How a measured site gradient compares with a modelled one

FDIS crews measured the channel's slope at the site (gradient_pct, and a second reading, gradient2_pct); every gradient limit the tuning publishes is a limit on a Freshwater Atlas average (link's segment gradient, or the whole Atlas line's). These measure how far apart the two are (#121), as data-raw/site_gradient_check.R reports them and the methodology article quotes them, from one code path.

Usage

knw_site_gradient_visits(visits_link)

knw_site_gradient_population(v, visits_species, species_code, stage)

knw_gradient_agree(measured, modelled)

knw_gradient_flips(d, limit, against, ladder)

Arguments

visits_link

0310's visits_link.csv.

v

from knw_site_gradient_visits().

visits_species

0310's visits_species.csv.

species_code, stage

the key.

measured, modelled

numeric gradients on one scale.

d

a population with meas and the modelled gradient column.

limit

numeric(1) the published limit.

against

the modelled gradient column ("segment_gradient", "fwa_gradient").

ladder

the key's gradient ladder (its tune table's grad_max).

Value

knw_site_gradient_visits(): the visits with meas, meas2, at_visit and stream_branch. knw_site_gradient_population(): the key's visits. knw_gradient_agree() and knw_gradient_flips(): a one-row tibble.

Details

knw_site_gradient_visits() reads the visit table onto the rules' scale (percent / 100, meas and meas2) and marks the tuning's stream branch (knw_tune_run()'s edge types and fished area, no river, wetland or lake). knw_site_gradient_population() is a key's scored visits on it: in the species' range, not blocked, no barrier downstream, and for rearing the groups that hold both outcomes (knw_tune_score()); a spawning key's is before the spawner-evidence and groups filters, which need tables not read here. knw_gradient_agree() is the agreement of two gradients: rank correlation over every pair; over pairs both above 0, the correlation of the logs and the measured/modelled ratio's p10, p50 and p90; the absolute difference's median and p90; the share where the model is higher. knw_gradient_flips() counts visits whose measured and modelled gradients lie on opposite sides of a limit (a reading on the limit is in, as a rule's <= has it), over every visit and within one step of the key's own ladder either side of the limit (band, mirrored at the ladder's ends; and the same window on the measured gradient).

Examples

knw_gradient_agree(c(0.01, 0.02, 0.05, 0.10), c(0.012, 0.03, 0.04, 0.08))
#> # A tibble: 1 × 10
#>   pairs zero_pairs spearman log_r ratio_p10 ratio_p50 ratio_p90 absdiff_p50
#>   <int>      <int>    <dbl> <dbl>     <dbl>     <dbl>     <dbl>       <dbl>
#> 1     4          0        1 0.971      0.72      1.04      1.25        0.01
#> # ℹ 2 more variables: absdiff_p90 <dbl>, model_higher_pct <dbl>
d <- data.frame(meas = c(0.10, 0.12, 0.05), segment_gradient = c(0.12, 0.10, 0.06))
knw_gradient_flips(d, 0.11, "segment_gradient", c(0.07, 0.09, 0.11, 0.13, 0.15))
#> # A tibble: 1 × 11
#>   against     visits flips_pct band  band_visits band_flips_pct meas_band_visits
#>   <chr>        <int>     <dbl> <chr>       <int>          <dbl>            <int>
#> 1 segment_gr…      3      66.7 (0.0…           2            100                2
#> # ℹ 4 more variables: meas_band_flips_pct <dbl>, meas_on_limit <int>,
#> #   meas_above_model_below <int>, meas_below_model_above <int>