Maize_Spread_Processes.Rmd

---
title: "Diffusion of Maize Agriculture to the Southwestern United States"
author: "Andrew Gillreath-Brown"
date: "`r format(Sys.time(), '%d %B %Y')`"
html_document:
  fig_width: 12
  toc: yes
toc_depth: 2
toc_float: yes
number_sections: no
code_folding: hide
output:
  html_document:
  df_print: paged
mainfont: Calibri
editor_options:
  chunk_output_type: console
keep_md: yes
---

```{r setup, include = FALSE, results = 'hide', echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide'}

# Run with R version 4.2.1
# See GitHub repo for more information: https://github.com/Archaeo-Programmer/cropSpreadR

# devtools::install()
# devtools::load_all()
library(cropSpreadR)
library(cropDiffusionR)
# Loading additional libraries as having an issue with knitting this rmd. 
library(raster)
library(tidyverse)

```

```{r load_data, include = FALSE, results = 'hide', echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide'} 

# Load temperature anomaly reconstruction from paleoMAT package.
temp_anom <-
  readr::read_csv(
    "https://raw.githubusercontent.com/Archaeo-Programmer/paleoMAT/master/vignettes/tables/supp_table_s2.csv",
    show_col_types = FALSE
  ) %>%
  dplyr::rename(
    year = 1,
    anom = 2,
    ci = 3,
    se = 4
  )

# Load location data for each sample. Site locations are not saved in this repository, as we want to protect archaeological site locations.
sf::sf_use_s2(FALSE)

MDB <- cropDiffusionR::maizeDB %>%
  dplyr::left_join(.,
                   readr::read_csv(
                     stringr::str_remove(
                       here::here("maize_database/Maize_Database_Locations_DONT_DISTRIBUTE.csv"),
                       "cropSpreadR/"
                     ),
                     col_types = readr::cols()
                   ),
                   by = "LabID") %>%
  dplyr::mutate(Long1 = Long, Lat1 = Lat) %>%
  # Convert to spatial object
  sf::st_as_sf(coords = c("Long1", "Lat1"),
               crs = 4326) %>%
  sf::st_make_valid() %>%
  # Join to counties data to get county and state information.
  sf::st_join(cropSpreadR::SWUS_counties) %>%
  dplyr::mutate(county_st = ifelse(!is.na(REGION) &
                                     !is.na(NAME), paste(NAME, REGION, sep = ", "), NA)) %>%
  sf::st_drop_geometry() %>% 
  suppressMessages()

# If you do not have your own location information, then you can run the code below using mostly coarse scale location data provided by the p3k14c package (which primarily will have county centroids). If needed, you can uncomment this code (lines 77-102) and run instead of lines 52-74 above. However, some of the samples may not be available in the p3k14c package, and locations could vary up to at least 10 km from the actual site. p3k14c_data has a column that details the location accuracy. Please refer to Table 2 in https://doi.org/10.1038/s41597-022-01118-7 for more information on the p3k14c location accuracy codes.
# sf::sf_use_s2(FALSE)
# 
# MDB <- cropDiffusionR::maizeDB %>%
#   dplyr::left_join(.,
#                    p3k14c::p3k14c_data %>% dplyr::select(LabID, Long, Lat),
#                    by = "LabID") %>%
#   # Remove any rows that do not have a location.
#   dplyr::filter(!is.na(Lat) & !is.na(Long)) %>%
#   # Currently, there is an error with coordinates for AA-3308, so correcting them here.
#   dplyr::mutate(
#     Long = dplyr::case_when(SiteName == "Coxcatlan" ~ -97,
#                             TRUE ~ Long),
#     Lat = dplyr::case_when(SiteName == "Coxcatlan" ~ 18,
#                            TRUE ~ Lat)
#   ) %>%
#   dplyr::mutate(Long1 = Long, Lat1 = Lat) %>%
#   # Convert to spatial object
#   sf::st_as_sf(coords = c("Long1", "Lat1"),
#                crs = 4326) %>%
#   sf::st_make_valid() %>%
#   # Join to counties data to get county and state information.
#   sf::st_join(cropSpreadR::SWUS_counties) %>%
#   dplyr::mutate(county_st = ifelse(!is.na(REGION) &
#                                      !is.na(NAME), paste(NAME, REGION, sep = ", "), NA)) %>%
#   sf::st_drop_geometry() %>%
#   suppressMessages()


# Get modern annual accumulated growing degree days (GDD) (°C) at each sample location. 
# For Mexico, we use WorldClim 30 sec. (~1 km) for GDD data.
MDB_gdd_wc <- MDB %>%
  sf::st_as_sf(coords = c("Long", "Lat"), crs = 4326) %>%
  dplyr::select(LabID) %>%
  cropDiffusionR::extract_GDD(gdd_raster = cropDiffusionR::WorldClim_annual_gdd,
                              sites = .)

# For USA, we use PRISM (~800 meters) for modern annual accumulated GDD (°C) data.
MDB_gdd_prism <- MDB %>%
  sf::st_as_sf(coords = c("Long", "Lat"), crs = 4326) %>%
  dplyr::select(LabID) %>%
  cropDiffusionR::extract_GDD(gdd_raster = cropSpreadR::PRISM_annual_gdd,  sites = .)

# Combine GDD together (i.e., coalesce), where PRISM takes precedence over WorldClim data.
MDB_gdd <-
  dplyr::left_join(MDB_gdd_prism, MDB_gdd_wc, by = "LabID") %>%
  dplyr::mutate(GDD = dplyr::coalesce(GDD.x, GDD.y)) %>%
  dplyr::select(-c(GDD.x, GDD.y))

# Add GDD and elevation to maize dataset.
MDB_env <- MDB %>%
  dplyr::select(-Elevation_masl) %>%
  dplyr::left_join(., MDB_gdd, by = "LabID") %>%
  # Add elevation to dataset.
  dplyr::left_join(., cropSpreadR::MDB_elev, by = "LabID")

# Get sites that are on the northern maize frontier through time from cropDiffusionR package for the SWUS.
temp <- tempfile()
download.file(
  "https://raw.githubusercontent.com/Archaeo-Programmer/cropDiffusionR/master/vignettes/output/maize_frontier_swus.rds",
  temp
)
maize_frontier_swus <-
  readr::read_rds(temp, "maize_frontier_swus.rds")
unlink(temp)

maize_frontier_swus <- MDB %>%
  dplyr::filter(LabID %in% maize_frontier_swus$LabID) %>%
  dplyr::mutate(Long1 = Long, Lat1 = Lat) %>%
  # Convert to spatial object
  sf::st_as_sf(coords = c("Long1", "Lat1"),
               crs = 4326) %>%
  sf::st_make_valid()

# Get sites that are on the northern maize frontier through time from cropDiffusionR package for Mexico and SWUS.
temp <- tempfile()
download.file(
  "https://raw.githubusercontent.com/Archaeo-Programmer/cropDiffusionR/master/vignettes/output/maize_frontier_mexico.rds",
  temp
)
maize_frontier_mexico <-
  readr::read_rds(temp, "maize_frontier_mexico.rds")
unlink(temp)

maize_frontier_mexico <- MDB_env %>%
  dplyr::filter(LabID %in% maize_frontier_mexico$LabID) %>%
  dplyr::mutate(Long1 = Long, Lat1 = Lat) %>%
  # Convert to spatial object
  sf::st_as_sf(coords = c("Long1", "Lat1"),
               crs = 4326) %>%
  sf::st_make_valid() %>% 
  dplyr::mutate(centroid_long = ifelse(is.na(centroid_long), Long, centroid_long),
                centroid_lat = ifelse(is.na(centroid_lat), Lat, centroid_lat))

```

```{r Figure_1, fig.cap="Figure 1. Frequency distribution of directly dated maize macrofossils in our maize database for sites that are north of 30° latitude and older than AD 1500 (n = 985), distinguishing between the earliest maize sample at each site (n = 275), and all other maize samples fitting the criteria above (n = 710). Two of the earliest maize sites, Las Capas and Old Corn, are individually labeled.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

# Get earliest maize samples from each site that are north of 30° latitude and less than AD 1500.
maize_earliest_distribution <- MDB_env %>%
  dplyr::filter(Lat > 30 & Date < 1500) %>%
  dplyr::group_by(SiteIDName) %>%
  dplyr::slice_min(Date, with_ties = FALSE) %>%
  dplyr::mutate(sample = "Earliest Maize Subset")

# Get all maize samples from each site that are north of 30° latitude and less than AD 1500.
full_distribution <- MDB_env %>%
  dplyr::filter(Lat > 30 & Date < 1500) %>%
  dplyr::mutate(sample = "Full Maize Database") %>%
  dplyr::filter(!LabID %in% maize_earliest_distribution$LabID)

# Combine data together, so that we can plot both sets as histograms on the same plot.
swus_date_distribution_stacked <- maize_earliest_distribution %>%
  dplyr::bind_rows(., full_distribution) %>%
  dplyr::ungroup() %>%
  dplyr::mutate(sample = factor(sample, levels = c(
    "Full Maize Database", "Earliest Maize Subset"
  ))) %>%
  ggplot(aes(x = Date, fill = sample)) +
  geom_histogram(
    binwidth = 100,
    color = "#000000",
    position = "stack",
    center = -3650
  ) +
  scale_fill_manual(values = c("cornflowerblue", "#DDCC77")) +
  xlab("Year BC/AD") +
  ylab("Count") +
  scale_x_continuous(breaks = seq(-4000, 1500, 500),
                     limits = c(-4000, 1500)) +
  scale_y_continuous(breaks = seq(0, 90, 10), limits = c(0, 90)) +
  theme_classic() +
  annotate(
    geom = "text",
    x = -3550,
    y = 15,
    label = "Las Capas",
    hjust = 0,
    angle = 45,
    size = 6.75
  ) +
  annotate(
    "segment",
    x = -3650,
    xend = -3550,
    y = 1,
    yend = 13,
    colour = "black"
  ) +
  annotate(
    "segment",
    x = -3450,
    xend = -3550,
    y = 1,
    yend = 13,
    colour = "black"
  ) +
  # Annotation for Old Corn.
  annotate(
    geom = "text",
    x = -2150,
    y = 15,
    label = "Old Corn",
    hjust = 0,
    angle = 45,
    size = 6.75
  ) +
  annotate(
    "segment",
    x = -2250,
    xend = -2150,
    y = 1,
    yend = 13,
    colour = "black"
  ) +
  theme(
    legend.title = element_blank(),
    legend.text = element_text(size = 14,
                               family = "Helvetica"),
    panel.spacing = unit(1, "lines"),
    strip.text.x = element_text(
      size = 16,
      family = "Helvetica",
      face = "bold"
    ),
    axis.ticks.length = unit(0.125, "inch"),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_text(
      size = 20,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 10,
        r = 5,
        b = 10,
        l = 10
      )
    ),
    axis.title.x = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    ),
    panel.grid.minor = element_blank(),
    axis.line = element_line(colour = "black"),
    plot.margin = margin(
      t = 5,
      r = 25,
      b = 5,
      l = 5
    )
  )

swus_date_distribution_stacked

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_1.png"),
#   plot = swus_date_distribution_stacked,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )

```

```{r Figure_2, fig.cap="Figure 2. Spread rates for maize beginning from Guilá Naquitz to sites in the SWUS on the MNF as defined in the text. A) Only change in northing contributes to the rates measured. B) As in A, except rates are expressed as Euclidean distances.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

# Get the maize northern frontier starting in Mexico, then calculate euclidean distance between sites and latitude North distance.
nodes <-
  maize_frontier_mexico %>%
  dplyr::mutate(
    SiteName = ifelse(SiteName == "Elsinore Burial", SiteID, SiteName),
    SiteName = ifelse(SiteName == "Guila Naquitz", "Guilá Naquitz", SiteName)
  ) %>%
  dplyr::arrange(Date) %>%
  dplyr::mutate(id = dplyr::row_number()) %>%
  # Calculate the distance between rows and store distance on the arrival location row.
  dplyr::mutate(
    # First, get the next row. Lag or row_number() - 1 does not work for geometry; thus, we had to use this to get the lead row.
    lead_row = geometry[dplyr::row_number() + 1],
    # Second, calculate the distance between the rows using euclidean distance. Then, we use lag so that it returns it for the later sample (arrival location).
    # We use lag since we cannot use it on the geometry and had to use lead instead.
    dist_deg = as.numeric(dplyr::lag(
      sf::st_distance(geometry, lead_row, by_element = TRUE, which = "Euclidean")
    )),
    # Convert degrees to km. We use the approximate 110.574 for the calculation (http://www.thegreenwichmeridian.org/tgm/articles.php?article=5).
    dist_km = dist_deg * 110.574,
    # Determine the change in degrees distance from the euclidean distance calculation above.
    delta_deg = dist_deg / (Date - dplyr::lag(Date)),
    # Determine the change in km distance from the euclidean distance calculation above.
    delta_km = dist_km / (Date - dplyr::lag(Date)),
    # Get the change in degrees latitude between sites, and limit the number of decimal points.
    delta_lat_deg = ((Lat - dplyr::lag(Lat))) / (Date - dplyr::lag(Date)),
    delta_lat_deg = as.numeric(sprintf("%.4f", delta_lat_deg)),
    # Get the change in km latitude between sites.
    delta_lat_km = delta_lat_deg * 110.574
  ) %>%
  dplyr::select(-lead_row) %>%
  sf::st_drop_geometry() %>%
  suppressWarnings()

# Keep only select columns that we will use for plotting.
nodes1 <- nodes %>%
  dplyr::select(
    c(
      "id",
      "Long",
      "Lat",
      "SiteName",
      "Date",
      "dist_deg",
      "dist_km",
      "delta_deg",
      "delta_km",
      "delta_lat_deg",
      "delta_lat_km"
    )
  )

# Change names of columns.
names(nodes1) <-
  c(
    "id",
    "lon",
    "lat",
    "name",
    "date",
    "dist_deg",
    "dist_km",
    "delta_deg",
    "delta_km",
    "delta_lat_deg",
    "delta_lat_km"
  )

# Get shapefiles for world boundaries.
country_shapes1 <-
  ggplot2::geom_polygon(
    aes(x = long, y = lat, group = group),
    data = map_data('world'),
    fill = "white",
    color = "black",
    size = 0.15
  )

# Get shapefiles for states.
country_shapes2 <-
  ggplot2::geom_polygon(
    aes(x = long, y = lat, group = group),
    data = map_data('state'),
    fill = "white",
    color = "black",
    size = 0.15
  )

# Set lat/long limits for plotting.
mapcoords1 <-
  ggplot2::coord_fixed(xlim = c(-114.814521,-92),
                       ylim = c(14.72716, 42.001167))

# Create an edge list.
edges2a <- data.frame(from = 1:13,
                     to = 2:14,
                     color = nodes1$delta_lat_km[!is.na(nodes1$delta_lat_km)])

# Join the edges together with the node data.
edges_for_plot2a <- edges2a %>%
  dplyr::inner_join(nodes1 %>% dplyr::select(id, lon, lat), by = c('from' = 'id')) %>%
  dplyr::rename(x = lon, y = lat) %>%
  dplyr::inner_join(nodes1 %>% dplyr::select(id, lon, lat), by = c('to' = 'id')) %>%
  dplyr::rename(xend = lon, yend = lat)

# Check that the edge list and the joins with node data still produce the same number of rows.
assertthat::assert_that(nrow(edges_for_plot2a) == nrow(edges2a))

# Figure 2a. Change in N latitude (°) per year. 
maize_frontier_mexico_delta_Lat <- ggplot(nodes1) +
  country_shapes1 +
  country_shapes2 +
  geom_curve(
    aes(
      x = x,
      y = y,
      xend = xend,
      yend = yend,
      color = color
    ),
    arrow = arrow(length = unit(0.02, "npc"), type = "closed"),
    data = edges_for_plot2a,
    curvature = 0.3,
    alpha = 1.0,
    size = 1.0
  ) +
  scale_color_distiller(palette = "RdYlGn",
                        direction = 1,
                        name = "\u0394 km N/year") +
  geom_point(
    aes(x = lon, y = lat),
    # draw nodes
    shape = 21,
    fill = 'white',
    color = 'black',
    stroke = 0.5,
    size = 3.5
  ) +
  ggrepel::geom_label_repel(
    aes(x = lon, y = lat, label = name),
    # draw text labels
    size = 2,
    color = "black",
    min.segment.length = 0.15,
    fill = alpha(c("white"), 0.7),
    force = 34,
    family = "Helvetica",
    max.overlaps = 15
  ) +
  mapcoords1 +
  theme_void() +
  theme(
    plot.margin = grid::unit(c(2, 2, 2, 2), "mm"),
    legend.justification = c(0,1),
    legend.box.spacing = unit(20.0, "pt"),
    legend.key.size = unit(0.55, "cm"),
    legend.text = element_text(size = 14, family = "Helvetica"),
    legend.title = element_text(
      size = 18,
      face = "bold",
      family = "Helvetica"
    )
  )

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_2a.png"),
#   plot = maize_frontier_mexico_delta_Lat,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )


# Figure 2b. Change in euclidean distance (km) per year.
edges2b <- data.frame(from = 1:13,
                     to = 2:14,
                     color = nodes1$delta_km[!is.na(nodes1$delta_km)])

edges_for_plot2b <- edges2b %>%
  dplyr::inner_join(nodes1 %>% dplyr::select(id, lon, lat), by = c('from' = 'id')) %>%
  dplyr::rename(x = lon, y = lat) %>%
  dplyr::inner_join(nodes1 %>% dplyr::select(id, lon, lat), by = c('to' = 'id')) %>%
  dplyr::rename(xend = lon, yend = lat)

assertthat::assert_that(nrow(edges_for_plot2b) == nrow(edges2b))

maize_frontier_mexico_euclidean_dist <- ggplot(nodes1) +
  country_shapes1 +
  country_shapes2 +
  geom_curve(
    aes(
      x = x,
      y = y,
      xend = xend,
      yend = yend,
      color = color
    ),
    arrow = arrow(length = unit(0.02, "npc"), type = "closed"),
    data = edges_for_plot2b,
    curvature = 0.3,
    alpha = 1.0,
    size = 1.0
  ) +
  scale_color_distiller(palette = "RdYlGn",
                        direction = 1,
                        name = "\u0394 km ED/year") +
  geom_point(
    aes(x = lon, y = lat),
    # draw nodes
    shape = 21,
    fill = 'white',
    color = 'black',
    stroke = 0.5,
    size = 3.5
  ) +
  ggrepel::geom_label_repel(
    aes(x = lon, y = lat, label = name),
    # draw text labels
    size = 2,
    color = "black",
    min.segment.length = 0.15,
    fill = alpha(c("white"), 0.7),
    force = 34,
    family = "Helvetica",
    max.overlaps = 15
  ) +
  mapcoords1 +
  theme_void() +
  theme(
    plot.margin = grid::unit(c(2, 2, 2, 2), "mm"),
    legend.justification = c(0,1),
    legend.box.spacing = unit(20.0, "pt"),
    legend.key.size = unit(0.55, "cm"),
    legend.text = element_text(size = 14, family = "Helvetica"),
    legend.title = element_text(
      size = 18,
      face = "bold",
      family = "Helvetica"
    )
  )

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_2b.png"),
#   plot = maize_frontier_mexico_euclidean_dist,
#   # width = 15.51,
#   # height = 10.32,
#   dpi = 600
# )

Figure_2 <-
  cowplot::plot_grid(
    maize_frontier_mexico_delta_Lat,
    maize_frontier_mexico_euclidean_dist,
    labels = "AUTO",
    label_x = 0.25,
    label_y = 0.99,
    align = "v",
    nrow = 2,
    ncol = 1
  )

Figure_2

# ggplot2::ggsave(
#   #Figure_2,
#   filename = here::here("vignettes/figures/Figure_2.png"),
#   width = 13.7,
#   height = 10.2,
#   dpi = 900
# )

```

```{r Table1, echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'asis'}

# Select some specific columns, which we will use to create Table 1 below (in the next code chunk).
maize_spread_table <- nodes %>%
  dplyr::arrange(Date) %>%
  dplyr::select(
    LabID,
    SiteName,
    NAME,
    REGION,
    Country,
    Date,
    Age,
    Error,
    MedianBP,
    GDD,
    elevation,
    dist_deg:delta_lat_km,
    dplyr::starts_with("TwoSigma")
  ) %>%
  dplyr::mutate(
    across(c(NAME, REGION), tools::toTitleCase),
    REGION = ifelse(SiteName == "Guilá Naquitz", "Oaxaca", REGION)
  )

# Pull out just LabID and the two sigma columns, so that we can extract the two-sigma range for each LabID, which we will include in Table 1.
sigma_range <-
  maize_spread_table %>%
  dplyr::select(LabID, dplyr::starts_with("TwoSigma")) %>%
  tidyr::pivot_longer(-LabID) %>%
  dplyr::filter(!is.na(value)) %>%
  dplyr::group_by(LabID) %>%
  dplyr::filter(dplyr::row_number() %in% c(1, dplyr::n())) %>%
  tidyr::separate(value, into = c("early", "late"), sep = " to ") %>%
  tidyr::pivot_longer(c(early, late), names_to = "date") %>%
  dplyr::mutate(range = paste0(max(value), " to ", min(value))) %>%
  dplyr::select(-c(name, date, value)) %>%
  dplyr::ungroup() %>%
  dplyr::distinct()

# Select particular columns and add two sigma data to table.
table1 <- maize_spread_table %>% 
  dplyr::select(LabID:elevation) %>%
  dplyr::left_join(., sigma_range, by = "LabID") %>% 
  dplyr::select(LabID:Country, Age, Error, range, MedianBP, Date, everything()) %>% 
  # We use paperpile to compile references in a google doc, so we just put a place holder for the column here.
  dplyr::mutate(References = "")

names(table1) <-
  c(
    "Lab ID",
    "Site Name",
    "County",
    "State",
    "Country",
    "14C Age",
    "14C Error",
    "95% Confidence Interval BP",
    "Median cal BP",
    "Median cal BC/AD",
    "GDD (°C)",
    "Elevation (m)",
    "References"
  )

# readr::write_excel_csv(table1,
#                        here::here("vignettes/tables/Table_1.csv"))

knitr::kable(table1,
             caption = "Table 1. Sites on the Maize Northern Frontier (MNF), beginning with Guilá Naquitz. Here, we include information on the uncalibrated and calibrated dates, as well as other information on elevation and GDD.")

```

```{r Table2, echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'asis'}

# Table 2 just focuses on the distance and change in distance data. 
table2 <- maize_spread_table %>% 
  dplyr::select(LabID:SiteName, Date, dist_deg:delta_lat_km) %>% 
  dplyr::mutate(dplyr::across(dist_deg:delta_lat_km, ~ sprintf("%.3f", .x)))

names(table2) <-
  c(
    "Lab ID",
    "Site Name",
    "Median cal BC/AD",
    "Euclidean Distance (Degrees)",
    "Euclidean Distance (km)",
    "\u0394 Degrees Euclidean Distance/Year",
    "\u0394 km Euclidean Distance/Year",
    "\u0394 Degrees North/Year",
    "\u0394 km North/Year"
  )

# readr::write_excel_csv(table2,
#                        here::here("vignettes/tables/Table_2.csv"))

knitr::kable(table2,
             caption = "Table 2. Distances and spread rates for sites on the Maize Northern Frontier (MNF), beginning with Guilá Naquitz. Distances and maize spread rates are between a previous row and a current row. For example, Guilá Naquitz is blank because the distance between Guilá Naquitz and Las Capas is provided in the Las Capas row.")
             
```

```{r Figure_3, fig.cap="Figure 3. Spread rates for maize beginning from Las Capas to sites in the SWUS on the MNF as defined in the text. A) Only change in northing contributes to the rates measured. B) As in A, except rates are expressed as Euclidean distances.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

# Use the same data as above for Figure 2; however, we remove Mexico sites from the frontier to focus just on the southwestern United States.
nodes2 <- nodes1 %>%
  dplyr::filter(name != "Guilá Naquitz") %>%
  dplyr::mutate(id = dplyr::row_number())

# Get state shapefiles for plotting.
country_shapes1 <-
  ggplot2::geom_polygon(
    aes(x = long, y = lat, group = group),
    data = map_data('state'),
    fill = "white",
    color = "black",
    size = 0.15
  )

# Set lat/long limits for plotting.
mapcoords1 <-
  ggplot2::coord_fixed(xlim = c(-114.814521, -105),
                       ylim = c(31.333254, 42))

# Figure 3a. Change in km North latitude per year.
edges3a <- data.frame(from = 1:12,
                      to = 2:13,
                      color = nodes2$delta_lat_km[-1])

# Create edge list.
edges_for_plot3a <- edges3a %>%
  dplyr::inner_join(nodes2 %>% dplyr::select(id, lon, lat), by = c('from' = 'id')) %>%
  dplyr::rename(x = lon, y = lat) %>%
  dplyr::inner_join(nodes2 %>% dplyr::select(id, lon, lat), by = c('to' = 'id')) %>%
  dplyr::rename(xend = lon, yend = lat)

assertthat::assert_that(nrow(edges_for_plot3a) == nrow(edges3a))

maize_frontier_SW_delta_Lat <- ggplot(nodes2) +
  country_shapes1 +
  geom_curve(
    aes(
      x = x,
      y = y,
      xend = xend,
      yend = yend,
      color = color
    ),
    arrow = arrow(length = unit(0.02, "npc"), type = "closed"),
    data = edges_for_plot3a,
    curvature = 0.3,
    alpha = 1.0,
    size = 1.0
  ) +
  scale_color_distiller(palette = "RdYlGn",
                        direction = 1,
                        name = "\u0394 km N/year") +
  geom_point(
    aes(x = lon, y = lat),
    # draw nodes
    shape = 21,
    fill = 'white',
    color = 'black',
    stroke = 0.5,
    size = 3.5
  ) +
  ggrepel::geom_label_repel(
    aes(x = lon, y = lat, label = name),
    # draw text labels
    size = 2.75,
    color = "black",
    min.segment.length = 0.15,
    fill = alpha(c("white"), 0.7),
    force = 34,
    family = "Helvetica"
  ) +
  mapcoords1 +
  theme_void() +
  theme(
    plot.margin = grid::unit(c(2, 2, 2, 2), "mm"),
    legend.box.spacing = unit(20.0, "pt"),
    legend.key.size = unit(0.6, "cm"),
    legend.text = element_text(size = 14, family = "Helvetica"),
    legend.title = element_text(
      size = 18,
      face = "bold",
      family = "Helvetica"
    )
  )

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/spread_rate.png"),
#   plot = maize_frontier_SW_delta_Lat,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )

# Figure 3b. Change in euclidean distance (km) per year.
edges3b <- data.frame(from = 1:12,
                      to = 2:13,
                      color = nodes2$delta_km[-1])

edges_for_plot3b <- edges3b %>%
  dplyr::inner_join(nodes2 %>% dplyr::select(id, lon, lat), by = c('from' = 'id')) %>%
  dplyr::rename(x = lon, y = lat) %>%
  dplyr::inner_join(nodes2 %>% dplyr::select(id, lon, lat), by = c('to' = 'id')) %>%
  dplyr::rename(xend = lon, yend = lat)

assertthat::assert_that(nrow(edges_for_plot3b) == nrow(edges3b))

maize_frontier_SW_delta_euclidean <- ggplot(nodes2) +
  country_shapes1 +
  geom_curve(
    aes(
      x = x,
      y = y,
      xend = xend,
      yend = yend,
      color = color
    ),
    arrow = arrow(length = unit(0.02, "npc"), type = "closed"),
    data = edges_for_plot3b,
    curvature = 0.3,
    alpha = 1.0,
    size = 1.0
  ) +
  scale_color_distiller(palette = "RdYlGn",
                        direction = 1,
                        name = "\u0394 km ED/year") +
  geom_point(
    aes(x = lon, y = lat),
    # draw nodes
    shape = 21,
    fill = 'white',
    color = 'black',
    stroke = 0.5,
    size = 3.5
  ) +
  ggrepel::geom_label_repel(
    aes(x = lon, y = lat, label = name),
    # draw text labels
    size = 2.75,
    color = "black",
    min.segment.length = 0.15,
    fill = alpha(c("white"), 0.7),
    force = 34,
    nudge_x = -0.51,
    nudge_y = 0.15,
    family = "Helvetica"
  ) +
  mapcoords1 +
  theme_void() +
  theme(
    legend.box.spacing = unit(20.0, "pt"),
    legend.key.size = unit(0.6, "cm"),
    legend.text = element_text(size = 14, family = "Helvetica"),
    legend.title = element_text(
      size = 18,
      face = "bold",
      family = "Helvetica"
    )
  )

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/spread_rate_euclidean.png"),
#   plot = maize_frontier_SW_delta_euclidean,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )

# Use cowplot to put the 2 figures together.
Figure_3 <-
  cowplot::plot_grid(
    maize_frontier_SW_delta_Lat,
    maize_frontier_SW_delta_euclidean,
    labels = "AUTO",
    label_x = 0.233,
    label_y = 0.99,
    align = "v",
    nrow = 2,
    ncol = 1
  )

Figure_3

# ggplot2::ggsave(
#   Figure_3,
#   filename = here::here("vignettes/figures/Figure_3.png"),
#   width = 13.7,
#   height = 10.2,
#   dpi = 900
# )

```

```{r Figure_4, fig.cap="Figure 4. Location of sites on the maize northern frontier and other sites mentioned in the text relative to A) elevation and B) annual accumulated GDDs (°C).", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

# GDD Map

# Get raster for the North American Southwest and convert to a dataframe.
gdd_cropped <-
  raster::crop(cropSpreadR::PRISM_annual_gdd, cropDiffusionR::swus_states)
gdd_cropped <-
  raster::mask(gdd_cropped, cropDiffusionR::swus_states)

# Convert to dataframe for plotting.
NASW_gdd_data <-
  raster::as.data.frame(gdd_cropped, xy = TRUE)
# Assign names to the columns.
names(NASW_gdd_data) <- c("long", "lat", "GDD")
#Transform the data for plotting in ggplot.
NASW_gdd_data_transformed <-
  usmap::usmap_transform(
    NASW_gdd_data,
    input_names = c("long", "lat"),
    output_names = c("x", "y")
  )

# We add Turkey Pen as an additional site to plot, but we distinguish between the Maize Northern Frontier and Other sites that we want to plot. 
maize_frontier_swus_tp <- maize_frontier_swus %>%
  dplyr::mutate(`Site Type` = "Maize Frontier") %>%
  dplyr::bind_rows(
    .,
    MDB %>%
      dplyr::filter(SiteName == "Turkey Pen Ruin") %>%
      dplyr::slice_min(Date) %>%
      dplyr::mutate(Long1 = Long, Lat1 = Lat) %>%
      # Convert to spatial object
      sf::st_as_sf(coords = c("Long1", "Lat1"),
                   crs = 4326) %>%
      sf::st_make_valid() %>%
      dplyr::mutate(`Site Type` = "Other")
  ) %>%
  dplyr::mutate(SiteName = ifelse(SiteName == "Elsinore Burial", SiteID, SiteName))

GDD_output <- ggplot2::ggplot(data = NASW_gdd_data_transformed) +
  geom_raster(aes(x = long, y = lat, fill = GDD),
              alpha = 0.8,
              na.rm = TRUE) +
  # For GDD raster in greyscale.
  # scale_fill_gradientn(
  #   colors = grey.colors(
  #     10,
  #     start = 0.15,
  #     end = 0.9,
  #     rev = TRUE
  #   ),
  #   na.value = "transparent",
  #   name = "Annual GDDs "
  # ) +
#For GDD raster in color.
scale_fill_gradientn(colors = colorRampPalette(rev(RColorBrewer::brewer.pal(11, "RdBu")))(255),
                     na.value = "transparent",
                     name = "Annual GDD (°C)") +
  coord_equal() +
  theme_bw() +
  geom_sf(
    data = cropDiffusionR::swus_states,
    color = "#2b2b2b",
    fill = "transparent",
    size = 0.4,
    inherit.aes = FALSE
  ) +
  geom_sf(
    data = maize_frontier_swus_tp,
    aes(geometry = geometry,
        shape = `Site Type`),
    inherit.aes = FALSE,
    size = 4,
    show.legend = "point"
  ) +
  scale_shape_manual(values = c(16, 17)) +
  ggrepel::geom_label_repel(
    data = maize_frontier_swus_tp,
    aes(label = SiteName, geometry = geometry),
    stat = "sf_coordinates",
    min.segment.length = 0.4,
    force = 47,
    label.size = 0,
    size = 4.5,
    inherit.aes = FALSE,
    max.overlaps = 20,
    fill = alpha(c("white"), 0.7)
  ) +
  theme(
    #legend.position = c(.15, .2),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_blank(),
    axis.text.x = element_blank(),
    axis.title = element_blank(),
    axis.line = element_blank(),
    axis.ticks = element_blank(),
    legend.text = element_text(size = 15, family = "Helvetica"),
    legend.title = element_text(
      size = 16,
      family = "Helvetica",
      face = "bold"
    )
  )


# Elevation Map

NASW_ned_data <-
  raster::as.data.frame(swus_NED, xy = TRUE)
# Assign names to the columns.
names(NASW_ned_data) <- c("long", "lat", "elev")
# Transform the data for plotting in ggplot.
swus_NED_transformed <-
  usmap::usmap_transform(
    NASW_ned_data,
    input_names = c("long", "lat"),
    output_names = c("x", "y")
  )

NED_output <- ggplot2::ggplot(data = swus_NED_transformed) +
  geom_raster(aes(x = long, y = lat, fill = elev),
              alpha = 0.8,
              na.rm = TRUE) +
  # For GDD raster in greyscale.
  # scale_fill_gradientn(
  #   colors = grey.colors(
  #     10,
  #     start = 0.15,
  #     end = 0.9,
  #     rev = TRUE
  #   ),
  #   na.value = "transparent",
  #   name = "Annual GDDs "
  # ) +
#For GDD raster in color.
scale_fill_gradientn(
  colors = grDevices::colorRampPalette(MetBrewer::met.brewer("VanGogh3"))(256),
  na.value = "transparent",
  name = "Elevation (m)"
) +
  coord_equal() +
  theme_bw() +
  geom_sf(
    data = cropDiffusionR::swus_states,
    color = "#2b2b2b",
    fill = "transparent",
    size = 0.4,
    inherit.aes = FALSE
  ) +
  geom_sf(
    data = maize_frontier_swus_tp,
    aes(geometry = geometry,
        shape = `Site Type`),
    inherit.aes = FALSE,
    size = 4,
    show.legend = "point"
  ) +
  scale_shape_manual(values = c(16, 17)) +
  ggrepel::geom_label_repel(
    data = maize_frontier_swus_tp,
    aes(label = SiteName, geometry = geometry),
    stat = "sf_coordinates",
    min.segment.length = 0.4,
    force = 47,
    label.size = 0,
    size = 4.5,
    inherit.aes = FALSE,
    max.overlaps = 20,
    fill = alpha(c("white"), 0.7)
  ) +
  theme(
    #legend.position = c(.15, .2),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_blank(),
    axis.text.x = element_blank(),
    axis.title = element_blank(),
    axis.line = element_blank(),
    axis.ticks = element_blank(),
    legend.text = element_text(size = 15, family = "Helvetica"),
    legend.title = element_text(
      size = 16,
      family = "Helvetica",
      face = "bold"
    )
  )

# Use cowplot to put the GDD and elevation maps in the same figure.
fig_envA <-
  cowplot::plot_grid(
    NED_output,
    GDD_output,
    labels = "AUTO",
    label_x = 0.26,
    align = "v",
    nrow = 2,
    ncol = 1
  )

fig_envA

# ggplot2::ggsave(
#   fig_envA,
#   filename = here::here("vignettes/figures/Figure_4.png"),
#   dpi = 600
# )

```

```{r Figure_5, fig.cap="Figure 5. Box plots showing elevation, GDD, and latitude through time for the earliest maize sample from each of 275 sites in our database (A. Gillreath-Brown and Kohler n.d. 2022), excluding 19 sites post-dating AD 1500 or south of 30° Latitude. Samples are binned by 500-year periods and plotted on the midpoint of each bin.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

# Limit box plot to just be greater than 30° N and less than AD 1500. We want only the oldest samples from each site, then we group in 500 year time bins for plotting the box plots.  
MDB_env2 <- MDB_env %>%
  dplyr::filter(Lat >= 30 & Date < 1500) %>%
  dplyr::group_by(SiteIDName) %>%
  dplyr::slice_min(Date, with_ties = FALSE) %>%
  dplyr::group_by(gr = cut(
    Date,
    breaks = seq(-4000, 1500, 500),
    labels = seq(-3750, 1250, 500)
  )) %>%
  dplyr::mutate(date2 = as.numeric(as.character(gr))) %>%
  dplyr::ungroup() %>%
  dplyr::select(-gr) %>%
  dplyr::arrange(date2) %>%
  dplyr::rename(Latitude = Lat, Elevation = elevation)

# Get the average for the 3 variables. 
env_summaries <- MDB_env2 %>%
  dplyr::summarise(dplyr::across(c(Latitude, GDD, Elevation), ~ mean(.x, na.rm = TRUE)))

# Transform the data into long format for plotting.
env_summaries_bp <-
  MDB_env2 %>%
  dplyr::select(date2, Latitude, GDD, Elevation) %>%
  dplyr::rename(
    `Latitude (°)` = Latitude,
    `Annual GDD (°C)` = GDD,
    `Elevation (m)` = Elevation
  ) %>%
  tidyr::pivot_longer(-date2) %>%
  ggplot(aes(
    x = factor(date2),
    y = value,
    fill = factor(name)
  )) +
  geom_boxplot() +
  facet_grid(vars(name), scales = "free") +
  labs(fill = "") +
  xlab("Years BC/AD") +
  ylab("Value") +
  theme_bw() +
  theme(
    strip.text.y = element_text(size = 12),
    legend.key.size = unit(1.5, "cm"),
    legend.text = element_text(size = 12),
    axis.ticks.length = unit(0.125, "inch"),
    axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1),
    axis.text = element_text(
      size = 26,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_blank(),
    axis.title.x = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    ),
    axis.line = element_line(colour = "black"),
    plot.margin = margin(
      t = 5,
      r = 25,
      b = 5,
      l = 5
    )
  )

env_summaries_bp

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_5.png"),
#   plot = env_summaries_bp,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )

```

```{r multiple_linear_regression, echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'asis'}

# Limit regressions to sites greater than 30° N and less than AD 1500, and only to the Maize Northern Frontier sites.
mlr <- MDB_env %>%
  dplyr::group_by(SiteIDName) %>%
  dplyr::slice_min(Date) %>%
  dplyr::filter(Lat >= 30 & Date < 1500) %>%
  dplyr::mutate(bin = cut(Date,
                          breaks = seq(-3700, 1500, 100))) %>%
  dplyr::group_by(bin) %>%
  dplyr::mutate(group = dplyr::cur_group_id()) %>%
  dplyr::filter(LabID %in% maize_frontier_swus$LabID) %>%
  dplyr::ungroup()

# Rescale the 4 variables.
mlr2 <- mlr %>%
  dplyr::select(Date, Lat, elevation, GDD) %>% 
  dplyr::mutate(dplyr::across(everything(), scales::rescale))

# Calculate the simple linear regressions for each combination of variables for plotting on the dag.
gdd_lat <- lm(GDD ~ Lat, data = mlr2)
gdd_elev <- lm(GDD ~ elevation, data = mlr2)
date_lat <- lm(Date ~ Lat, data = mlr2)
date_elev <- lm(Date ~ elevation, data = mlr2)
date_gdd <- lm(Date ~ GDD, data = mlr2)

# Extract all the statistics for the linear regressions.
stats_table <-
  purrr::map_dfr(list(gdd_lat, gdd_elev, date_lat, date_elev, date_gdd), function(x) {
    x %>%
      summary() %>%
      {
        data.frame(
          formula = as.character(.$call)[[2]],
          slope = formatC(.$coefficients[2, 1], digits = 3, format = "f"),
          r2 = formatC(.$r.squared, digits = 2, format = "f"),
          pvalue = formatC(.$coefficients[2, 4],
                           digits = 3,
                           format = "f"),
          Fstatistic = formatC(.$fstatistic[[1]],
                               digits = 2,
                               format = "f")
        )
      } })

knitr::kable(stats_table,
             caption = "Statistics Table. Summary of statistics for the linear models shown in Figure 6.")

```

```{r Figure_6, fig.cap="Figure 6. A directed acyclic graph of the relationships between the ages of maize on the MNF and the latitude, elevation, and GDD for the sites with this maize. Beta weights on each arrow indicate the size of the effect of the independent variable on the dependent. Solid lines indicate a significant effect (at p < 0.01); the dashed line indicates an effect that is not significant.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

# Create layout for the dag.
dag_coords <-
  tibble::tibble(
    name = c("Latitude", "Elevation", "GDD", "Date"),
    x    = c(1, 2, 1.5, 1.5),
    y    = c(3, 3, 2, 1)
  )

# Create connections in the dag.
dag_out <- ggdag::dagify(GDD ~ Latitude + Elevation,
                         Date ~ GDD,
                         Date ~ Latitude + Elevation,
                         coords = dag_coords) %>%
  # Put in a tidy format.
  ggdag::tidy_dagitty() %>%
  # Use a custom function to shorten the dag arrows on both ends.
  cropSpreadR::shorten_dag_arrows(proportion = 0.27)

# Add the slope (i.e., beta weights) to the tidy tibble, so that we can also plot the beta weights on the dag.
dag_out$data <-
  dplyr::left_join(
    x = dag_out$data,
    stats_table %>%
      tidyr::separate(formula, into = c("to", "name"), sep = " ~ ") %>%
      dplyr::mutate(across(
        c(to, name),
        ~ dplyr::case_when(.x == "Lat" ~ "Latitude",
                           .x == "elevation" ~ "Elevation",
                           TRUE ~ .x)
      )) %>%
      dplyr::mutate(sig = pvalue <= 0.05) %>%
      dplyr::select(name, to, slope, sig),
    by = c("name", "to")
  ) %>%
  dplyr::rowwise() %>%
  dplyr::mutate(
    slope = sprintf("%.2f", as.numeric(slope)),
    x_label = ((xstart - xend) / 2) + xend,
    y_label = ((ystart - yend) / 2) + yend,
    edge_linetype = ifelse(sig == TRUE, "solid", "dashed")
  ) %>%
  dplyr::ungroup()

# Plot the dag.
dag_figure <- dag_out %>%
  ggplot(aes(
    x = x,
    y = y,
    xend = xend,
    yend = yend
  )) +
  ggdag::geom_dag_point(color = "cornflowerblue",
                        alpha = 1 / 4,
                        size = 45) +
  ggdag::geom_dag_text(color = "cornflowerblue",
                       size = 7,
                       fontface = "bold") +
  ggdag::geom_dag_edges(
    aes(
      x = xstart,
      y = ystart,
      edge_linetype = edge_linetype
    ),
    edge_color = "cornflowerblue",
    arrow_directed = grid::arrow(length = grid::unit(12, "pt"), type = "closed"),
    edge_width = 1.35
  ) +
  ggdag::geom_dag_label_repel(
    aes(x = x_label, y = y_label, label = slope),
    segment.size = 0,
    force = 0,
    size = 6,
    label.size = NA
  ) +
  scale_x_continuous(NULL, breaks = NULL, expand = c(0.1, 0.1)) +
  scale_y_continuous(NULL, breaks = NULL, expand = c(0.2, 0.2)) +
  theme_bw() +
  theme(panel.grid = element_blank(),
        panel.border = element_blank())

dag_figure

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_6.png"),
#   plot = dag_figure,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )

```

```{r Figure_7, fig.cap="Figure 7. Regression of the rate of maize dispersal per year along the MNF on the estimated summer temperature anomaly (°C) (Andrew Gillreath-Brown, Bocinsky, and Kohler n.d. 2022) (r2 = 0.16; p = 0.20). Summer temperature estimates increase from left to right on the x-axis. The gray band represents a 95% confidence interval.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

nodes7 <-
  maize_frontier_swus %>%
  dplyr::mutate(SiteName = ifelse(SiteName == "Elsinore Burial", SiteID, SiteName)) %>%
  dplyr::arrange(Date) %>%
  dplyr::mutate(id = dplyr::row_number()) %>%
  dplyr::select(id, Long, Lat, SiteName, Date) %>%
  sf::st_drop_geometry() %>%
  dplyr::mutate(delta = ((Lat - dplyr::lag(Lat)) * 110.574) / (Date - dplyr::lag(Date)),
                delta = as.numeric(sprintf("%.2f", delta)))

anom_point <- temp_anom %>%
  dplyr::select(year, anom) %>%
  dplyr::mutate(
    anom_25 = dplyr::lag(anom, 25),
    anom_50 = dplyr::lag(anom, 50),
    anom_75 = dplyr::lag(anom, 75),
    anom_100 = dplyr::lag(anom, 100),
    anom_200 = dplyr::lag(anom, 200)
  )

lags_point <- nodes7 %>%
  dplyr::filter(SiteName != "Las Capas") %>%
  dplyr::left_join(., anom_point, by = c("Date" = "year"))

# Get fit, and make a variable with labels for points outside conf. interval
fit <- lm(delta ~ anom_25, data = lags_point)
dat <- predict(fit, interval = "confidence")

lags_point$inside <-
  ifelse(
    lags_point$delta < dat[, "upr"] &
      lags_point$delta > dat[, "lwr"],
    "",
    as.character(lags_point$SiteName)
  )

temp_spreadrate <- lags_point %>%
  ggplot2::ggplot(aes(x = anom_25, y = delta)) +
  geom_point(size = 4) +
  stat_smooth(method = "lm", size = 2) +
  scale_x_continuous(breaks = seq(-1.5, 0, 0.25)) +
  scale_y_continuous(breaks = seq(-0.4, 1.0, 0.2)) +
  ggrepel::geom_text_repel(
    aes(label = inside),
    cex = 6.5,
    show.legend  = F,
    min.segment.length = unit(0.1, 'lines'),
    box.padding = 1
  ) +
  theme_bw() +
  xlab("Temperature Anomaly (°C)") +
  ylab("\u0394 km N/year") +
  theme(
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_text(
      size = 26,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 10,
        r = 5,
        b = 10,
        l = 10
      )
    ),
    axis.title.x = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    ),
    panel.grid.minor = element_blank(),
    axis.line = element_line(colour = "black"),
    plot.margin = margin(
      t = 5,
      r = 25,
      b = 5,
      l = 5
    )
  )

temp_spreadrate

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_7.png"),
#   plot = temp_spreadrate,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )

```

```{r Figure_8, fig.cap="Figure 8. The 12 counties showing A) number of radiocarbon samples and B) number of sites with radiocarbon samples in the four-state SWUS with a site on the MNF (2 sites fall within Daggett County, Utah). The northern frontier of maize expansion was located almost entirely within the western Southwest. We still show Delta County here as it is on the MNF; however, we have excluded it from Figure 9 due to sample size.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

sw_rc <- cropSpreadR::sw_rc

#frontier_counties <- unique(maize_frontier_swus$county_st)[!grepl("delta, colorado", unique(maize_frontier_swus$county_st))]
frontier_counties <-
  c(
    "sevier, utah",
    "delta, colorado",
    "daggett, utah",
    "catron, new mexico",
    "navajo, arizona",
    "salt lake, utah",
    "pima, arizona",
    "kane, utah",
    "montrose, colorado",
    "uintah, utah",
    "box elder, utah",
    "san juan, utah"
  )

sw_rc_subset <- sw_rc %>%
  dplyr::filter(county_st %in% frontier_counties) %>% 
  dplyr::group_by(county_st) %>% 
  dplyr::summarise(no_samples = dplyr::n(),
                   no_sites = length(unique(SiteIDName)))

# Get sites that are on the northern maize frontier through time from cropDiffusionR package.
temp <- tempfile()
download.file(
  "https://raw.githubusercontent.com/Archaeo-Programmer/cropDiffusionR/master/vignettes/output/maize_frontier_swus.rds",
  temp
)
maize_frontier_swus <-
  readr::read_rds(temp, "maize_frontier_swus.rds") %>%
  dplyr::left_join(., MDB %>% dplyr::select(LabID, Long, Lat), by = "LabID")
unlink(temp)

# Create new column for fill, which will be used for plotting. Grey if the county occurs on the maize northern frontier, and blank if it does not.
SWUS_counties_fill <- cropSpreadR::SWUS_counties %>%
  dplyr::left_join(., sw_rc_subset, by = "county_st") %>% 
  dplyr::mutate(
    #fill = ifelse(county_st %in% maize_frontier_swus$county_st, "grey", NA),
    label = ifelse(
      county_st %in% maize_frontier_swus$county_st,
      tools::toTitleCase(NAME),
      NA
    )
  )

# Plot.
spd_counties_samples <- ggplot(data = SWUS_counties_fill) +
  geom_sf(color = "black",
          aes(fill = no_samples),
          size = 0.25,
          na.rm = TRUE
          ) +
  scale_fill_continuous(low="#FFC20A", high="#0C7BDC",
                       guide="colorbar",na.value="transparent",
                       name = "no. radiocarbon dates") +
  ggrepel::geom_label_repel(
    aes(label = label, geometry = geom),
    stat = "sf_coordinates",
    min.segment.length = 0.1,
    size = 2.5
  ) +
coord_sf() +
  theme_bw() +
  theme(
    legend.position = "bottom",
    legend.key.width=unit(0.05,"npc"),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_blank(),
    axis.text.x = element_blank(),
    axis.title = element_blank(),
    axis.line = element_blank(),
    axis.ticks = element_blank(),
    legend.text = element_text(size = 12, family = "Helvetica"),
    legend.title = element_text(
      size = 14,
      family = "Helvetica",
      face = "bold"
    )
  )

spd_counties_sites <- ggplot(data = SWUS_counties_fill) +
  geom_sf(color = "black",
          aes(fill = no_sites),
          size = 0.25,
          na.rm = TRUE
          ) +
  scale_fill_continuous(low="#FFC20A", high="#0C7BDC",
                       guide="colorbar",na.value="transparent",
                       name = "no. sites with \nradiocarbon dates") +
  ggrepel::geom_label_repel(
    aes(label = label, geometry = geom),
    stat = "sf_coordinates",
    min.segment.length = 0.1,
    size = 2.5
  ) +
coord_sf() +
  theme_bw() +
  theme(
    legend.position = "bottom",
    legend.key.width=unit(0.05,"npc"),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_blank(),
    axis.text.x = element_blank(),
    axis.title = element_blank(),
    axis.line = element_blank(),
    axis.ticks = element_blank(),
    legend.text = element_text(size = 12, family = "Helvetica"),
    legend.title = element_text(
      size = 14,
      family = "Helvetica",
      face = "bold"
    )
  )

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_8.png"),
#   plot = spd_counties,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )

Figure_8 <-
  cowplot::plot_grid(
    spd_counties_samples,
    spd_counties_sites,
    labels = "AUTO",
    label_x = 0.36,
    label_y = 0.99,
    align = "v",
    nrow = 2,
    ncol = 1
  )

Figure_8

# ggplot2::ggsave(
#   Figure_8,
#   filename = here::here("vignettes/figures/Figure_8.png"),
#   width = 13.7,
#   height = 10.2,
#   dpi = 900
# )

```

```{r Figure_9, fig.cap="Figure 9. The MNF relative to the Monte-Carlo simulation test for SPDs for the counties in the four-state southwestern United States that are on the MNF at some point in time (see Gillreath-Brown and Kohler n.d., Figure 6 and site locations in Figure 2). Red points locate the earliest known maize in each county. Sample sizes for the SPDs shown in parentheses and the p value for the global significance test for the summed probabilities. The grey band (or envelope) represents the 95% confidence interval, which is where we would expect 95% of the radiocarbon data to be. Note variability in scales on y axis.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

sw_rc <- cropSpreadR::sw_rc

#frontier_counties <- unique(maize_frontier_swus$county_st)[!grepl("delta, colorado", unique(maize_frontier_swus$county_st))]
frontier_counties <-
  c(
    "sevier, utah",
    "daggett, utah",
    "catron, new mexico",
    "navajo, arizona",
    "salt lake, utah",
    "pima, arizona",
    "kane, utah",
    "montrose, colorado",
    "uintah, utah",
    "box elder, utah",
    "san juan, utah"
  )

sw_rc_subset <- sw_rc %>%
  dplyr::filter(county_st %in% frontier_counties)

nsim = 100

sw.caldates <- sw_rc_subset %>%
  dplyr::select(county_st, Age, Error, calib) %>%
  dplyr::group_by(county_st) %>%
  tidyr::nest() %>%
  dplyr::mutate(monte_carlo = purrr::map(data, function(x) {
    sw.caldates <-
      rcarbon::calibrate(x = x$Age,
                         errors = x$Error,
                         calCurves = 'intcal20')
    rcarbon::modelTest(
      sw.caldates,
      errors = x$Error,
      nsim = nsim,
      timeRange = c(7000, 0),
      model = "exponential",
      runm = 100,
      method = c("calsample")
    )
  }))

monte_carlo_sim <- sw.caldates %>%
  dplyr::mutate(
    pvalue = purrr::map_dbl(monte_carlo, ~ .x$pval),
    n = purrr::map_dbl(monte_carlo, ~ .x$n),
    pvalue = ifelse(pvalue < 0.01, paste0("p < 0.01"), paste0("p = ", sprintf("%.2f", pvalue))),
    monte_carlo = purrr::map(
      monte_carlo,
      ~ .x$result %>% dplyr::mutate(Date = rcarbon::BPtoBCAD(calBP))
    )
  ) %>%
  dplyr::select(-data) %>%
  tidyr::unnest(monte_carlo) %>%
  dplyr::mutate(
    boom_bust = dplyr::case_when(
      PrDens < hi & PrDens > lo ~ "medium",
      PrDens > hi ~ "indianred",
      PrDens < lo ~ "royalblue",
      TRUE ~ "medium"
    )
  )

maize_oldest <- maize_frontier_swus %>%
  sf::st_drop_geometry() %>%
  dplyr::mutate(SiteNameID = dplyr::coalesce(SiteName, SiteID)) %>%
  dplyr::select(county_st, SiteNameID, Date) %>%
  dplyr::mutate(date_maize = Date) %>%
  dplyr::group_by(county_st) %>%
  dplyr::slice_min(Date) %>%
  dplyr::ungroup() %>%
  dplyr::arrange(Date) %>%
  dplyr::filter(county_st %in% frontier_counties)

ckde_test3 <- monte_carlo_sim %>%
  dplyr::left_join(., maize_oldest, by = c("county_st", "Date")) %>%
  dplyr::mutate(
    county_st2 = county_st,
    county_st = tools::toTitleCase(county_st),
    county_st = paste0(county_st, " (n = ", n, "; ",  pvalue, ")")
  ) %>%
  dplyr::mutate(SiteNameID = ifelse(SiteNameID == "Elsinore Burial", "42SV2111", SiteNameID))

county_order <- cropSpreadR::SWUS_counties %>%
  sf::st_drop_geometry() %>%
  dplyr::select(county_st2 = county_st, centroid_lat) %>%
  dplyr::arrange(desc(centroid_lat)) %>%
  dplyr::filter(county_st2 %in% frontier_counties) %>%
  dplyr::left_join(
    .,
    ckde_test3 %>% dplyr::ungroup() %>% dplyr::select(county_st, county_st2) %>% dplyr::distinct(),
    by = "county_st2"
  ) %>%
  dplyr::pull(county_st)

condensed <- monte_carlo_sim %>%
  dplyr::group_by(county_st) %>%
  dplyr::mutate(change = cumsum(boom_bust != lag(boom_bust, def = first(boom_bust)))) %>%
  dplyr::group_by(change, .add = TRUE) %>%
  dplyr::filter(dplyr::row_number() %in% c(1, dplyr::n()) &
                  boom_bust != "medium") %>%
  dplyr::mutate(time = dplyr::row_number()) %>% 
  dplyr::ungroup() %>% 
  tidyr::complete(county_st, change, time) %>%
  tidyr::fill(dplyr::everything()) %>% 
  dplyr::select(-time) %>% 
  dplyr::group_by(grp = stringr::str_c('Column', rep(1:2, length.out = dplyr::n()))) %>%
  dplyr::mutate(rn = dplyr::row_number()) %>%
  dplyr::ungroup() %>%
  dplyr::select(county_st, Date, boom_bust, grp, rn) %>%
  tidyr::pivot_wider(names_from = grp, values_from = Date) %>%
  dplyr::select(-rn,
                xmin = Column1,
                xmax = Column2,
                fill = boom_bust) %>%
  dplyr::mutate(ymin = -Inf, ymax = Inf) %>%
  dplyr::left_join(
    .,
    ckde_test3 %>% dplyr::ungroup() %>% dplyr::select(county_st, county_st2) %>% dplyr::distinct(),
    by = c("county_st" = "county_st2")
  ) %>%
  dplyr::select(-county_st, county_st = county_st.y) %>%
  dplyr::mutate(county_st = factor(county_st, levels = county_order))

monte_carlo_plot <-
  ggplot(data = ckde_test3, aes(y = PrDens, x = Date)) +
  geom_ribbon(aes(ymin = lo, ymax = hi),
              alpha = 0.4,
              fill = "grey70") +
  geom_rect(
    aes(
      xmin = xmin,
      xmax = xmax,
      ymin = ymin,
      ymax = ymax,
      fill = I(fill)
    ),
    alpha = 0.4,
    inherit.aes = FALSE,
    data = condensed
  ) +
  geom_line() +
  geom_point(
    aes(y = PrDens, x = date_maize),
    size = 2.75,
    color = "red",
    inherit.aes = FALSE
  ) +
  ggrepel::geom_label_repel(
    aes(y = PrDens, x = date_maize, label = SiteNameID),
    color = "red",
    cex = 5,
    show.legend  = F,
    min.segment.length = unit(0.1, 'lines'),
    box.padding = 1,
    nudge_x = -550,
    nudge_y = 0.35,
    fill = alpha(c("white"), 0.6),
    label.size = NA
  ) +
  theme_bw() +
  scale_x_continuous(breaks = seq(-4000, 2000, 500),
                     limits = c(-4100, 2000)) +
  scale_y_continuous(
    breaks = function(x)
      seq(0, max(x), length.out = 3),
    labels = function(y)
      sprintf("%.2f", y)
  ) +
  facet_wrap(~ factor(county_st, levels = county_order),
             scales = "free_y",
             ncol = 1) +
  xlab("Years BC/AD") +
  ylab("Summed Probability") +
  theme(
    legend.position = "none",
    panel.spacing = unit(1, "lines"),
    strip.text.x = element_text(
      size = 16,
      family = "Helvetica",
      face = "bold"
    ),
    axis.ticks.length = unit(0.125, "inch"),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_text(
      size = 20,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 10,
        r = 5,
        b = 10,
        l = 10
      )
    ),
    axis.title.x = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    ),
    panel.grid.minor = element_blank(),
    axis.line = element_line(colour = "black"),
    plot.margin = margin(
      t = 5,
      r = 25,
      b = 5,
      l = 5
    )
  )

monte_carlo_plot

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_9.png"),
#   plot = monte_carlo_plot,
#   width = 16,
#   height = 18,
#   dpi = 600
# )

```

```{r estimating_maize_departure, echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'asis'}

maize_spread_rates <- readr::read_csv(here::here("vignettes/tables/Table_2.csv"), show_col_types = FALSE) %>% 
  dplyr::select(`Lab ID`, `Site Name`, delta_kmN = `Δ km North/Year`)

max_SW_spread <- maize_spread_rates %>% 
  dplyr::filter(!`Site Name` %in% c("Guilá Naquitz", "Las Capas")) %>% 
  dplyr::slice_max(delta_kmN)

max_spread <- max_SW_spread %>% 
  dplyr::slice(2) %>% 
  dplyr::pull(delta_kmN)

LC_ST_rate <- maize_frontier_swus %>% 
  dplyr::filter(SiteName %in% c("Las Capas", "Six Toe Shelter")) %>% 
  dplyr::arrange(Date) %>% 
  dplyr::mutate(delta_lat_deg = ((Lat - dplyr::lag(Lat))) / (Date - dplyr::lag(Date)), 
                delta_lat_deg = as.numeric(sprintf("%.4f", delta_lat_deg)), 
                delta_lat_km = delta_lat_deg * 110.574) %>% 
  dplyr::filter(!is.na(delta_lat_km)) %>% 
  dplyr::pull(delta_lat_km)

LC_WM_rate <- maize_frontier_swus %>% 
  dplyr::filter(SiteName %in% c("Las Capas", "Willard Mounds")) %>% 
  dplyr::arrange(Date) %>% 
  dplyr::mutate(delta_lat_deg = ((Lat - dplyr::lag(Lat))) / (Date - dplyr::lag(Date)), 
                delta_lat_deg = as.numeric(sprintf("%.4f", delta_lat_deg)), 
                delta_lat_km = delta_lat_deg * 110.574) %>% 
  dplyr::filter(!is.na(delta_lat_km)) %>% 
  dplyr::pull(delta_lat_km)

balsas_dissolve <- cropSpreadR::Balsas_Basin %>% 
  dplyr::summarize()

balsas_centroid <- balsas_dissolve %>% 
  sf::st_centroid() %>% 
  sf::st_coordinates()

dist <- (maize_frontier_swus %>% 
  dplyr::filter(SiteName == "Las Capas") %>% 
  dplyr::pull(Lat) - balsas_centroid[1,2]) * 110.574
  
out <- data.frame(type = c("Max SWUS from MNF", "Las Capas to Six Toe Shelter", "Las Capas to Willard Mounds"),
                  rate = c(max_spread, LC_ST_rate, LC_WM_rate)) %>% 
  dplyr::mutate(time = dist / rate)

names(out) <- c("Spread Rate", "Rate (km N/Year)", "Time Elapsed (Years)")

knitr::kable(out,
             caption = "Table S2. Summary for the spread rates of maize. Here, we show the maximum rate between any two sites on the MNF, then we show the rate between Las Capas and two other sites, Six Toe Shelter and Willard Mounds. Time elapsed is the amount of time that it would take to reach the Balsas River Basin at the given spread rate. We use the distance between Las Capas to the centroid of the Balses River Basin to calculate the time elapsed.")

```

# Supplemental Figures

```{r Figure_S1, fig.cap="Figure S1. Spread rates for maize beginning from Las Capas to sites in the SWUS on the MNF as defined in the text. However, here we show a route splitting at Three Fir Shelter into an eastern and western trajectories along which new rates are calculated. A) Only change in northing contributes to the rates measured. B) As in A, except rates are expressed as Euclidean distances.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

# Get sites that are on the northern maize frontier through time from cropDiffusionR package for the SWUS.
temp <- tempfile()
download.file(
  "https://raw.githubusercontent.com/Archaeo-Programmer/cropDiffusionR/master/vignettes/output/maize_frontier_swus.rds",
  temp
)
maize_frontier_swus <-
  readr::read_rds(temp, "maize_frontier_swus.rds")
unlink(temp)

maize_frontier_swus <- MDB %>%
  dplyr::filter(LabID %in% maize_frontier_swus$LabID) %>%
  dplyr::mutate(Long1 = Long, Lat1 = Lat) %>%
  # Convert to spatial object
  sf::st_as_sf(coords = c("Long1", "Lat1"),
               crs = 4326) %>%
  sf::st_make_valid()

nodesS1 <-
  maize_frontier_swus %>%
  dplyr::mutate(
    SiteName = ifelse(SiteName == "Elsinore Burial", SiteID, SiteName),
    SiteName = ifelse(SiteName == "Guila Naquitz", "Guilá Naquitz", SiteName)
  ) %>%
  dplyr::mutate(
    grp = dplyr::case_when(
      SiteName == "Las Capas" ~ 1,
      SiteName == "Old Corn" ~ 1,
      SiteName == "Three Fir Shelter" ~ 1,
      SiteName == "Eagles Watch" ~ 2,
      SiteName == "42SV2111" ~ 2,
      SiteName == "Hot Spring Lake Site" ~ 2,
      SiteName == "Willard Mounds" ~ 2,
      SiteName == "Six Toe Shelter" ~ 3,
      SiteName == "Cottonwood Cave" ~ 3,
      SiteName == "Eagle Rock Shelter" ~ 3,
      SiteName == "Juniper Ledge Shelter" ~ 3,
      SiteName == "Sonders Basket" ~ 3,
      SiteName == "Finch Draw Shelter" ~ 3
    )
  ) %>%
  dplyr::bind_rows(
    .,
    maize_frontier_swus %>% dplyr::filter(SiteName == "Three Fir Shelter") %>% dplyr::mutate(grp = 2),
    maize_frontier_swus %>% dplyr::filter(SiteName == "Three Fir Shelter") %>% dplyr::mutate(grp = 3)
  ) %>%
  dplyr::arrange(grp, Date) %>%
  dplyr::select(grp, Long, Lat, SiteName, Date, LabID) %>%
  dplyr::group_by(grp) %>%
  dplyr::rename(geoms = geometry) %>% 
  dplyr::mutate(
    lead_row = geoms[dplyr::row_number() + 1],
    dist_deg = as.numeric(dplyr::lag(
      sf::st_distance(geoms, lead_row, by_element = TRUE, which = "Euclidean")
    )),
    dist_km = dist_deg * 110.574,
    delta_deg = dist_deg / (Date - dplyr::lag(Date)),
    delta_km = dist_km / (Date - dplyr::lag(Date)),
    delta_lat_deg = ((Lat - dplyr::lag(Lat))) / (Date - dplyr::lag(Date)),
    delta_lat_deg = as.numeric(sprintf("%.4f", delta_lat_deg)),
    delta_lat_km = delta_lat_deg * 110.574
  ) %>%
  dplyr::select(-lead_row) %>%
  sf::st_drop_geometry() %>%
  dplyr::arrange(grp, Date) %>%
  dplyr::ungroup() %>%
  dplyr::mutate(id = dplyr::row_number())


names(nodesS1) <-
  c(
    "grp",
    "lon",
    "lat",
    "name",
    "date",
    "lab_id",
    "dist_deg",
    "dist_km",
    "delta_deg",
    "delta_km",
    "delta_lat_deg",
    "delta_lat_km",
    "id"
  )

nodesS1 <- nodesS1 %>%
  dplyr::arrange(id) %>%
  dplyr::select(id, everything())

country_shapes1 <-
  ggplot2::geom_polygon(
    aes(x = long, y = lat, group = group),
    data = map_data('state'),
    fill = "white",
    color = "black",
    size = 0.15
  )

mapcoords1 <-
  ggplot2::coord_fixed(xlim = c(-114.814521, -105),
                       ylim = c(31.333254, 42))

legend.breaks = 3

edges1 <- data.frame(
  from = c(1:2, 4:7, 9:14),
  to = c(2:3, 5:8, 10:15),
  color = nodesS1$delta_lat_km[!is.na(nodesS1$delta_lat_km)]
)

nodes_dups <- nodesS1 %>%
  dplyr::filter(!dplyr::row_number() %in% c(4, 9))

edges_for_plot1 <- edges1 %>%
  dplyr::inner_join(nodesS1 %>% dplyr::select(id, lon, lat), by = c('from' = 'id')) %>%
  dplyr::rename(x = lon, y = lat) %>%
  dplyr::inner_join(nodesS1 %>% dplyr::select(id, lon, lat), by = c('to' = 'id')) %>%
  dplyr::rename(xend = lon, yend = lat)

assertthat::assert_that(nrow(edges_for_plot1) == nrow(edges1))

fig_s1A <-
ggplot(nodesS1) +
  country_shapes1 +
  geom_curve(
    aes(
      x = x,
      y = y,
      xend = xend,
      yend = yend,
      color = color
    ),
    arrow = arrow(length = unit(0.02, "npc"), type = "closed"),
    data = edges_for_plot1,
    curvature = 0.3,
    alpha = 1.0,
    size = 1.0
  ) +
  scale_color_distiller(palette = "RdYlGn",
                        direction = 1,
                        name = "\u0394 km N/year") +
  geom_point(
    aes(x = lon, y = lat),
    # draw nodes
    shape = 21,
    fill = 'white',
    color = 'black',
    stroke = 0.5,
    size = 3.5
  ) +
  ggrepel::geom_label_repel(
    aes(x = lon, y = lat, label = name),
    # draw text labels
    size = 2.75,
    color = "black",
    min.segment.length = 0.15,
    fill = alpha(c("white"), 0.7),
    force = 34,
    nudge_x = -0.51,
    nudge_y = 0.15,
    family = "Helvetica"
  ) +
  mapcoords1 +
  theme_void() +
  theme(
    plot.margin = grid::unit(c(2, 2, 2, 2), "mm"),
    legend.box.spacing = unit(20.0, "pt"),
    legend.key.size = unit(0.6, "cm"),
    legend.text = element_text(size = 14, family = "Helvetica"),
    legend.title = element_text(
      size = 18,
      face = "bold",
      family = "Helvetica"
    )
  )

# ggplot2::ggsave(
#   fig_s1A,
#   filename = here::here("vignettes/figures/supplemental/Figure_S1A.png"),
#   width = 13.7,
#   height = 10.2,
#   dpi = 900
# )


edges1 <- data.frame(
  from = c(1:2, 4:7, 9:14),
  to = c(2:3, 5:8, 10:15),
  color = nodesS1$delta_km[!is.na(nodesS1$delta_km)]
)

nodes_dups <- nodesS1 %>%
  dplyr::filter(!dplyr::row_number() %in% c(4, 9))

edges_for_plot1 <- edges1 %>%
  dplyr::inner_join(nodesS1 %>% dplyr::select(id, lon, lat), by = c('from' = 'id')) %>%
  dplyr::rename(x = lon, y = lat) %>%
  dplyr::inner_join(nodesS1 %>% dplyr::select(id, lon, lat), by = c('to' = 'id')) %>%
  dplyr::rename(xend = lon, yend = lat)

assertthat::assert_that(nrow(edges_for_plot1) == nrow(edges1))

fig_s1B <-
ggplot(nodesS1) +
  country_shapes1 +
  geom_curve(
    aes(
      x = x,
      y = y,
      xend = xend,
      yend = yend,
      color = color
    ),
    arrow = arrow(length = unit(0.02, "npc"), type = "closed"),
    data = edges_for_plot1,
    curvature = 0.3,
    alpha = 1.0,
    size = 1.0
  ) +
  scale_color_distiller(palette = "RdYlGn",
                        direction = 1,
                        name = "\u0394 km ED/year") +
  geom_point(
    aes(x = lon, y = lat),
    # draw nodes
    shape = 21,
    fill = 'white',
    color = 'black',
    stroke = 0.5,
    size = 3.5
  ) +
  ggrepel::geom_label_repel(
    aes(x = lon, y = lat, label = name),
    # draw text labels
    size = 2.75,
    color = "black",
    min.segment.length = 0.15,
    fill = alpha(c("white"), 0.7),
    force = 34,
    nudge_x = -0.51,
    nudge_y = 0.15,
    family = "Helvetica"
  ) +
  mapcoords1 +
  theme_void() +
  theme(
    plot.margin = grid::unit(c(2, 2, 2, 2), "mm"),
    legend.box.spacing = unit(20.0, "pt"),
    legend.key.size = unit(0.6, "cm"),
    legend.text = element_text(size = 14, family = "Helvetica"),
    legend.title = element_text(
      size = 18,
      face = "bold",
      family = "Helvetica"
    )
  )

# ggplot2::ggsave(
#   fig_s1B,
#   filename = here::here("vignettes/figures/supplemental/Figure_S1B.png"),
#   width = 13.7,
#   height = 10.2,
#   dpi = 900
# )

Figure_S1 <-
  cowplot::plot_grid(
    fig_s1A,
    fig_s1B,
    labels = "AUTO",
    label_x = 0.233,
    label_y = 0.99,
    align = "v",
    nrow = 2,
    ncol = 1
  )

Figure_S1

# ggplot2::ggsave(
#   Figure_S1,
#   filename = here::here("vignettes/figures/supplemental/Figure_S1.png"),
#   width = 13.7,
#   height = 10.2,
#   dpi = 900
# )

```

```{r TableS1, echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'asis'}

maize_spread_table <- nodesS1 %>%
  dplyr::select(-c(id, Long, Lat)) %>%
  dplyr::filter(!(SiteName == "Three Fir Shelter" & is.na(dist_deg))) %>% 
  dplyr::mutate(dplyr::across(dist_deg:delta_lat_km, ~ sprintf("%.3f", .x))) %>% 
  dplyr::relocate(LabID, .before = "SiteName") %>% 
  dplyr::arrange(grp, Date)

names(maize_spread_table) <-
  c(
    "Group",
    "Lab ID",
    "Site Name",
    "Median cal BC/AD",
    "Euclidean Distance (Degrees)",
    "Euclidean Distance (km)",
    "\u0394 Degrees Euclidean Distance/Year",
    "\u0394 km Euclidean Distance/Year",
    "\u0394 Degrees North/Year",
    "\u0394 km North/Year"
  )

# readr::write_excel_csv(maize_spread_table, here::here("vignettes/tables/supplemental/Table_S1.csv"))

knitr::kable(maize_spread_table,
             caption = "Table S1. Distances and spread rates for sites on the Maize Northern Frontier (MNF), beginning with Guilá Naquitz. However, the rates are for a split route with new rates calculated, where it splits at Three Fir Shelter into an eastern and western route. Distances and maize spread rates are between a previous row and a current row. For example, Guilá Naquitz is blank because the distance between Guilá Naquitz and Las Capas is provided in the Las Capas row. See also Figure S1.")

```

```{r Figure_S2, fig.cap="Figure S2. Distribution of the number of samples at each site in each county on the MNF.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

sw_rc <- cropSpreadR::sw_rc

#frontier_counties <- unique(maize_frontier_swus$county_st)[!grepl("delta, colorado", unique(maize_frontier_swus$county_st))]
frontier_counties <- c("sevier, utah", "daggett, utah", "catron, new mexico",
                       "navajo, arizona", "salt lake, utah", "pima, arizona",
                       "kane, utah", "montrose, colorado", "uintah, utah", "box elder, utah",
                       "san juan, utah")

ckde_test1 <- sw_rc %>%
  dplyr::filter(county_st %in% frontier_counties) %>%
  dplyr::select(SiteIDName, county_st, Age, Error, calib)

# Inspect for outlying sites (i.e., sites that have a lot of samples for when number of sites overall is relatively low).
samples_site_county <- cropSpreadR::sw_rc %>%
  dplyr::filter(county_st %in% c(frontier_counties, "delta, colorado")) %>%
  dplyr::group_by(SiteIDName, county_st, SiteName) %>%
  dplyr::summarise(count = dplyr::n()) %>%
  dplyr::group_by(county_st) %>%
  dplyr::mutate(id = dplyr::row_number(),
                county_st = tools::toTitleCase(county_st)) %>%
  ggplot(aes(x = count)) +
  geom_histogram(binwidth = 10) +
  scale_x_continuous(breaks = seq(0, 150, 25)) +
  xlab("Number of Samples/Site") +
  ylab("Count") +
  facet_wrap(. ~ county_st, ncol = 3) +
  theme_bw() +
  theme(
    strip.text.x = element_text(
      size = 14,
      family = "Helvetica",
      face = "bold"
    ),
    axis.text = element_text(
      size = 16,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_text(
      size = 18,
      family = "Helvetica",
      margin = margin(
        t = 10,
        r = 5,
        b = 10,
        l = 10
      )
    ),
    axis.title.x = element_text(
      size = 18,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    )
  )

samples_site_county

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/supplemental/Figure_S2.png"),
#   plot = samples_site_county,
#   dpi = 600
# )

```

```{r Figure_S3, fig.cap="Figure S3. The MNF relative to the SPDs for the counties in the four-state southwestern United States that are on the MNF through time (see Gillreath-Brown and Kohler n.d., Figure 6 and site locations in Figure 2). Red points locate the earliest known maize in each county. Sample sizes for the SPDs shown in parentheses. Note variability in scales on y axis.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

MDB3A <- MDB %>%
  dplyr::filter(Country == "USA") %>%
  dplyr::select(LabID, Long, Lat, SiteIDName, MedianBP) %>%
  dplyr::mutate(SiteIDName = ifelse(SiteIDName == "Willard Mounds", "42BO3", SiteIDName))

maize_oldest <- maize_frontier_swus %>% 
  sf::st_drop_geometry() %>%
  dplyr::mutate(SiteNameID = dplyr::coalesce(SiteName, SiteID)) %>% 
  dplyr::select(LabID, county_st, SiteNameID, Date, MedianBP) %>% 
  dplyr::mutate(date_maize = Date) %>% 
  dplyr::group_by(county_st) %>% 
  dplyr::slice_min(Date) %>% 
  dplyr::ungroup() %>% 
  dplyr::arrange(Date) %>% 
  dplyr::mutate(county_st = tools::toTitleCase(county_st))

county_centroid <- cropSpreadR::SWUS_counties %>%
  sf::st_drop_geometry() %>%
  dplyr::select(county_st, centroid_lat) %>%
  dplyr::mutate(county_st = tools::toTitleCase(county_st))

by_county_changes <- cropSpreadR::SPD_county %>%
  dplyr::filter(county_st %in% maize_frontier_swus$county_st) %>%
  dplyr::mutate(county_st = tools::toTitleCase(county_st)) %>%
  dplyr::left_join(maize_oldest, ., by = "county_st") %>%
  dplyr::left_join(
    .,
    cropSpreadR::SWUS_counties %>%
      sf::st_drop_geometry() %>%
      dplyr::select(county_st, centroid_lat) %>%
      dplyr::mutate(county_st = tools::toTitleCase(county_st)),
    by = "county_st"
  ) %>%
  dplyr::arrange(desc(centroid_lat)) %>%
  # Remove all columns with word site, except for SiteNameID, and remove name and region columns.
  dplyr::select(-c(starts_with("Site", vars = names(.)[stringr::str_detect(names(.), "SiteNameID", negate = TRUE)]), NAME, REGION, LabID)) %>%
  tidyr::unnest(cols = c(spds)) %>%
  dplyr::mutate(
    MedianBP = ifelse(MedianBP == calBP, MedianBP, NA),
    SiteNameID = ifelse(!is.na(MedianBP), SiteNameID, NA)
  ) %>%
  dplyr::mutate(county_st = paste0(county_st, " (n = ", n, ")")) %>% 
  dplyr::mutate(date_maize = rcarbon::BPtoBCAD(MedianBP),
                date_spd = rcarbon::BPtoBCAD(calBP)) %>% 
  dplyr::mutate(SiteNameID = ifelse(SiteNameID == "Elsinore Burial", "42SV2111", SiteNameID))


spd_plot <-
  ggplot(data = by_county_changes, aes(y = PrDens, x = date_spd)) +
  geom_line() +
  geom_point(
    aes(y = PrDens, x = date_maize),
    size = 2.75,
    color = "red",
    inherit.aes = FALSE
  ) +
  ggrepel::geom_label_repel(
    aes(y = PrDens, x = date_maize, label = SiteNameID),
    color = "red",
    cex = 5,
    show.legend  = F,
    min.segment.length = unit(0.1, 'lines'),
    box.padding = 1,
    nudge_x = -550,
    nudge_y = 0.35,
    fill = alpha(c("white"), 0.6),
    label.size = NA
  ) +
  theme_bw() +
  scale_x_continuous(breaks = seq(-4000, 2000, 1000), limits = c(-4100, 2000)) +
  scale_y_continuous(
    breaks = function(x)
      seq(0, max(x), length.out = 3),
    labels = function(y)
      sprintf("%.2f", y)
  ) +
  facet_wrap( ~ factor(county_st, levels = unique(by_county_changes$county_st)),
              scales = "free_y",
              ncol = 1) +
  xlab("Years BC/AD") +
  ylab("Summed Probability") +
  theme(
    legend.position = "none",
    panel.spacing = unit(1, "lines"),
    strip.text.x = element_text(
      size = 16,
      family = "Helvetica",
      face = "bold"
    ),
    axis.ticks.length = unit(0.125, "inch"),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_text(
      size = 20,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 10,
        r = 5,
        b = 10,
        l = 10
      )
    ),
    axis.title.x = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    ),
    panel.grid.minor = element_blank(),
    axis.line = element_line(colour = "black"),
    plot.margin = margin(
      t = 5,
      r = 25,
      b = 5,
      l = 5
    )
  )

spd_plot

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/supplemental/Figure_S3.png"),
#   plot = spd_plot,
#   width = 16,
#   height = 18,
#   dpi = 600
# )

```

```{r Figure_S4, fig.cap="Figure S4. The MNF sites (red points) relative to the composite kernel density estimates (CKDE) for the counties in the four-state southwestern United States, which contain the MNF sites (see Gillreath-Brown and Kohler n.d., Figure 6 and site locations in Figure 2). Sample sizes for the CKDEs shown in parentheses. Note variability in scales on y axis. We do not include Delta County, Colorado as the sample size was small and dominated by samples from Eagle Rock Shelter (5DT813, containing 51 of the 67 samples in the county).", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

sw_rc <- cropSpreadR::sw_rc

#frontier_counties <- unique(maize_frontier_swus$county_st)[!grepl("delta, colorado", unique(maize_frontier_swus$county_st))]
frontier_counties <- c("sevier, utah", "daggett, utah", "catron, new mexico",
                       "navajo, arizona", "salt lake, utah", "pima, arizona",
                       "kane, utah", "montrose, colorado", "uintah, utah", "box elder, utah",
                       "san juan, utah")

ckde_test1 <- sw_rc %>%
  dplyr::filter(county_st %in% frontier_counties) %>%
  dplyr::select(SiteIDName, county_st, Age, Error, calib)

sites_to_subsample <- c("42BO1", "42SV1686")

ckde_test2 <- ckde_test1 %>%
  dplyr::filter(SiteIDName %in% sites_to_subsample) %>%
  dplyr::group_by(SiteIDName) %>%
  dplyr::slice_sample(prop = 0.2, replace = TRUE) %>%
  dplyr::bind_rows(ckde_test1 %>% 
                     dplyr::filter(!SiteIDName %in% sites_to_subsample),
                   .,
  ) %>%
  dplyr::select(-SiteIDName) %>%
  dplyr::group_by(county_st) %>%
  dplyr::mutate(n = dplyr::n()) %>%
  dplyr::group_by(n, .add = TRUE) %>%
  tidyr::nest() %>%
  dplyr::mutate(ckde = purrr::map(
    data,
    ~ cropSpreadR::calc_ckde_dates(
      .x,
      interval = 0.95,
      timeRange_max = 9000,
      timeRange_min = 0
    )
  )) %>%
  dplyr::select(-data) %>%
  tidyr::unnest(ckde)

maize_oldest <- maize_frontier_swus %>% 
  sf::st_drop_geometry() %>%
  dplyr::mutate(SiteNameID = dplyr::coalesce(SiteName, SiteID)) %>% 
  dplyr::select(county_st, SiteNameID, Date) %>% 
  dplyr::mutate(date_maize = Date) %>% 
  dplyr::group_by(county_st) %>% 
  dplyr::slice_min(Date) %>% 
  dplyr::ungroup() %>% 
  dplyr::arrange(Date)

ckde_test3 <- ckde_test2 %>%
  dplyr::left_join(., maize_oldest, by = c("county_st", "Date")) %>%
  dplyr::mutate(county_st2 = county_st,
                county_st = tools::toTitleCase(county_st),
                county_st = paste0(county_st, " (n = ", n, ")")) %>%
  dplyr::mutate(SiteNameID = ifelse(SiteNameID == "Elsinore Burial", "42SV2111", SiteNameID))

county_order <- cropSpreadR::SWUS_counties %>%
  sf::st_drop_geometry() %>%
  dplyr::select(county_st2 = county_st, centroid_lat) %>%
  dplyr::arrange(desc(centroid_lat)) %>%
  dplyr::filter(county_st2 %in% frontier_counties) %>%
  dplyr::left_join(., ckde_test3 %>% dplyr::ungroup() %>% dplyr::select(county_st, county_st2) %>% dplyr::distinct(), by = "county_st2") %>%
  dplyr::pull(county_st)

ckde_plot <-
  ggplot(data = ckde_test3, aes(y = avg, x = Date)) +
  geom_ribbon(aes(ymin=lo, ymax=hi),
              alpha=0.4,
              fill="grey70") +
  geom_line() +
  geom_point(
    aes(y = avg, x = date_maize),
    size = 2.75,
    color = "red",
    inherit.aes = FALSE
  ) +
  ggrepel::geom_label_repel(
    aes(y = avg, x = date_maize, label = SiteNameID),
    color = "red",
    cex = 5,
    show.legend  = F,
    min.segment.length = unit(0.1, 'lines'),
    box.padding = 1,
    nudge_x = -550,
    nudge_y = 0.35,
    fill = alpha(c("white"), 0.6),
    label.size = NA
  ) +
  theme_bw() +
  scale_x_continuous(breaks = seq(-4000, 2000, 1000), limits = c(-4100, 2000)) +
  scale_y_continuous(
    breaks = function(x)
      seq(0, max(x), length.out = 3),
    labels = function(y)
      sprintf("%.4f", y)
  ) +
  facet_wrap( ~ factor(county_st, levels = county_order),
              scales = "free_y",
              ncol = 1) +
  xlab("Years BC/AD") +
  ylab("Composite Kernel Density Estimates") +
  theme(
    legend.position = "none",
    panel.spacing = unit(1, "lines"),
    strip.text.x = element_text(
      size = 16,
      family = "Helvetica",
      face = "bold"
    ),
    axis.ticks.length = unit(0.125, "inch"),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_text(
      size = 20,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 10,
        r = 5,
        b = 10,
        l = 10
      )
    ),
    axis.title.x = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    ),
    panel.grid.minor = element_blank(),
    axis.line = element_line(colour = "black"),
    plot.margin = margin(
      t = 5,
      r = 25,
      b = 5,
      l = 5
    )
  )

ckde_plot

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/Figure_S4.png"),
#   plot = ckde_plot,
#   width = 16,
#   height = 18,
#   dpi = 900
# )

```

```{r Figure_S5, fig.cap="Figure S5. The MNF relative to the random mark permutation test for SPDs for the counties in the four-state southwestern United States that are on the MNF through time (see Gillreath-Brown and Kohler n.d., Figure 6 and site locations in Figure 2). Red points locate the earliest known maize in each county. Sample sizes for the SPDs shown in parentheses and the global p value, which means that most of the data falls outside of the envelope. Note variability in scales on y axis.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

sw_rc <- cropSpreadR::sw_rc

#frontier_counties <- unique(maize_frontier_swus$county_st)[!grepl("delta, colorado", unique(maize_frontier_swus$county_st))]
frontier_counties <-
  c(
    "sevier, utah",
    "daggett, utah",
    "catron, new mexico",
    "navajo, arizona",
    "salt lake, utah",
    "pima, arizona",
    "kane, utah",
    "montrose, colorado",
    "uintah, utah",
    "box elder, utah",
    "san juan, utah"
  )

sw_rc_subset <- sw_rc %>%
  dplyr::filter(county_st %in% frontier_counties)

cal.emedyd = rcarbon::calibrate(x = sw_rc_subset$Age,
                                errors = sw_rc_subset$Error,
                                normalised = FALSE)

nsim = 100

perm.emedyd = rcarbon::permTest(
  x = cal.emedyd,
  marks = sw_rc_subset$county_st,
  timeRange = c(7000, 0),
  nsim = nsim,
  runm = 50
)

pvalues <- data.frame(pvalue = perm.emedyd[["pValueList"]]) %>%
  tibble::rownames_to_column("county_st")

sample_size <- dplyr::bind_rows(perm.emedyd[["metadata"]]) %>%
  tidyr::pivot_longer(everything(), names_to = "county_st", values_to = "n") %>%
  dplyr::distinct()

perm_spd_output <-
  dplyr::bind_cols(dplyr::bind_rows(perm.emedyd[["observed"]], .id = "county_st"),
                   dplyr::bind_rows(lapply(perm.emedyd[["envelope"]], as.data.frame))) %>%
  dplyr::rename(lo = V1, hi = V2) %>%
  dplyr::mutate(Date = rcarbon::BPtoBCAD(calBP)) %>%
  tibble::tibble() %>%
  dplyr::left_join(., pvalues, by = "county_st") %>%
  dplyr::left_join(., sample_size, by = "county_st") %>%
  dplyr::mutate(
    pvalue = ifelse(pvalue < 0.01, paste0("p < 0.01"), paste0("p = ", sprintf("%.2f", pvalue))),
    boom_bust = dplyr::case_when(
      PrDens < hi & PrDens > lo ~ "medium",
      PrDens > hi ~ "indianred",
      PrDens < lo ~ "royalblue",
      TRUE ~ "medium"
    )
  )

maize_oldest <- maize_frontier_swus %>%
  sf::st_drop_geometry() %>%
  dplyr::mutate(SiteNameID = dplyr::coalesce(SiteName, SiteID)) %>%
  dplyr::select(county_st, SiteNameID, Date) %>%
  dplyr::mutate(date_maize = Date) %>%
  dplyr::group_by(county_st) %>%
  dplyr::slice_min(Date) %>%
  dplyr::ungroup() %>%
  dplyr::arrange(Date) %>%
  dplyr::filter(county_st %in% frontier_counties)

ckde_test3 <- perm_spd_output %>%
  dplyr::left_join(., maize_oldest, by = c("county_st", "Date")) %>%
  dplyr::mutate(
    county_st2 = county_st,
    county_st = tools::toTitleCase(county_st),
    county_st = paste0(county_st, " (n = ", n, "; ",  pvalue, ")")
  ) %>%
  dplyr::mutate(SiteNameID = ifelse(SiteNameID == "Elsinore Burial", "42SV2111", SiteNameID))

county_order <- cropSpreadR::SWUS_counties %>%
  sf::st_drop_geometry() %>%
  dplyr::select(county_st2 = county_st, centroid_lat) %>%
  dplyr::arrange(desc(centroid_lat)) %>%
  dplyr::filter(county_st2 %in% frontier_counties) %>%
  dplyr::left_join(
    .,
    ckde_test3 %>% dplyr::ungroup() %>% dplyr::select(county_st, county_st2) %>% dplyr::distinct(),
    by = "county_st2"
  ) %>%
  dplyr::pull(county_st)

condensed <- perm_spd_output %>%
  dplyr::group_by(county_st) %>%
  dplyr::mutate(change = cumsum(boom_bust != lag(boom_bust, def = first(boom_bust)))) %>%
  dplyr::group_by(change, .add = TRUE) %>%
  dplyr::filter(dplyr::row_number() %in% c(1, dplyr::n()) &
                  boom_bust != "medium") %>%
  dplyr::mutate(time = dplyr::row_number()) %>% 
  dplyr::ungroup() %>% 
  tidyr::complete(county_st, change, time) %>%
  tidyr::fill(dplyr::everything()) %>% 
  dplyr::select(-time) %>% 
  dplyr::group_by(grp = stringr::str_c('Column', rep(1:2, length.out = dplyr::n()))) %>%
  dplyr::mutate(rn = dplyr::row_number()) %>%
  dplyr::ungroup() %>%
  dplyr::select(county_st, Date, boom_bust, grp, rn) %>%
  tidyr::pivot_wider(names_from = grp, values_from = Date) %>%
  dplyr::select(-rn,
                xmin = Column1,
                xmax = Column2,
                fill = boom_bust) %>%
  dplyr::mutate(ymin = -Inf, ymax = Inf) %>%
  dplyr::left_join(
    .,
    ckde_test3 %>% dplyr::ungroup() %>% dplyr::select(county_st, county_st2) %>% dplyr::distinct(),
    by = c("county_st" = "county_st2")
  ) %>%
  dplyr::select(-county_st, county_st = county_st.y) %>%
  dplyr::mutate(county_st = factor(county_st, levels = county_order))

perm_spd_plot <-
  ggplot(data = ckde_test3, aes(y = PrDens, x = Date)) +
  geom_ribbon(aes(ymin = lo, ymax = hi),
              alpha = 0.4,
              fill = "grey70") +
  geom_rect(
    aes(
      xmin = xmin,
      xmax = xmax,
      ymin = ymin,
      ymax = ymax,
      fill = I(fill)
    ),
    alpha = 0.4,
    inherit.aes = FALSE,
    data = condensed
  ) +
  geom_line() +
  geom_point(
    aes(y = PrDens, x = date_maize),
    size = 2.75,
    color = "red",
    inherit.aes = FALSE
  ) +
  ggrepel::geom_label_repel(
    aes(y = PrDens, x = date_maize, label = SiteNameID),
    color = "red",
    cex = 5,
    show.legend  = F,
    min.segment.length = unit(0.1, 'lines'),
    box.padding = 1,
    nudge_x = -550,
    nudge_y = 0.35,
    fill = alpha(c("white"), 0.6),
    label.size = NA
  ) +
  theme_bw() +
  scale_x_continuous(breaks = seq(-4000, 2000, 1000),
                     limits = c(-4100, 2000)) +
  scale_y_continuous(
    breaks = function(x)
      seq(0, max(x), length.out = 3),
    labels = function(y)
      sprintf("%.2f", y)
  ) +
  facet_wrap(~ factor(county_st, levels = county_order),
             scales = "free_y",
             ncol = 1) +
  xlab("Years BC/AD") +
  ylab("Summed Probability") +
  theme(
    legend.position = "none",
    panel.spacing = unit(1, "lines"),
    strip.text.x = element_text(
      size = 16,
      family = "Helvetica",
      face = "bold"
    ),
    axis.ticks.length = unit(0.125, "inch"),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    axis.text = element_text(
      size = 20,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 10,
        r = 5,
        b = 10,
        l = 10
      )
    ),
    axis.title.x = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    ),
    panel.grid.minor = element_blank(),
    axis.line = element_line(colour = "black"),
    plot.margin = margin(
      t = 5,
      r = 25,
      b = 5,
      l = 5
    )
  )

perm_spd_plot

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/supplemental/Figure_S5.png"),
#   plot = perm_spd_plot,
#   width = 16,
#   height = 18,
#   dpi = 600
# )

```

```{r Figure_S6, fig.cap="Figure S6. Box plots showing elevation, GDD, and latitude through time for the earliest maize sample from each of 275 sites in our database (Gillreath-Brown and Kohler n.d. 2022), excluding 19 sites post-dating AD 1500 or south of 30° Latitude. Samples are binned by 100-year periods and plotted on the midpoint of each bin. This is the same as Figure 5 except here we bin the data into 100 year time bins.", echo = FALSE, warning = FALSE, message = FALSE, error = FALSE, results = 'hide',fig.keep = 'all', fig.align = "left"}

MDB_env2_Figure_S6 <- MDB_env %>%
  dplyr::filter(Lat >= 30 & Date < 1500) %>%
  dplyr::group_by(SiteIDName) %>%
  dplyr::slice_min(Date, with_ties = FALSE) %>%
  dplyr::group_by(gr = cut(
    Date,
    breaks = seq(-4000, 1500, 100),
    labels = seq(-3950, 1450, 100)
  )) %>%
  dplyr::mutate(date2 = as.numeric(as.character(gr))) %>%
  dplyr::ungroup() %>%
  dplyr::select(-gr) %>%
  dplyr::arrange(date2) %>%
  dplyr::rename(Latitude = Lat, Elevation = elevation)

env_summaries_Figure_S6 <- MDB_env2_Figure_S6 %>%
  dplyr::summarise(dplyr::across(c(Latitude, GDD, Elevation), ~ mean(.x, na.rm = TRUE)))

Figure_S6 <-
  MDB_env2_Figure_S6 %>%
  dplyr::select(date2, Latitude, GDD, Elevation) %>%
  dplyr::rename(
    `Latitude (°)` = Latitude,
    `Annual GDD (°C)` = GDD,
    `Elevation (m)` = Elevation
  ) %>%
  tidyr::pivot_longer(-date2) %>%
  ggplot(aes(
    x = factor(date2),
    y = value,
    fill = factor(name)
  )) +
  geom_boxplot() +
  facet_grid(vars(name), scales = "free") +
  labs(fill = "") +
  xlab("Years BC/AD") +
  ylab("Value") +
  theme_bw() +
  theme(
    strip.text.y = element_text(size = 12),
    legend.key.size = unit(1.5, "cm"),
    legend.text = element_text(size = 12),
    axis.ticks.length = unit(0.125, "inch"),
    axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1),
    axis.text = element_text(
      size = 26,
      colour = "black",
      family = "Helvetica"
    ),
    axis.title.y = element_blank(),
    axis.title.x = element_text(
      size = 28,
      family = "Helvetica",
      margin = margin(
        t = 5,
        r = 10,
        b = 10,
        l = 10
      )
    ),
    axis.line = element_line(colour = "black"),
    plot.margin = margin(
      t = 5,
      r = 25,
      b = 5,
      l = 5
    )
  )

Figure_S6

# ggplot2::ggsave(
#   filename = here::here("vignettes/figures/supplemental/Figure_S6.png"),
#   plot = Figure_S6,
#   width = 15.51,
#   height = 10.32,
#   dpi = 600
# )

```