In ktoddbrown/RIDS: Read In Data Scripts

library(tidyverse) # manage data
library(EML) # read in EML meta data
library(knitr) # make pretty tables using kable
library(RIDS)

This file identifies, downloads and intakes two data sets that both have soil mositure measurements.

Download and read datasets

Groffman.ls <- RIDS::readGroffman2020(dataDir = 'data/Groffman2012', verbose=FALSE)
Forrester.ls <- RIDS::readForrester2019(dataDir = 'data/Forrester2019')

Groffman 2012

ABSTRACT: r Groffman.ls$studyInfo$abstract

Groffman.ls$data <- Groffman.ls$data %>%
  mutate_at(c('Project', 'Season', 'Horizon', 'Sample'), as.factor) %>%
  select(-DEA) #No variation in enzyme activity so remove it here

The microbial biomass, resperation, and water content appear to have strong horizon dependencies.

knitr::kable(summary(Groffman.ls$data))
knitr::kable(Groffman.ls$meta)

plot.df <-  Groffman.ls$data %>%
  pivot_longer(BIOC:H2O) %>%
  left_join(Groffman.ls$meta, by='name')

ggplot(plot.df) +
  geom_histogram(aes(x=value)) +
  facet_grid(Horizon~name + description + unit, scales = 'free')

Forrester 2019

ABSTRACT: r Forrester.ls$studyInfo$abstract

Forrester.ls$data1 <- Forrester.ls$data1 %>%
  mutate_at(c('site', 'time_period', 'treatment', 'location'), as.factor)

Forrester.ls$data2 <- Forrester.ls$data2 %>%
  mutate_at(c('site', 'treatment', 'location'), as.factor) %>% 
  mutate(snow_depth = as.numeric(gsub('>', '', snow_depth)))

The microbial biomass, resperation, and water content appear to have strong horizon dependencies.

knitr::kable(summary(Forrester.ls$data1 ))
knitr::kable(summary(Forrester.ls$data2))
knitr::kable(Forrester.ls$meta)

ggplot(Forrester.ls$data1) +
  geom_histogram(aes(x=VWC))+
  facet_wrap(~treatment)

ggplot(Forrester.ls$data2) +
  geom_histogram(aes(x=snow_depth)) +
  facet_wrap(~treatment)

Harmonizing data

merge Forrester

# transfer probe type to the meta data
Forrester.ls$meta$description[Forrester.ls$meta$name == 'probe'] <- paste0(Forrester.ls$meta$description[Forrester.ls$meta$name == 'probe'], '. Probe type is:', unique(Forrester.ls$data1$probe))
# remove probe type from data file
Forrester.ls$data1$probe <- NULL

temp1 <- Forrester.ls$data1 %>%
  pivot_longer(cols = c('VWC'), values_to = 'value_numerical')

temp2 <- Forrester.ls$data2 %>%
  select(-comments) %>%
  pivot_longer(cols=c('snow_depth'), values_to = 'value_numerical')

Forrester.ls$merged_data <- Forrester.ls$data2 %>%
  select(-snow_depth) %>%
  pivot_longer(cols=c('comments'), values_to = 'value_text') %>%
  bind_rows(temp2) %>%
  bind_rows(temp1)

merge Forrester with Groffman

Make Groffman long data format

Groffman.ls$long_data <- Groffman.ls$data %>%
  pivot_longer(cols=BIOC:H2O, values_to = 'value_numerical') %>%
  rename('date' = 'Date')

Merge Groffman with Forrester

soilMoistureData <- Groffman.ls$long_data %>%
  bind_rows(Forrester.ls$merged_data)

#create a new metadata for this merged dataframe.

metaData <- Forrester.ls$meta %>%
  bind_rows(Groffman.ls$meta)

temp <- soilMoistureData %>%
  left_join(metaData, by = 'name') %>%
  filter(!is.na(value_numerical) | (value_text != 'NaN' & !is.na(value_text)))

temp2 <- metaData %>%
  filter(!(name %in% c(unique(temp$name), 'DEA', 'Date')))

merged.ls <- list(data = temp, meta = temp2)

Data visulization

Common between the two data sets are water content measurements. Groffman has a "Gravimetric water content gram water Per gram wet soil" and Forrester has "volumetric water content" volumne water per volumne wet soil

plot.df <- soilMoistureData %>% 
  filter(name %in% c('H2O', 'VWC'))

ggplot(plot.df) +
  geom_point(aes(x=lubridate::yday(date), y=value_numerical)) +
  facet_grid(treatment~name, scales='free')