Validate projected range shifts against resurveys (hindcasting)
Source:R/range-shift-validate.R
range_shift_validate.RdA range-shift model earns trust when it predicts shifts that have already
happened. range_shift_validate() compares the shifts observed between a
historical survey and a resurvey (range_shift_resurvey()) with the
shifts a model predicted for the same species and period, run from the
climate of the first survey to that of the second (hindcasting).
For each measure (lower edge, upper edge, midpoint, optimum) it reports:
bias: mean of predicted minus observed shift (m);mae,rmse: mean absolute and root mean square error (m);r,rho: Pearson and Spearman correlation across species;slope: calibration slope of observed on predicted (1 is ideal; below 1, the model overstates differences between species);direction: share of species whose shift has the predicted sign;tracking: mean observed shift divided by mean predicted shift, the share of the expected shift that species realised (1 = full tracking; below 1, a lag), with a bootstrap interval resampling species.
range_shift_expected() gives the simplest prediction, the isotherm
shift \(\Delta T / \gamma\) for warming \(\Delta T\) and lapse rate
\(\gamma\) (Colwell et al. 2008; Freeman and Freeman 2014). It is the
baseline any model must beat.
Usage
range_shift_validate(
observed,
predicted,
drop_truncated = TRUE,
nboot = 999,
conf = 0.95,
seed = NULL
)
range_shift_expected(warming, lapse_rate)Arguments
- observed
Result of
range_shift_resurvey(), or a data frame withspeciesand observedshift_*columns.- predicted
Data frame with
speciesand predicted shifts, in columns named like those ofobserved(shift_lower,shift_upper,shift_midpoint,shift_optimum), or a single columnshiftused for every measure.- drop_truncated
Leave out edges truncated at the ends of the sampled gradient (needs the
truncated_*columns ofobserved).- nboot
Bootstrap resamples for the tracking interval.
- conf
Confidence level.
- seed
Optional random seed.
- warming
Warming in degrees C between the surveys (or scenarios).
- lapse_rate
Temperature lapse rate in degrees C per km.
Value
range_shift_validate(): a tibble with one row per measure.
range_shift_expected(): the expected upslope shift in metres.
References
Colwell, R. K., Brehm, G., Cardelús, C. L., Gilman, A. C. and Longino, J. T. (2008). Global warming, elevational range shifts, and lowland biotic attrition in the wet tropics. Science, 322, 258-261.
Freeman, B. G. and Class Freeman, A. M. (2014). Rapid upslope shifts in New Guinean birds illustrate strong distributional responses of tropical montane species to global warming. PNAS, 111, 4490-4494.
Examples
# Virtual resurvey: species realise 60% of a 200 m expected shift
before <- png_frogs
after <- png_frogs
after$elevation <- after$elevation + 120
obs <- range_shift_resurvey(before, after, gradient = elevation, abundance = abundance)
pred <- data.frame(species = obs$species, shift = range_shift_expected(1, 5))
range_shift_validate(obs, pred, nboot = 199, seed = 1)
#> # A tibble: 4 × 14
#> measure n observed predicted bias mae rmse r rho slope direction
#> <chr> <int> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 lower 42 120 200 80 80 80 NA NA NA 1
#> 2 upper 53 120 200 80 80 80 NA NA NA 1
#> 3 midpoi… 55 120 200 80 80 80 NA NA NA 1
#> 4 optimum 55 120 200 80 80 80 NA NA NA 1
#> # ℹ 3 more variables: tracking <dbl>, tracking_lower <dbl>,
#> # tracking_upper <dbl>