Using the geoconnex Reference Feature Server with R

Introduction

Below we demonstrates how to use the Geoconnex Reference Feature server API with R’s spatial packages, particularly sf. We’ll explore various use cases for working with hydrological and related spatial data. These use cases can be groups into three categories:

1. Finding identifiers and spatial geometries for real-world features
2. Finding features that are hydrologically related to a feature of interest (if that information is available)
3. Finding datasets related to a feature of interest

This API conforms to the OGC-API Features specification, and its full API documentation is available here for those who wish to explore its capabilities further than what is demonstrated below.

Setup

First, let’s load the necessary libraries and set up our API base URL.

Code

library(sf)
library(dplyr)
library(httr)
library(mapview)
library(jsonlite)
library(knitr)
library(DT)

base_url <- "https://reference.geoconnex.us"

Use Case 1: Identifying features and their spatial geometry

The Geoconnex Reference Feature server is a source for identifiers and geometries for real-world features that many organizations may collect and publish data about. The attributes of these features vary, but all include a “uri” which serves as a persistent identifer. First, let’s discover what kinds of features are available:

Code

url <- paste0(base_url, "/collections?f=json")
collections <- jsonlite::fromJSON(url)

datatable(collections$collections)

We see a number of options available, including watershed boundaries like the Hydrologic Unit Codes, administrative boundaries like counties, states, and public water systems, and hydrologic features like mainstems (rivers) and aquifers, and hydrometric features such as dams and gages. The reference feature server lets us find features according to attribute and spatial queries. In general, this is accomplished by passing queries of the form https://reference.geoconnex.us/collections/<collectionId>/items?filter= Let’s see how to use these collections!

Attribute Queries

Fuzzy text match

Let’s say we’re interested in a specific river, one we think is called the “Animas river”. We can pass an attribute filter query to the mainstems collection”, and then use the R sf package to read into a dataframe, and the mapview package to visualize.

Code

# construct a query for a river name that includes the string "animas"
query <- URLencode("name_at_outlet ILIKE '%animas%'") 
url <- paste0(base_url, "/collections/mainstems/items?f=json&filter=", query)

# Read the data into an sf object
animas_rivers <- st_read(url, quiet = TRUE) 

Warning in CPL_read_ogr(dsn, layer, query, as.character(options), quiet, : GDAL
Message 1: GDAL was built against curl 7.79.1, but is running against 7.77.0.
Runtime failure is likely !

Code

# Display the results
animas_rivers |> 
  select(uri, name_at_outlet, outlet_drainagearea_sqkm) |>
  datatable()

Code

# Map the results
mapview(animas_rivers |> 
  select(uri, name_at_outlet), zcol = "name_at_outlet")

There are evidently 3 rivers that include the word “Animas”. Let’s say we were interested in the “Animas River”, shown on the map in Green. We find that it’s Geoconnex URI is https://geoconnex.us/ref/mainstems/35394.

Logical and quantitative

We can also do filters based on logical and quantitative filters on attributes. Let’s say we wanted to find all rivers with drainage areas (in this reference dataset, the attribute is outlet_drainagearea_sqkm) greater than 1,000,000 square kilometers:

Code

# construct a query for a river with outlet_drainagearea_sqkm > 600,000
query <- URLencode("outlet_drainagearea_sqkm > 500000")
url <- paste0(base_url, "/collections/mainstems/items?f=json&filter=", query)

# Read the data into an sf object
large_mainstems <- st_read(url, quiet = TRUE)

# Display the results
large_mainstems |> 
  select(uri, name_at_outlet, outlet_drainagearea_sqkm) |>
  datatable()

Code

# Map the results
mapview(large_mainstems, zcol = "name_at_outlet")

Combining Queries

Queries over multiple attributes can also be made, combining with ‘AND’ or ‘OR’. For example, let’s find all Dams that include the name “Hoover”, but then also filter to only the ones with a drainage area of more than 100,000 square kilometers:

Code

# Step 1: Find all dams named "Hoover"
query_hoover <- URLencode("name LIKE '%Hoover%'")
url_hoover <- paste0(base_url, "/collections/dams/items?f=json&filter=", query_hoover)
hoover_dams <- st_read(url_hoover, quiet = TRUE)

cat("Number of dams named 'Hoover':", nrow(hoover_dams), "\n")

Number of dams named 'Hoover': 39 

Code

# Create an interactive table of all Hoover dams
datatable(
  hoover_dams |>
    st_drop_geometry() |>
    select(uri, name, drainage_area_sqkm) |>
    arrange(desc(drainage_area_sqkm)),
  options = list(pageLength = 10, scrollX = TRUE),
  caption = "All Dams Named 'Hoover'",
  rownames = FALSE
)

Code

# Step 2: Query for large Hoover dams using a combined filter
query_large_hoover <- URLencode("name LIKE '%Hoover%' AND drainage_area_sqkm > 100000")
url_large_hoover <- paste0(base_url, "/collections/dams/items?f=json&filter=", query_large_hoover)
large_hoover_dams <- st_read(url_large_hoover, quiet = TRUE)

cat("\nNumber of large Hoover dams (Drainage Area > 100,000 sq km):", nrow(large_hoover_dams), "\n")

Number of large Hoover dams (Drainage Area > 100,000 sq km): 1 

Code

# Create an interactive table of large Hoover dams
datatable(
  large_hoover_dams |>
    st_drop_geometry() |>
    select(uri, name, drainage_area_sqkm) |>
    arrange(desc(drainage_area_sqkm)),
  options = list(pageLength = 10, scrollX = TRUE),
  caption = "Large Dams Named 'Hoover' (Drainage Area > 100,000 sq km)",
  rownames = FALSE
)

Code

# Create a map view of all Hoover dams, highlighting the large ones
m <- mapview(hoover_dams |>
    select(uri, name, drainage_area_sqkm), layer.name = "All Hoover Dams", label="name")
m + mapview(large_hoover_dams |>
    select(uri, name, drainage_area_sqkm), color = "red", col.regions="red", layer.name = "Large Hoover Dams", lwd=2, cex=15, label="Hoover")

We found 39 Dams in the US named “Hoover”, but only 1 with a large drainage area, the famous one near Las Vegas, NV.

Spatial Queries

Bounding box

We can also do spatial queries, using bounding box queries (min lon, min lat, max lon, max lat) or by passing WKT-encoded geometry. Let’s say we want to find all Public Water Systems within a bounding box around the four-corners region.

Code

# Define the bounding box for the Four Corners area
# Format: (min Longitude, min Latitude, max Longitude, max Latitude)
bbox <- c(-109.5, 36.5, -107.5, 37.5)

# Construct the URL with the bbox parameter
url <- paste0(base_url, "/collections/pws/items?f=json&bbox=", paste(bbox, collapse = ","))

# Read the data into an sf object
four_corners_pws <- st_read(url, quiet = TRUE)

# Display summary of the results
cat("Number of Public Water Systems found:", nrow(four_corners_pws), "\n")

Number of Public Water Systems found: 75 

Code

# Create an interactive table of the results
four_corners_pws |>
  st_drop_geometry() |>
  select(uri, pws_name, population_served_count) |>
  arrange(desc(population_served_count)) |>
  datatable(
    options = list(pageLength = 5, scrollX = TRUE),
    caption = "Public Water Systems in the Four Corners Area",
    rownames = FALSE
  )

Code

# Create a map view of the results
m <- mapview(four_corners_pws, zcol = "population_served_count", layer.name = "Population Served", label= "pws_name")

# Add the bounding box to the map
bbox_poly <- st_as_sf(st_as_sfc(st_bbox(c(xmin = bbox[1], ymin = bbox[2], xmax = bbox[3], ymax = bbox[4]), crs = 4326)))
m + mapview(bbox_poly, col.region = "red", alpha.regions=0, color="red", lwd=2, layer.name = "Query Bounding Box")

Intersection

When it comes to spatial queries, we are not restricted to bounding box queries. We can pass any spatial predicate along with WKT geometries to a collection filter. Let’s say we have several field sites near Farmington, NM, and we want to identify which HUC10 watersheds they fall within. We’ll use the point-in-polygon query capability of the INTERECTS spatial; predicate to find this information:

Code

# Define our field site (example coordinate near Farmington, NM)
site_lon <- -108.2186
site_lat <- 36.7280

# Construct the query
query <- sprintf("INTERSECTS(geometry, POINT(%f %f))", site_lon, site_lat) |> URLencode()
url <- paste0(base_url, "/collections/hu10/items?f=json&filter=", query)

# Make the API call
huc10 <- st_read(url, quiet = TRUE) |>
  select(id,uri,name)



# Display the results table
datatable(huc10)

Code

# Create a map
site_point <- st_point(c(site_lon, site_lat)) |> 
  st_sfc(crs = 4326) |>
  st_sf()

mapview(huc10, zcol = "name", layer.name = "HUC10 Watershed") +
  mapview(site_point, col.regions = "red", layer.name = "Field Site")

Here we see that our field site is in the HUC10 1408010505, which has the associated Geoconnex URI https://geoconnex.us/ref/hu10/1408010505. This identifier can be used if we were to publish data about our site, following Geoconenx guidance and best practices.

Intersection by URL reference

Complex features can have many coordinates, and thus requests via ?filter to the API can be too long to format as URL. To get around this, the API supports a special intersection process that involves passing a URL for any GeoJSON feature to a collection. Let’s say we want to know which counties intersect the Animas River (https://geoconnex.us/ref/mainstems/35394).

Code

# Define the process endpoint
process_url <- "https://reference.geoconnex.us/processes/intersector/execution"

# Define the input parameters
input_params <- list(
  inputs = list(
    url = "https://geoconnex.us/ref/mainstems/35394",
    collection = "counties"
  )
)

# Execute the process
response <- POST(
  url = process_url,
  body = toJSON(input_params, auto_unbox = TRUE),
  add_headers("Content-Type" = "application/json"),
  encode = "json"
)


  # Convert the result to an sf object
  intersecting_counties <- st_read(response, quiet = TRUE)
  
  
  # Create an interactive table of the results
  intersecting_counties |>
    st_drop_geometry() |>
    select(name, uri) |>
    datatable(
      options = list(pageLength = 5, scrollX = TRUE),
      caption = "Counties Intersecting the Animas River"
    )

Code

  # Fetch the Animas River geometry
  animas_river <- st_read("https://geoconnex.us/ref/mainstems/35394", quiet = TRUE)
  
  # Create a map view of the results
  mapview(intersecting_counties, zcol = "name", layer.name = "Intersecting Counties") +
    mapview(animas_river, color = "blue", layer.name = "Animas River")

Note that of the three counties intersecting the Animas River, two are named “San Juan”, https://geoconnex.us/ref/counties/08111 in Colorado, and https://geoconnex.us/ref/counties/35045 in New Mexico, highlighting the importance of unique identifiers and the usefulness of HTTP identifiers that direct to spatial/data representations of a given feature.

Use Case 2: Finding hydrologically related features

One of Geoconnex’ main use cases is to support the discovery of hydrologically related monitoring locations. That’s why one of the key uses of the Geoconnex API is to serve relationships between key hydrometric locations like Dams and Streamggages and mainstem rivers. Both the dams and gages collections include the attribute mainstem_uri that allows one to find all such features that are on a given mainstem. In addition, dams and gages have the attribute subjectof that directs to the URL of a source data system’s record about the dam or streamgage. Below we demonstrate finding all dams and gages on the Animas River:

Code

# Animas River mainstem URI
animas_uri <- "https://geoconnex.us/ref/mainstems/35394"

# Function to query features by mainstem URI
query_by_mainstem <- function(collection, mainstem_uri) {
  query <- URLencode(sprintf("mainstem_uri = '%s'", mainstem_uri))
  url <- paste0(base_url, "/collections/", collection, "/items?f=json&filter=", query)
  st_read(url, quiet = TRUE)
}

# Query dams and gages
animas_dams <- query_by_mainstem("dams", animas_uri)
animas_gages <- query_by_mainstem("gages", animas_uri)

# Function to create a hyperlink in HTML
create_link <- function(url) {
  ifelse(!is.na(url) & url != "",
         sprintf('<a href="%s" target="_blank">Link</a>', url),
         "N/A")
}

# Create interactive tables
animas_dams <- animas_dams |>
#  select(uri, name, subjectof) |>
  mutate(subjectof = sapply(subjectof, create_link)) 


datatable(animas_dams, options = list(pageLength = 5, scrollX = TRUE),
            caption = "Dams on the Animas River",
            escape = FALSE)  # Allow HTML in the table

Code

animas_gages <- animas_gages |>
  select(uri, name, subjectof) |>
  mutate(subjectof = sapply(subjectof, create_link))

datatable(animas_gages,options = list(pageLength = 5, scrollX = TRUE),
            caption = "Gages on the Animas River",
            escape = FALSE)  # Allow HTML in the table

Code

# Fetch the Animas River geometry
animas_river <- st_read(animas_uri, quiet = TRUE)

# Create a map view of the results
mapview(animas_river, color = "blue", layer.name = "Animas River") +
  mapview(animas_dams, col.regions = "red", layer.name = "Dams") +
  mapview(animas_gages, col.regions = "green", layer.name = "Gages")

Use Case 3: Finding datasets realted to hydrologic features

The most important use case of the Geoconnex system for data users is to discover datasets related to a given hydrologic feature. This functionality is the reason for implementing a Knowledge Graph approach. There are two ways we currently offer users to discover datasets:

1. Convenience references to “datasets” within Reference Feature Collections (currently implemented only for mainstems)
2. Offering full SPARQL access to the graph database at https://graph.geoconnex.us

Convenience “dataset” reference in API

Each mainstem’s GeoJSON response includes a nested array within the “datasets” attribute which can be extracted. In the example below, 58 datasets are available about the Animas River. These can be filtered downstream of the API call. For example, in the interactive data table, we can use the search bar to searchf or mention of “temperature”. Of the 58, 6 datasets are regarding the variableMeasured “Temperature”, and these appear to be 3 pairs of USGS datasets from 3 distinct USGS monitoring locations, with each pair having a 2 different download options.

Code

# Animas River mainstem URI
animas_uri <- "https://geoconnex.us/ref/mainstems/35394"

# Fetch the Animas River data
response <- GET(animas_uri, query = list(f = "json"))


animas_data <- content(response, "text") |> fromJSON()
  
# Extract datasets
datasets <- animas_data$properties$datasets
datatable(datasets,
          options = list(pageLength = 5, scrollX = TRUE, 
                         search = list(regex=TRUE, caseInsensitive = TRUE, search = 'temperature')
                         )
          )

SPARQL

For those who wish to make SPARQL queries directly, the endpoint is https://graph.geoconnex.us/repositories/iow. For example, one would query the graph for all datasets schema:about monitoring locations that are on (<hyf:referencedPosition/hyf:HY_IndirectPosition/hyf:linearElement>) the Animas River, that have a value for the schema:variableMeasured that includes the string “temperature”.

Code

# SPARQL endpoint
endpoint <- "https://graph.geoconnex.us/repositories/iow"

# Revised SPARQL query
query <- 'PREFIX schema: <https://schema.org/> 
PREFIX gsp: <http://www.opengis.net/ont/geosparql#> 
PREFIX hyf: <https://www.opengis.net/def/schema/hy_features/hyf/> 

SELECT DISTINCT ?monitoringLocation ?siteName ?datasetDescription ?type ?url
       ?variableMeasured ?variableUnit ?measurementTechnique ?temporalCoverage 
       ?distributionName ?distributionURL ?distributionFormat ?wkt
WHERE {
  VALUES ?mainstem { <https://geoconnex.us/ref/mainstems/35394> }
  
  ?monitoringLocation hyf:referencedPosition/hyf:HY_IndirectPosition/hyf:linearElement ?mainstem ;
                      schema:subjectOf ?item ;
                      hyf:HydroLocationType ?type ;
                      gsp:hasGeometry/gsp:asWKT ?wkt .
  
  ?item schema:name ?siteName ;
        schema:temporalCoverage ?temporalCoverage ;
        schema:url ?url ;
        schema:variableMeasured ?variableMeasured .
  
  ?variableMeasured schema:description ?datasetDescription ;
                    schema:name ?variableMeasuredName ;
                    schema:unitText ?variableUnit ;
                    schema:measurementTechnique ?measurementTechnique .
  
  OPTIONAL {
    ?item schema:distribution ?distribution .
    ?distribution schema:name ?distributionName ;
                  schema:contentUrl ?distributionURL ;
                  schema:encodingFormat ?distributionFormat .
  }
  
  # Filter datasets by the desired variable description
  FILTER(REGEX(?datasetDescription, "temperature", "i"))
}
ORDER BY ?siteName
'

# Execute the SPARQL query
response <- GET(
  url = endpoint,
  query = list(query = query),
  accept("application/sparql-results+json")
)

# Parse the JSON response
  result <- content(response, "text", encoding = "UTF-8") %>% fromJSON()
  
  # Extract the results
  datasets <- as.data.frame(result$results$bindings) |>
    mutate(across(everything(), ~.$value))
  
  datatable(datasets)