How it's built

One static page, the whole public record

This atlas gathers every public water-quality observation for the Carson River Basin into one map, with the streamflow, snowpack, population, news, and published science around it. There is no server behind it and no database: a data pipeline in R does the work ahead of time, the browser reads small pre-built files, and the live dashboard asks the agencies directly.

Where the data comes from

• Water Quality Portal (EPA and USGS), via the WQX 3.0 services. Every organization that has ever published a sample in the basin appears here: federal, state, tribal, and volunteer programs.
• USGS NHDPlus for the stream network and waterbodies, and the Watershed Boundary Dataset for the basin and its subwatersheds.
• USGS streamflow at every gauge in the basin and NRCS SNOTEL and snow courses back to the 1940s, for the time-lapses and the live dashboard.
• Nevada Administrative Code 445A for the numeric standards that apply at each control point, parsed from the published code and drawn on every chart where a site sits on a reach with a table; criteria protect the beneficial uses, the requirements to maintain existing higher quality are stricter. On the Pyramid Lake Paiute Reservation the Tribe's EPA-approved standards govern instead, from its Water Quality Control Plan, for the lower Truckee and the lake itself. California sites carry the Lahontan Basin Plan objectives: the region-wide pH, oxygen, and turbidity objectives, plus the site-specific tables for the Truckee, Lake Tahoe, Carson, and Walker hydrologic units.
• EPA ATTAINS for the states' assessments: which reaches and lakes are listed as impaired, for what, and whether a TMDL exists, from the latest integrated reports. The USGS Network Linked Data Index ties each gauge to its reach for the live flow view.
• USGS groundwater levels: every static depth-to-water reading at every well the Survey has measured, so a well opens to its water table through time and its trend.
• Census block counts for 2000, 2010, and 2020, assigned to subwatersheds, and county estimates back to 1970 for the people layer. NIFC fire perimeters, the USGS Publications Warehouse, and the National Weather Service for the stories and the dashboard.
• Census TIGER for counties and towns, OpenFreeMap for the basemap, and AWS Terrain Tiles for the hillshade and the flight. All of it public domain or openly licensed, none of it behind an API key.

The pipeline

1. 1
   Pull and preserve. An R pipeline built on targets pulls the record one characteristic at a time, subdividing requests until the portal answers, and stores each pull as a dated Parquet snapshot. Raw values stay as text: the archive is testimony, and typing is interpretation applied later, once, visibly.
2. 2
   Archive. Snapshots copy to append-only object storage with content-hash verification. Nothing is ever overwritten. Package versions are pinned with renv, so any machine rebuilds the same result.
3. 3
   Shape. A second script joins measurements to sites, computes per-site statistics and yearly medians, dissolves the stream network for labels, and writes one small JSON record per monitored site. QGIS is the cartographic workbench alongside it.
4. 4
   Show the record, then test on request. A site opens to every sample as measured, nondetects drawn hollow at their limits, nothing removed and nothing fitted. Only when asked does a trend engine run the Seasonal Kendall test with a Sen slope on every site and parameter with at least 20 samples over 10 years: month is the season, so seasonal cycles never masquerade as trends, and nondetects enter as tied lowest ranks rather than being dropped. Every result carries its p-value, method, and period, and the map can be colored by it. The stories use the same discipline: sites downstream of a fire, a mine, or a restoration are traced through the stream network and compared against sites elsewhere, before and after, and the page writes its verdict from the numbers.
5. 5
   Tile and ship. Every layer packs into a single PMTiles archive. The browser fetches only the byte ranges it needs, so the whole basin loads from one file on any static host.

Why it is built this way

• It costs almost nothing to run and it never goes down. Static files on a CDN, no server to patch, no monthly bill to justify.
• Every view is a link. The map's state lives in the URL, so a site, a parameter, and a zoom can be shared, cited, or embedded in a report.
• Nothing here knows it is the Carson. The watershed is a configuration file. The same pipeline and the same page build the atlas for any basin with a public record.

Stack

R · targets · renv · dataRetrieval · sf · nhdplusTools · arrow · QGIS · tippecanoe · PMTiles · MapLibre GL JS · plain HTML, CSS, and JavaScript · Census PL 94-171 · NRCS AWDB · USGS Water Data API · NWS · NIFC