Skip to contents

Create Transition Rates Table

Source: R/trans_rates_t.R
trans_rates_t.Rd

Creates a trans_rates_t table that stores transition rates (probabilities) for each transition type in each time period. Historical rates are calculated from observed transitions, and future rates are extrapolated using linear regression.

Usage

as_trans_rates_t(x)

get_obs_trans_rates(self)

extrapolate_trans_rates(obs_rates, periods, coord_count = NA_integer_)

# S3 method for class 'trans_rates_t'
print(x, nrow = 10, ...)

trans_rate_areas(lulc_data, rates, trans_meta)

Arguments

x

A list or data.frame coercible to a data.table

self

a DB instance

obs_rates

A trans_rates_t table of observed transition rates for historical periods

periods

A periods_t table with is_extrapolated = TRUE for future periods

coord_count

Optional integer specifying the number of coordinates (cells) for normalization

nrow

see data.table::print.data.table

...

passed to data.table::print.data.table

lulc_data

A lulc_data_t for a single id_run; the areas of its last period are the state the replay starts from.

rates

A trans_rates_t table for a single id_run.

trans_meta

A trans_meta_t table, resolving id_trans to a pair of classes.

Value

A data.table of class "trans_rates_t" with columns:

  • id_run: Foreign key to runs_t

  • id_period: Foreign key to periods_t

  • id_trans: Foreign key to trans_meta_t

  • count: Absolute number of transitioning cells for (id_trans, id_period)

  • rate: Transition rate: count of transitions in (id_trans, id_period) over count of cells of id_lulc_anterior in id_period

trans_rate_areas() returns a data.table with id_lulc, id_period and area; id_period is the period whose state the area describes, so the initial state carries the last period of lulc_data.

Methods (by generic)

  • print(trans_rates_t): Print a trans_rates_t object, passing params to data.table print

Functions

  • get_obs_trans_rates(): Calculate observed transition rates from historical data. For each period and transition type, calculates the rate as the proportion of id_lulc_anterior cells that transitioned to id_lulc_posterior.

  • extrapolate_trans_rates(): Return future transition rates using linear regression. For each id_run + id_trans, fits a linear model of rate vs period number and extrapolates to future periods. Negative predicted rates are set to 0.

  • trans_rate_areas(): Replay a rate table forward from an observed state to recover the class areas it implies. Transitions not recorded in rates are implied to be zero, so the residual 1 - sum(rate) of each class persists. This is what makes a solved trajectory recoverable from a trans_rates_t alone, and therefore comparable against the areas an allocation run actually realised.