Large Data Pulls from Water Quality Portal - A Pipeline-Based Approach

A pipeline-based approach for making large data pulls from Water Quality Portal

Background

The Water Quality Portal (WQP ) database aggregates and standardizes discrete water quality data from numerous federal, state, tribal, and other monitoring agencies. The WQP enables the access and retrieval of over 297,000,000 water quality records (Read et al. 2017) through web services and an application programming interface (API) that can be called programmatically using the dataRetrieval package in R. Downloading data from the WQP represents a common pattern across USGS data teams.

In this post, we highlight an example data pipeline to increase the reusability, reproducibility, and efficiency of WQP data workflows. This post is an alternative method to the script-based workflow presented in Large sample pulls using dataRetrieval . We’ve designed it with large-scale data pulls in mind, but this example pipeline will work at any spatial or temporal scale.

Why targets?

The workflow described here uses the targets package to leverage modular functions, dependency tracking, and automated workflows to inventory and download data from the Water Quality Portal. Using targets allows the user to develop a maintainable pipeline that tracks changes over time and will only re-run portions of the workflow that are out of date due to those changes.

The basic ingredient of a targets workflow is a script file named _targets.R. This file is used to define and configure all of the steps in an analysis pipeline and their connections to each other. After connecting the individual analysis steps (also known as targets) in the _targets file, the targets package can then track the relationships between each connection. Analysis steps that precede any given step of interest are considered “upstream” while the analysis steps that follow are considered “downstream”. Once these connections have been established, targets can also visualize the relationships in a network graph like the one shown below. The WQP pipeline is structured so that various inputs - including the date range, spatial extent, parameters of interest, and/or specific arguments to pass along to WQP queries - can all be modified within the _targets.R file.

Pulling data from the Water Quality Portal

The WQP targets pipeline is divided into three phases that divide the workflow:

1. Inventory what sites and records are available in the WQP
2. Download the inventoried data
3. Harmonize , or clean, the downloaded data to prepare the dataset for further analysis

This post will focus on how to carry out bulk data pulls using the first two phases of the pipeline.

Decide which characteristic names to query

The first step is to decide which water quality parameters should be included in the data pull. One challenge posed by the aggregated WQP database is that many characteristic names may refer to the same water quality parameter (e.g., temperature records can take on different values for CharacteristicName, including "Temperature" or "Temperature, water"). The WQP pipeline includes a configuration file to help map various WQP characteristics onto more commonly-used parameter groups (1_inventory/cfg/wqp_codes.yml ).

Here we are interested in compiling temperature data, so we define this parameter group within _targets.R .

The characteristic names belonging to “temperature” are parsed in p1_char_names to create a vector of CharacteristicName values that are then used as input to the WQP query. Which values of CharacteristicName to include may vary depending on the needs of a specific project and which entries are considered valid in WQP, which can change over time. To accommodate this variability, the configuration file can be edited to omit certain characteristic names or include new ones. For example, we might think of a new characteristic name that contains temperature records and add that to the configuration file. This simple example illustrates a decision that users of WQP data must make, but that can sometimes be difficult to do so with confidence (i.e., is “temp” really a valid characteristic name?). Two pipeline features are designed to assist the user when making these decisions. First, all requested characteristic names from the configuration file are checked against a list of valid entries in WQP, and will notify the user if a characteristic name is not valid.

So we can see that our newly-added characteristic name “temp” is not a valid entry and can be omitted from the configuration file. Second, a user may wonder whether there are other characteristic names they might have missed. The pipeline includes a target p1_similar_char_names_txt that uses fuzzy string matching to check for valid characteristics that are similar to the requested parameters, and return an output file that can be evaluated by the user.

In p1_char_names we end up with a list of characteristic names to use in the query:

> tar_load(p1_char_names)
> p1_char_names
[1] "Temperature"               "Temperature, sample"       "Temperature, water"        "Temperature, water, deg F"
>

Define the area of interest

The next step is to define the spatial extent of the data pull. In this example, we are requesting data for a triangular “watershed” northwest of Philadelphia, PA, which we’ve specified in _targets.R using a set of latitude and longitude coordinates.

# Specify coordinates that define the spatial area of interest
# lat/lon are referenced to WGS84
coords_lon <- c(-77.063, -75.333, -75.437)
coords_lat <- c(40.547, 41.029, 39.880)

Although we use spatial coordinates, a user could also easily use other, predefined boundaries by replacing the targets p1_AOI and p1_AOI_sf with targets that download and read in an external shapefile:

# Download a shapefile containing the Delaware River Basin boundary
  tar_target(
    p1_shp_zip,
    {
      fileout <- "1_inventory/out/drbbnd.zip"
      utils::download.file("https://www.state.nj.us/drbc/library/documents/GIS/drbbnd.zip",
                  destfile = fileout,
                  mode = "wb", quiet = TRUE)
      fileout
    },
    format = "file"
  ),

# Unzip the shapefile and read in as an sf polygon object
  tar_target(
    p1_AOI_sf,
    {
      savedir <- tools::file_path_sans_ext(p1_shp_zip)
      unzip(zipfile = p1_shp_zip, exdir = savedir, overwrite = TRUE)
      sf::st_read(paste0(savedir,"/drb_bnd_arc.shp"), quiet = TRUE) %>%
        sf::st_cast(.,"POLYGON")
    }
  ),

In this example, the area of interest is relatively small and we could probably request all of the temperature data within the boundary of our triangular watershed without issue. However, what if we wanted to download WQP data for the full state of Pennsylvania? This would result in a much bigger request! The WQP pipeline is built around the central idea that smaller queries to the WQP are more likely to succeed and therefore, most workflows that pull WQP data would benefit from dividing larger requests into smaller ones.

One way we break up larger data pulls is by using a set of grid cells to define the spatial extent of multiple, smaller queries that, when combined, represent a data inventory for the full area of interest. The size of each grid cell can be customized by the user, but using 1 degree cell sizes results in a set of five grid cells that overlap our example watershed:

Inventory the data before downloading

We use targets “branching” capabilities to apply (or map) our data inventory and download functions over each grid cell that overlaps our area of interest. The user can quickly reference the number of sites and records that were returned in the inventory by referencing a saved log file. If using the pipeline along with git for version control, this log file also allows a user to readily track changes to the data over the time.

Little by little: break up the inventoried sites to prepare for download

The initial data inventory lets us know how many sites and records we can expect for each unique CharacteristicName in our query. Before actually downloading the data, we bin the inventoried sites within each grid cell into distinct download groups so that the total number of records in any given download group does not exceed a user-specified maximum threshold (defaults to 250,000 records per download group ). This binning step acts as another safeguard against timeout issues with large data requests, but also allows us to take advantage of targets dependency tracking to efficiently build or update the data pipeline. For example, we do not have to re-download all of the data records from WQP just because we added a new characteristic name or because new sites were recently uploaded to WQP and detected in our inventory. targets will only update those data subsets that become “outdated” by an upstream change.

Download the data!

We’re finally ready to download the data from the Water Quality Portal and do so by mapping the function fetch_wqp_data() over each unique download group recombining the data in p2_wqp_data_aoi. As a check on our downloaded data, we compare the expected number of sites and records from p1_wqp_inventory_summary_csv against the number of sites and records that were actually downloaded, and inform the user of the result:

Updating the data pull

As we mentioned above, targets dependency tracking allows us to efficiently update our pipeline or expand our region of interest without re-pulling grids that have already been queried. To illustrate, say we decide to expand an analysis to include areas west of our initial focal watershed. The map below shows that the area of interest now overlaps six grid cells instead of five, but the five original grids are still included in our query. Using common scripting workflows we would usually just re-pull all of the data again even though that is time-consuming.

However, targets recognizes that data has already been inventoried for five of these grid cells and will only query data for the newly-added grid, 46902:

Customizing the pipeline

Our goal is for you to take this example pipeline and tailor it to your own projects. Users can customize the spatial extent, the date range, the list of water quality parameters of interest, and add new functions for harmonizing data for various water quality constituents.

Disclaimer

Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

References

Read, E. K., Carr, L., De Cicco, L., Dugan, H. A., Hanson, P. C., Hart, J. A., Kreft, J., Read, J. S., and Winslow, L. A. (2017), Water quality data for national-scale aquatic research: The Water Quality Portal, Water Resour. Res., 53, 1735– 1745, doi:10.1002/2016WR019993 .