An introduction to the rnrfa package

Claudia Vitolo and Ilaria Prosdocimi

2024-06-19

Introduction

The UK National River Flow Archive serves daily streamflow data, spatial rainfall averages and information regarding elevation, geology, land cover and FEH related catchment descriptors.

There is currently an API under development that in future should provide access to the following services: metadata catalogue, catalogue filters based on a geographical bounding-box, catalogue filters based on metadata entries, gauged daily data for about 400 stations available in WaterML2 format, the OGC standard used to describe hydrological time series.

The information returned by the first three services is in JSON format, while the last one is an XML variant.

The rnrfa package aims to achieve a simpler and more efficient access to data by providing wrapper functions to send HTTP requests and interpret XML/JSON responses.

Dependencies

R package dependencies can be installed running the following code:

# these should normally already be installed with rnrfa
install.packages(c("curl", "ggmap", "ggplot2", "httr", "jsonlite",
                   "lubridate", "parallel", "sf", "tibble", "zoo"))

This demo makes also use of external libraries. To install and load them run the following commands:

packs <- c("devtools", "DT", "leaflet", "dygraphs")
install.packages(packs, repos = "https://cloud.r-project.org")
lapply(packs, require, character.only = TRUE)

Installation

The stable version of the rnrfa package is available from CRAN:

install.packages("rnrfa")

Or you can install the development version from Github with remotes:

remotes::install_github("ilapros/rnrfa")

Now, load the rnrfa package:

library(rnrfa)

Functions

List of station identification numbers

The function station_ids() returns a vector of all NRFA station identifiers.

# Retrieve station identifiers:
allIDs <- station_ids()
head(allIDs)
## [1] 1001 2001 2002 3001 3002 3003

List of monitoring stations

The function catalogue() retrieves information for monitoring stations. The function, used with no inputs, requests the full list of gauging stations with associated metadata. The output is a tibble containing one record for each station and as many columns as the number of metadata entries available.

# Retrieve information for all the stations in the catalogue:
allStations <- catalogue()
head(allStations)
## # A tibble: 6 × 114
##      id name              `catchment-area` `grid-reference`$ngr easting northing
##   <int> <chr>                        <dbl> <chr>                  <dbl>    <dbl>
## 1  1001 Wick at Tarroul               162. ND2620254915          326202   954915
## 2  2001 Helmsdale at Kil…             551. NC9983918163          299839   918163
## 3  2002 Brora at Bruachr…             434. NC8915603916          289156   903916
## 4  3001 Shin at Lairg                 495. NC581062              258100   906200
## 5  3002 Carron at Sgodac…             241. NH4913692152          249136   892152
## 6  3003 Oykel at Easter …             331. NC4030500145          240305   900145
## # ℹ 110 more variables: `grid-reference`$easting <dbl>, $northing <dbl>,
## #   `lat-long` <df[,3]>, latitude <dbl>, longitude <dbl>, river <chr>,
## #   location <chr>, `station-level` <dbl>, `measuring-authority-id` <chr>,
## #   `measuring-authority-station-id` <chr>, `hydrometric-area` <int>,
## #   opened <chr>, closed <chr>, `station-type` <chr>, `bankfull-flow` <dbl>,
## #   `structurefull-flow` <dbl>, sensitivity <dbl>, `nrfa-mean-flow` <lgl>,
## #   `nrfa-peak-flow` <lgl>, `feh-pooling` <lgl>, `feh-qmed` <lgl>, …

The columns are briefly described below (see also API documentation):

Station filtering

The same function catalogue() can be used to filter stations based on a bounding box or any of the metadata entries.

# Define a bounding box:
bbox <- list(lon_min = -3.82, lon_max = -3.63, lat_min = 52.43, lat_max = 52.52)
# Filter stations based on bounding box
x <- catalogue(bbox)
dim(x); range(x$latitude); range(x$longitude)
## [1]   9 114
## [1] 52.43854 52.47522
## [1] -3.73172 -3.68999
# Filter based on minimum recording years
x <- catalogue(min_rec = 100)
dim(x); range(lubridate::year(x$`gdf-end-date`) - lubridate::year(x$`gdf-start-date`))
## [1]   6 114
## [1] 100 143
# Filter stations measuring a certain river
x <- catalogue(column_name="river", column_value="Wye")
dim(x); unique(x$river)
## [1]  11 114
## [1] "Wye"
# Filter based on bounding box & metadata strings
x <- catalogue(bbox, column_name="river", column_value="Wye")
dim(x); unique(x$river)
## [1]   2 114
## [1] "Wye"
# Filter stations based on threshold
x <- catalogue(bbox, column_name="catchment-area", column_value=">1")
dim(x); range(x$`catchment-area`)
## [1]   8 114
## [1]  1.08 10.55
# Filter based on minimum recording years
x <- catalogue(bbox, column_name = "catchment-area",
          column_value = ">1",
          min_rec = 30)
dim(x)
## [1]   7 114
# Filter stations based on identification number
x <- catalogue(column_name="id", column_value="== c(3001,3002,3003)")
x$id
## [1] 3001 3002 3003
# Other combined filtering
someStations <- catalogue(bbox,
                          column_name = "id",
                          column_value = "==c(54022,54090,54091,54092,54097)",
                          min_rec = 35)

Conversions

The rnrfa package allows convenient conversion between UK grid reference and more standard coordinate systems. The function osg_parse(), for example, converts the string to easting and northing in the BNG coordinate system (EPSG code: 27700), as in the example below:

# Where is the first catchment located?
someStations$`grid-reference`$ngr[1]
## [1] "SN853872"
# Convert OS Grid reference to BNG
osg_parse("SN853872")
## $easting
## [1] 285300
## 
## $northing
## [1] 287200

The same function can also convert from BNG to latitude and longitude in the WSGS84 coordinate system (EPSG code: 4326) as in the example below.

# Convert BNG to WSGS84
osg_parse(grid_refs = "SN853872", coord_system = "WGS84")
## $lon
## [1] -3.689987
## 
## $lat
## [1] 52.47065

osg_parse() also works with multiple references:

osg_parse(grid_refs = someStations$`grid-reference`$ngr)
## $easting
## [1] 285300 284233 284296 284514 283083
## 
## $northing
## [1] 287200 287655 287733 287266 286939

Get time series data

The first column of the table someStations contains the id number. This can be used to retrieve time series data and convert waterml2 files to time series object (of class zoo).

The National River Flow Archive mainly serves two types of time series data: gauged daily flow and catchment mean rainfall.

These time series can be obtained using the functions gdf() and cmr(), respectively. Both functions accept three inputs:

Other time series data are made available via the API and can be retrieved using get_ts() - for details on the time series which can be retrieved see ?get_ts.

Here is how to retrieve mean rainfall (monthly) data for Shin at Lairg (id = 3001) catchment.

# Fetch only time series data from the waterml2 service
info <- cmr(id = "3001")
plot(info)

One can also ask to retrieve some metadata with useful information:

# Fetch time series data and metadata from the waterml2 service
info <- cmr(id = "3001", metadata = TRUE)
info$meta
##             timestamp         interval station.id  station.name station.easting
## 1 2024-06-19T16:41:49 R816/1950-01/P1M       3001 Shin at Lairg          258100
##   station.northing station.latitude station.longitude data.type.id
## 1           906200         58.02222         -4.404049          cmr
##               data.type.name data.type.parameter data.type.units
## 1 Catchment Monthly Rainfall            Rainfall              mm
##   data.type.measurement.type data.type.period
## 1               Accumulation              P1M
plot(info$data, 
     main = paste("Monthly rainfall data for the", 
                info$meta$station.name,"catchment"), 
     xlab = "", ylab=info$meta$data.type.units)

Here is how to retrieve (daily) flow data for Shin at Lairg (id = 3001) catchment.

# Fetch only time series data
# info <- gdf(id = "3001")
# plot(info)

# Fetch time series data and metadata from the waterml2 service
info <- gdf(id = "3001", metadata = TRUE)
plot(info$data, 
     main = paste0("Daily flow data for the ", info$meta$station.name,
                   " catchment (",info$meta$data.type.units, ")"), 
     ylab = info$meta$data.type.name)

Multiple sites

By default, the functions gdf() and cmr() functions can be used to fetch time series data from multiple site in a sequential mode (using 1 core):

# Search data/metadata
s <- cmr(c(3002,3003), metadata = TRUE)

# s is a list of 2 objects (one object for each site)
plot(s[[1]]$data, 
     main = paste(s[[1]]$meta$station.name, "and", s[[2]]$meta$station.name), 
     sub = "Catchment monthly rainfall", ylab = s[[1]]$meta$data.type.units)
lines(s[[2]]$data, col = "green")

s <- get_ts(c(3002, 3003), type = "gdf", metadata = TRUE)
plot(s[[1]]$data, 
     main = paste(s[[1]]$meta$station.name, "and", s[[2]]$meta$station.name), 
     sub = "Gauged daily flow", ylab=s[[1]]$meta$data.type.units)
lines(s[[2]]$data, col="pink2")

Interoperability

Upgrade your data.frame to a data.table:

library(DT)
datatable(catalogue(column_name = "river", column_value = "Thames", all = FALSE))

Create interactive maps using leaflet:

library(leaflet)

leaflet(data = someStations) %>% addTiles() %>%
  addMarkers(~longitude, ~latitude, popup = ~as.character(paste(id,name)))

Interactive plots using dygraphs:

library(dygraphs)
dygraph(info$data) %>% dyRangeSelector()

Sequential vs Concurrent requests: a simple benchmark test

library(parallel)
# Use detectCores() to find out many cores are available on your machine
cl <- makeCluster(getOption("cl.cores", detectCores()))

# Filter all the stations within the above bounding box
someStations <- catalogue(bbox)

# Get flow data with a sequential approach
system.time(s1 <- gdf(someStations$id, cl = NULL))

# Get flow data with a concurrent approach (using `parLapply()`)
system.time(s2 <- gdf(id = someStations$id, cl = cl))

stopCluster(cl)

The measured flows are expected to increase with the catchment area. Let’s show this simple regression on a plot:

# Linear model
library(ggplot2)
ggplot(allStations[!is.na(allStations$qmed),], 
       aes(x = as.numeric(`catchment-area`), y = qmed)) +
  geom_point() +
  stat_smooth(formula = y ~ x, method = "lm", col = "red") +
  xlab(expression(paste("Catchment area [Km^2]", sep=""))) +
  ylab(expression(paste("Mean flow [m^3/s]", sep="")))