emp_spatial_cov function

Computes empirical spatial covariance using a dataframe or a stars object

Computes empirical spatial covariance by removing trends and examining residuals. It can compute lag-0 or log-1 empirical covariance either by latitude or longitude. You can split up the spatial domain by latitude or longitude and plot the covariance for each longitudinal/latitudinal strips.

emp_spatial_cov(
  x,
  lat_or_lon_strips = "lon",
  quadratic_time = FALSE,
  quadratic_space = FALSE,
  num_strips = 1,
  lag = 0,
  ...
)

## S3 method for class 'data.frame'
emp_spatial_cov(
  x,
  lat_or_lon_strips = "lon",
  quadratic_time = FALSE,
  quadratic_space = FALSE,
  num_strips = 1,
  lag = 0,
  lat_col,
  lon_col,
  t_col,
  z_col,
  ...
)

## S3 method for class 'stars'
emp_spatial_cov(
  x,
  lat_or_lon_strips = "lon",
  quadratic_time = FALSE,
  quadratic_space = FALSE,
  num_strips = 1,
  lag = 0,
  ...
)

## S3 method for class 'spatialcov'
autoplot(object, xlab = "Latitude", ...)

Arguments

• x: A stars object or a dataframe. Arguments differ according to the input type.

• lat_or_lon_strips: Takes the values lat or lon. The value lat produces latitudinal strips, i.e., covariance plots over longitude for different latitudinal strips. The value lon produces longitudinal strips, i.e., covariance plots over latitude for different longitudinal strips.

• quadratic_time: If TRUE a linear model with quadratic time is fitted and residuals computed. If FALSE

  the model is fitted with linear space and time coefficients.

• quadratic_space: If TRUE a linear model with quadratic space is fitted and residuals computed. If FALSE

  the model is fitted with linear space and time coefficients.

• num_strips: The number of latitudinal/longitudinal strips to produce. This is used when plotting using autoplot.

• lag: Lag can be either 0 or 1.

• ...: Other arguments currently ignored.

• lat_col: For dataframes: the column or the column name giving the latitude. The y coordinate can be used instead of latitude.

• lon_col: For dataframes: the column or the column name giving the longitude. The x coordinate can be used instead of longitude.

• t_col: For dataframes: the time column. Time must be a set of discrete integer values.

• z_col: For dataframes: the The quantity of interest that will be plotted. Eg. temperature.

• object: For autoplot: the output of the function `emp_spatial_cov'.

• xlab: For autoplot: the label for x-axis.

Returns

A spatialcov object with empirical covariance data organised spatially according to the number of strips and the lagged covariance.

Examples

# Dataframe example
library(dplyr)
data(NOAA_df_1990)
Tmax <- filter(NOAA_df_1990,
  proc == "Tmax" &
  month %in% 5:6 &
  year == 1993)
Tmax$t <- Tmax$julian - min(Tmax$julian) + 1
emp_df <- emp_spatial_cov(Tmax,
                lat_col = "lat",
                lon_col = "lon",
                t_col ="t",
                z_col = "z",
                lat_or_lon_strips = "lon",
                num_strips = 4,
                lag = 1)
autoplot(emp_df)

# Stars example
library(stars)
# Create a stars object from a data frame
precip_df <- NOAA_df_1990[NOAA_df_1990$proc == 'Precip', ] %>%
  filter(date >= "1992-02-01" & date <= "1992-02-05")
precip <- precip_df[ ,c('lat', 'lon', 'date', 'z')]
st_precip <- st_as_stars(precip, dims = c("lon", "lat", "date"))
emp_spatial_cov(st_precip)