Sampling"
In forestmangr: Forest Mensuration and Management

knitr::opts_chunk$set(collapse = T, comment = "#>")
knitr::opts_chunk$set(fig.width=7, fig.height=5)
options(tibble.print_min = 6L, tibble.print_max = 6L)
library(forestmangr)

First we'll load up the package and some data:

library(forestmangr)
data(exfm1)
data(exfm2)
data(exfm3)
data(exfm4)
data(exfm5)

data_acs_pilot <- as.data.frame(exfm3)
data_acs_def <- as.data.frame(exfm4)
data_ace_pilot <- as.data.frame(exfm1)
data_ace_def <- as.data.frame(exfm2)
data_as <- as.data.frame(exfm5)

Simple Random Sampling

The objective of this example is to survey an area of 46.8 ha using the simple random sampling method. The aimed error is 20%. 10 plots of 3000 m² each were measured for a pilot inventory. The data collected is shown below:

data_acs_pilot

Now we'll calculate the inventory variables for a 20% error, considering a finite population with the sprs function. Area values must be inserted in square meters, and total area values must be in hectares:

sprs(data_acs_pilot, "VWB", 3000, 46.8,error = 20, pop = "fin")

With these results, we can see that in order to meet the desired error, we'll need 15 more samples. After a new survey was done, this are the new data:

data_acs_def

Now the definitive inventory can be done:

sprs(data_acs_def, "VWB", 3000, 46.8, error = 20, pop = "fin")

The desired error was met.

The area values can also be inserted as variables:

sprs(data_acs_def, "VWB", "PLOT_AREA", "TOTAL_AREA", 
     error = 20, pop = "fin")

It's also possible to run multiple simple random sampling inventories. To demonstrate this, we'll use the example dataset for stratified sampling, but running simple random statistics. We'll still use the sprs function, but use the .groups argument to run a simple random sampling inventory for each stratum:

sprs(data_ace_def, "VWB", "PLOT_AREA", "STRATA_AREA",
     .groups = "STRATA" ,error = 20, pop = "fin")

Stratified Random Sampling

The objective of this example is to survey an area using the stratified random sampling method. The area was divided into 3 strata: one with 14.4 ha and 7 plots, another with 16.4 ha and 8 plots, and another with 14.2 ha and 7 plots. The plots have an area of 1000 square meters. In total, 22 plots were sampled for the pilot inventory. The data is shown below:

data_ace_pilot

We'll calculate the statistics with an aimed error of 5%, considering a finite population using the strs function. Area values can be inserted as a numeric vector, or as a variable. The plot area must be inserted in square meters, and strata area must be in hectares:

strs(data_ace_pilot, "VWB", 3000, c(14.4, 16.4, 14.2), 
     strata = "STRATA", error = 5, pop = "fin")

Analyzing the first table, we can see that in order to achieve the desired error, we must sample 24 additional plots. 4 in stratum 1, 8 in stratum 2 and 12 in stratum 3.

After a new survey, the new data is shown below:

data_ace_def

Now we'll run the inventory again, this time with the definitive data:

strs(data_ace_def, "VWB", "PLOT_AREA", "STRATA_AREA", 
     strata = "STRATA", error = 5, pop = "fin")

The desired error was met.

Systematic Sampling

Now we'll survey an area of 18 hectares in which 18 plots of 200 m² each were systematically sampled:

data_as

First, let's see what error we would get, if we used the simple random sampling method:

sprs(data_as, "VWB", 200, 18)

We got a 22.2% error. Now, let's calculate the sampling error using the method of successive differences, with the ss_diffs function. To use this function, the data must be set in the measured order, the plot area must be in square meters, and the total area value must be in hectares.