The short version
Nitrogen that entered groundwater years ago keeps travelling toward streams and bays, so the load reaching the water today reflects past land use as well as present sources. USGS scientists Kalle Jahn and Donald Walter asked how much the answer depends on the way groundwater nitrogen loads are estimated. Using a 120-year model of the Long Island, New York aquifer (1900 to 2019), they compared a simple method with no groundwater lag against two simulations that include it, for four water bodies in and around the Peconic Estuary. The methods agreed within about 10% in some places and decades, and differed by as much as ±100% in others, depending on how long groundwater takes to reach each water body and on how nitrogen sources changed over time. In the model, cutting land-based nitrogen cut loads to the water only gradually: reaching a 70% reduction took from 19 to more than 61 years, depending on the water body.
What the researchers did
The authors built one groundwater model of Long Island and ran it three ways, so that differences in results came from the method and not from different data.
- Nitrogen inputs for every year from 1900 to 2019 came from earlier USGS work (Monti et al., 2024) on six sources: septic systems, air deposition, home and golf course fertilizer, farm fertilizer, livestock waste and pet waste. Each was reduced by set fractions for losses at the surface, such as plant uptake.
- The instantaneous method adds up the nitrogen applied on the land that drains to each water body in a given year, with no travel time through the aquifer.
- The steady-state flow simulation moves nitrogen through the aquifer with one fixed average flow pattern, which captures travel time but not year-to-year changes in rainfall recharge and well pumping.
- The transient flow simulation moves nitrogen through an aquifer whose flow changes every year with recharge and pumping, the most complete of the three.
- The water bodies were the Peconic River, Great Peconic Bay, the eastern part of Long Island Sound and the eastern South Shore Estuary (Shinnecock, Moriches and Narrow Bays). The model assumed the aquifer is mostly oxygenated, so nitrate moves through it without being broken down.
- The simulated Peconic River loads were checked against stream gage data (available from the 1970s on). Two future scenarios ran from 2020 to 2119: 2019 sources continuing, and all land-based sources removed (air deposition kept) as an extreme case to show the fastest possible response.
What they found
- Sometimes close, sometimes far apart. Depending on earlier nitrogen and hydrologic conditions, historical loads from the numerical simulations differed from the instantaneous estimates by less than 10% in some periods and by as much as ±100% in others (Jahn & Walter, 2026). For Great Peconic Bay and the South Shore Estuary the methods typically agreed within 10% from 1920 to 1980, when the share of nitrogen going to short and long groundwater paths stayed steady.
- Long travel times pull loads below the simple estimate. In eastern Long Island Sound, where agriculture keeps adding nitrogen and more than half of the input goes to groundwater paths taking over 20 years (about 20% over 100 years), the steady-state simulation ran up to 50% below the instantaneous estimate.
- Short travel times can push loads above it. From 1930 to 1970 the Peconic River got large nitrogen inputs, including from duck farms, on land close to the river, and much of it reached the river within 10 years. Because older nitrogen was still arriving too, the steady-state simulation gave loads well above the instantaneous estimate. The peak was 683 kg/day against 313 kg/day.
- The aquifer keeps filling. Simulated nitrogen stored in the Long Island aquifer nearly doubled from 1900 to 2019. It kept growing after nitrogen inputs levelled off around 1980 and sewers cut septic inputs, because groundwater travel times are longer than the 120-year simulation.
- Year-to-year swings need the transient model. Only the transient simulation reproduced the yearly ups and downs in measured Peconic River loads. After 1970 both simulations broadly matched the gage data, with average errors of 30.6% (transient) and 35.2% (steady-state). The authors note earlier work linking yearly changes in groundwater flow to brown tide blooms.
- Loads lag sources by decades. If 2019 sources simply continued, simulated loads kept rising for a century and approached, but mostly stayed below, the instantaneous 2019 load. By 2119 they reached about 89% (Peconic River), 88% (Long Island Sound) and 82% (South Shore Estuary) of it, and 435 kg/day for Great Peconic Bay against 431.
- Recovery times differ by water body. With all land-based sources removed, a 70% load reduction took 19 years in the South Shore Estuary, 48 in the Peconic River, 51 in Great Peconic Bay and more than 61 in Long Island Sound. Half the reduction took about 13 years in the Peconic River and Great Peconic Bay and 32 in Long Island Sound. This removal case is unrealistic and shows the fastest possible response.
- Loads track the history of algal blooms. All three methods followed the broad record of coastal problems on Long Island, including 1950s green tides tied to duck farming, brown tides in Great Peconic Bay from the mid-1980s and red tides and algal blooms from 2002 to 2006.
What it means for Waquoit Bay today
This study was done on Long Island, New York, in a different watershed from Waquoit Bay, so its numbers do not transfer to the Bay. What carries over is the lesson on timing. When a water body gets much of its nitrogen from groundwater, sewering or other cuts on land show up in the water only as older groundwater is replaced, and the delay depends on how long groundwater takes to reach the water. The authors also note that simple, no-lag load estimates have been used to set nitrogen limits for Cape Cod estuaries, where groundwater travel times are short and the aquifer is thin, and that such estimates tend to hold up better in that kind of setting. They describe Long Island, with its oxygenated aquifer, as an end-member case, so the delays seen there are likely at the long end.
Cape Cod work has looked at this timing before. A USGS model of Mashpee's Popponesset Bay watershed examined how fast sewers would reduce nitrogen loads, EPA scientists modeled the slow groundwater pipeline to Cape Cod's bays, and USGS tracked a Falmouth neighborhood after it was sewered. Read with those, the Long Island comparison is a reminder to judge Falmouth and Mashpee projects over decades, and to expect that water bodies fed by longer groundwater paths will respond more slowly.
Read the paper
Jahn, K. L., & Walter, D. A. (2026). Modeling legacy nitrogen transport under instantaneous, steady-state, and transient groundwater flow conditions. *Water Resources Research*, 62(7), e2025WR041077. https://doi.org/10.1029/2025wr041077
License: The article is open access under a Creative Commons Attribution (CC BY) license, and it is a U.S. Government work in the public domain in the USA. The PDF is hosted here under CC BY. This summary is based on the full paper. The card image is a title card made from the citation.
Drafted with AI assistance from the linked source.
