model_ndvi: Model NDVI time series

In robitalec/irg: Instantaneous Rate of Green Up

Model NDVI time series

Description

Fit double logistic model to NDVI time series given parameters estimated with model_params.

Usage

model_ndvi(DT, observed = TRUE)

Arguments

DT	data.table of model parameters (output from model_params).
observed	boolean indicating if a full year of fitted values should be returned (observed = FALSE) or if only observed values will be fit (observed = TRUE)

Value

Model parameter data.table appended with 'fitted' column of double logistic model of NDVI for a full year. Calculated at the daily scale with the following formula from Bischoff et al. (2012).

fitted = \frac{1}{1 + \exp{\frac{xmidS - t}{scalS}}} - \frac{1}{1 + \exp{\frac{xmidA - t}{scalA}}}

(See the "Getting started with irg vignette" for a better formatted formula.)

References

https://www.journals.uchicago.edu/doi/abs/10.1086/667590

See Also

Other model: model_params(), model_start()

Examples

# Load data.table
library(data.table)

# Read in example data
ndvi <- fread(system.file("extdata", "sampled-ndvi-MODIS-MOD13Q1.csv", package = "irg"))

# Filter and scale NDVI time series
filter_ndvi(ndvi)
scale_doy(ndvi)
scale_ndvi(ndvi)

# Guess starting parameters for xmidS and xmidA
model_start(ndvi)

## Two options: fit to full year or observed data
# Option 1 - returns = 'models'

# Double logistic model parameters
#   given global starting parameters for scalS, scalA
#   and output of model_start for xmidS, xmidA
mods <- model_params(
  ndvi,
  returns = 'models',
  xmidS = 'xmidS_start',
  xmidA = 'xmidA_start',
  scalS = 0.05,
  scalA = 0.01
)

# Fit to the whole year (requires assignment)
fit <- model_ndvi(mods, observed = FALSE)

# Option 2 - returns = 'columns'
model_params(
  ndvi,
  returns = 'columns',
  xmidS = 'xmidS_start',
  xmidA = 'xmidA_start',
  scalS = 0.05,
  scalA = 0.01
)

# Fit double logistic curve to NDVI time series for the observed days
model_ndvi(ndvi, observed = TRUE)

robitalec/irg documentation built on Sept. 19, 2023, 9:33 p.m.