Timothy Gilbert (tsgilbert@arizona.edu)
This Shiny app uses National Ecological Observatory Network (NEON data portal) to compare analyte concentrations using Surface Water Chemistry (SWC) sampling across various NEON sites across the United States. Start by selecting a date range, aquatic field site, and two analytes to examine, then click Process Selection(s). More information on the data collection process can be found by clicking on SWC anywhere in this document. More information on the NEON project can by clicking here.
This app allows you to select an Aquatic Field site, dates, and two analytes to compare, one 'main' and another a secondary component. The data is pulled directly from the NEON database from the selected date range and plotted through time on a dual y-axis plot. The plot auto scales to allow for pattern detection between the various analytes. The plot is interactive and provides the exact sampling dates and concentrations.
The Correlation Report tab uses the 'main' analyte chosen and calculates correlation compared to every other analyte independently.
# list of analyes being evaluated
alist<- c('ANC','Br','Ca','Cl','CO3','conductivity','DIC','DOC','F','Fe','HCO3','K','Mg','Mn','Na','NH4 - N','NO2 - N','NO3+NO2 - N','Ortho - P','pH','Si','SO4','TDN','TDP','TDS','TN','TOC','TP','TPC','TPN','TSS','TSS - Dry Mass','UV Absorbance (250 nm)','UV Absorbance (280 nm)')
# empty table to store correlation results
correlation <- data.table(Analyte=character(),Correlation=numeric())
# main analyte selected data
analyte_data<- External.data %>%
filter(analyte == select.analyte) %>%
select(analyte, analyteConcentration, collectDate) %>%
arrange(collectDate)
i=1
while (i < length(alist)+1 ) {
# goes through all analytes and saves data
analyte_dataB<- External.data %>%
filter(analyte == alist[i]) %>%
select(analyte, analyteConcentration, collectDate) %>%
arrange(collectDate)
# joins analyte data together by collect date
analytes<- left_join(analyte_data, analyte_dataB, by= 'collectDate')
# saves correlation # into table
correlation[[i,2]]<- cor(analytes$analyteConcentration.x, analytes$analyteConcentration.y)
# saves 2nd analyte being evaluated into table
correlation[[i,1]]<- paste0(analytes$analyte.y[1])
i=i+1
}
# organizing data
correlation<- correlation %>%
arrange(desc(Correlation))
i=1
# if no value is produced...
while (i < length(alist)+1 ) {
if(is.na(correlation[[i,2]])) {
correlation[[i,2]]<- ('Not enough data')
}
i=i+1
}
# take out 'main' analyte that is being compared to the rest
correlationA<- correlation %>%
filter(Analyte != select.analyte)
Some sites may not have enough data points for certain analytes and do not produce a correlation result. Therefore the table will print 'Not enough data'
The Analysis tab plots out the two selected analyte concentrations through linear regression (lm). Below that the summary is taken for the relationship between the two analytes and the R^2 and P-value is printed out. Data points are paired by collection date of samples taken.
# portion of code for plotting linear regression of two analytes
sct_base<-ggplot(analytes,aes(y = analyteConcentration.y,x = analyteConcentration.x))
d.plot<- sct_base+geom_point()+
geom_smooth(method = "lm",se = F, color = "Red", show.legend = T, formula = 'y ~ x', na.rm = F)+
geom_smooth(color = "Grey", show.legend = T, inherit.aes = T)+
theme_classic()+
ggtitle(paste0(analytes$analyte.y[1],' vs ',analytes$analyte.x[1]))+
xlab(analytes$analyte.x[1])+
ylab(analytes$analyte.y[1])
# portion of code for summary of linear regression (for P-value)
mdl_1<-lm(analyteConcentration.y ~ analyteConcentration.x,data = analytes)
summary(mdl_1)
