Similar Soil Contrasts — similar

A method for applying standardized, customizable "similar soils" rules to arbitrary site-level data.

Usage

similar_soils(
  x,
  mapping,
  condition = NULL,
  idname = "id",
  thresh = NULL,
  thresh_single = 2,
  thresh_all = 3,
  absolute = TRUE,
  verbose = TRUE
)

Arguments

x: A data.frame or a SoilProfileCollection.
mapping: named list of functions. List element names refer to site-level data columns of x. Each function performs a conversion of the raw data in x to a value used in the "similar soils" distance calculation.
condition: integer or character. For default (NULL) the value is calculated internally based on the dominant condition of intersection of mapping results in x. You may specify an integer row ID in x for specific similar soil contrasts, or you can specify a character dominant condition equivalent to the name assigned by interaction() e.g. "4.3" for a two rating mapping result where the first rating has value 4 and the second has value 3.
idname: character. ID column name, default "id".
thresh: integer. Deprecated. If used, the same value is passed to thresh_all and thresh_single.
thresh_single: integer. hreshold difference in any one property required to be dissimilar. Default 2. See details for discussion on the default calculation for similarity.
thresh_all: integer. Threshold sum of differences across all properties required to be dissimilar. Default 3. See details for discussion on the default calculation for similarity.
absolute: logical. Report absolute difference? Default: TRUE. Absolute difference is always used for comparison against thresh_single and thresh_all, but the sign of the difference can be relevant to interpretation of relative limitations of soils.
verbose: Default: TRUE message about selected condition.

Value

A data.frame containing inputs and three new columns: similar_single (maximum difference in any one property, relative to condition), similar_dist (cumulative sum of differences relative to condition), similar (logical; soil is similar to condition)

Details

The sum of differences across conditions (specified by the intersection of output of the functions in mapping) is used as the "distance" of a soil relative to condition. A threshold value is used to decide which are "similar" and which are not. The functions in mapping can be customized to use alternate thresholds: thresh_single defines the maximum distance in any one property for soils considered to be similar, whereas thresh_all defines the maximum sum of distances across all properties. Defaults reflect thresholds used by Norfleet and Eppinette (1993) and replicated in most local similarity guides.

References

Norfleet, M.L. and Eppinette, R.T. (1993), A Mathematical Model for Determining Similar and Contrasting Inclusions for Map Unit Descriptions. Soil Survey Horizons, 34: 4-5. https://doi.org/10.2136/sh1993.1.0004

Examples


data(loamy, package = "SOILmilaR")

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 ), pattern =
  "R|Cr|Cd|kk|m", hzdesgn = aqp::guessHzDesgnName(x, required = TRUE),
                            ...) {
  res <- cut(x, c(0, breaks))
  factor(res, levels = levels(res), labels = names(breaks), ordered = TRUE)
}

rate_pscs_clay <- function(x,
                           breaks = c(18, 27, 40, 60, 100)) {
  res <- cut(x, c(0, breaks))
  factor(res, levels = levels(res), ordered = TRUE)
}

m <- list(taxpartsize = rate_taxpartsize, depth = rate_depthclass,
pscs_clay = rate_pscs_clay)

s <- similar_soils(loamy, m)
#> comparing to dominant reference condition (`13.moderately deep.(0,18]` on 5 rows)
head(s)
#>   id taxpartsize           depth pscs_clay similar_dist similar_single
#> 1 A1           7 moderately deep   (27,40]            8              6
#> 2 B1           7            deep   (18,27]            8              6
#> 3 C1           7 moderately deep   (27,40]            8              6
#> 4 D1           7            deep   (27,40]            9              6
#> 5 E1           5            deep    (0,18]            9              8
#> 6 F1           5         shallow    (0,18]            9              8
#>                       group similar
#> 1 7.moderately deep.(27,40]   FALSE
#> 2            7.deep.(18,27]   FALSE
#> 3 7.moderately deep.(27,40]   FALSE
#> 4            7.deep.(27,40]   FALSE
#> 5             5.deep.(0,18]   FALSE
#> 6          5.shallow.(0,18]   FALSE

# inspect distances using agglomerative clustering+dendrogram
d <- cluster::agnes(s[, 5, drop = FALSE], method="gaverage")
d$height <- d$height + 0.2 # fudge factor for 0-distance
plot(stats::as.dendrogram(d), center=TRUE, type="triangle")


# allow relative contrast ratings to be negative # (i.e. ordinal factors, concept of "limiting")
# absolute value is still used for "similar" threshold
s2 <- similar_soils(loamy, m, absolute=FALSE)
#> comparing to dominant reference condition (`13.moderately deep.(0,18]` on 5 rows)

# inspect distances using agglomerative clustering+dendrogram
d2 <- cluster::agnes(s2[, 5, drop = FALSE], method="gaverage")
d2$height <- d2$height + 0.2 # fudge factor for 0-distance
plot(stats::as.dendrogram(d2), center=TRUE, type="triangle")