collapse = TRUE,
  echo     = FALSE,
  cache    = FALSE,
  comment  = "#>"
)

suppressPackageStartupMessages(library(ggplot2))
suppressPackageStartupMessages(library(shapefiles))
suppressPackageStartupMessages(library(STRIPSyield))
suppressPackageStartupMessages(library(reshape2))

describe_shapefile <- function(shape) {
  describe_column <- function(shape, columnName) {
    column <- shape$dbf$dbf[, columnName]
    data.frame(
      name     = columnName,
      class    = class(column)[1],
      nFactors = if (is.factor(column)) { length(unique(column)) } else { NA },
      min      = if (is.factor(column)) { NA } else { min(column)  },
      mean     = if (is.factor(column)) { NA } else { mean(column) },
      max      = if (is.factor(column)) { NA } else { max(column)  },
      example  = trimws(as.character(column[1]))
    )
  }

  do.call(
    rbind,
    lapply(colnames(shape$dbf$dbf), describe_column, shape = shape)
  )
}

describe_yield <- function(yield) {
  describe_column <- function(yield, columnName) {
    column <- yield[, columnName]
    data.frame(
      name     = columnName,
      class    = class(column)[1],
      nFactors = if (is.factor(column)) { length(unique(column)) } else { NA },
      min      = if (is.factor(column)) { NA } else { min(column)  },
      mean     = if (is.factor(column)) { NA } else { mean(column) },
      max      = if (is.factor(column)) { NA } else { max(column)  },
      example  = trimws(as.character(column[1]))
    )
  }

  do.call(
    rbind,
    lapply(colnames(yield), describe_column, yield = yield)
  )
}

Accessing data

How to access the datasets

The followings are the datasets contained in this R Package:

• Point-level yield monitor data:
  • R data frames: after the package is loaded, two objects yield and yieldExtra containing point-coordinate level yield data frames become available. These are described in Table \ref{tab:yieldExtra-description} and displayed in Figures \ref{fig:data-visualization-soybeans} and \ref{fig:data-visualization-corn}.
  • Shapefiles: datasets in shapefile format are available as discussed in Section \ref{sec:curation-protocol}.
• PNAS paper: to access the dataset from the PNAS paper [1], load the package and run pnas_data().

dYield <- describe_yield(yieldExtra)
units  <- c(
  NA, NA, NA, "Hectare", NA, "Hectare", NA, "% (hundreds)", NA, 
  "% (hundreds)", NA, NA, "Feet", "Integer", NA, 
  NA, NA, "LatLong", "LatLong", "UTM", "UTM", NA, "Feet", "MPH", "Degrees", 
  "Feet", "Seconds", "Pounds per second", "% (hundreds)", 
  "Pounds (dry)", "Pounds (std)", "Pounds (wet)",
  "Kilograms (dry)", "Kilograms (std)", "Kilograms (wet)",
  "Bushels per acre (dry)", "Bushels per acre (std)", "Bushels per acre (wet)",
  "Megagrams per hectare (dry)", "Megagrams per hectare (std)", 
  "Megagrams per hectare (wet)"
)
extra <- ifelse(colnames(yieldExtra) %in% colnames(yield), "Yes", "No")
colNames <- c(
  "Name", "Class", "# of levels", "Min", "Mean", "Max", "Example"
)

tab <- cbind(dYield, units = units, "extra only" = extra)
colnames(tab) <- c(colNames, "Measurement unit", "yieldExtra only")

k <- knitr::kable(
  tab,
  format = "latex", booktabs = TRUE,
  label = "yieldExtra-description",
  caption = "Structure and content of the yield and yieldExtra datasets.",
  digits = 2, format.args = list(scientific = FALSE)
)

kableExtra::kable_styling(k, latex_options = c("scale_down"))

Terminology

The dataset adheres to the terminology used in [2] to describe the experimental design.

• Site: three locations within the Neal Smith National Wildlife Refuge (NSNWR) in central Iowa (namely Basswood, Interim, and Orbweaver).
• Blocks: there are four blocks (namely BasswoodA, BasswoodB, Interim, and Orbweaver).
• Watershed: twelve experimental units (Basswood 1 to 6, Interim 1 to 3, Orbweaver 1 to 3).
• Field: the portion of the whatershed planted with either row crops or perennial vegetation.
• Treatment: four watershed-scale treatments having different proportions and topographic positions of PFS (no PFS, 10% PFS at toeslope position, 10% PFS distributed on toe and contour strips, and 20% PFS distributed on toe and contour strips).
• Coordinate point: each of the spatial coordinate units with recorded information (number and position of the points vary per year and treatment).

Curation protocol

\label{sec:curation-protocol}

Folder structure

In STRIPSyield v0.2.0, the datasets are stored in the folders: data-raw\source\YYYY-site.ext.

• legacy: CSV existing in STRIPSYield v0.1.1 that were produced by an unspecified methodology.
• original: shapefiles as received from the Coop. Besides modifying the file name, file structure and content are kept intact.
• curated: new shapefiles originating from the curation protocol. Note that this process modifies both the datasets structure and content.

Curation protocol

Because not all the datasets have the same structure and measurement units, we create a curation protocol. We identify two patterns in the data sources, namely Template I (2007-2010 and 2012) and Template II (2013-2019 and 2011). We read the shapefiles from the original folder, apply the modifications mentioned below, and store the new shapefiles in the curated folder. These editing rules may be broadly classified into five actions:

• Rename: we modify the name of the variable but not the content.
• Reformat: we modify the name of the variable and the content (e.g. change of measurement unit).
• Drop: we discard some content if it is not present in every shapefiles across the years and sites.

Although keeping both the original and the curated shapefiles result in significant storage redundancy, this procedure guarantees that no original data is lost in the process.

Naming convension

File naming convention:

• data-raw/yield_original/YYYY-site.ext
• data-raw/yield_curated/YYYY-site.ext
• Note that we use hyphen to separate words, and site names are lowercase.

Column naming convention:

• Use camelCase (e.g. prairiePosition). Note that the starting letter is lowercase.
• No measurement units in the column names. For measurement units, see this vignette.

Data structure convention:

• All strings as factors.
• All strings start with uppercase. (ex. Soybeans, Orbweaver).
• Dates and timestamps are Date and POSIXct objects respectively.
• Use NA for missing data.

Original shapefiles (2007-2010, 2012)

The PROJ4 string defining the CRS of the coordinates recorded in these shapesfiles is "+proj=longlat +datum=WGS84 +no_defs +ellps=WGS84 +towgs84=0,0,0". Since the 2007-2010 and 2012 shapefiles are consistent, 2009-basswood original shapefile structure is described in Table \ref{tab:2009-basswood-description}. To homogenenize measurement units, we rescale the columns distance and swath width from inches to foot.

shape2009  <- read.shapefile("../data-raw/yield_original/2009-basswood")
dShape2009 <- describe_shapefile(shape2009)

tab <- dShape2009
colnames(tab) <- colNames
k <- knitr::kable(
  tab, 
  format = "latex", booktabs = TRUE,
  digits = 2, format.args = list(scientific = FALSE),
  caption = "Structure and content of the Basswood 2015 original shapefile.",
  label = "2009-basswood-description"
)

kableExtra::kable_styling(k, latex_options = c("scale_down"))

Original shapefiles (2013-2019, 2011)

The PROJ4 string defining the CRS of the coordinates recorded in these shapesfiles is "+proj=longlat +datum=WGS84 +no_defs +ellps=WGS84 +towgs84=0,0,0". Since the 2013-2019 and 2011 shapefiles are consistent, the 2015-basswood original shapefile structure is described in Table \ref{tab:2015-basswood-description}.

shape2015  <- read.shapefile("../data-raw/yield_original/2015-basswood")
dShape2015 <- describe_shapefile(shape2015)

tab <- dShape2015
colnames(tab) <- colNames
k <- knitr::kable(
  tab, 
  format = "latex", booktabs = TRUE,
  digits = 2, format.args = list(scientific = FALSE),
  label = "2015-basswood-description",
  caption = "Structure and content of the Basswood 2015 original shapefile."
)

kableExtra::kable_styling(k, latex_options = c("scale_down"))

Curated shapefile (all years)

The PROJ4 string defining the CRS of the coordinates recorded in these shapesfiles is "+proj=longlat +datum=WGS84 +no_defs +ellps=WGS84 +towgs84=0,0,0" (no projections were needed). As the shapefile structure and content does not vary across the years and sites, the structure described in Table \ref{tab:curated-shapefile-description} is valid for every curated shapefile.

shape2015  <- read.shapefile("../data-raw/yield_curated/2015-basswood")
dShape2015 <- describe_shapefile(shape2015)
units      <- c(
  NA, NA, "Feet", "Integer", "Integer", NA, "WGS84", "WGS84", "Feet above sea", 
  "Miles per hour", "Arc degrees", "Feet", "Seconds", NA, "Pounds per second", 
  "% (hundreds)", "Dry bushels per acre"
)

tab <- cbind(dShape2015[, 1:2], units)
colnames(tab) <- c("Name", "Class", "Measurement unit")
k <- knitr::kable(
  tab,
  format = "latex", booktabs = TRUE,
  digits = 2, format.args = list(scientific = FALSE),
    caption = "Structure a curated shapefile.",
  label = "curated-shapefile-description"
)

kableExtra::kable_styling(k)

To build our consolidated shapefiles, we decided to keep only those variables recorded for every site and year (i.e. columns that were present in every source file). The only exceptions is direction, available for years 2013-2015 and 2011 only, which we kept as partial information may be relevant for our future research.

References

[1] Lisa A. Schulte, Jarad B. Niemi, Matthew J. Helmers, Matt Liebman, J. G. Arbuckle, David E. James, Randall K. Kolka, Matthew E. O’Neal, Mark D. Tomer, John C. Tyndall, Heidi Asbjornsen, Pauline Drobney, Jeri Neal, Gary Van Ryswyk, and Chris Witte (2017). “Prairie strips improve biodiversity and the delivery of multiple ecosystem services from corn-soybean croplands” Proceedings of the National Academy of Sciences, 114(42), 11247-11252. (url)

[2] Xiaobo Zhou, Matthew J. Helmers, Heidi J. Asbjornsen, Randy Kolka, and Mark D. Tomer (2010). "Perennial filter strips reduce nitrate levels in soil and shallow groundwater after grassland-to-cropland conversion" Journal of environmental quality, 39(6), 2006-2015.

\clearpage

Data visualization

\clearpage

plot_grid <- function(df, title = NULL) {
  p <- ggplot(df) +
  geom_violin(aes(x = year, y = value), draw_quantiles = c(0.5)) +
  facet_grid(
    rows = vars(variable),
    cols = vars(site),
    scales = "free"
  ) +
  labs(
    title = title,
    x     = "Year",
    y     = "Values"
  ) +
  theme_bw() +
  theme(
    panel.grid.major.y = element_blank(),
    axis.text.x        = element_text(angle = 90, hjust = 1)
  )

  suppressWarnings(print(p))
}

keys <- c("site", "year", "crop", "record")
cols <- c(
  "swath", "elevation", "speed", "distance",
  "cycle", "flow", "moisture", "yieldDryBuAc"
)

yieldExtraLong <- melt(
  data = yieldExtra,
  id.vars = keys,
  measure.vars = cols
)

``r Visualization of some selectedyieldExtra` data frame variables.", warning = FALSE} plot_grid( df = yieldExtraLong[ yieldExtraLong$site != "OffBounds" & yieldExtraLong$crop == "Soybeans", ], title = "Soybeans" )

```r Visualization of some selected `yieldExtra` data frame variables.", warning = FALSE}
plot_grid(
  df = yieldExtraLong[
    yieldExtraLong$site != "OffBounds" & 
    yieldExtraLong$crop == "Corn", 
  ],
  title = "Maize"
)

\clearpage

Tables

\clearpage

ISU-STRIPS/STRIPSyield documentation built on Jan. 31, 2021, 10:16 a.m.