(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Estimated reduction of nitrogen in streams of the Chesapeake Bay in areas with agricultural conservation practices [1]

['Andrew J. Sekellick', 'U.S. Geological Survey', 'Maryland-Delaware-Dc Water Science Center', 'Baltimore', 'Maryland', 'United States Of America', 'Scott W. Ator', 'Olivia H. Devereux', 'Devereux Consulting', 'Inc.']

Date: 2023-05

Spatial data provided by the U.S. Department of Agriculture National Resource Conservation Service representing implementation at the field-level for a selection of agricultural conservation practices were incorporated within a spatially referenced regression model to estimate their effects on nitrogen loads in streams in the Chesapeake Bay watershed. Conservation practices classified as “high-impact” were estimated to be effective (p = 0.017) at reducing contemporary nitrogen loads to streams of the Chesapeake Bay watershed in areas where groundwater ages are estimated to be less than 14-years old. Watershed-wide, high-impact practices were estimated to reduce nitrogen loads to streams by 1.45%, with up to 60% reductions in areas with shorter groundwater ages and larger amounts of implementation. Effects of “other-impact” practices and practices in areas with groundwater ages of 14 years or more showed less evidence of effectiveness. That the discernable impact of high-impact practices was limited to areas with a median groundwater age of less than 14 years does not imply that conservation practices are not effective in areas with older groundwater ages. A model recalibrated using high-impact agricultural conservation practice data summarized by county suggests effects may also be detectable using implementation data available at such coarser resolution. Despite increasing investment, effects of agricultural conservation practices on regional water quality remain difficult to quantify due to factors such as groundwater travel times, varying modes-of-action, and the general lack of high-quality spatial datasets representing practice implementation.

Funding: This work was funded by the U.S. Geological Survey (USGS) through the Priority Ecosystem Studies program and the Chesapeake Bay Program. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

In this paper, we estimate regional effects of “high-impact” and “other-impact” agricultural conservation practices on nitrogen fluxes in streams of the Chesapeake Bay watershed and include groundwater lag times as a component to the analysis. Spatially referenced regression (SPARROW) modeling [ 37 , 38 ] was used to relate spatially detailed records of agricultural conservation practices acquired from the U.S. Department of Agriculture (USDA) along with known sources and other watershed conditions to mean-annual flow-normalized nitrogen fluxes estimated at selected locations on watershed tributaries. The model was then used to estimate effects of conservation practices on nitrogen fluxes to the bay and in each of more than 80,000 tributary stream reaches within the bay watershed. An additional SPARROW model calibrated to agricultural conservation practices averaged over individual counties supports inferences about the importance of the spatial resolution of available information on conservation practices to such regional estimates. Implications of model estimates and results for adaptive management in the bay watershed and elsewhere are discussed.

Adaptive management in support of bay restoration under the 2010 total maximum daily load (TMDL) is complicated by uncertainty about the regional effects of agricultural management practices on water quality. In response to ecological and economic degradation since the mid-20 th century [ 15 – 19 ], the U.S. Environmental Protection Agency established a TMDL in 2010 under the Clean Water Act to limit nitrogen flux to Chesapeake Bay from its watershed [ 20 , 21 ]. A substantial fraction of the nitrogen input to the bay watershed and subsequently transported to Chesapeake Bay is attributable to agriculture [ 22 – 25 ]. Management practices intended to limit effects of agriculture on water quality have expanded substantially in the bay watershed in recent decades [ 6 , 14 , 26 ] and have been effective in some cases [ 12 , 27 – 29 ] but less so in others [ 15 , 30 – 33 ]. Although the net effects of such conservation actions suggest substantial expected declines in nitrogen fluxes in the watershed in recent decades [ 25 , 26 , 34 ], observed flow-normalized nitrogen trends in bay tributaries have been mixed in recent decades [ 35 ], and changes in the estimated flow-normalized flux of nitrogen from agriculture to the bay between the early 1990s and the early 2010s were minimal [ 8 , 36 ].

Multivariate empirical models have been used to relate selected agricultural conservation practices to nutrient fluxes in streams over large regions [ 1 , 8 – 10 ]. Such model applications may be complicated, however, by the difficulty in consistently estimating conservation practices over large areas and by the paucity of long-term monitoring data available for small streams at which effects of such practices may be most apparent [ 11 ]. In addition, the expected effects of conservation practices implemented widely in the Chesapeake Bay watershed to limit nitrogen losses from farmland to surface waters have not been fully realized in watershed streams. This is largely because a major transport pathway of nitrogen to streams is in the form of nitrate traveling through groundwater, which can take several decades in many parts of the watershed. Although satellite remote sensing may be used to map certain agricultural practices over large areas, data gathering has traditionally relied primarily on farmer surveys for more comprehensive accounting of conservation practices [ 12 , 13 ]. Information about agricultural practices for which the government provides financial assistance to private individuals generally is not available for specific locations but rather aggregated over counties or other relatively large areas in the United States [ 14 ]. The relatively coarse spatial resolution of such data may complicate efforts to isolate effects of conservation practices from those of source inputs, natural soil and hydrogeologic conditions, and other factors affecting nitrogen fluxes and concentrations in surface waters.

Effects of agricultural conservation practices on nitrogen in surface waters are often difficult to isolate or estimate. Even where monitoring data are available, isolating effects of specific practices on observed water-quality trends may be complicated by the substantial time periods (years to decades) often required for nitrogen transport from uplands to surface waters and by concurrent changes in land use, climate, nitrogen applications, and other watershed conditions [ 1 – 5 ]. Understanding effects of conservation practices on regional water quality over large areas presents additional challenges. Effects of conservation practices and other watershed changes are often particularly difficult to identify in large streams or over large heterogeneous regions and generally are most apparent in local streams or shallow groundwater draining small areas [ 1 , 5 ]. Monitoring to provide direct observation of such effects is uncommon [ 6 ] and generally of limited (< 4 year) duration [ 7 ].

Methods

Agricultural conservation practice data Agricultural conservation practice data were provided by the USDA National Resource Conservation Service (NRCS) in 2019 using methods described previously [14,39]. These NRCS data were taken from the National Planning and Agreements Database (NPAD). The NPAD contains data from Toolkit, ProTracts, National Easements Staging Tool (NEST), Service Center Information Management System (SCIMS), Integrated Data for Enterprise Analysis (IDEA), and the Performance Results System (PRS). The sharing of these data was made possible by the signing of a Conservation Cooperator Agreement between USGS and USDA (Farm Service Agency and Natural Resource Conservation Service) that allows access to federal conservation data for farm locations throughout the Chesapeake Bay watershed while ensuring the privacy of farmers as mandated under Section 1619 of the 2008 Farm Bill. Aggregated totals of the implementation data can be released to the public if five or more farmers are enrolled in the same reported practice or category of practices within the aggregated area. In this case, data were used as an input to SPARROW models, and model results showing the expected effect of the conservation practices are released. The dataset included conservation practices implemented between 2006 and 2012 that were expected to be functioning in 2012. Data included those practices where NRCS or FSA provided financial assistance, conservation enhancements, and those practices where NRCS provided only conservation technical assistance. Conservation technical assistance is any practice that is recommended by NRCS, meets NRCS technical standards, and is not funded by USDA. The conservation practice data were summarized and combined for management practices that are expected to have the most impact on reducing nitrogen applications and/or runoff to streams. These “high-impact” practices are those that convert cropped land to grass-based agriculture, various riparian buffer practices, enhanced nutrient management plans, and wetland creation, enhancement, and restoration practices. Other management practices designed or expected to have some impact on reducing nitrogen loads to streams were classified into an “other-impact” category. These practices include actions like cover crops, animal and grazing controls, certain types of runoff controls, and other agricultural management practices. The complete list of the “high-impact” and “other-impact” agricultural management practices used in this study are in Table 1. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Table 1. High-impact and other-impact U.S. Department of Agriculture agricultural conservation practices potentially affecting nitrogen transport in the Chesapeake Bay watershed, 2012. https://doi.org/10.1371/journal.pwat.0000108.t001 The high-impact practices are selected as the most effective for reducing nitrogen delivered to streams based on the Chesapeake Bay Program partnership’s Chesapeake Assessment Scenario Tool (CAST) 19 model analysis of the BMP pounds reduced and costs (conducted May 25, 2020 and published on the CAST website at https://cast.chesapeakebay.net/Documentation/CostProfiles) [40]. The analysis estimates the load reduction for each of the more than 200 best management practices (BMPs) that CAST models. An expert panel report establishes the reduction amount and method for each BMP. These recommendations are reviewed and adopted by the Chesapeake Bay Program partnership. The CAST model analysis isolates the effect of each BMP to determine the typical pounds reduced in an average hydrologic year. The analysis assumes there are interaction effects among the BMPs. These CAST BMPs were intersected with the NRCS conservation practices and the practices with the greatest direct impact were selected. These practices were implemented by 16,181 farmers on nearly 650 square kilometers throughout the Chesapeake Bay watershed. The average conservation practice intensity was determined for each NHDPlus Version 2 catchment [41]. Intensity was calculated as the sum of implementation in a catchment divided by the area of 2011 agricultural land using the National Land Cover Dataset (NLCD) [42]. More than one practice can be present on the same acre of land so the calculated intensity could be greater than one. For use in this analysis, intensity was capped at one, which was exceeded 2.0% of the time for the high-impact practice selection and 7.8% of the time for the other-impact practice selection. Urban BMPs were not included, and the intensity calculation is only represented by USDA certified agricultural BMPs that are measured in units of area or those that can be converted to area. Practices measured in units of length were converted to an area using the NRCS practice standard minimum width. Practices measured in “number” are not included. This excludes Animal mortality facility (practice code 316). Some Wetland restoration (practice code 657) is also measured in units of number and are excluded.

Spatial distribution of USDA data Implementation of the conservation practices selected for this study vary across the Chesapeake Bay watershed. This can be due to the types of agriculture in the region, physical landscape characteristics, and jurisdictional priorities [26]. Implementation of practices not included in the USDA NRCS-provided dataset also vary spatially for similar reasons in addition to different reporting standards and consistency [26]. Considering only the area within the Chesapeake Bay watershed, Delaware has the highest implementation intensity of high-impact category conservation practices, with 0.011 acres of implementation per acre of agricultural land cover. West Virginia has the largest amount of other-impact category BMPs per square kilometer of agricultural land cover, with an intensity value of 0.184 (Table 2). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Table 2. Agricultural land cover and U.S. Department of Agriculture agricultural conservation practices implementation in selected regions within the Chesapeake Bay watershed, 2012. https://doi.org/10.1371/journal.pwat.0000108.t002 Implementation of high-impact and other-impact BMPs also varies across the sub-watersheds of the estuary. Sub-watersheds are defined as areas upstream of USGS River Input Monitoring (RIM) stations [43]. The Appomattox watershed had the most implementation of both high-impact and other-impact management practices per square kilometer of agricultural land cover. Some watersheds, such as the Choptank, had relatively high intensities of high-impact practices and lower intensities of other-impact BMPs. Conversely, the Mattaponi watershed had a large amount of other-impact BMPs per square kilometer of agricultural land cover with a low amount of high-impact BMP intensity. The high-impact and other-impact conservation practices used in this analysis are implemented on agricultural lands, however, not all agricultural areas implement these practices. Effects of individual explanatory terms may be difficult to isolate in regression models (such as SPARROW) in cases where those terms are correlated or colinear with other terms [37]. Agricultural land cover in the bay watershed is weakly correlated with high-impact BMP implementation intensity (Spearman’s ρ = 0.238) and moderately correlated with other-impact BMP implementation intensity (ρ = 0.435), however, which suggests that collinearity should not overly complicate inclusion and interpretation of such terms in the models.

The SPARROW watershed model framework SPARROW is a spatially explicit nonlinear regression model used to identify and quantify linkages between in-stream water-quality, point and non-point contaminant sources, and other landscape characteristics within a watershed [37,38]. Calibration data, typically mean-annual contaminant loads at monitoring stations throughout a stream network, are empirically related to explanatory data representing sources and landscape characteristics affecting fate and transport of the constituent from uplands to streams. The SPARROW modeling framework is a useful tool for analyzing sources, fate, and transport of nutrients from uplands to streams and has been used to infer the effects of management actions on nitrogen loads in streams [1,8–10,22]. NHDPlus Version 2 was used for the digital stream network and consists of 85,620 stream reaches and associated local catchments within the Chesapeake Bay watershed [41]. The stream network is used to locate input data and route predicted in-stream contaminant loads downstream. A SPARROW model previously calibrated to quantify nitrogen sources and transport within the northeastern United States [22] was modified for use in this study. Most notably, the study area for this analysis was limited to streams that drain to the Chesapeake Bay estuary. Model specifications were also revised to improve analysis of BMP effects on nitrogen in streams specifically and to account for the smaller study area.

SPARROW model input data Nitrogen sources were specified in the model to include the mass of nitrogen from wastewater point sources, septic system effluent, atmospheric deposition of nitrogen, and developed, cropland, and pasture area [42,44,45]. Model coefficients for source terms expressed as a mass of nitrogen can be interpreted as a fraction of the average annual nitrogen mass from that source that reaches local streams [22,37,38]. For source terms represented as an area, model coefficients represent an annual average yield, or mass of nitrogen per unit area, from that source [22,37,38]. Using agricultural land cover (cultivated crops and pasture/hay) as sources allows the effect of BMPs on nitrogen yields from agricultural lands to be examined, improving interpretability for this specific analysis. The agricultural land cover classes were further divided based on the predominance (greater than or equal to 50 percent of area) of carbonate geology within a catchment because of fast groundwater travel-time in these areas relative to most other parts of the watershed. Areas underlain with carbonate rock have been associated with increased nitrogen yields [8,22,23,46–49]. Explanatory data representing factors that increase or decrease delivery of nitrogen from upland sources to local streams can be specified to interact with a subset of sources [8,37,38]. These land-to-water terms are log-transformed and mean-adjusted to improve interpretability of model calibration results [37]. Negative coefficient values for these variables indicate reduced transport of nitrogen from uplands to local streams; positive values indicate enhanced transport of nitrogen from upland to local streams [37,38]. Forest or wetland land cover [42], soil thickness, and runoff [45] were significant in the previous northeastern model [22] and were included in this model, along with high-impact implementation intensity in one model run and other-impact BMP implementation intensity in another model run (Fig 1). Tillage practices and cover crops were also significant in the northeastern model [22] but were omitted from these models in favor of the USDA implementation data. Cover crop implementation is included in the other-impact BMP classification; conservation tillage and no-till practices have been associated with increased delivery of nitrogen to streams and are not included in either classification [22,50]. Additionally, the land-to-water term representing areas underlain by carbonate rock was removed because these areas are accounted for in the source variables. Finally, mean-annual air temperature was not a significant predictor in model testing and was removed from consideration. This may be due to reduced temperature variability in the Chesapeake Bay watershed in comparison to the larger northeastern United States. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Watersheds showing implementation intensity of U.S. Department of Agriculture high-impact and other-impact best management practices (BMPs) by NHDPlus Version 2 catchment in 2012. https://doi.org/10.1371/journal.pwat.0000108.g001 Aquatic decay of nitrogen within streams and in impoundments are specified the same as in the northeastern SPARROW model. These decay terms are comprised of estimates of travel time in small streams, travel time in large streams, and inverse hydraulic load in impoundments [22].

Model specification, calibration and evaluation Separate models were specified to evaluate and quantify the importance of “high-impact” and “other-impact” BMPs on regional nitrogen flux within the Chesapeake Bay watershed. Flow-normalized and detrended estimates of annual nitrogen loads during 2012 under long-term average hydrologic conditions were used for calibration of the northeastern model [22] and for calibration of these models. Monitoring sites outside of the Chesapeake Bay watershed were removed. During construction of the Chesapeake Bay model, a calibration site from the northeastern SPARROW model was removed because it was suspected that point source loads in that monitoring station’s watershed were overestimated due to an incorrect wastewater treatment location. The models were specified to evaluate the importance of conservation practices to regional stream-water quality only in areas receiving relatively young groundwater contributions (Fig 2). Both high-impact and other conservation practice intensities were insignificant (p > 0.2) as land-to-water terms in initial model runs. After examination of model diagnostics and consideration of conservation practice dynamics within a watershed, a dataset of groundwater age estimates was incorporated into the model to account for lags between conservation practice implementation and resulting changes in water quality [40,51,52]. Groundwater lags are understood to be an important factor affecting the identification of conservation practice impacts on nitrogen in streams, delaying possible improvements or obscuring their effect over long periods of time when coupled with other watershed changes. Assuming that effects of conservation practices on contemporary stream water quality would be most apparent in areas with generally shorter travel times from upland application areas, implementation of high-impact and other-impact practices were limited in the models to areas with median groundwater ages of < 14 and < 6 years (respectively). These thresholds were selected based on the fit of successive model runs using different median groundwater age limits and account for 72.2 percent of the watershed for the model with high-impact practices. This captures a wide range of landscapes and types of agriculture. Only 14.4 percent of the watershed is accounted for in the model with other-impact practices (S1 Table). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Estimated median groundwater age and calibration stations used for model calibration [ Estimated median groundwater age and calibration stations used for model calibration [ 40 ]. https://doi.org/10.1371/journal.pwat.0000108.g002

County summarization model The site-specific conservation practice data used in this study are not publicly available [39]. Although methods for down-scaling county-level data to smaller watersheds or catchments exist, doing so with conservation practice data can be complicated due to the variety of practices and their modes of action [11]. An additional SPARROW model was calibrated with conservation practice high-impact intensity summarized by county to evaluate the importance of spatial resolution in conservation practice data to the detection of regional effects on water quality.

[END]
---
[1] Url: https://journals.plos.org/water/article?id=10.1371/journal.pwat.0000108

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.