The object loamy is a sample data set representing 3
“synthetic transects”.
We assume a similar “soil forming function” for these transects
across each hypothetical “delineation” being transected. Here the values
have been customized so that taxonomic particle size class is
fine-loamy, but coarser textured in higher fragment material. Soil depth
varies from shallow to very deep, uniform but centered around moderately
deep. pscs_* quantities provided for example numeric
quantities that can be used.
Next we define some rating functions for particle size class and depth class.
rate_taxpartsize <- function(x) {
dplyr::case_match(x,
c("sandy-skeletal") ~ 1,
c("sandy") ~ 3,
c("loamy", "coarse-loamy", "coarse-silty") ~ 5,
c("fine-loamy", "fine-silty") ~ 7,
c("clayey", "fine") ~ 9,
c("very-fine") ~ 11,
c("loamy-skeletal", "clayey-skeletal") ~ 13,
"fragmental" ~ 15)
}
rate_depthclass <- function(x,
breaks = c(
`very shallow` = 25,
`shallow` = 50,
`moderately deep` = 100,
`deep` = 150,
`very deep` = 1e4
),
...) {
res <- cut(x, c(0, breaks))
factor(res, levels = levels(res), labels = names(breaks))
}The above rating functions can be combined in a list (m)
which will be use used as the mapping argument to
similar_soils(). The similar_soils() function
applies the rating functions to the columns of the input data
x. Target column names in the data match the names of
m, providing the “mapping” of data to rating functions.
Now we will demonstrate iterative filtering and application of similar soils criteria.
m <- list(taxpartsize = rate_taxpartsize,
depth = rate_depthclass)
x <- loamy
res0 <- similar_soils(x, m)
#> comparing to dominant reference condition (`7.deep` on 7 rows)
res0
#> id taxpartsize depth similar_dist similar_single
#> 1 A1 7 moderately deep 1 1
#> 2 B1 7 deep 0 0
#> 3 C1 7 moderately deep 1 1
#> 4 D1 7 deep 0 0
#> 5 E1 5 deep 2 2
#> 6 F1 5 shallow 4 2
#> 7 G1 5 moderately deep 3 2
#> 8 H1 13 deep 6 6
#> 9 I1 13 moderately deep 7 6
#> 10 J1 13 moderately deep 7 6
#> 11 A2 7 deep 0 0
#> 12 B2 7 deep 0 0
#> 13 C2 7 deep 0 0
#> 14 D2 7 deep 0 0
#> 15 E2 5 shallow 4 2
#> 16 F2 5 moderately deep 3 2
#> 17 G2 5 deep 2 2
#> 18 H2 13 moderately deep 7 6
#> 19 I2 15 moderately deep 9 8
#> 20 J2 13 moderately deep 7 6
#> 21 A3 7 deep 0 0
#> 22 B3 5 shallow 4 2
#> 23 C3 7 moderately deep 1 1
#> 24 D3 7 moderately deep 1 1
#> 25 E3 5 deep 2 2
#> 26 F3 5 moderately deep 3 2
#> 27 G3 5 deep 2 2
#> 28 H3 13 deep 6 6
#> 29 I3 15 deep 8 8
#> 30 J3 13 moderately deep 7 6
#> group similar
#> 1 7.moderately deep TRUE
#> 2 7.deep TRUE
#> 3 7.moderately deep TRUE
#> 4 7.deep TRUE
#> 5 5.deep FALSE
#> 6 5.shallow FALSE
#> 7 5.moderately deep FALSE
#> 8 13.deep FALSE
#> 9 13.moderately deep FALSE
#> 10 13.moderately deep FALSE
#> 11 7.deep TRUE
#> 12 7.deep TRUE
#> 13 7.deep TRUE
#> 14 7.deep TRUE
#> 15 5.shallow FALSE
#> 16 5.moderately deep FALSE
#> 17 5.deep FALSE
#> 18 13.moderately deep FALSE
#> 19 15.moderately deep FALSE
#> 20 13.moderately deep FALSE
#> 21 7.deep TRUE
#> 22 5.shallow FALSE
#> 23 7.moderately deep TRUE
#> 24 7.moderately deep TRUE
#> 25 5.deep FALSE
#> 26 5.moderately deep FALSE
#> 27 5.deep FALSE
#> 28 13.deep FALSE
#> 29 15.deep FALSE
#> 30 13.moderately deep FALSEIdentifying soils similar to "13.3" (moderately deep,
skeletal), corresponds to a major, dominant, component in a map
unit.
We might consider selecting a different reference condition manually
after inspection. If we were to do that we could set, for example,
condition="7.4" to compare against condition
"7.4" rather than "13.3" that was
automatically selected in this example.
Next, let’s take the remaining dissimilar soils, and re-apply the similarity criteria based on the next-most-dominant condition.
y <- subset(x, !res0$similar, select = c("id", "taxpartsize", "depth"))
res1 <- similar_soils(y, m)
#> comparing to dominant reference condition (`13.moderately deep` on 5 rows)
res1
#> id taxpartsize depth similar_dist similar_single
#> 5 E1 5 deep 9 8
#> 6 F1 5 shallow 9 8
#> 7 G1 5 moderately deep 8 8
#> 8 H1 13 deep 1 1
#> 9 I1 13 moderately deep 0 0
#> 10 J1 13 moderately deep 0 0
#> 15 E2 5 shallow 9 8
#> 16 F2 5 moderately deep 8 8
#> 17 G2 5 deep 9 8
#> 18 H2 13 moderately deep 0 0
#> 19 I2 15 moderately deep 2 2
#> 20 J2 13 moderately deep 0 0
#> 22 B3 5 shallow 9 8
#> 25 E3 5 deep 9 8
#> 26 F3 5 moderately deep 8 8
#> 27 G3 5 deep 9 8
#> 28 H3 13 deep 1 1
#> 29 I3 15 deep 3 2
#> 30 J3 13 moderately deep 0 0
#> group similar
#> 5 5.deep FALSE
#> 6 5.shallow FALSE
#> 7 5.moderately deep FALSE
#> 8 13.deep TRUE
#> 9 13.moderately deep TRUE
#> 10 13.moderately deep TRUE
#> 15 5.shallow FALSE
#> 16 5.moderately deep FALSE
#> 17 5.deep FALSE
#> 18 13.moderately deep TRUE
#> 19 15.moderately deep FALSE
#> 20 13.moderately deep TRUE
#> 22 5.shallow FALSE
#> 25 5.deep FALSE
#> 26 5.moderately deep FALSE
#> 27 5.deep FALSE
#> 28 13.deep TRUE
#> 29 15.deep FALSE
#> 30 13.moderately deep TRUEAt this second step, "5.4" (coarse-loamy, deep) is the
dominant condition, also identified as similar are "5.3"
(loamy, moderately deep)
One might consider which one of these is the best representative condition for the map unit (including unobserved areas) regardless of what is “dominant” in the observation data.
If there are issues with dissimilar soils being included in the same
groups, consider revising the rating functions to ensure dissimilar
properties have a distance greater than the set threshold
(thresh_single). With similar_soils(), you can
specify an alternate condition to compare against, or a
thresh_single value higher or lower than
2.
z <- subset(x, !x$id %in% c(res0$id[res0$similar], res1$id[res1$similar]),
select = c("id", "taxpartsize", "depth"))
res2 <- similar_soils(z, m)
#> comparing to dominant reference condition (`5.deep` on 4 rows)
res2
#> id taxpartsize depth similar_dist similar_single
#> 5 E1 5 deep 0 0
#> 6 F1 5 shallow 2 2
#> 7 G1 5 moderately deep 1 1
#> 15 E2 5 shallow 2 2
#> 16 F2 5 moderately deep 1 1
#> 17 G2 5 deep 0 0
#> 19 I2 15 moderately deep 11 10
#> 22 B3 5 shallow 2 2
#> 25 E3 5 deep 0 0
#> 26 F3 5 moderately deep 1 1
#> 27 G3 5 deep 0 0
#> 29 I3 15 deep 10 10
#> group similar
#> 5 5.deep TRUE
#> 6 5.shallow FALSE
#> 7 5.moderately deep TRUE
#> 15 5.shallow FALSE
#> 16 5.moderately deep TRUE
#> 17 5.deep TRUE
#> 19 15.moderately deep FALSE
#> 22 5.shallow FALSE
#> 25 5.deep TRUE
#> 26 5.moderately deep TRUE
#> 27 5.deep TRUE
#> 29 15.deep FALSEApplying the similar soils criteria again, we find "7.4"
is next most dominant, and "7.3" is similar to it.
We are left with a few soils that are not similar to any of the prior 3 sets:
subset(res2, !similar)
#> id taxpartsize depth similar_dist similar_single
#> 6 F1 5 shallow 2 2
#> 15 E2 5 shallow 2 2
#> 19 I2 15 moderately deep 11 10
#> 22 B3 5 shallow 2 2
#> 29 I3 15 deep 10 10
#> group similar
#> 6 5.shallow FALSE
#> 15 5.shallow FALSE
#> 19 15.moderately deep FALSE
#> 22 5.shallow FALSE
#> 29 15.deep FALSEWe see that all 3 of the remaining soils are "5.2"
(loamy, shallow).
Let’s construct a data.frame, with the 4 groups of similar soils each identified with a greek letter. We will see which is the most prevalent overall based on the whole data set. We could also assess prevalence within individual transects.
fin <- do.call('rbind', list(
data.frame(component = greekletters[[1]][1], subset(res0, similar)),
data.frame(component = greekletters[[1]][2], subset(res1, similar)),
data.frame(component = greekletters[[1]][3], subset(res2, similar)),
data.frame(component = greekletters[[1]][4], subset(res2, !similar))
))
# label any unassigned observations
una <- subset(res0, !res0$id %in% fin$id)
if (nrow(una) > 0) {
fin <- rbind(fin, data.frame(component = "unassigned", una))
}
# put in original order of dataset
fin <- fin[match(x$id, fin$id), ]We can tabulate the assignments we made and see how that corresponds with our concept for the relative abundance of the soils on the landscape in the typical delineation.
res <- sort(prop.table(table(fin$component)), decreasing = TRUE)
res
#>
#> Alpha Beta Gamma Delta
#> 0.3666667 0.2333333 0.2333333 0.1666667In this case, we see Alpha, Beta and Gamma as major components, and Delta as a lesser component. However, Delta is dissimilar (and shallow, which is likely strongly contrasting) compared to all of the prior soils, so if we accept the observation proportions as map unit component percentages we would have a four major component map unit.
It may be that there are other miscellaneous areas, or more-rarely-observed, unique, contrasting soils that were not captured as distinct by the rating functions. These could be split off either by creating a rating to capture them, or just noting their presence as minor components.
# TODO: abstract this concept
cmp <- subset(fin, component == names(res[1]))
ref <- names(tail(sort(table(
interaction(cmp$taxpartsize, cmp$depth)
)), 1))
fin_sim <- similar_soils(x, m, ref)
#> comparing to dominant reference condition (`7.deep` on 7 rows)
# transfer similarity distance and similar ranking
fin$similar_dist <- fin_sim$similar_dist
fin$similar <- fin_sim$similar # similarity to the dominant condition within Beta
# original sort order
fin
#> component id taxpartsize depth similar_dist similar_single
#> 1 Alpha A1 7 moderately deep 1 1
#> 2 Alpha B1 7 deep 0 0
#> 3 Alpha C1 7 moderately deep 1 1
#> 4 Alpha D1 7 deep 0 0
#> 5 Gamma E1 5 deep 2 0
#> 6 Delta F1 5 shallow 4 2
#> 7 Gamma G1 5 moderately deep 3 1
#> 8 Beta H1 13 deep 6 1
#> 9 Beta I1 13 moderately deep 7 0
#> 10 Beta J1 13 moderately deep 7 0
#> 11 Alpha A2 7 deep 0 0
#> 12 Alpha B2 7 deep 0 0
#> 13 Alpha C2 7 deep 0 0
#> 14 Alpha D2 7 deep 0 0
#> 15 Delta E2 5 shallow 4 2
#> 16 Gamma F2 5 moderately deep 3 1
#> 17 Gamma G2 5 deep 2 0
#> 18 Beta H2 13 moderately deep 7 0
#> 19 Delta I2 15 moderately deep 9 10
#> 20 Beta J2 13 moderately deep 7 0
#> 21 Alpha A3 7 deep 0 0
#> 22 Delta B3 5 shallow 4 2
#> 23 Alpha C3 7 moderately deep 1 1
#> 24 Alpha D3 7 moderately deep 1 1
#> 25 Gamma E3 5 deep 2 0
#> 26 Gamma F3 5 moderately deep 3 1
#> 27 Gamma G3 5 deep 2 0
#> 28 Beta H3 13 deep 6 1
#> 29 Delta I3 15 deep 8 10
#> 30 Beta J3 13 moderately deep 7 0
#> group similar
#> 1 7.moderately deep TRUE
#> 2 7.deep TRUE
#> 3 7.moderately deep TRUE
#> 4 7.deep TRUE
#> 5 5.deep FALSE
#> 6 5.shallow FALSE
#> 7 5.moderately deep FALSE
#> 8 13.deep FALSE
#> 9 13.moderately deep FALSE
#> 10 13.moderately deep FALSE
#> 11 7.deep TRUE
#> 12 7.deep TRUE
#> 13 7.deep TRUE
#> 14 7.deep TRUE
#> 15 5.shallow FALSE
#> 16 5.moderately deep FALSE
#> 17 5.deep FALSE
#> 18 13.moderately deep FALSE
#> 19 15.moderately deep FALSE
#> 20 13.moderately deep FALSE
#> 21 7.deep TRUE
#> 22 5.shallow FALSE
#> 23 7.moderately deep TRUE
#> 24 7.moderately deep TRUE
#> 25 5.deep FALSE
#> 26 5.moderately deep FALSE
#> 27 5.deep FALSE
#> 28 13.deep FALSE
#> 29 15.deep FALSE
#> 30 13.moderately deep FALSE| component | id | taxpartsize | depth | similar_dist | similar_single | group | similar | |
|---|---|---|---|---|---|---|---|---|
| 2 | Alpha | B1 | 7 | deep | 0 | 0 | 7.deep | TRUE |
| 4 | Alpha | D1 | 7 | deep | 0 | 0 | 7.deep | TRUE |
| 11 | Alpha | A2 | 7 | deep | 0 | 0 | 7.deep | TRUE |
| 12 | Alpha | B2 | 7 | deep | 0 | 0 | 7.deep | TRUE |
| 13 | Alpha | C2 | 7 | deep | 0 | 0 | 7.deep | TRUE |
| 14 | Alpha | D2 | 7 | deep | 0 | 0 | 7.deep | TRUE |
| 21 | Alpha | A3 | 7 | deep | 0 | 0 | 7.deep | TRUE |
| 1 | Alpha | A1 | 7 | moderately deep | 1 | 1 | 7.moderately deep | TRUE |
| 3 | Alpha | C1 | 7 | moderately deep | 1 | 1 | 7.moderately deep | TRUE |
| 23 | Alpha | C3 | 7 | moderately deep | 1 | 1 | 7.moderately deep | TRUE |
| 24 | Alpha | D3 | 7 | moderately deep | 1 | 1 | 7.moderately deep | TRUE |
| 5 | Gamma | E1 | 5 | deep | 2 | 0 | 5.deep | FALSE |
| 17 | Gamma | G2 | 5 | deep | 2 | 0 | 5.deep | FALSE |
| 25 | Gamma | E3 | 5 | deep | 2 | 0 | 5.deep | FALSE |
| 27 | Gamma | G3 | 5 | deep | 2 | 0 | 5.deep | FALSE |
| 7 | Gamma | G1 | 5 | moderately deep | 3 | 1 | 5.moderately deep | FALSE |
| 16 | Gamma | F2 | 5 | moderately deep | 3 | 1 | 5.moderately deep | FALSE |
| 26 | Gamma | F3 | 5 | moderately deep | 3 | 1 | 5.moderately deep | FALSE |
| 6 | Delta | F1 | 5 | shallow | 4 | 2 | 5.shallow | FALSE |
| 15 | Delta | E2 | 5 | shallow | 4 | 2 | 5.shallow | FALSE |
| 22 | Delta | B3 | 5 | shallow | 4 | 2 | 5.shallow | FALSE |
| 8 | Beta | H1 | 13 | deep | 6 | 1 | 13.deep | FALSE |
| 28 | Beta | H3 | 13 | deep | 6 | 1 | 13.deep | FALSE |
| 9 | Beta | I1 | 13 | moderately deep | 7 | 0 | 13.moderately deep | FALSE |
| 10 | Beta | J1 | 13 | moderately deep | 7 | 0 | 13.moderately deep | FALSE |
| 18 | Beta | H2 | 13 | moderately deep | 7 | 0 | 13.moderately deep | FALSE |
| 20 | Beta | J2 | 13 | moderately deep | 7 | 0 | 13.moderately deep | FALSE |
| 30 | Beta | J3 | 13 | moderately deep | 7 | 0 | 13.moderately deep | FALSE |
| 29 | Delta | I3 | 15 | deep | 8 | 10 | 15.deep | FALSE |
| 19 | Delta | I2 | 15 | moderately deep | 9 | 10 | 15.moderately deep | FALSE |
This process of iteratively and grouping soils into similarity groups
can be automated using the design_mapunit() function.