sfn_data classes

In sapfluxnetr: Working with 'Sapfluxnet' Project Data

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

sfn_data

sfn_data is an S4 class designed for store and interact with sap flow data at individual plant level (not raw data), primarily from Sapfluxnet project sites data and metadata.

S4 Slots

sfn_data class has twelve different slots:

1. sapf_data: Tibble containing the sap flow data, each column representing an individual tree, without any TIMESTAMP variable.

2. env_data: Tibble containing the environmental data, each column an environmental variable, without any TIMESTAMP variable. It must have the same nrow thet sapf_data, in order to be able to combine both in further analyses or aggregations.

3. sapf_flags: Tibble with the same dimensions as sapf_data, containing the flags (special remarks indicating possible outliers or any annotation of interest) for each observation in sapf_data.

4. env_flags: Tibble with the same dimensions as env_data, containing the flags for each observation in env_data.

5. si_code: Character vector of length 1 with the site code. Useful for further analyses or aggregations in order to identify the site when working with more then one.

6. timestamp: POSIXct vector of length equal to nrow(sapf_data) with the timestamp values.

7. solar_timestamp: POSIXct vector of length equal to nrow(sapf_data) with the apparent solar timestamp.

8. site_md: Tibble with the site metadata. See sfn_vars_to_filter() for a list of possible metadata variables. This variables are not mandatory, and new ones can be added.

9. stand_md: Tibble with the stand metadata. See sfn_vars_to_filter() for a list of possible metadata variables. This variables are not mandatory, and new ones can be added.

10. species_md: Tibble with the species metadata. See sfn_vars_to_filter() for a list of possible metadata variables. This variables are not mandatory, and new ones can be added.

11. plant_md: Tibble with the plant metadata. See sfn_vars_to_filter() for a list of possible metadata variables. This variables are not mandatory, and new ones can be added.

12. env_md: Tibble with the environmental metadata. See sfn_vars_to_filter() for a list of possible metadata variables. This variables are not mandatory, and new ones can be added.

sfn_data design

The schematics of the class are summarised in Fig 1.

This design have some characteristics:

• Dimensions: Dimensions must comply:

  nrow(sapf_data) == nrow(env_data) == nrow(sapf_flags) == nrow(env_flags) == length(timestamp) == length(solar_timestamp)

  ncol(sapf_data) == ncol(sapf_flags)

  ncol(env_data) == ncol(env_flags)

  In the case of differences in the dimensions between environmental and sap flow data, NAs must be added in the corresponding places when equalising both TIMESTAMPs (i.e. if env data is longer then sapf data, rows are added to the later with NAs to make them equal, and flags ("NA_ADDED") are raised to indicate the adding).

• si_code slot: This slot indicates the site. This is useful for identifying the site in the metrics functions results.

• Compartmentalization: timestamp and solar_timestamp are isolated from data (sapf and env) in their own slot. This made easier subsetting the object and open the possibility of working with the timestamps only.

• Flags: Two slots are used for storing data flags. As the flag slots have the same dimensions as the corresponding dataset slot, selection of flagged data is straightforward:

  sapf_data[flag_data == "flag", ]

Methods

Some methods are included in the class:

• show: calling a sfn_data object shows info about it

library(sapfluxnetr)
data('ARG_TRE', package = 'sapfluxnetr')
ARG_TRE

• get methods: methods to get any slot outside the object. In the case of sapf and env data, TIMESTAMP is added (based on the solar argument) in order to obtain fully functional datasets. See ?sfn_get_methods for more details.

get_sapf_data(ARG_TRE, solar = FALSE)
get_sapf_data(ARG_TRE, solar = TRUE)
get_env_data(ARG_TRE) # solar default is FALSE
get_sapf_flags(ARG_TRE) # solar default is FALSE
get_env_flags(ARG_TRE) # solar default is FALSE
get_si_code(ARG_TRE)
get_timestamp(ARG_TRE)[1:10]
get_solar_timestamp(ARG_TRE)[1:10]
get_site_md(ARG_TRE)
get_stand_md(ARG_TRE)
get_species_md(ARG_TRE)
get_plant_md(ARG_TRE)
get_env_md(ARG_TRE)

• assignation: methods for "<-" are implemented, allowing transformation of the slots. This is used internally in some functions but is not recommended except for updating metadata slots. In data slots this changes are not reflected in the flags, so the reproducibility and traceability will be lost if used without care. See ?sfn_replacement_methods for more details.

# extraction and modification
foo_site_md <- get_site_md(ARG_TRE)
foo_site_md[['si_biome']]
foo_site_md[['si_biome']] <- 'Temperate forest'
# assignation
get_site_md(ARG_TRE) <- foo_site_md
# check it worked
get_site_md(ARG_TRE)[['si_biome']]

• validation: a validation method is implemented in order to avoid creation of invalid sfn_data objects (i.e objects with different data dimensions, objects without metadata...)

# get sap flow data
foo_bad_sapf <- get_sapf_data(ARG_TRE)
# pull a row, now it has diferent dimensions then
foo_bad_sapf <- foo_bad_sapf[-1,]
# try to assign the incorrect data fails
get_sapf_data(ARG_TRE) <- foo_bad_sapf[,-1] # ERROR
# try to build a new object also fails
sfn_data(
  sapf_data = foo_bad_sapf[,-1], # remember to remove timestamp column
  env_data = get_env_data(ARG_TRE)[,-1],
  sapf_flags = get_env_flags(ARG_TRE)[,-1],
  env_flags = get_env_flags(ARG_TRE)[,-1],
  si_code = get_si_code(ARG_TRE),
  timestamp = get_timestamp(ARG_TRE),
  solar_timestamp = get_solar_timestamp(ARG_TRE),
  site_md = get_site_md(ARG_TRE),
  stand_md = get_stand_md(ARG_TRE),
  species_md = get_species_md(ARG_TRE),
  plant_md = get_plant_md(ARG_TRE),
  env_md = get_env_md(ARG_TRE)
)

Utilities

sapfluxnetr package offers some utilities to visualize and work with sfn_data objects:

sfn_plot

This function allows to plot sfn_data objects. See ?sfn_plot for more details.

library(ggplot2)

sfn_plot(ARG_TRE, type = 'env') +
  facet_wrap(~ Variable, ncol = 3, scales = 'free_y') +
  theme(legend.position = 'none')

sfn_plot(ARG_TRE, formula_env = ~ vpd) +
  theme(legend.position = 'none')

sfn_filter

This function emulates filter function from dplyr package for sfn_data objects. Useful to filter by some specific timestamp. Be advised, using this function to filter by sap flow or environmental variables can create TIMESTAMP gaps. See sfn_filter for more details.

library(lubridate)
library(dplyr)

# get only the values for november
sfn_filter(ARG_TRE, month(TIMESTAMP) == 11)

sfn_mutate

This function allows mutation of data variables inside the sfn_data object. Useful when you need to transform a variable to another units or similar. A flag ('USER_MODF') will be added to all values in the mutated variable. See sfn_mutate for more details.

At this moment, mutate does not allows creating new variables, only mutate existing variables

# transform ws from m/s to km/h
foo_mutated <- sfn_mutate(ARG_TRE, ws = ws * 3600/1000)
get_env_data(foo_mutated)[['ws']][1:10]

sfn_mutate_at

This function mutates all variables declared with the function provided. Useful when you need to conditionally transform the data, i.e. converting to NA sap flow values when an environmental variable exceeds some threshold. See sfn_mutate_at for more details.

foo_mutated_2 <- sfn_mutate_at(
  ARG_TRE,
  vars(one_of(names(get_sapf_data(ARG_TRE)[,-1]))),
  list(~ case_when(
    ws > 25 ~ NA_real_,
    TRUE ~ .
  ))
)

# see the difference between ARG_TRE and foo_mutated_2
get_sapf_data(ARG_TRE)
get_sapf_data(foo_mutated_2)

*_metrics functions

Family of functions to aggregate and summarise the site data. See ?metrics for more details.

foo_daily <- daily_metrics(ARG_TRE)
foo_daily[['sapf']][['sapf_gen']]

For full control of metrics and custom aggregations see vignette('custom-aggregation', package = 'sapfluxnetr')

sfn_data_multi

sfn_data_multi is an S4 class designed to store multiple sfn_data objects. It inherits from list so, in a nutshell, sfn_data_multi is a list of sfn_data objects.

Methods

sfn_data_multi has the following methods declared.

• show: it shows the number, codes and combined timestamp span of sites in the multi object.

# creating a sfn_data_multi object
data(ARG_MAZ, package = 'sapfluxnetr')
data(AUS_CAN_ST2_MIX, package = 'sapfluxnetr')
multi_sfn <- sfn_data_multi(ARG_TRE, ARG_MAZ, AUS_CAN_ST2_MIX)

# show method
multi_sfn

• get methods: they work the same as in the sfn_data objects, except they return a list, with the corresponding data for the sites in the sfn_data_multi object as elements.

# get sap flow data
get_sapf_data(multi_sfn)
# get plant metadata
get_plant_md(multi_sfn)
# with metadata, we can collapse
get_plant_md(multi_sfn, collapse = TRUE)

Utilities

All the utilities thet exists for sfn_data work for sfn_data_multi objects, executing the function for all the sites contained in the sfn_data_multi object:

sfn_plot

multi_plot <- sfn_plot(multi_sfn, formula = ~ vpd)
multi_plot[['ARG_TRE']] + theme(legend.position = 'none')
multi_plot[['AUS_CAN_ST2_MIX']] + theme(legend.position = 'none')

sfn_filter

multi_filtered <- sfn_filter(multi_sfn, month(TIMESTAMP) == 11)
get_timestamp(multi_filtered[['AUS_CAN_ST2_MIX']])[1:10]

sfn_mutate

multi_mutated <- sfn_mutate(multi_sfn, ws = ws * 3600/1000)
get_env_data(multi_mutated[['AUS_CAN_ST2_MIX']])[['ws']][1:10]

sfn_mutate_at

vars_to_not_mutate <- c(
  "TIMESTAMP", "ta", "rh", "vpd", "sw_in", "ws",
  "precip", "swc_shallow", "ppfd_in", "ext_rad"
)

multi_mutated_2 <- sfn_mutate_at(
  multi_sfn,
  vars(-one_of(vars_to_not_mutate)),
  list(~ case_when(
    ws > 25 ~ NA_real_,
    TRUE ~ .
  ))
)

multi_mutated_2[['ARG_TRE']]

*_metrics

multi_metrics <- daily_metrics(multi_sfn)
multi_metrics[['ARG_TRE']][['sapf']]

sapfluxnetr documentation built on Feb. 16, 2023, 7:52 p.m.