knitr::opts_chunk$set(echo = FALSE, message = FALSE, warning = FALSE, fig.align = "centre", fig.width = 16.6/2.54)
htmltools::img(src = knitr::image_uri("../man/figures/hexsticker.png"), alt = 'logo', style = paste('position:absolute', 'top:0', 'right:0', 'padding:10px', 'width:150px', 'height:160px', 'background:none', 'border:none', 'box-shadow:none', sep = ";"))
knitr::include_graphics("schematic.png")
ThermStats
is designed for biologists using thermography to quantify thermal heterogeneity. It uses the Thermimage
package [@tattersall_thermimage:_2017] to batch process data from FLIR thermal cameras, and takes inspiration from FRAGSTATS [@fragstats_2012], SDMTools
[@sdmtools_2014], @faye_toolbox_2016 and @shi_framework_2016 to facilitate the calculation of various metrics of thermal heterogeneity for any gridded temperature data.
The package is available to download from GitHub using devtools
:
devtools::install_github("rasenior/ThermStats")
library(ThermStats)
Once loaded, the code below can be followed step-by-step.
Data are extracted from FLIR images using batch_extract
. This is a batch implementation of the readflirJPG
function from Thermimage
. It requires only the path to the directory of FLIR thermal images, and the freely available external software 'ExifTool'. Besides raw data, this step also retrieves camera-specific calibration parameters which are required later to convert raw data to temperature values.
# Batch extract thermal images included in ThermStats installation flir_raw <- batch_extract(in_dir = system.file("extdata", package = "ThermStats"), write_results = FALSE)
Raw data are encoded in each thermal image as a 16 bit analog-to-digital signal, which represents the radiance received by the infrared sensor. The function batch_convert
converts these raw data to temperature values using equations from infrared thermography, via a batch implementation of the function raw2temp
in Thermimage
. It uses the calibration constants extracted in batch_extract
and environmental parameters defined by the user:
# Define raw data raw_dat <- flir_raw$raw_dat # Define camera calibration constants dataframe camera_params <- flir_raw$camera_params # Define metadata metadata <- flir_metadata # Create vector denoting the position of each photo within metadata photo_index <- match(names(raw_dat), metadata$photo_no) # Batch convert flir_converted <- batch_convert( raw_dat = raw_dat, # Emissivity = mean of range in Scheffers et al. 2017 E = mean(c(0.982,0.99)), # Object distance = hypotenuse of right triangle where # vertical side is 1.3 m (breast height) & angle down is 45° OD = (sqrt(2))*1.3, # Apparent reflected temperature & atmospheric temperature = # atmospheric temperature measured in the field RTemp = metadata$atm_temp[photo_index], ATemp = metadata$atm_temp[photo_index], # Relative humidity = relative humidity measured in the field RH = metadata$rel_humidity[photo_index], # Calibration constants from 'batch_extract' PR1 = camera_params[,"PlanckR1"], PB = camera_params[,"PlanckB"], PF = camera_params[,"PlanckF"], PO = camera_params[,"PlanckO"], PR2 = camera_params[,"PlanckR2"], # Whether to write results or just return write_results = FALSE)
Statistics can be calculated for individual thermal images (in a matrix or raster format), or across multiple images within a specified grouping. The latter is useful for sampling designs where multiple images are collected at each sampling event to capture temperature across a wider sampling unit, such as a plot. In either case, statistics can include summary statistics specified by the user -- for example, mean, minimum and maximum -- as well as thermal connectivity [based on the climate connectivity measure of @mcguire_achieving_2016] and spatial statistics for hot and cold spots, identified using the G* variant of the Getis-Ord local statistic [@getis_local_1996].
For an individual image, get_stats
requires the user to specify the image and the desired statistics. Statistics can be calculated for geographic temperature data, in which case the user should also define the extent and projection of the data.
flir_stats <- get_stats( # The temperature dataset img = flir_converted$`8565`, # The ID of the dataset id = "8565", # Whether or not to calculate thermal connectivity calc_connectivity = FALSE, # Whether or not to identify hot and cold spots patches = TRUE, # The image projection (only relevant for geographic data) img_proj = NULL, # The image extent (only relevant for geographic data) img_extent = NULL, # The data to return return_vals = c("df", # Temperature data as dataframe "patches", # Patch outlines "pstats"), # Patch statistics dataframe # The summary statistics of interest sum_stats = c("median", "SHDI", "perc_5", "perc_95"))
For grouped images, stats_by_group
requires the user to supply a list of matrices or a raster stack, and (optionally) the metadata and the name of the variable in the metadata that defines the grouping. Table 1 shows the metadata used in the code snippet, where photo number ('photo_no') defines individual temperature matrices, and the replicate identity ('rep_id') defines the grouping of photos. There are two replicates, 'T7P1' and 'T7P2', and each has two associated photos.
tab1 <- " Table: Table 1: Example metadata denoting the grouping ('rep_id') of different thermal images. Statistics can be calculated over multiple images within a group, using the function `stats_by_group`. | photo_no|rep_id | atm_temp| rel_humidity| |--------:|:------|--------:|------------:| | 8565|T7P1 | 24.00| 96| | 8583|T7P1 | 24.00| 96| | 8589|T7P2 | 23.25| 98| | 8613|T7P2 | 23.50| 96| " cat(tab1)
flir_stats_group <- stats_by_group( # A dataframe denoting the grouping metadata = metadata, # List of images img_list = flir_converted, # Variable denoting the ID of unique images idvar = "photo_no", # Variable denoting the grouping grouping_var = "rep_id", # Desired precision of data round_val = 0.5, # The data to return return_vals = c("df", # Temperature data as dataframe "patches", # Patch outlines "pstats"), # Patch statistics dataframe # The summary statistics of interest sum_stats = c("median", "SHDI", "perc_5", "perc_95"))
By default, both get_stats
and stats_by_group
return a dataframe with patch statistics (Table 2) for each image or group, respectively.
tab2 <- " Table: Table 2: A snippet of hot spot patch statistics returned by `stats_by_group`, which implements `get_stats` within groups. | img_median| img_perc_5| img_perc_95| img_SHDI| hot_shape_index| hot_aggregation| |----------:|----------:|-----------:|------------:|---------------:|---------------:| | 23.5| 23| 24.5| 1.16| 7.54| 0.895| | 24.0| 23| 25.0| 1.68| 7.80| 0.855| " cat(tab2)
In addition to patch statistics, get_stats
can return (1) the temperature dataset in a dataframe format, and (2) a SpatialPolygonsDataFrame
of its hot and cold spots. The function plot_patches
can then recreate the original thermal image overlaid with outlines of hot and cold spots, as well as the temperature distribution if plot_distribution = TRUE
(Figure 2).
plot_patches( # The raw temperature data df = flir_stats$df, # The patch outlines patches = flir_stats$patches)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.