Field study · Trinity stormwater pond · Centreville, VA
Every Saturday from May through August, 2023–2026, we fished the Trinity retention pond off Route 29 and recorded what we caught, hour by hour. Paired with public NOAA weather records, that log becomes a small but real dataset about how an engineered pond responds to summer heat.
38.84°N, 77.43°W · Sat 11:00–17:00 · 61 sessions logged
Abstract
Small urban stormwater retention ponds collect runoff from neighborhood roads and rooftops. Through summer they swing in temperature and water chemistry as that runoff arrives, which can stress the fish that keep the pond functioning. These basins are rarely instrumented, so the data needed to understand them usually does not exist.
Rather than wait for sensors that were never installed, we built the record by hand: a structured catch census taken on a fixed schedule, then joined to the nearest official weather station (NOAA, Washington Dulles). The result is a reproducible look at how daily conditions move catch composition and rate.
How to read this site
The honest limit of a backyard study is that you can't measure everything. So this site marks the difference everywhere it appears, in one consistent visual language:
No in-pond water sensor was ever available for this basin. Where you see the dashed amber treatment, the number is a model output we built and intend to validate — never a measurement dressed up as one. That distinction is the point of the project.
Skills demonstrated
Defined a fixed-schedule sampling protocol (same site, same hours, same method) so the record stays comparable across three years.
Cleaned 427 hourly observations and joined them to 1,270 days of NOAA station data on date keys.
Tested catch against temperature, rainfall, wind, and time of day; reported the real relationships, including the null ones.
Built a transparent, clearly-labeled model to estimate pond water temperature where direct measurement wasn't possible.
Built and deployed as a static site (HTML / CSS / Vercel). See the research and the data →