Non-axis-aligned rectangle stand from GPS data"
In SamsaRaLight: Simulate Tree Light Transmission Using Ray-Tracing

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

library(SamsaRaLight)
library(dplyr)

Introduction

In previous tutorials, inventories were assumed to be defined directly in a local Cartesian coordinate system (meters). In practice, however, forest inventories are often collected with GPS data.

In this vignette, we how to convert an inventory based on GPS coordinates where tree positions are given as longitude and latitude. We will convert the geographic coordinates into planar and axis-align the inventory zone with the virtual plot axes.

Context and data

We first use the example inventory IRRES1, stored in the package as SamsaRaLight::data_IRRES1. This inventory was collected in Belgian Ardennes by Gauthier Ligot in the scope of the IRRES project, which investigates the transition from even-aged to uneven-aged forest management. The stand is dense in a sloppy terrain and is mainly composed of Norway spruce and Douglas-fir, with a coppice stool of beech at its center and a few silver fir and larch trees.

trees_irres <- SamsaRaLight::data_IRRES1$trees

str(trees_irres)

Running check_inventory() on this dataset immediately fails, with an error indicating that columns x and y are missing. This is expected: tree positions are provided as longitude (lon) and latitude (lat), expressed in degrees. To illustrate why this is problematic, we (incorrectly) rename longitude and latitude to x and y and attempt to plot the inventory. It leads to a meaningless plot as angular coordinates (degrees) are incompatible with crown dimensions expressed in meters.

SamsaRaLight::plot_inventory(
  trees_irres %>% rename(x = lon, y = lat)
)

Convert the coordinates

Before creating a stand, coordinates must be converted into a planar Cartesian system with metric units. Tree coordinates can be converted from the global WGS84 reference system (EPSG:4326) to a projected coordinate system expressed in meters such as the Universal Transverse Mercator (UTM) system. It is a grid-based, metric coordinate system mapping the Earth using 60 longitudinal zones (each of them being 6-degree between 80°S and 84°N latitude), for each of the two hemispheres North and South.

The EPSG needed to convert coordinates depends on the plot coordinates. However, the UTM zone can be automatically inferred from the mean longitude ($zone = floor((lon + 180) / 6) + 1$ and hemisphere inferred from the mean latitude ($hemisphere = 32600$ if latitude is positive or $hemisphere = 32700$ if latitude is negative). Thus, EPSG code can be automatically computed as $EPSG = hemisphere + zone$. The function SamsaRaLight::create_xy_from_lonlat() allows to automatically convert a data.frame containing lon/lat coordinates into planar XY coordinates determining the appropriate UTM system.

trees_irres_xy <- SamsaRaLight::create_xy_from_lonlat(trees_irres)

str(trees_irres_xy)

After this conversion, tree positions are expressed in meters and the inventory can now be validated and visualized correctly.

SamsaRaLight::check_inventory(trees_irres_xy$df)
plot_inventory(trees_irres_xy$df)

Define the inventory zone

In the IRRES1 example dataset, the trees were inventoried within a rectangular inventory zone. However, the vertices are also expressed in a lon/lat coordinate system and therefore need to be converted.

polygon_irres_xy <- SamsaRaLight::create_xy_from_lonlat(
  SamsaRaLight::data_IRRES1$core_polygon
)

str(polygon_irres_xy)

We can verify that the trees and the inventory polygon are both expressed in metres within the same coordinate system. To do so, we can use the same plot_inventory() function as above but adding the core polygon data.frame as a second argument.

SamsaRaLight::plot_inventory(
  trees_irres_xy$df,
  polygon_irres_xy$df
)

We can use the function SamsaRaLight::check_polygon() to check that the core polygon is geometrically correct and encompasses all the inventoried trees. If it does not, the function tries to correct it by making minimal changes, such as converting the polygon into a valid one (e.g. if the vertices are not in the correct order) or adding a small buffer to the polygon in an attempt to include all the trees (e.g. if some trees are close to the border, small rounding errors can lead to the polygon excluding them computationally). Thus, the function returns the minimally corrected polygon and specifies this with a message if the polygon has been modified; otherwise, it throws an error.

polygon_irres_xy$df <- SamsaRaLight::check_polygon(
  polygon_irres_xy$df,
  trees_irres_xy$df
)

Create the virtual stand

Fortunately, the SamsaRaLight package allows you to provide both tree inventory and core polygon tables with only longitude/latitude coordinates to the create_sl_stand() function, which automatically performs system conversions. The stand creation process will also handle coordinate shifts into a relative coordinate system starting at 0.

Because the projected coordinates follow a conventional GIS orientation (Y axis pointing North), we set north2x = 90, meaning that geographic North corresponds to the positive Y direction.

stand_irres <- SamsaRaLight::create_sl_stand(
  trees_inv = SamsaRaLight::data_IRRES1$trees,
  cell_size = 5,

  latitude = SamsaRaLight::data_IRRES1$info$latitude,
  slope    = SamsaRaLight::data_IRRES1$info$slope,
  aspect  = SamsaRaLight::data_IRRES1$info$aspect,
  north2x = 90,

  core_polygon_df = SamsaRaLight::data_IRRES1$core_polygon
)

plot(stand_irres)

The stand dimensions are chosen as the smallest grid (in number of cells) that fully contains the inventory zone:

stand_irres$geometry$n_cells_x
stand_irres$geometry$n_cells_y

This corresponds to a stand size of:

stand_irres$geometry$n_cells_x * stand_irres$geometry$cell_size
stand_irres$geometry$n_cells_y * stand_irres$geometry$cell_size

Then, tree coordinates are shifted to a local coordinate system starting at zero:

stand_irres$transform$shift_x
stand_irres$transform$shift_y

Set the axis-aligned rectangle option

At this stage, the inventory is not yet axis-aligned. This is not a technical issue with this package, and the light computation can be run using the virtual non-axis-aligned stand created above. However, as can be seen in the above plot, the area surrounding the rectangular inventory zone is empty, which could affect light interception. Therefore, in most cases, it is preferable to work with a rectangle aligned with the simulation axes. To do so, we have to recreate the virtual stand by setting modify_polygon = "aarect" (for axis-aligned rectangle), which:

compute the minimum bounding rectangle of the inventory polygon (in this case, as our inventory zone is already a rectangle, it does not change anything),
rotate the entire stand (trees and polygon) so that this rectangle becomes axis-aligned (the rotation counter-clockwise in degrees applied to the stand is stored internally in transform$rotation_ccw$)
update the north2x value accordingly (and can be seen in the compass of the plot() function)

stand_irres_aarect <- SamsaRaLight::create_sl_stand(
  trees_inv = SamsaRaLight::data_IRRES1$trees,
  cell_size = 5,

  latitude = SamsaRaLight::data_IRRES1$info$latitude,
  slope    = SamsaRaLight::data_IRRES1$info$slope,
  aspect  = SamsaRaLight::data_IRRES1$info$aspect,
  north2x = 90,

  core_polygon_df = SamsaRaLight::data_IRRES1$core_polygon,
  modify_polygon = "aarect"
)

plot(stand_irres_aarect)

As we can see, rotating the stand results in a rectangular inventory zone (shown in yellow) that may not cover the entire virtual stand area. This creates empty spaces around the borders of the virtual stand and reduces the total basal area per hectare (due to a larger area with the same number of trees). This could slightly bias the light computation, even though the small empty areas on the borders could be negligible for tree light interception. This can also be avoided by:

Setting the cell size to a smaller value to reduce the empty space, which would result in much higher computation time.
Alternatively, the rectangle could be set up without being axis-aligned and the 'fill_around' argument could be used. This will be explained in the second and third examples of this tutorial and involves filling around the inventory zone with virtual trees, but introduces stochasticity to the stand virtualisation.

Run SamsaraLight

As shown in the previous tutorials, monthly radiation data are retrieved using the geographic location of the stand.

data_radiations_irres <- SamsaRaLight::get_monthly_radiations(
  latitude  = SamsaRaLight::data_IRRES1$info$latitude,
  longitude = SamsaRaLight::data_IRRES1$info$longitude
)

And the simulation is run using run_sl() (here run with the axis-aligned inventory zone).

output_irres_aarect <- SamsaRaLight::run_sl(
  sl_stand = stand_irres_aarect,
  monthly_radiations = data_radiations_irres
)

plot(output_irres_aarect)