River Basin Export Reduction Optimization Support Tool (RBEROST) User Guide, v1.15


EPA/600/B-21/170
December 2021
www.epa.gov/ord

River Basin Export Reduction Optimization Support Tool

(RBEROST) User Guide, v1.15

Atlantic Coastal Environmental Sciences Division
Environmental Measurement and Modeling Division
Narragansett, Rhode Island 02882

Office of Research and Development
Center for Environmental Measurement and Modeling

-------
River Basin Export Reduction Optimization Support Tool
(RBEROST) User Guide, vl.15

12/08/2021

Table of Contents

1	Contributors and Contacts	3

2	List of Acronyms	4

3	Model Framework	6

3.1	Optimization Method	8

3.2	Optimization Variables and Constraints	8

3.3	Obj ecti ve Functi on	10

3 .4 Optimization Parameters	11

4	Model Use	22

4.1	Getting Started	22

4.2	Preprocessing	28

4.2.1	User Specifications	29

4.2.2	Preprocess Data Inputs	33

4.2.3	Write AMPL Model Files	35

4.3	NEOS Server	38

4.4	Postprocessing	45

4.4.1	Necessary Files	46

4.4.2	Preview Files	46

4.4.3	Display Results	46

4.4.4	Download Detailed Results	52

5	Model Sensitivity	53

6	Data Dictionary	66

References	83

2

-------
1 Contributors and Contacts

Catherine Chamberlin (ORISE - EPA): contributor to RBEROST vl and Upper Connecticut
Basin case study; chamberlin.catherine@epa.gov

Naomi Detenbeck (EPA): lead PI; detenbeck.naomi@epa.gov

Marilyn ten Brink (EPA): co-PI; tenbri.nk.mariIvn@epa.gov

Alyssa Le (ICF): contributor to WMOST Scaled-up Optimization vl and Upper Connecticut
Basin case study

Kate Munson (ICF): contributor to WMOST Scaled-up Optimization vl and Upper Connecticut
Basin case study

Isabelle Morin (ICF): contributor to WMOST Scaled-up Optimization vl
Miranda Marks (ICF): contributor to WMOST Scaled-up Optimization vl
Yishen Li (ICF): contributor to Upper Connecticut Basin case study

3

-------
2 List of Acronyms

ACESD - Atlantic Coastal Environmental Sciences Division
ACRE - Agricultural Conservation Reduction Estimator
AMPL - A Mathematical Programming Language
BMP - Best Management Practice

CEMM - Center for Environmental Measurement and Modeling
COMID - Common Identifier

CPLEX - Optimizer based on the simplex method as implemented in the C programming
language

CSV - Comma Separated Values

DELFRAC - Delivery Fraction

EPA - Environmental Protection Agency

EQIP - Environmental Quality Incentives Program

HDMA - Hydrologic Derivatives for Modeling and Analysis

HSG - Hydrologic Soil Group

HUC - Hydrologic Unit Code

HUC10 - 10-digit Hydrologic Unit Code

HUC 12 - 12-digit Hydrologic Unit Code

HUC8 - 8-digit Hydrologic Unit Code

LOESS - locally estimated scatterplot smoothing

MA - Massachusetts

MS4 - Municipal Separate Stormwater Sewer
N - Nitrogen

NATSGO - National Soil Survey Geographic Database

NEIWPCC - New England Interstate Water Pollution Control Commission

NH - New Hampshire

NHD - National Hydrography Dataset

NHDPlus - National Hydrography Dataset Plus

NLCD - National Land Cover Dataset

4

-------
NPDES - National Pollution Discharge Elimination System

NRCS - Natural Resources Conservation Service

ORD - Office of Research and Development

ORISE - Oak Ridge Institute for Science and Education

P - Phosphorus

PI - Principal Investigator

R - Programming language

RBEROST - River Basin Export Reduction Optimization Support Tool
RShiny - R package providing graphical user interface for user inputs and outputs
Se - Standard Error

SPARROW - Spatially Referenced Regressions On Water

TDEP - Total Nitrogen Deposition

TN - Total Nitrogen

TP - Total Phosphorus

USGS - United States Geological Survey

UVM - University of Vermont

VT - Vermont

WEDB - Watershed and Estuarine Diagnostic Branch
WMOST - Watershed Management Optimization Support Tool
WQv - Water Quality Volume
WWTP - Wastewater Treatment Plant
XML - Extensible Markup Language

5

-------
3 Model Framework

The River Basin Export Reduction Optimization Support Tool (RBEROST) is a decision support
tool designed to support integrated, regional, watershed planning. The tool is designed to help
managers reduce nutrient loading to targeted waterbodies for the least financial cost. This tool
optimizes costs for meeting targets for nutrient export at the annual scale and is designed to be
used as a screening tool for large watersheds (e.g., HUC 6 - HUC 8 scale). The tool is
mathematically similar to the Watershed Management Optimization Support Tool, or WMOST
(Detenbeck, ten Brink, et al., 2018; Detenbeck, Piscopo, et al., 2018). WMOST is built for
optimization at the HUC 10 or HUC 12 scale and can be run on daily or monthly timesteps.
RBEROST was developed as a regional screening tool to overcome computational challenges
with running WMOST at larger spatial scales. All RBEROST scripts are written in R 4.0.5. and
may display incompatibilities with other versions. Execution of the RBEROST application is
recommended with RStudio version 1.4.1106 and may display incompatibilities with other
versions. Additional R packages may need to be installed by the user. The R code and
instructions on how to execute the code necessary for these installations are included in the
documentation below.

RBEROST allows users to screen which locations and choices of Best Management Practices
(BMPs) will meet annual loading targets for the least financial cost. There are three main steps
within RBEROST including a preprocessing step, an interaction with an online server, and a
postprocessing step. Additional work may be necessary before beginning RBEROST to collect
and format the necessary data. The preprocessing step combines medium-resolution National
Hydrography Dataset Plus (NHDPlus v2; McKay et al. (2012)) reach lengths, NHDPlus v2
catchment-level annual nutrient loading, land use data, hydrologic soil group data, nitrogen (N)
deposition data, user-defined loading targets, user-defined agricultural, urban, point source and
riparian buffer BMPs, and data on BMP-specific costs and nutrient removal efficiencies. The
preprocessing step then uses this information to write three program files in A Mathematical
Programming Language, or AMPL, including a model, data, and command file. There are two
options available for the preprocessing step, i.e., creating AMPL files with or without uncertainty
information included. The AMPL files describe a model that defines a cost-minimization
optimization problem subject to meeting downstream annual loading targets. These files are then
sent to a free online CPLEX server hosted by the Network-Enabled Optimization System
(NEOS; University of Wisconsin in Madison (2021)). CPLEX is a linear solver that will solve
the optimization problem defined by the user inputs of loading targets and selected BMPs for the
least cost. Once the solution is optimized, the model outputs its decisions of which BMPs to
implement, and where to implement them. The final step of RBEROST is the postprocessing
step. This step parses the output from NEOS into a summary report describing which BMPs were
implemented. It also provides csv files for download that describe the BMP choices on the
NHDPlus v2 catchment-level scale. If the user chooses to run the model with uncertainty, the
postprocessor will provide information on the expected cost range, as well as the likelihood of
meeting each loading target. When run with uncertainty, the model can display multiple
scenarios with increasing cost and likelihood for meeting targets. Figure 3.1 shows a diagram of
the model framework.

-------
User specified
BMP specs (csv)

User specified
loading targets
(csv)

BMP Costs: EQIP,
WMOST, WWTP
Retrofits Report

AMPL Model File
(option for uncertainty)

Legend:

Data



Optimization Support File
User Input/Interaction
Result
Model

Summary
Report

BMP Efficiencies: ACRE,

WMOST, WWTP
Retrofits Report, Green
Credits Report for
Riparian Buffers,
NATSGO HSG grids,
HDMA slope grids

Baseline Loadings:

2012 Regional
SPARROW model,

TDEPTN
depositional grids,
NPDES loading
reports

Land Use: StreamCat
Cropland and
Imperviousness,
Building footprint
impervious area,
NLCD summaries of
riparian buffers

Preprocessor
(option for
uncertainty)

(option for
uncertainty)

Optimization
Solver

BMP selection
at minimized
cost

Postprocessor

AMPL Command File
(option for uncertainty)

CSV files of BMP
implementation in
each catchment

Figure 3.1: Schematic overview of RBEROST, made up of data, support file, user input, result, and model components

-------
3.1 Optimization Method

RBEROST uses the IBM ILOG CPLEX Optimizer for linear programming that solves a
mathematical problem written in AMPL to minimize the total annualized cost of selected BMPs.
RBEROST interacts with this solver through the NEOS server, hosted by the University of
Wisconsin-Madison. The CPLEX Optimizer was chosen because it accepts XML calls so that
RBEROST can interact with it directly, but users also have the option of manually interacting
with the server through the webpage at https://ri.eos-
server, org/neos/sol vers/lp: C P LEX/AM PI ,.html.

3.2 Optimization Variables and Constraints

RBEROST includes four categories of BMPs:

1. Point Source BMPs. Point source BMP options include water treatment plant low-cost
retrofits to facilitate nutrient removal in wastewater effluent.

2. Urban BMPs. Urban BMP options include practices applied to developed land and may
serve functions such as increasing evaporation from standing surface water, infiltration of
ponded water into soil media, percolation of infiltrated water into groundwater, filtration
of particulate matter, denitrification, or outflow through an orifice or weir, among others.

3. Agricultural BMPs. Agricultural BMP options include practices applied to agricultural
land and may serve functions such as slowing runoff flow velocities from cropland areas,
increasing infiltration into underlying soils, routing runoff through pools and basins, and
adjusting fertilizer application or other farming practices to slow and reduce nutrient
transport to waterbodies, among others. RBEROST only treats row crop area with
agricultural BMPs, and 'ag' throughout the code and documentation refers only to row
crop area.

4. Riparian Buffer BMPs. Riparian Buffer BMP options include the conversion of land
within riparian areas to either herbaceous/grassed or forested land. Such practices slow
water as it approaches the stream or river and increases infiltration. Nutrients may then be
removed by soil processes.

RBEROST can consider removal of both total nitrogen (TN) and total phosphorus (TP)
simultaneously. Table 3.1 summarizes the optimization variables included in the model and their
associated constraints. Constraints on the point source BMPs are binary, such that the model
chooses to implement or not to implement point source BMPs (in this case, low-cost wastewater
treatment plant retrofits) based on the associated cost and removal efficiency. Constraints on the
urban and agricultural BMPs relate to the fraction of each land area that is treated. The model
chooses to treat a fraction of urban or agricultural land area with each BMP based on the
associated cost and removal efficiency, however the sum of all fractions must be between 0 and
1. No agricultural or urban land can be treated by two BMPs. Additional constraints exist on
infiltration-based BMPs, as the model will not implement them in catchments with very low
infiltration values, and on porous pavement BMPs, as the model will only implement these on
roadways and parking lots. Constraints on riparian buffer BMPs relate to the length (in ft) of
stream bank available to be treated, where riparian BMPs can only be implemented on currently
un-buffered stream lengths. Initial conditions for the optimization model reflect current practices,
where the selected point source, urban, and agricultural BMPs have not yet been implemented

-------
and the load is set to the baseline annual nutrient load delivered by the upstream reaches to the
target waterbodies.

Table 3.1: A summary ofRBEROST model variables, constraints, and initial conditions

Optimization

Optimization

Constraint

Constraint

Initial

Variable

Variable Name

Description



Conditions

Description

in AMPL Model
File







Per-catchment

agBMP bin

Choose

Binary (0 or 1)

0

agricultural



whether to





BMP selection



implement





Per-catchment

urbanBMP bin

Choose

Binary (0 or 1)

0

urban BMP



whether to





selection



implement





Point Source

pointdec

Choose

Binary (0 or 1)

0

BMP



whether to
implement





Urban BMP

urban frac

Fraction of
urban and
treated

Fraction >= User
Specified Min * Fraction
urban area that is suitable.
Fraction <= User
Specified Max * Fraction
urban area that is suitable.
Sum Fractions <= 1. Sum
Fractions for pavement
BMPs <= Fraction urban
area that is roads

0

Agricultural

agfrac

Fraction of

Fraction >= User

0

BMP



row crop
land treated

Specified

Min. Fraction <= User
Specified Max. Sum
Fractions <= 1



Riparian Buffer

ripbuflength

Length of

Length <= Unbuffered

0

BMP



stream reach
treated

Stream length * User
specified max fraction.
Removal along all Length
<= Riparian loads. Sum
all lengths <= total
unbuffered bank length



9

-------
3.3 Objective Function

The objective function minimizes BMP implementation costs while achieving a reduction in
baseline nutrient loadings to the specified targets. Minimized costs are a function of the costs to
implement point source, urban, agricultural, and riparian buffer BMPs in the catchments
included in the optimization model (Eq. 3.1).

Cost-Minimized

In Eq 3.1 and following equations, i = I...n are NHDPlus catchments, j = 1 ...p are agricultural
BMPs, k = l...q are urban BMPs, I = l...r are Riparian Buffer BMPs, all costs are in 2019 USA
dollars, and individual parts of Eq. 3.1 are defined below in Eq. 3.2, Eq. 3.3, Eq. 3.4, Eq. 3.5, and
Eq. 3.6 .

CostAgricultural,I,j = (Fraction TreatedLand,ij) * (Agricultural Land Areat) *

Ag Cost Adjustment * (Capital Costsij + O&M Costsij)	(3.2)

CostAgricuituranj describes the costs of implementing BMP j in row crop agricultural fields.
Fraction TreatedLand,i,j ranges 0-1 and is treated as a variable in the optimization problem,
Agricultural Land Areat is the amount of rowcrop area in catchment i, Ag Cost Adjustment
reflects the difference in Environmental Quality Incentives Program (EQIP) base payments and
actual costs of agricultural BMPs, Capital Costsij represent the annualized capital costs of BMP
j, and O&M Costsij represent the annual operations and maintenance costs of BMP j. Costs may
differ by catchment i.

CostUrbanik = (Capital Costsk + O&M Costsk) * Urban Cost Adjustment * WQvk

CostUrbanii k describes the cost of treating stormwater runoff from urbanized land in catchment i
with stormwater BMP k. Captial Costsk and O&M Costsk are base costs for urban BMPs and do
not differ by catchment. The Urban Cost Adjustment value scales urban costs between 1- and 3-
times base costs based on the expected amount of retrofitting as determined by the intensity of
urban development in catchment i (Hill, et al., 2016; Voorhees, 2016; Yang, et al., 2018). WQvk
is the water quality treatment volume (in cubic feet) of runoff that can be treated by BMP k
defined as

WQvk = Urban design depthk * Rvt * (Urban Land Area^ * {Fraction TreatedLand,i,k)

Urban design depthk is user-specified and determines the treatment capacity of stormwater BMP
k. Urban design depthkis defined by the user in inches and is converted to feet within
RBEROST. Rvi is the runoff coefficient for catchment i calculated as 0.05 + 0.009 *

Percent Site Imperviousnessi (Schueler, 1987; Vermont Agency of Natural Resources, 2017).

^ (CostdflTicitlturalji,./) _ (^05^f/rban,i,fc)

+ Costp0intsource,i

(3.1)

(3.3)

(3.4)

10

-------
Urban Land Area, is the area of urbanized land in catchment i (in ft2), and FractionrreatedLandj.k
ranges 0-1 and is treated as a variable in the optimization problem.

CostPointSourcei = (0,1); * (Capital CostSi + 0&.M CostSi)	(3.5)

CostPointSource,i describes the total costs of retrofitting WWTPs given a binary decision (0,1), of
whether to implement. This binary decision is treated as a variable in the optimization model.
Capital Costs and O&M Costs refer to annual operations and maintenance costs, and differ by
WWTP in catchment i.

Costpiparian Buffers,i.i Length Treated Bank,i,i * Cost Adjustment * (Capital CostSi i +

O&M CostSi i)	(3.6)

CostRiparian Buffers,ui describes the costs of converting riparian zones into forested or grassed
buffers, LengthrreatedBankxiis a variable in the optimization and ranges from 0 ft to twice the
length of stream reach (reflecting total bank length in ft) in catchment i. Capital CostSi,i and
O&M CostSij are base costs from EQIP, and are adjusted with Ag Cost Adjustment as in Eq. 3.2.

3.4 Optimization Parameters

Sources for BMP costs (and other parameters) in the Upper Connecticut case study are given in
Table 3.2. The water quality volume (WQV), or treated volume, reflects Vermont Water Quality
Treatment Standards (Vermont Agency of Natural Resources, 2017). Table 3.2 summarizes
parameters in RBEROST and their sources for the case study described later (Ator, 2019;
Detenbeck, ten Brink, et al., 2018; Detenbeck, Piscopo, et al., 2018; Heris et al., 2020; Hill et al.,
2015; Houle et al., 2019; Jin et al., 2019; JJ Environmental, 2015; McKay et al., 2012; National
Atmospheric Deposition Program, 2021; New Hampshire Department of Environmental
Services, 2020a, 2020b; Soil Survey Staff, 2020a, 2020b; U.S. Department of Agriculture,
Natural Resources Conservation Service, n.d.; U.S. Department of Agriculture Staff, 2021; UVM
Spatial Analysis Lab, 2019; Verdin, 2017; Vermont Agency of Natural Resources, 2017;
Voorhees, 2016; White et al., 2019; Yang et al., 2018). Table 3.3 summarizes additional
parameters used for uncertainty analysis and their data sources (Ator, 2019; Dell et al., 2016; Hill
et al., 2015; Houle et al., 2019; Jin et al., 2019; McKay et al., 2012; New Hampshire Department
of Environmental Services, 2020a, 2020b; Schueler, 1987; Soil Survey Staff, 2020a, 2020b; U.S.
Department of Agriculture Staff, 2021; U.S. Geological Survey, National Geospatial Program,
2020; U.S. Geological Survey & U.S. Department of Agriculture, Natural Resources
Conservation Service, 2013; Verdin, 2017; Vermont Agency of Natural Resources, 2017; White
et al., 2019; Wickham et al., 2017).

11

-------
Table 3.2: A summary ofRBEROST parameters.

Optimization Parameter	Optimization Parameter Name in AM PL

Description	Model and Data Files

Nitrogen baseline annual average baseloads_N 1.. .baseloadsNn
loading data per catchment to
each TN target

Phosphorus baseline annual	baseloadsP 1.. .baseloadsPn

average loading data per
catchment to each TP

Nitrogen baseline annual riparian riparianload N 1... riparianload Nn
loading data per catchment to
each TN target

Phosphorus baseline annual	riparianload P 1...riparianload Pn

riparian loading data per
catchment to each TP target

Available urban and row crop area
area per catchment

Percent of Urban land that is urbanbmpimplementationpotential
suitable for BMP implementation

12

Data Source

Northeastern Regional SPARROW model (Ator, 2019) annual
average loadings output (ne_sparrow_model_output_tn.txt)
modified with pollutant discharge data from NPDES Permit
No. NHO100200 (New Hampshire Department of
Environmental Services 2020a)

Northeastern Regional SPARROW model (Ator, 2019) annual
average loadings output (ne_sparrow_model_output_tp.txt)

Methodology from Pollutant Removal Credits for Buffer
Restoration in MS4 Permits Final Panel Report (Houle et al.
2019) using land cover data from the National Landcover
Database (Jin et al. 2019, Yang et al. 2018) and river reach
shape files from NHDPlus V2 (McKay et al. 2012)

Methodology from Pollutant Removal Credits for Buffer
Restoration in MS4 Permits Final Panel Report (Houle et al.
2019) using land cover data from the National Landcover
Database (Jin et al. 2019, Yang et al. 2018) and river reach
shape files from NHDPlus V2 (McKay et al. 2012)

Urban area: Northeastern Regional SPARROW Model (Ator,
2019) input data (ne_sparrow_model_input.txt) - "urban_km2"
field. Row crop/Agricultural area: Percent cropland data from
the 2011 NLCD database (catchment-specific data per state
downloaded from StreamCat; Hill et al. 2015) -
"PctCrop201 ICat" field

Hydrologic soil group: gNATSGO (Soil Survey Staff, 2020a;
Soil Survey Staff, 2020b) and USDA NRCS 2009. River Reach
shape files from NHD plus v2 (McKay et al. 2012). Building
footprints from a national dataset of rasterized building
footprints for the U.S. (Heris et al., 2020). High resolution
impervious area from the Vermont Base Landcover 2015
database (UVM Spatial Analysis Lab, 2019)

-------
Optimization Parameter	Optimization Parameter Name in AM PL

Description	Model and Data Files

Length of streambank that is not unbuffered_banklength
already buffered per catchment

Total stream bank length (2 x total banklength
reach length) per catchment

Capital costs associated with ag costs capital
implementing agricultural BMPs

Operations and maintenance	ag costs operations

costs associated with
implementing agricultural BMPs

Costs associated with
implementing wastewater
treatment plant retrofits for
nutrient removal

Costs associated with
implementing urban BMPs

Capital costs associated with
implementing riparian buffer
BMPs

Operations and maintenance
costs associated with
implementing riparian buffer
BMPs

pointcosts

urbancosts
ri pbufcostscapital

ri pbufcostsoperati on s

13

Data Source

Land cover data from the National Landcover Database (Jin
et al. 2019, Yang et al. 2018) and river reach shape files
from NHDPlusV2 (McKay etal. 2012)

River reach shape files from NHDPlus V2 (McKay et al. 2012)

U.S. Department of Agriculture; Environmental Quality
Incentives Program state-by-state cost sheets from 2020. For
the case study, Massachusetts costs were used when New
Hampshire or Vermont costs were not available (U.S.
Department of Agriculture Staff, 2021)

U.S. Department of Agriculture; Environmental Quality
Incentives Program state-by-state cost sheets from 2020. For
the case study, Massachusetts costs were used when New
Hampshire or Vermont costs were not available (U.S.
Department of Agriculture Staff, 2021)

NE1WPCC Final Report - Low Cost Retrofits for Nitrogen
Removal at Wastewater Treatment Plants in the Upper Long
Island Sound Watershed (JJ Environmental, 2015)

WMOST Theoretical Documentation (Detenbeck, ten Brink et
al., 2018; Table 7-2)

U.S. Department of Agriculture; Environmental Quality
Incentives Program state-by-state cost sheets from 2020.

U.S. Department of Agriculture; Environmental Quality
Incentives Program state-by-state cost sheets from 2020.

-------
Optimization Parameter	Optimization Parameter Name in AM PL

Description	Model and Data Files

Calculated urban volumetric	runoff coeff urban

runoff coefficient, based on
percent impervious area per
catchment

Adjustments to urban BMP costs urbancostadjustmentcoef

based on the expected amount of :

retrofitting (proportional to

relative intensity of urban land

per catchment)

Agricultural nitrogen removal
efficiency per BMP option and
HUC12

WWTP-specific nitrogen
removal efficiency

Urban nitrogen removal
efficiency per BMP option and
catchment

Agricultural phosphorus removal
efficiency per BMP option and
HUC12

WWTP-specific phosphorus
removal efficiency

ag_effic_N

point_effic_N

urban effic N

agefficP

pointefficP

14

Data Source

Imperviousness data from National Land Cover Database,
2011 (catchment-specific data per state downloaded from
StreamCat; Hill et al. 2015) - "Pctlmp201 ICat" field. Data
used to calculate urban runoff coefficient, based on Vermont
Stormwater Manual equation (0.05+0.009*PctTmp201 ICat;
Vermont Agency of Natural Resources, 2017)

Development intensity data by catchment from the National
Land Cover Database, 2011 (catchment-specific data per state
downloaded from StreamCat; Hill et al. 2015) -
"PctUrbOp201 ICat," "PctUrbLo201 ICat,"

"PctUrbMd201 ICat," and "PctUrbHi201 ICat" fields. Cost
adjustment factors from Opti-Tool documentation (Voorhees,
2016)

"Fert_20" and "Manurelnjection" BMPs: Efficiencies
provided by EPA via NRCS. Remaining agricultural BMPs:
Efficiencies summarized by HUC12 (or if unavailable, HUC10
and/or HUC8) based on ACRE database (White et al., 2019)

NE1WPCC Final Report - Low Cost Retrofits for Nitrogen
Removal at Wastewater Treatment Plants in the Upper Long
Island Sound Watershed (JJ Environmental, 2015)

New Hampshire MS4 Permit BMP Performance Curves (New
Hampshire Department of Environmental Services, 2020b).
Hydrologic soil group: gNATSGO (Soil Survey Staff, 2020a;
Soil Survey Staff, 2020b; USDANRCS 2009). Land cover
data from the National Landcover Database (Jin et al. 2019,
Yang et al. 2018)

"Fert_20" and "Manure lnjection" BMPs: Efficiencies
provided by EPA via NRCS. Remaining agricultural BMPs:
Efficiencies summarized by HUC12 (or if unavailable, HUC10
and/or HUC8) based on ACRE database (White et al., 2019)

Not applicable

-------
Optimization Parameter	Optimization Parameter Name in AM PL

Description	Model and Data Files

Urban phosphorus removal	urbanefficP

efficiency per BMP option and

catchment

Total nitrogen removal by	riparianremoval N 1.. .riparianremoval Nn

riparian buffers per catchment for :
each TN target

Total phosphorus removal by ; riparianremovalP 1... riparianremovalPn
riparian buffers per catchment for j
each TP target

Total nitrogen load allowed after
reduction for all TN targets

Total phosphorus load allowed
after reduction for all TP targets

loads_lim_N 1.. ,loads_lim_Nn
loads lim PI...loads lim Pn

15

Data Source

New Hampshire MS4 Permit BMP Performance Curves (New
Hampshire Department of Environmental Services, 2020b).
Hydrologic soil group: gNATSGO (Soil Survey Staff, 2020a;
Soil Survey Staff, 2020b; USDANRCS 2009). Land cover
data from the National Landcover Database (Jin et al. 2019,
Yang et al. 2018)

Annual removal rates are a function of riparian loading and
nutrient removal efficiency of riparian buffers. Loading is
dependent on the amount of urbanized land within a 400 ft
riparian zone, and efficiency is dependent on slope and
hydrologic soil group (Houle et al. 2019). Land cover data
from the National Landcover Database (Jin et al. 2019, Yang et
al. 2018). River reach shape files from NHDPlus V2 (McKay
et a.l 2012). Hydrologic soil group: gNATSGO (Soil Survey
Staff, 2020a; Soil Survey Staff, 2020b). Slope data from
Hydrologic Derivatives for Modeling and Analysis by the
USGS (Verdin, 2017)

Annual removal rates are a function of riparian loading and
nutrient removal efficiency of riparian buffers. Loading is
dependent on the amount of urbanized land within a 400 ft
riparian zone, and efficiency is dependent on slope and
hydrologic soil group (Houle et al. 2019). Land cover data
from the National Landcover Database (Jin et al. 2019, Yang et
al. 2018). River reach shape files from NHDPlus V2 (McKay
et al. 2012). Hydrologic soil group: gNATSGO (Soil Survey
Staff, 2020a, 2020b). Slope data from Hydrologic Derivatives
for Modeling and Analysis by the USGS (Verdin, 2017)

User supplied

User supplied

-------
Optimization Parameter
Description

The sum of SPARROW model
annual nitrogen average loads
delivered to each TN target that
are not associated with point
sources, urban area, or row crop
area (e.g., loads from septic
sources)

The sum of SPARROW model
annual phosphorus average loads
delivered to each TN target that
are not associated with point
sources, urban area, or row crop
area (e.g., loads from bedrock
leaching)

Agricultural BMP costs reflected
in ag costs represent base
payment costs (75% of actual
costs). The agcost frac parameter
reflects 100% of costs (100/75 =
1.33).

Conversion from acre-ft to cubic
ft

Conversion for precipitation:
inches to ft

Parameter to ensure that
agricultural BMPs are
implemented on at least some
minimum acreage. Currently set
to 0.

Minimum allowed percent
implementation of urban BMPs
per catchment

Optimization Parameter Name in
Model and Data Files

other loads Nl...other loads Nn

other loads PI...other loads Pn

agcost frac

acfttoft3

pep

agBMPminarea

urban frac min

Data Source

Northeastern Regional SPARROW model (Ator, 2019) annual
average loadings output (ne_sparrow_model_output_tn.txt)
modified with N depositional data from the National
Atmospheric Deposition Program (National Atmospheric
Deposition Program, 2021)

Northeastern Regional SPARROW model (Ator, 2019) annual
average loadings output (ne_sparrow_model_output_tp.txt)

Correspondence with EPA and NRCS state conservationist for
Vermont

Not applicable

Not applicable

Not applicable

User supplied

-------
Optimization Parameter
Description

Maximum allowed percent
implementation of urban BMPs
per catchment

Minimum allowed percent
implementation of agricultural
BMPs per catchment

Maximum allowed percent
implementation of agricultural
BMPs per catchment

Minimum allowed percent
implementation of riparian buffer
BMPs per catchment

Maximum allowed percent
implementation of riparian buffer
BMPs per catchment

The specified design depth for
each urban BMP

Optimization Parameter Name
Model and Data Files

urbanfracmax

agfracmin

agfracmax

ri pb uffracmi n

ripbuffracmax

urbandesigndepth

Data Source

User supplied

User supplied
User supplied
User supplied
User supplied
User supplied

-------
Table 3.3: A summary ofRBEROST parameters used for uncertainty analyses.

Optimization Parameter Optimization Parameter Name in AM PL Model
Description	and Data Files

Standard error of nitrogen baseloads N l _se.. .baseloadsNn se
baseline annual average
loading data per catchment
to each TN target

Standard error of nitrogen baseloadsP l_se.. .baseloadsPnse
baseline annual average
loading data per catchment
to each TP target

Standard error of available area se
urban and row crop area per
catchment

Standard error of capital ag costs capital se
costs associated with
implementing agricultural
BMPs

Standard error of operations ag costs operations se

and maintenance costs

associated with

implementing agricultural

BMPs

Standard error of capital ripbuf costs capital se
costs associated with
implementing riparian
buffer BMPs

18

Data Source

Northeastern Regional SPARROW model (Ator, 2019) annual
average loadings output (ne_sparrow_model_output_tn.txt)
modified with standard error of regression of pollutant
discharge over time from NPDES Permit No. NHO100200
(New Hampshire Department of Environmental Services
2020a)

Northeastern Regional SPARROW model (Ator, 2019) annual
average loadings output (ne_sparrow_model_output_tp.txt)

Coefficient of variation for NLCD data and StreamCat values
from Wickham et al., 2017 and Hill et al., 2015. Uncertainty in
Incremental area from U.S. Geological Survey and U.S.
Department of Agriculture, Natural Resources Conservation
Service, 2013, McKay et al., 2012, and U.S. Geological
Survey, National Geospatial Program, 2020.

U.S. Department of Agriculture; Environmental Quality
Incentives Program state-by-state cost sheets from 2018 -
2021. For the case study, Massachusetts costs were used when
New Hampshire or Vermont costs were not available (U.S.
Department of Agriculture Staff, 2021)

U.S. Department of Agriculture; Environmental Quality
Incentives Program state-by-state cost sheets from 2018 -
2021. For the case study, Massachusetts costs were used when
New Hampshire or Vermont costs were not available (U.S.
Department of Agriculture Staff, 2021)

U.S. Department of Agriculture; Environmental Quality
Incentives Program state-by-state cost sheets from 2018 - 2021
(U.S. Department of Agriculture Staff, 2021)

-------
Optimization Parameter
Description

Standard error of operations
and maintenance costs
associated with
implementing riparian
buffer BMPs

Standard error of the
calculated urban volumetric
runoff coefficient, based on
percent impervious area per
catchment

Standard error of
adjustments to urban BMP
costs based on the expected
amount of retrofitting
(proportional to relative
intensity of urban land per
catchment)

Standard error of
agricultural nitrogen
removal efficiency per
BMP option and HUC 12

Standard error of urban
nitrogen removal efficiency
per BMP option and
catchment

Optimization Parameter Name in AM PL Model
and Data Files

ripbufcostsoperationsse

runoff coefif urban se

urbancostadj ustmentcoefse

ag_effic_N_se

urban effic N se

Standard error of	ag effic P se

agricultural phosphorus
removal efficiency per
BMP option and HUC 12

19

Data Source

U.S. Department of Agriculture; Environmental Quality
Incentives Program state-by-state cost sheets from 2018 - 2021
(U.S. Department of Agriculture Staff, 2021)

Standard error around the slope and intercept of the equation
presented by Vermont Agency of Natural Resources, 2017
calculated from the data originally published by Schueler,
1987.

Sampled from development intensity data by catchment from
the National Land Cover Database, 2011 (catchment-specific
data per state downloaded from StreamCat; Hill et al. 2015) -
"PctUrbOp201 ICat," "PctUrbLo201 ICat,"

"PctUrbMd201 ICat," and "PctUrbHi201 ICat" fields.

Cost adjustment factors from Opti-Tool documentation
(Voorhees, 2016)

"Manurelnjection" BMP: Dell et al. 2016.

Remaining agricultural BMPs: Standard error of efficiencies
summarized by HUC 12 (or if unavailable, HUC 10 and/or
HUC8) based on ACRE database (White et al., 2019)

Sampled from values calculated according to New Hampshire
MS4 Permit BMP Performance Curves (New Hampshire
Department of Environmental Services, 2020b). Hydrologic
soil group: gNATSGO (Soil Survey Staff, 2020a; Soil Survey
Staff, 2020b). Land cover data from the National Landcover
Database (Jin et al. 2019, Yang et al. 2018)

Manure lnjection" BMP: Dell et al. 2016.

Remaining agricultural BMPs: Standard error of efficiencies
summarized by HUC12 (or if unavailable, HUC10 and/or
HUC8) based on ACRE database (White et al., 2019)

-------
Optimization Parameter Optimization Parameter Name in AM PL Model
Description	and Data Files

Standard error of urban urbanefficPse
phosphorus removal
efficiency per BMP option
and catchment

Standard error of total	riparianremoval N l_se... riparianremoval Nn se

nitrogen removal by

riparian buffers per

catchment for each TN

target

Standard error of total	riparianremovalP l_se... riparianremovalPnse

nitrogen removal by

riparian buffers per

catchment for each TP

target

20

Data Source

Sampled from values calculated according to New Hampshire
MS4 Permit BMP Performance Curves (New Hampshire
Department of Environmental Services, 2020b). Hydrologic
soil group: gNATSGO (Soil Survey Staff, 2020a; Soil Survey
Staff, 2020b). Land cover data from the National Landcover
Database (Jin et al. 2019, Yang et al. 2018)

Sampled values of annual removal rates. Annual removal rates
are a function of riparian loading and nutrient removal
efficiency of riparian buffers. Loading is dependent on the
amount of urbanized land within a 400 ft riparian zone, and
efficiency is dependent on slope and hydrologic soil group
(Houle et al. 2019). Land cover data from the National
Landcover Database (Jin et al. 2019, Yang et al. 2018). River
reach shape files from NHDPlus V2 (McKay et al. 2012).
Hydrologic soil group: gNATSGO (Soil Survey Staff, 2020a;
Soil Survey Staff, 2020b). Slope data from Hydrologic
Derivatives for Modeling and Analysis by the USGS (Verdin,
2017)

Sampled values of annual removal rates. Annual removal rates
are a function of riparian loading and nutrient removal
efficiency of riparian buffers. Loading is dependent on the
amount of urbanized land within a 400 ft riparian zone, and
efficiency is dependent on slope and hydrologic soil group
(Houle et al. 2019). Land cover data from the National
Landcover Database (Jin et al. 2019, Yang et al. 2018). River
reach shape files from NHDPlus V2 (McKay et al. 2012).
Hydrologic soil group: gNATSGO (Soil Survey Staff, 2020a,
2020b). Slope data from Hydrologic Derivatives for Modeling
and Analysis by the USGS (Verdin, 2017)

-------
Uncertainty in several additional parameters are not included as parameters in the model, but
rather are hard coded into the command scripts. These include uncertainty around pointcosts,
point effic N and pointefficP, other loads N* and otherloadsP*, and agcostfrac.
Uncertainty around point costs were estimated based on the predicted contractor markup, and
assuming that additional engineering services or detailed designs would have a similar markup
(JJ Environmental, 2015). Point efficiencies assumed a coefficient of variation of 20% (JJ
Environmental, 2015). Other loads for nitrogen and phosphorus were derived from SPARROW
models (Ator, 2019). As other nitrogen loads were adjusted by changes in N deposition,
uncertainty in the change in N deposition was also accounted for according to a modified version
of the qualitative measure presented by (Walker et al., 2019). The ratio of actual agricultural
costs to base payments (agcost frac) was modeled as a uniform distribution between 1 and 1.67
(i.e., base payment costs ranging from 60% - 100% of actual costs).

21

-------
4 Model Use

In order to run RBEROST, the user first defines specifications, including the load reduction
goals and the potential BMPs to be considered for the optimization model. The user
specifications are then run through the R preprocessing code, which develops and formats the
AMPL files required for the NEOS server optimization run. The user can choose to generate
AMPL files that either include uncertainty or do not include uncertainty analyses. The user
uploads the AMPL files to the NEOS server and accesses the optimization results using
information e-mailed to the user. Finally, the user runs the R postprocessing code to summarize
results of the NEOS run. Figure 4.1 summarizes this workflow.

Figure 4.1: User workflow of RBEROST.

4.1 Getting Started

Users download a .zip file that contains RBEROST program files and data required for
preprocessing (creation of AMPL files) and postprocessing (viewing of NEOS optimization
results). It is necessary for the file structure of the zip file as well as naming conventions for
Preprocessing inputs and R files to be maintained.

For the default pathways in RBEROST to work, you will need to open the file via the
RBEROST. Rpnoj file. Double click the file to open the project in RStudio. Figure 4.2 shows the
file in the unzipped folder. The project should open in RStudio, and the files can be seen on the
bottom right. RBEROST can be executed through the RunRBEROST. Rmd file, which is opened by
clicking on it in the Files pane of RStudio (Figure 4.3)

-------
BP- RBEROST

E3 Extra large icons B Large icons as Medium i<

V f P > This PC > Documents > RBEROST

¦J-	Downloads	# '

A	WMOST	y

P	TimeCards	*

P	Tier_1_Optimization-SSWF t

I	Inputs

P	Inputs

P	MockMAdatainputs

P	Outputs

|n| Environmental Protection Ager
P Detenbeck, Naomi - WMOST
P Puget Sound Regional Optim

A OneDrive - Environmental Prot
P EZ Records - Private
P EZ Records - Shared
P Profile

,^i This PC
10 items I

P	Postprocessing

P	Preprocessing

*	R

8	gitignore

I	codeinventory.json

9	LICENSE.md

f	RBEROST.Rproj

Q	RBEROST-User-Guide.pdf
9 README.md

#	RunRBEROST.Rmd

Figure 4.2: RBEROST Rproject file

~ x

/v ®

[¦] Group by '	~ item check boxes

in Add columns -	@ File name extensions

Size all columns to fit ~ Hidden items

Hide selected Options
items

V O P Search RBEROST

Status

Date modified

Type

Size

O A

10/28/2021 4:23 PM

File folder



©A

8/30/2021 5:14 PM

File folder



O A

8/30/2021 5:15 PM

File folder



o

8/30/2021 5:15 PM

Text Document

1 KB

© A

8/30/2021 5:15 PM

JSON File

10 KB

© A

8/30/2021 5:15 PM

MD File

35 KB

O A

11/4/2021 9:51 AM

R Project

1 KB

©

8/30/2021 5:15 PM

Adobe Acrobat D...

3,177 KB

©A

8/30/2021 5:15 PM

MD File

8KB

© P,

10/28/2021 3:49 PM

RMD File

5 KB

I a

23

-------
RBEROST was developed in R version 4.0.5 and can be run through RStudio. RBEROST was
developed with RStudio 1.4.1106. The folder structure includes R scripts to preprocess the data
and postprocess optimization results. All files are accessed through the RunRBEROST. Rmd file.
RunRBEROST. Rmd will source the necessary scripts to run each part of RBEROST. Scripts used in
RBEROST are in the . / R folder, where . / refers to the file path of the unzipped folder. All
scripts must retain their location and naming to be called by RBEROST. Files include
Optimization_HelpenFunctions. R which includes functions used in RBEROST,
01_Optimization_Pnepnocessing_gateway. R that routes RBEROST to either preprocessing
with or without uncertainty, 01_Optimization_Pnepnocessing. R which creates AMPL files
without uncertainty, and 01_Optimization_Pnepnocessing+Uncentainty. R which creates
AMPL files with uncertainty. Running RBEROST with uncertainty will also automatically
produce AMPL scripts without uncertainty as well, and the user is free to ignore or save these
files without uncertainty. R scripts for the postprocessor include 02_Optimization_RunShiny. R
which sources Optimization_SenvenFile. R,
Optimization_SenvenFunctions_Postprocessor.R,

Optimization_UserInterfaceFile.R and Optimization_UI_Postprocessor.Rto build the
Shiny application.

RBEROST relies on several R packages to run. The application was developed with tidyverse
1.3.1, reshape2 1.4.4, data.table 1.14.0, stringr 1.4.0, foreach 1.5.1, shiny 1.6.0, shinycssloaders
1.0.0, tidygraph 1.2.0, and bit64 4.0.5. If these packages are already installed, it is recommended
to update. RBEROST provides code necessary to install and/or update all packages in the
RunRBEROST. Rmd file. This code will only need to be run once, the first time RBEROST is
opened, by clicking the green triangle on the first code chunk (indicated by the red circle in
Figure 4.4). While the code is running, the green triangle will change to a red square (Figure
4.5). The icon will change back to a green triangle when the code is finished. During package
installation, you may be prompted with a question "Do you want to install from sources the
package which needs compilation?" (Figure 4.6). Select "Yes."

-------
© RBEROST - master - RStudio

File Edit Code View

Plots Session

Build Debug Profile Tools Help











• +D« *!

s1 LP

1 o

UM " Addins *







Q RBEROST -



Console Terminal

Jobs





0

Environment

History Connections Git Tutorial



-/RBEROST/







y

5 Diff 61

Commit * O '

7. master * C *

Natural language support but running in an English locale

R is a collaborative project with many contributors.

Type 'contributorsO' for more information and

'citationO' on how to cite R or R packages in publications.

Type 'demoO' for some demos, 'helpO' for on-line help, or
'help.startO" for an html browser interface to help.

Type 'qO' to quit R.

R version 4.0.5 (2021-03-31) — "shake and Throw"

Copyright (c) 2021 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO warranty.
You are welcome to redistribute it under certain conditions.
Type 1 licenseO' or 'licence()' for distribution details.

Natural language support but running in an English locale

R is a collaborative project with many contributors.

Type 'contributors()' for more information and

'citationO' on how to cite R or R packages in publications.

Type 'demoO' for some demos, 'helpO' for on-line help, or
'help.startO' for an html browser interface to help.

Type 'qO' to quit R.



Files

Plots Packages Help Viewer



=~

tM New Folder KJ Delete l!j Rename

f More -

c

¦ /|| Home > RBEROST Q ...



Name

Size

Modified



t.





¦

3 .gitignore

42 B

Aug 30, 2021, 5:1 f

¦

^ codeinventory.json

9.4 KB

Aug 30, 2021, 5:15

¦

S LICENSE.md

34.3 KB

Aug 30, 2021, 5:15

¦

Ml Postprocessing





¦

Preprocessing





¦

R





¦

3 RBEROST-User-Guide.pdf

3.1 MB

Aug 30, 2021, 5:15

¦

@ RBEROST.Rproj

203 B

Nov 12, 2021, 10:(

¦

3 README.md

7.7 KB

Aug 30, 2021, 5:1!

m

3 RunRBEROST.Rmd

4.9 KB

Oct 28, 2021, 3:49

Figure 4.3: Where to find the file 'RunRBEROST.Rmd' to open RBEROST though RStudio

25

-------
0 RBEROST - master - RStudio

File Edit Code View Plots Session Build	Debug Profile Tools Help

H f |,<+ J I ^ " 1? d	Go to file/function jjlJ <

3 RunRBEROST.Rmd

^ HjM V " * 1 f ^ Knit - # -

24	## configure R

25

26	This code only needs to be run once, the first time you run RBEROST.
It will install several R packages that RBEROST uses. Once run, you
will not need to run this code again unless you update R, RStudio, or
there are problems with installation. To run the code, click the
green triangle below, if no green arrow appears, and instead there is
a red square, click the red square to stop any ongoing processes and
a green triangle should appear in its place. When clicked, the green
triangle will turn into a red square until the code is finished.

Copyright Information a
Configure R
configureR
Run Preprocessor
runpreprocessor
Run Postprocessor
runpostprocessor

Environment History Connections Git Tutorial a Q
3 Diff 0 Commit ~	Q ft * master -

Staged Status Path

28- ~ '{r configureR}

29

30	packages <- c("tidyverse", "reshape2", "data.table",
"foreach", "shiny", "shinycssloaders", "tidygraph", '

31	invisible(lapply(packages, install.packages))

stri ngr
bit64")



Files Plots Packages Help Viewer



=>~

tSi New Folder *M Delete IS Rename

jt:ft More *

c

¦ ft Home > RBEROST Q ...

* Name

Size

Mod

54:27 C Chunk 2: runpreprocessor

R Markdown



Console Terminal

Jobs

e= n

-/RBEROST/



y

'citationO' on how to cite R or R packages in publications.

Type ' demo()' for some demos, 'helpO' for on-line help, or
'help.start()' for an HTML browser interface to help.

Type 'q()' to quit R.

3 .gitignore
ft codeinventory.json
& LICENSE.md
M Postprocessing
tfl Preprocessing
R

A RBEROST-llser-Guide.pdf
6 RBEROST.Rproj
9 README.md
¦3 RunRBEROST.Rmd

Figure 4.4: Run package installation the first time RBEROST is opened.

26

-------
© RBEROST - master - RStudio

File Edit Code View Plots Session Build Debug Profile Tools Help

+J • I,+ | 'I « -	§	Go to file/function	- I §§ * Addins -

3 RunRBEROST.Rmd

U"' • I Knit * * •	°« ' t * ¦

a green triangle should appear in its place, when clicked, the green
triangle will turn into a red square until the code is finished.

27	iT1N

28- "' {r configureR}	&

29|

301 packages <- c("tidyverse", "reshape2", "data.table", "stringr",

packages <- cC'tidyverse", "reshape2", "data.table", "stringr"
"foreach", "shiny", "shinycssloaders", "tidygraph", "bit64")
invisible(lapply(packages, install.packages))



Copyright Information e
Configure R
configureR
Run Preprocessor
rurtpreprocessor
Run Postprocessor
runpostprocessor

Environment History Connections Git Tutorial a Q
S Diff 0 Commit Hi ~	e * ?. master -

Staged Status	Path

Files Plots Packages Help Viewer

New Folder	Delete 15 Rename ^ More *

¦ ft Home > RBEROST

Name	Size

5427 c Chunk 2: runpreprocessor



R Markdown



Console Terminal Jobs



a n

-/RBEROST/ *





Type "qO" to quit R.





>

>	packages <- cC'tidyverse", "reshape2", "data.table",
inycssloaders", "tidygraph", "bit64")

>	invisible(lapply(packages, install.packages))

"stringr", "foreach",

"shiny", "sh

3	.gitignore
codeinventory.json

SB LICENSE.md

4	Postprocessing
¦I Preprocessing

m r

A RBEROST-User-Guide.pdf
9 RBEROST.Rproj
9 README.md
3 RunRBEROST.Rmd

Figure 4.5: An image of what RStudio displays while code is running

27

-------
Question

X

e

Do you want to install from sources the package which needs
compilation?

Yes

No	Cancel

Figure 4.6: A message users may encounter while installing packages.

4.2 Preprocessing

This section details the processes required to develop the AMPL model files that are run through
the NEOS server to solve the optimization. The user is required to update the Run Preprocessor
section of the code in RunRBEROST. Rmd as well as the two csv files 01_UserSpecs_BlviPs. csv
and 01_UserSpecs_loadingtargets.csv, which are in the ./Preprocessing/Inputs folder
(Figure 4.7). To run the preprocessing step, first edit the UserSpecs files and the available
options within RunRBEROST. Rmd, Then, click the green triangle in the upper right of the
preprocessing code chunk. The green triangle will turn into a red square until the code is done
running, at which point the red square will return to a green triangle.

¦ Q | T Inputs

Home Share View

1 | x™

09 Copy path

Pin to Quick Copy Paste
access

Q Paste shortcut to T to -

Move Copy Delete Rename New

I ^ New item T
^ Easy access

¦ E

x ¦ Open T | Select all
J Edit	22 Select none

History Invert selectic

^ it « TierlOptimization > TTer_1_Optimization-SSWR_5_3_2 Preprocessing Inputs

Search I,

|| TimeCards jt

A

Name

Status

Date modified

Type

Size

P Tier_1_Optim #



^ 01_UserSpec5_BMP5,csv

O

5/3/2021 5:39 PM

Microsoft Excel C..

4KB

H Images

DS 01_UserSpecsJoadingtargets.csv

O

4/30/2021 2:11 PM

Microsoft Excel C...

1 KB

|| Inputs



~ ACRE_HUC12_HRU_Summary.csv

• ft

5/3/2021 5:39 PM

Microsoft Excel C...

227 KB

H Outputs



3 AgBMPEffic_FertManure.csv

• ft

11/23/2020 8:43 AM

Microsoft Excel C...

1 KB

J| References



3 EQIPcosts_overyears.csv

• ft

5/3/2021 5:39 PM

Microsoft Excel C...

2KB



3 LengthinBuffer_2016.csv

• ft

5/3/2021 5:39 PM

Microsoft Excel C...

1,035 KB

p|$ Environmental Prt



^ NdepChange_2012_2019.csv

• ft

3/23/2021 2:29 PM

Microsoft Excel C...

251 KB

A OneDrive - Enviro



*1 ne_sparrow_model_input.csv

• ft

7/1/20204:27 PM

Microsoft Excel C...

84,174 KB





3 ne_sparrow_model_output_tn.csv

• ft

4/21/202010:21 AM

Microsoft Excel C...

67,679 KB

^ This PC



31 ne_sparrow_model_output_tp.csv

• ft

11/23/2020 8:43 AM

Microsoft Excel C...

57,903 KB

^l 3D Objects



NH_streamcat_2011_cropland.csv

• ft

5/6/202011:48 AM

Microsoft Excel C...

3,287 KB

¦1 Desktop



3 NH_streamcat_2011 Jmperv.csv

• ft

5/6/202011:49 AM

Microsoft Excel C...

609 KB

9 Documents



^ NHD+infiltrationrates.csv

• ft

5/3/2021 5:39 PM

Microsoft Excel C...

1,011 KB

^ Downloads



RiparianEfficiencies.csv

• ft

5/3/2021 5:39 PM

Microsoft Excel C...

64,757 KB



9 RiparianLoadinqs.csv

	#ft	

5/3/2021 5:39 PM	

Microsoft Excel C...

397 KB

J' Music

V

< |







| >

22 items I 2 items selected 3,58 KB I Available on this device I

	1—1	-»—->	 J	r 		-»—n—-»	







i

Figure 4.7: Location of the 2 User Specification csv files used in RBEROST.

28

-------
4.2.1 User Specifications

There are eight parameters that users can change in the RunRBEROST. Rmd file. These include
InPath, which should point to the folder with the Preprocessing inputs, and OutPath which
should point to the folder that the AMPL files will be written to. If RBEROST is opened from
the RBEROST. Rpnoj file, and if file structure is left intact, these parameters do not need to be
changed. Other parameters include horizon which is the planning horizon used to annualize
costs. The default value is 15 years. The variable intenest_nate is the expected interest for the
project, and the default value is 0.03, or 3%. AgBMPcompanison can be set as "No Practice" or
"Baseline" and determines how the efficiencies of agricultural BMPs will be determined. "No
Practice" represents the efficiency of a practice in isolation compared to conventional
practices, and "Baseline" represents the efficiency of removing existing conservation practices
and replacing them with a certain practice. Includellncentainty is a logical parameter and can
be set as TRUE or FALSE. If TRUE, RBEROST will create AMPL scripts that include uncertainty
calculations. If FALSE, it will not. Making AMPL scripts that include uncertainty may take
several minutes. Variables n. scenarios and scenaniostepchange only apply when running
RBEROST with uncertainty. The variable n. scenarios sets the number of scenarios you would
like to view, and scenaniostepchange sets the relative difference between these scenarios. The
different scenarios describe increasingly restrictive loading targets to be applied after the first
scenario is solved with the user specifications from 01_UsenSpecs_loadingtangets. csv. Each
successive scenario decreases the loading targets by the percent given in scenaniostepchange.
This variable scenaniostepchange is equivalent to the additional margin of safety applied to
each scenario. The default is to compute three scenarios with a 1% change in loading targets
between scenarios. The lines of code to edit are marked with red stars in Figure 4.8.

Users specify the BMP options that will run through RBEROST within the
01_UsenSpecs_BMPs. csv file (Figure 4.9). BMP options are specified by placing a capitalized
"X" in the BMP_Selection field of the 01_UsenSpecs_BMPs. csv file. Users can also choose a
minimum and/or maximum implementation of each BMP as a fraction of total possible
implementation. 01_UsenSpecs_BMPs .csv includes all the BMPs available for model
optimization per BMP category (agricultural, urban, point source, and riparian buffer BMPs),
state-specific capital and operations & maintenance costs to implement the BMP, units of the
costs, user-specified design runoff depths for urban BMPs and buffer widths for riparian buffer
BMPs.

The data provided covers the geographic extent of the Upper Connecticut River case study. To
expand the geographic range of optimization, data files will need to be expanded to include
ComlDs located in additional areas.

-------
G TierJ_Optimization-55WR_5_3_2 - Development - RStudio

File Edit Code View Plots Session Build Debug Profile Tools Help

/~ 60 to file/function

•3 Run RBEROST. Rmd

0 Tier_1_Optimization-SSWR_5_3_2 — TierlOptirnizatian •

Environment History Connections Git

17- ## Run Preprocessor

18 To run the preprocessor, change any lines of code necessary and click the green triangle below. If no green arrow appears, and
instead there is a red square, click the red square to stop any ongoing processes and a green triangle should appear in its
place. When clicked, the green triangle will turn into a red square until the code is finished.

19 I

20- ~~~{r runpreprocessor} # ~ ~

22 # This variable represents the location of input data files used to develop AMPL model files. If you opened RBEROST through the
Tier_l_0ptimization-SSWR_5_3_2.Rproj file, the represents the folder where the Tier_l_0ptimization-SSWR_5_3_2.Rproj file is
located. If you did not open this file through the R project you may need to write out the full path to your RBEROST inputs.]

2', lnPath<-pasteO(" ./Preprocessing/Inputs/")

25 # This variable represents the location where AMPL model files will be printed for use on the NEOS server. If you opened RBEROST
through the Tier_l_opti mi zati on-sswR_5_3_2.Rproj file, the represents the folder where the

Tier_l_Optimization-SSWR_5_3_2.Rpro} file is located. If you did not open this file through the R project you may need to write
out the full path to your RBEROST inputs.

.26 OutPath<-pasteO("./Preprocessing/Outputs/")

2T. # This variable represents the planning horizon in terms of years. Default is 15.

29 horizon <- 15

31', # This variable represents to expected interest rate. Default is 0.03, or 3%.
interest_rate <- 0.03

34 # This variable lets the user decide if they wish to consider agricultural BMP efficiencies of practices versus "No Practice" (a
default of no conservation practices) or "Baseline" (a default of practices currently in place, and implementing a BMP will
remove practices currently in place).

55 AgBMPcomparison <- "No Practice"

37 # This variable decides if you will run RBEROST with or without uncertainty. It can be TRUE or FALSE. Running with uncertainty
takes longer them running without.

3c IncludeUncertainty <- TRUE

40 # If running RBEROST with uncertainty, how many scenarios would you like to view? Default is 3.
n.scenarios <- 3

# If running RBEROST with uncertainty, how different would you like the scenarios to be? Each scenario will be solved for
loading limits that are a certain percentage lower than the previous. Default is 0.01, or 1%.
scenariostepchange <- 0.01

Q Chunk 2: runpreprocessor

Figure 4.8: Lines of code for the user to edit in the preprocessing step.

-------
AutoSave (• Off) 57 01_UserSpecs_BMPs.csv - Excel _P Search

File Home Insert Draw Page Layout Formulas Data Review View Help

frh " Cut Calibri "111 A" A" = = •=?: &/¦ - Wrap Text I&

idr Qlcopy -

. BIU-ffiv^vA. S- = -E _E S Merge&Center - $

V Format Painter
Clipboard G

v c° Wrap Text
E __E Zj Merge & Center
Alignment

Formatting v Table v Styles v
Styles

S3 S3<

l I I I M : ¦

Chamberlin, Catherine
X AutoSum v Ar-?

m Fin - z '

Sort & Find &
-------
4.2.1.1 BMP Costs

Implementation of the majority of the BMPs requires both capital and operations & maintenance
costs, except for several agricultural BMPs. All default costs are scaled to 2019$. The user can
override these costs if they choose costs that are representative of costs per treated area for
agricultural and riparian buffer BMPs, costs per treated volume for urban BMPs, and costs per
retrofit implementation for point source BMPs. Urban costs are specified assuming
implementation in new developments rather than with retrofitting. RBEROST will adjust these
costs internally based on the expected amount of retrofitting and complexity of installations in
each catchment as judged by development intensity.

If cost data from additional states beyond the geographic extent of the Upper Connecticut River
case study must be included, these are added to the file 01_UsenSpecs_BMPs. csv. For each
additional state, two columns must be added to this file, following the naming convention of
"captial_[two letter state code]" and "openations_[two letter state code]". The
units for these costs must be the same across all states included, and units are indicated in the
"capital_units" and "openations_units" fields. Possible units are "ac", "ft2", "yd2" and
"km2" for agricultural and riparian buffer BMPs, "ft 3" for urban BMPs, and "flat" for point
source BMPs. If optimizations that include uncertainty are performed for geographic ranges that
extend to states beyond the ones in the case study, additional columns must also be added to the
file EQIPcosts_ovenyeans. csv. Column headings should follow the naming conventions of
"capital_[two letter state code]_[yean]" or "openations_[two letter state
code]_[yean]", and cost units must be consistent across all columns.

4.2.1.2 Urban BMP Runoff Depth and Riparian Buffer Width

Nutrient removal efficiency resulting from implementation of urban BMPs depends on the BMP
design runoff depth. The 01_UsenSpecs_BMPs. csv file specifies default runoff depths for
BMPs, but the user may specify alternate runoff depths in the UsenSpec_RD_in field. The
Min_RD_in and Max_RD_in fields specify the minimum and maximum runoff depths (i.e., upper
and lower limit) that the user can specify (in inches) for the urban BMP UsenSpec_RD_in field.
RBEROST does not use the Min_RD_in and Max_RD_in fields for any calculations or data
modifications, the fields are just for reference. Specification of riparian buffer width also affects
the amount of nutrients removed by the buffers. Users can specify buffer widths in the same
column, UsenSpec_RD_in. A note is provided that, though presented to the user in the same
column, the units for riparian buffer BMPs are different than for urban BMPs.

4.2.1.3 Load Reduction Goal

Users can specify loading targets in the 01_UserSpecs_loadingtargets. csv file (Figure 4.10).
Target reductions are described as a percent reduction from current loading. Users can specify
whether targets are for Total Nitrogen (TN) or Total Phosphorus (TP). If targets are not within
the river network, they can be flagged with a capital X in Outof NetwonkFlag_X. If the target is
the terminal end of a watershed, it can be flagged with a capital X in TenmFlag_X. COMID refers
to the NHDPlus V2 catchment that contains the loading target.

-------
Al

- i X ¦/ fx \

A/aterbody_Name

D E

_
1

Waterbody_Name

IcomID

Watershed_HUC

Percent_Reduction TN_or_TP

OutofNetworkFlag_X

TermFlag^X

CT River at MA border

9332552

10801

0.1 TN

Back Lake (NHLAK801010203-01-01)

4592401

10801010203

0.12 TP

Forest Lake (NHLAK802010401-01-01)

4594723

10801030101

0.02 TP

Figure 4.10: A screenshot of the loading targets User Specs file with the specifications for the
Upper Connecticut Rive case study.

4.2.2 Preprocess Data Inputs

The data preprocessing section of the code adjusts and formats the input data for AMPL model
file development. Input data include baseline nutrient loads, land use information, watershed
characteristics, BMP costs, and BMP nutrient removal efficiencies. The provided data cover the
geographic extent of the Upper Connecticut River case study only.

4.2.2.1 Baseline Nutrient Loading and Land Use Data

Recent regional SPARROW model outputs from the U.S. Geological Survey provide the basis
for RBEROST baseline nutrient loading conditions (Ator, 2019). These models provide
catchment-level nutrient loads and land use data. Catchments are specified using common
identifiers (COMIDs) based on the National Hydrography Dataset (NHD) Plus Version 2 reach
network (McKay et al., 2012). RBEROST also includes 2011 cropland and imperviousness land
use data from the National Land Cover Database (Yang et al., 2018) provided at the NHDPlus
V2 catchment level via StreamCat (Hill et al., 2015). For modeling that includes states outside of
the region presented in the case study, StreamCat files for additional states can be added to the
Preprocessing/Inputs folder and renamed according to the same convention as the provided files.

Data modifications entail classifying incremental baseline SPARROW loadings as point source,
urban, agricultural, or "other" loads. Incremental loads in SPARROW are stream loads that
originate from the NHDPlus V2 catchment. This is differentiated from total loads in SPARROW
which is the sum of incremental load for each catchment and delivered load from upstream.

Other loads include loadings from sources that cannot be classified as point, urban, and
agricultural (e.g., septic systems or atmospheric deposition). The model code adjusts baseline
point source loading to account for changes in wastewater nutrient effluent and atmospheric
nitrogen deposition since 2012. This is done within RBEROST so that the user can potentially
include additional locations for baseline loadings adjustments where changes are known to have
occurred. This would be done by editing the input files

WWTP_BaselineRemoval_Finengnain.csv, WWTP_COMIDs_BslnRemoval.csv, and
NdepChange_2012_2019. csv to include information for additional COMIDs. Details of these
files are available in Table 6.1.

The optimization depends on agricultural and urban land area available for BMP
implementation. Urban acreage data are derived from the SPARROW regional models, while
agricultural acreage is calculated as the product of incremental acreage available per catchment
(from SPARROW; Ator (2019)) and the percent of area defined as cropland per catchment (from
National Land Cover Data/StreamCat; Hill et al. (2015)).

-------
Urban BMP costs depend on the volume of water being treated. RBEROST relies on a water
quality volume equation from the Vermont Stormwater Management Manual (Vermont Agency
of Natural Resources, 2017). Water quality volume depends on a volumetric runoff coefficient,
equal to 0.05 + 0.009 * Imperviousness. The National Land Cover Dataset (NLCD) provides
imperviousness data at the catchment level via StreamCat (Hill et al., 2015).

Riparian loading depends on the amount of impervious surface cover within the river corridor, as
per Houle et al. (2019). These estimates must be derived prior to modeling with RBEROST.

They are provided in the RipanianLoadings. csv file. Riparian loading is a subset of urban, ag,
point, or other loads and does not describe an additional source. Users who wish to use different
methodologies to estimate riparian loads (e.g., including agricultural loads) may do so and
overwrite the RipanianLoadings. csv file with a new file with the same naming conventions
and format.

4.2.2.2 Watershed and Target Specifications

The 01_UsenSpecs_loadingtangets. csv file describes the locations of loading targets that
will be included in RBEROST. The COMID identifiers are used to identify upstream
contributing reaches to each target. These contributing reaches are included in the input baseline
loading and land use datasets used in calculating inputs to each target waterbody.

The SPARROW regional models specify a delivery fraction (DELFRAC) that reflects the
fraction of the incremental nutrient loads that are delivered to a flowline's terminal reach (in
most cases, the ocean). RBEROST revises this delivery fraction to recognize each loading target
as the specified terminal reach. For instance, if the specified target COMID has a delivery
fraction of 92% in the SPARROW regional model dataset, indicating that 92% of the
incremental loads associated with the target reach are delivered to the terminal reach, the revised
delivery fraction at this location is 100%, indicating that 100% of the incremental loads
associated with the target are delivered to the corresponding target reach.

The delivery fractions for all reaches upstream of the target are revised based on the ratio of the
revised target delivery fraction to the original terminal reach delivery fraction. For instance, if
the SPARROW regional model data indicates that an upstream reach delivers 83% of its
incremental load to the terminal reach and that the target reach delivers 92% of its incremental
load to the terminal reach, then the revised delivery fraction for the upstream reach to the target
reach is (100%/92%)*(83%) = 90%. Delivery fractions typically differ between TN and TP, and
these calculations are performed separately for each nutrient target. RBEROST allows nested
targets as well, in which case a stream reach that contributes to more than one loading target will
be assigned more than one DEL FRAC value.

4.2.2.3 BMP Costs and Efficiencies

BMP costs specified in the 01_UsenSpecs_BMPs. csv file are adjusted to per-acre units for
agricultural BMPs, and per-square foot units for riparian buffer BMPs. Because point source
BMPs are location-specific, the model code merges the wastewater treatment plant (WWTP)
costs with each plants' reach location (COMID) prior to developing the AMPL model files.
Default urban BMP capital costs represent new development costs. RBEROST will adjust costs
for expected amounts of retrofits based on retrofit costs specified in Table 7-2 of the WMOST v3
Theoretical Documentation (Detenbeck, ten Brink, et al., 2018). This adjustment is a weighted

-------
average of the new development cost (assigned to open and low development), retrofit costs
(assigned to medium development) and difficult retrofit costs (assigned to high development).
The development data originate from the NLCD (Yang et al., 2018) and are summarized by
catchment (Hill et al., 2015). As in the WMOST model, default urban BMP operations and
maintenance costs are assumed to be 5% of capital costs, but this can be modified by the user if
more accurate estimates are available.

RBEROST assumes a default planning horizon of 15 years and an interest rate for capital costs
of 3%. The user can adjust the horizon and interest rate variables within the RunRBEROST. Rmd
file.

While some BMP efficiencies are BMP-specific, others are also location-specific. For
agricultural BMPs with nutrient removal efficiencies that vary based on HUC12, the model code
identifies the individual reaches that fall within each HUC12 and assigns the HUC12-specific
efficiency to each reach. There are two available agricultural BMP cost datasets that differ based
on how they compare the export of nutrients. "No Practice" compares the export of nutrients
with a given BMP versus the export of nutrients under conventional practices. "Baseline"
compares the export of nutrients with only one given BMP versus the export of nutrients
occuring under existing conservation practices. The "Baseline" approach assumes existing
conservation practices are removed and only the one BMP is implemented. Point source BMP
efficiencies are specific to each WWTP upgrade and urban BMP efficiencies are calculated
based on the user-specified runoff depth. As some urban BMP efficiencies are dependent on the
infiltration rate of the soil underlying developed land, the urban BMP efficiency curves are
selected based on the average infiltration rate. This information is provided in
NHD+infiltnationnates. csv and is derived from gNATSGO soil layers (Soil Survey Staff,
2020b, 2020a).

Riparian buffer BMP efficiencies similarly depend on average infiltration rate, as well as the
slope of the riparian buffer. The efficiency curves (as a function of buffer widths) of different
riparian buffer BMPs for different nutrients at five representative buffer widths are provided in
RipanianEff iciences. csv. These efficiency curves have been assigned based on infiltration
rate and slope of different buffer widths in each COMID. RBEROST will use the efficiency
curve based on the closest match to the user-selected buffer width.

4.2.3 Write AMPL Model Files

Running the optimization model through the NEOS server depends on the development of three
AMPL model files: the command file (.amp), the data file (.dat) and the model file (.mod). The
command file contains instructions for the optimization server to solve the mathematical problem
and defines what results should be displayed afterwards. The data file contains all of the data
used in the optimization, including values for each parameters across their geographic range. The
model file defines the mathematical problem to be optimized and contains a list of constraints
placed on the model during optimization. After running the preprocessing code, the AMPL
model files will by default be written to the . /Preprocessing/Outputs folder, unless otherwise
specified in the RunRBEROST. Rmd file (Figure 4.11). If the user chooses to run the model with
uncertainty, these command, data and model files will be designated by "_uncentainty" in their
name.

-------
Several messages appear while running RBEROST preprocessor (Figure 4.12, as highlighted in
the red box). The first indicates when RBEROST has begun and when it has finished
determining the list of COMlDs upstream from the loading targets. This step usually takes a
considerable amount of time, so the messages are intended to assure the user that the code is
running. The next few indicate whether there are more COMIDs in the StreamCat datasets or in
the SPARROW dataset of the contributing reaches to each target. The number of messages will
match the number of targets. SPARROW COMIDs that are not in StreamCat, or StreamCat
COMIDs that are not in SPARROW contribute to the other loads parameters. The next messages
will indicate when RBEROST has completed writing AMPL scripts. When running RBEROST
with uncertainty analysis, AMPL scripts without uncertainty will be written first as part of the
preprocessing. These AMPL files without uncertainty may be ignored or saved for further use.

6 items I | E3

Figure 4.11: Outputs of running RBEROST preprocessor with uncertainty.

4.2.3.1 AMPL Command File

The command file specifies display characteristics of the NEOS optimization results. If the
model is being run with uncertainty, the command file specifies the number of scenarios to run
and alters the loading limits accordingly for each successive solve. The command file also
calculates 200 'bootstraps,' of cost and total loading based on the solved model's suggested suite
of BMPs. These bootstraps are 200 calculations of cost and total loading based on random
sampling of values for all the necessary parameters in RBEROST. For most parameters, this is a
normal distribution, though several parameters are described as uniform distributions. The mean
value used for the resampling is the value used by RBEROST to solve the optimization problem,
and the standard errors are additional parameters written to the STdata_undertainty.dat file.

-------
4.2.3.2 AMPL Data File

The data file specifies the catchment- and BMP-specific data required for the NEOS
optimization run, including:

Baseline loadings, efficiency, and cost data formatted by the R code;

Total reduced loading values (loadsjim parameters) calculated based on the
specified load reduction percentage;

Total loading value for loads specified as "other loads" (otherjoads parameters)
that are not available for BMP loading reduction;

The fraction of agricultural costs (agcost_frac) that reflect base payment versus
actual agricultural BMP costs;

The adjustment factor of urban costs (urban_cost_adjustment_coef); and

Conversion factors for urban BMP treated water volume (acfttoft3) and
precipitation (pep).

These parameters are described in greater detail in Table 3.2. Additional parameters written
when modeling with uncertainty are described in detail in Table 3.3.

Depending on the geographic extent of optimization, the data file with uncertainty may become
larger than the allowed upload size for NEOS. If this occurs, it is recommended to proceed using
the AMPL scripts that do not contain the uncertainty analysis.

4.2.3.3 AMPL Model File

The model file specifies the following:

• COMIDs, BMPs, cost types, and load types that are included in the optimization;

• The bases on which parameters vary (e.g., agricultural efficiencies vary by both
reach and BMP);

• Constraints on parameters and variables (e.g., whether a variable represents binary
conditions or a fraction less than or equal to 1);

• The total load reduction functions; and

• The cost minimization objective function.

The model file with uncertainty includes the same specifications for parameters that describe
standard errors.

-------
4.3 NEOS Server

RBEROST uses the CPLEX Optimizer available through NEOS to solve the optimization.
CPLEX will allow XML calls from RBEROST to submit AMPL files to the server, however at
the time of current release, only a manual interaction option with the server is available. Users
will click on the following link and upload the AMPL model, data, and commands files to the
Web Submission Form: https://neos-server.Org/neos/solvers/lp:CPLEX/AMPL.html (Figure
4.13). Before submitting, the user will specify the e-mail address to receive an update when the
model run is complete. Users are also highly encouraged to include a descriptive note of the
model in the comments section (Figure 4.14). Please refer to the WMOST User Guide
(Detenbeck, Piscopo, et al., 2018) for additional information on how to run optimization models
on the NEOS server. All AMPL files must be of the same type - either with uncertainty, or
without uncertainty. Mixing scripts of the two types will produce errors and the model will not
solve. If you are running RBEROST with uncertainty analysis, you only need to submit the
AMPL files with " uncertainty". The other files that were created, which describe the same
problem without the uncertainty analysis, may be ignored or saved for reference.

Generally, NEOS can solve models without uncertainty in less than five minutes, while models
with uncertainty may take nearer to 20 minutes. Exact times may vary, perhaps greatly,
depending on the complexity of the specified problem and the current demands on NEOS and on
the CPLEX optimizer. Once the optimization has solved, users will receive an email from the
server (Figure 4.15) with a link, a job number, and a password. Users can retrieve their results
from the server with this information (Figure 4.16). Users save the NEOS server optimization
results by selecting all text within the results window (CTRL+A), pasting it into a text file (via
text editor such as Notepad), and saving it as a text file (.txt) to a location that makes sense for
them (Figure 4.17). Example output files are provided in . \Postpnocesson\Input. If NEOS
was unable to optimize the model, the returned result may be very short, simply stating that
optimization failed. This result file may still be submitted to the RBEROST Postprocessor.

-------
[1] "RBEROST is now determining which reaches are included in the watersheds
of the identified loading targets."

[1] "RBEROST has now determined which reaches are included in each
watershed."

Note: For the 1st TN target, there are fewer reaches in the provided
StreamCat datasets than are included in SPARROW.

Only the reaches that are included in both datasets will be available
for BMP optimization. Loads the remaining reaches will be included in
the "other_loads' parameter.

Note: For the 1st TP target, there are the same number of reaches in SPARROW
and Streamcat
subsetted datasets.

All reaches available for BMP optimization

Note: For the 2nd TP target, there are the same number of reaches in SPARROW
and Streamcat
subsetted datasets.

All reaches available for BMP optimization

[1] "Urban base costs are assumed to be the same across states. RBEROST is
using values from vt."

[1] "RBEROST has finished writing AMPL scripts without uncertainty at 2021-
11-12 12:36:45"

[1] "RBEROST is now creating AMPL files with uncertainty analysis at 2021-11-
12 12:36:46"

[1] "RBEROST has finished writing AMPL scripts with uncertainty at 2021-11-12
12:38:36"

Figure 4.12: Messages displayed by RBEROST preprocessor

-------
fl NEOS Seiven CPLEX X +

O i neos-server.org/neos/solvers/lp:CPLfX/AMPLhtml
U: Apps ¦ EPA pages ¦ R Resources ¦ AMPL ¦ ArcGIS ¦ RSPARROW Q GrtHub ¦ WMOST Recources © NWS NOAA Snow... ¦ Puget Sound » Gentle Night Rain 1.„ ^ CT - Hartford | Wee...

^ 9 Contact © Help

O - O x
~ * i :

NEOS Interfaces to CPLEX

Sample Submissions
WWW Form - XML-RPC

CPLEX

The NEOS Server offers the IBM ILOG CPLEX Optimizer for the solution of linear programming (LP), mixed-integer linear programming (MILP), and second-order conic
programming (SOCP) problems. Acceptable input formats for CPLEX on the NEOS server include AMPL, GAMS, LP, MPS, and NL formats.

Details on CPLEX can be found on the IBM CPLEX website. Additional information on all IBM software available to academics can be found on the IBM Academic Resources
webpage.

Using the NEOS Server for CPLEX/AMPL

The user must submit a model in AMPL format. Examples are provided in the examples section of the AMPL website.

The problem must be specified in a model file. A data file and commands files may also be provided. If the commands file is specified, it must contain the AMPL solve
command: however, it must not contain the model or data commands. The model and data files are renamed internally by NEOS.

The commands file may include option settings for the solver. To specify solver options, add

option cplex_options "OPTIONS'j

where OPTIONS is a list of one or more of the available solver options for AMPL.

Note: An email address is required for any submissions that use CPLEX. This email address will be forwarded to IBM and may be used by IBM for promotional purposes.
When using the XML-RPC interface, you must add the following line into the XML file that is sent to NEOS:

your. addressglemail. edu

Web Submission Form

Model File

Enter the location of the AMPL model file (local file)
| Choose File | No file chosen

Data File

Enter the location of the AMPL data file (local file)
Choose File | No file chosen

Figure 4.13: The CPLEX Optimizer hosted by NEOS.

-------

MPLhtml

GIS | RSPARROW

O GrtHub ¦ WMOST Recources <

S) NWS NQAA Snow... | Puget Sound

~ Gentle Night Rain 1.„ vvu CT ~ Hartford | Wee...

# Contact

O Help

> Sign In 1 I

When using the XML-RPC interface, you must add the following line into the XML file that is sent to NEOS:

< email >you r. add ressjStemail. edu

Web Submission Form

Model File

Enter the location of the AMPL model file (local file)

| Choose File | STmodel.mod

Data File

Enter the location of the AMPL data file (local file)

| Choose File [ STdata.dat

Commands File

Enter the location of the AMPL commands file (local file)

| Choose File | STcommand.amp

Comments

RBEROST demonstration

Upper CT river, 3 loading targets.

Porous Pavement w/ subsurface infiltration excluded, 10%

cap on ag ponds

No uncertainty

Additional Settings

~ Dry run: generate job XML instead of submitting it to NEOS

~ Short Priority: submit to higher priority queue with maximum CPU time of 5 minutes
E-Mail address: |chamberlin catherine@epa govj "~|

Please do not click tne 'Submit to NEOS'button more than once.

Clear this Form

By submitting a job, you have accepted the Terms of Use

Figure 4.14: User inputs to CPLEX to run RBEROST.

-------
- - P Search
Home Send / Receive

H a, ®|"»™

IE5 Clean Up »

L2U 43

New New ~

Email Items- >'.Junk

Help

e Archive

All Unread

By Date v T

V Today

support@neos-serve...

NEOS Results for Job #104...

8:01 AM

Your Job results were too

support@neos-serve...

NEOS Results for Job #104...

7:47 AM

Your job results were too

r\7] Ld Meeting [_J Action Required To Manager

Cq IM ~ ® Team Email Done

Replv Reply Forward i _ . _ _ . t2

All Eft "tore- Reply B. Delete r Create New

Respond Quick Steps

NEOS Results for Job #10459757

support@neos-server.org
To Chamberlin, Catherine
Retention Policy 10 years (Capstone approach) (10 years)

Si 0 iQ £i > ¦ P

Move Rules OneNote Unread/ Categorize Follow
» v Read » Up"

Search People
H Address Book
"y3 Filter Email-
Find

Expires 5/5/2031

A')

Read Get
Aloud Add-ins

vj)

Insights

Reply Reply All

Forward

Fri 5/7/2021 8:01 AM

Your job results were too long for email. The results may be downloaded from: https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fneos-server.orE%2Fneos%2Fadmin.html&data=04%7C01%7CChamberlin.Catherir\e%40epa.gov%
7C4b5405elQQ834b0bbl0aO8d9114fc888%7CS8b378b367484867acf976aacbeca6a7%7C0%7CO%7C637559S56637812570%7CUnknown%7CTWFpbGZsb3d8evJWIioiMC4wUAwMDAiLQQIioiV2luMzliLCIBTil6lklhaWwiLQXVCI6Mn0%3D%
7C1000&sdata-JZNDlz%2FJAxHBxMlt30eitaYpKGQpt%2FQLsPbXYUJxDog%3D&reserved=0
Job Number=10459757, Password=aZmYhHRi

All folders are up to date. Connected to: Microsoft Exchange [43] Display Settings Q S® — -

Figure 4.15: Example email from NEOS.

-------
fj" NEOS Server Job Administration X +

C i neos-server.org/neos/admin.html

p| # Contact O Help

O ~ o x
~ * A i

o CT - Hartford | Wee...

> Sign In 1

SERVER

Enter the job number and the password of the job you wish to kill/view.
You can leave these blank if viewing the queue.

Enter the job number: 110459757
Enter the password for job: | aZmYhHRi
O View Job Queue
® View Job Results
O Kill or Dequeue Job

Submit

We want to keep our services as available and free as possible. Please consider making a con: 'button to help us keep the optimizations flowing.

©WISCONSIN -jL WISCONSIN ^4*hQb

v "H—"«"¦¦¦ ¦¦¦¦¦.." INSTIIUll fOR OISCOVTRV V 1 1 w

Terms of Use • Acknowledgements • Questions and Comments
Copyright © 2021, Wisconsin Institutes for Discovery at the University of Wisconsin, Madison

Figure 4.16: Retrieving results through the NEOS Job Administration page.

-------
fj1 NEOS Results for Job #10459757 X +

G A neos-server.org/neos/cgi-bin/nph-neos-solver.cgi
ii: Apps ¦ EPA pages ¦ R Resources ¦ AMPl | ArcGIS ¦ RSPARROW Q GitHub ¦ WMOST Recources

rieos results

SERVER

**************** *******************************

NEOS Server Versior

Dob#

Password

User

Solver

Start

End

Host

Disclaimer:

10459757
aZmYbHRi

lp:CPLEX
2021-05
2021-05-
prod-sub

This information is
implied warranty
of any kind concern
information for any

Announcements:J

We will require an
on the NEOS Server
Please see:

*********************:
3ob 10459757 has finis
You are using the solv

%% comments mmma

RBEROST demonstration
Upper CT river, 3 loac
Porous Pavement w/ sufc
With uncertainty

%%%%%%%%%%%%%%%%%%%%%3l
Checking ampl.mod for
Checking ampl.com for
Executing AMPL.
processing data,
processing commands.

2 *Untitled - Notepad
File Edit Format View Help

'9332520'
'9332526'
'9332528'
'9332532'
'9332534'
'9332536'
'9332538*
'9332542'
'9332546'
'9332548'
*9332550'
*9332552*
'9332558*
*9332560*
*9332564'
'9332566'
'9332568'
'9332570'
'9332582'
'9332584'
'9336412'
'9336426*
*9336430*
*9336432*
*9336434*
*9336440*
'9336444*
*9336446*

0
0
0
0
0

0.9

0.9
0

0.9

NEOS Server Home

0.9
0.9
0.9
0.9
0.9

0.9
0.9
0.9

0.9
0.9

0.9
0.9
1

0
0
0
0
0

0.9

m » «l

4; «r 0

Figure 4.17: Saving NEOS results as a text file.

O ~ ~ X
~ * .1 :

@ NWS NOAA Snow... | Puget Sound ~ Gentle Night Rain 1... CT - Hartford | Wee...

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

I n 7753? Cnl 17 1(¥1% Winiinw; t"f Bl Pi | lTF-ft

/N 9:03 AM

-------
4.4 Postprocessing

The postprocessing R code combines optimization results from the NEOS server with input data
to develop result summaries. The code uses R-Shiny, an R package that produces a summary of
RBEROST either within the RStudio interface or in the default web browser.

To run the postprocessing code, users can click the green triangle in the Run Postprocessor
section of the RunRBEROST. Rmd file (Figure 4.18). More than one result file can be viewed per
session, however the Shiny App will not automatically close if the viewing window is closed. To
stop the Shiny App after closing the browser window, click the red square in the Run
Postprocessor section, and a green triangle will appear that will allow the user to launch the
application again.

If RBEROST postprocessor is running within a browser, you may print the results screen to a pdf
file for your records.

3 RunRBEROST.Rmd

«¦* «s f'Knit - £ - % H* 4-QCE

without.

35 Includeuncertainty <- TRUE

37 # If running RBEROST with uncertainty, how many scenarios would you like to
view? Default is 3.

38 n.scenarios <- 3

40 # If running RBEROST with uncertainty, how different would you like the
scenarios to be? Each scenario will be solved for loading limits that are a
certain percentage lower than the previous. Default is 0.01, or 1%.

41 scenariostepchange <- 0.01

43* # Do not change anything below this line

45 source("./R/01_Optimization_Preprocessing_gateway.R", local = TRUE)

46-

48 - ## Run Postprocessor

49 To run the preprocessor. click the green triangle below. A window will pop up
either in RStudio or in your default web browser. If no green arrow appears,
and instead there is a red square, click the red square to stop any ongoing
processes and a green triangle should appear in its place, when clicked, the
green triangle will turn into a red square until the code is finished. Be aware
that simply closing the browser window will not stop the Shiny app. You must
also click the red square below if you wish to relaunch the app.

51* ~~ {r runpostprocessor} ' ~

52 source("./R/02_optimization_Runshiny.R", local = TRUE)

53»

24:1 c Chunk 2: runpreprocessor

Figure 4.18: Run the RBEROST postprocessor

-------
4.4.1 Necessary Files

The postprocessor requires upload of at least five files and will display warning messages if files
are not uploaded to the correct options (Figure 4.19). The upload of StreamCat cropland files
accepts all the state cropland files used in the optimization, which may be more than one file.
Printing to pdf is the current recommended way of capturing metadata regarding which files are
provided to the postprocessor. Required files include a result file from a NEOS optimization run,
the list of WWTPs by their COMID locations, the SPARROW input file, and the StreamCat files
for all states included in the optimization.

4.4.2 Preview Files

The preview step is not necessary but may be helpful for users to ensure they have selected and
uploaded the intended files. To preview files, click the Preview Uploads button and navigate to
the File Preview tab (Figure 4.20). At any point users can upload new files and preview them
by selecting Preview Uploads again.

4.4.3 Display Results

To view results, select View NEOS Results and navigate to the View Results tab (Figure
4.21). Several panels will appear including an option to select the scenario the user wishes to
view, the comments entered by the user to NEOS, a summary of the total cost calculated by
RBEROST, and a summary of BMP implementations by category. For RBEROST runs without
uncertainty, only one scenario is available. For RBEROST runs with uncertainty, more than one
scenario may be available, and only one will be displayed at a time. Additional panels displayed
for models with uncertainty include a panel showing the expected probability density distribution
of costs (Figure 4.22) and a panel showing the total annual expected loading to the user-specified
loading targets (Figure 4.23). If the model was unable to solve, a message will appear with the
following explanation:

"RBEROST may fail to optimize models for a variety of reasons. Often, this is a result
of loading targets that are too restrictive, or a result of not including enough BMPs for
the model to use. It may be informative to run RBEROST again with less restrictive
loading targets and/or including more BMPs in the optimization. Please check your
user inputs and refer to the Model Sensitivity section of the documentation for more
information."

In addition, in the event of failure due to AMPL syntax, errors from CPLEX will be displayed.
These errors can occasionally occur for a variety of reasons, including if the users have uploaded
the AMPL files in the wrong order during job submission.

-------
Choose Results File

Browse... RBEROSTdemo_uncertainty.txt

Upload complete

Choose the corresponding 01_UserSpecs_loadingtargets.csv' file

B rowse... 01 _U serSpecs I oadi ngtargets. csv

Upload compfete

Choose WWTP File

Browse... WWTP COM IDs.csv

Upload complete

Choose SPARROW Inputs File

Browse... ne_sparrow_model_input.csv

Upload complete

Choose State Cropland Streamcat Files

Browse...

2 files

Upload complete

Important: Selecting 'Preview Uploads' or 'View NEOS Results' before all
uploads have completed may produce unexpected results. Previews can be
found under the 'File Preview' tab and generated reports can be found under
the 'View Results' tab.

Preview Uploads View NEOS Results

Figure 4.19: Files necessary to run RBEROSTpost-processor

-------
Choose Results File

Browse. RBEROST demojjncertainty txt

Choose the corresponding "01 JJserSpecsJoadingtargets.csv' file

Browse 01_UserSpecs_loadingtargets csv

Postprocessing Step

Choose WWTP File

Browse... WWTP_COMIDs.csv

Choose SPARROW Inputs File

Browse.. ne_sparrow_modelJnput.csv

Choose State Cropland Streamcat Files

Browse

2 files

Upload complete

Important: Selecting 'Preview Uploads' or 'View NEOS Results' before all
uploads have completed may produce unexpected results. Previews can be
found under the 'File Preview' tab. and generated reports can be found under
the View Results' tab

View NEOS Results

RBEROST Results Postprocessing Step

^ >

File Preview l View Results

File Previews

NEOS Results File :

NEOS Server Version 6.0
Job# 10702977
Password: unbPLyoe
User

User Inputs File (loading targets):

Waterbody_Name ComID Watershed_HUC Percent_Reduction TN_or_TP OutofNetworkFlag_X TermFlag_X

9332552 0.00 0.10 TN NA X

CT River at MA
border

Back Lake 4592401 0.00

(NHLAK801010203-

01-01)

Forest Lake 4594723 0.00

(NHLAK802010401 -

01-01)

WWTP File :

State Plant_Name NPDESJD COMID

012 TP

Figure 4.20: Previewing files for RBEROST postprocessor.

-------
Choose Results File

Browse RBEROSTdemo_uncertaintytxt

Upload complete

Choose the corresponding "01_UserSpecs_loadingtargets.csv" file

Browse 01_UserSpecs_loadingtargets.csv

Upload complete

Choose WWTP File

Browse WWTP_COMIDs csv

Upload complete

Choose SPARROW Inputs File

Browse . ne_sparrow_modelJnput.csv

Upload complete

Choose State Cropland Streamcat Files

Browse...

2 files

Upload complete

Important: Selecting 'Preview Uploads' or 'View NEOS Results' before all
uploads have completed may produce unexpected results. Previews can be
found under the 'File Preview' tab. and generated reports can be found under
the "View Results' tab.

Preview Uploads View NEOS Results

Figure 4.21: Viewing RBEROST results.

Postprocessing Step

RBEROST Results Postprocessing Step

Scaled-up Optimization Results

To save this report, open the RBEROST postprocessor in a web browser and use the print to pdf functionality.
User Notes:

RBEROST demonstration
Upper CT river, 3 loading targets.

Porous Pavement w/ subsurface infiltration excluded, 10% cap on ag ponds
With uncertainty

Your model has successfully solved.

The total cost to reduce loads to the limit you provided is $22,246,497.32.

Estimates of Total Annualized Cost ranged from $21,914,809 54 to $22,673,807.02. RBEROST costs are
usually most influenced by the amount of retrofitting necessary to install urban BMPs. More retrofitting leads to
higher costs.

Below is the probability density distribution of cost estimates for this scenario

-------
Estimates of Total Annualized Cost ranged from $21,914,809.54 to $22,673,807.02. RBEROST
costs are usually most influenced by the amount of retrofitting necessary to install urban BMPs.
More retrofitting leads to higher costs.

Below is the probability density distribution of cost estimates for this scenario.

(/)
C
-------
This suggested plan has an estimated minimum of 20% likelihood of meeting each of the specified loading targets UteHioods of meeting each specified loading target individually are
jted in the table betow. "Likelihood" refers to the percentage of the probability density of estimated total annual load that fa Is below Hie user specified loading targe:.

95% CI of Estimated Annual Total Likelihood of Meeting Your Loading

Loading Target Your Specified Target Load Loading Target

CT River at MA border 10% reduction in TN; or 6,2t 8,700.0 kg 8. 1 58.828.3 - 6.318.925.9 kg N.'yr 42%

Wyr

Back Lake (NHLAK801010203-01-01) 12% reduction in TP: or 3.8 kg Ptyr 1.1 - 18.1 kg P/yr 20.5%

Forest Lake (NHLAK802010401 -01- 2% reduction in TP; or 2.7 kg Pfyr 0.7 - 7 4 kg Ptyr 41.5%

Ot)

Below are the probability denssty distributions of totaJ annual load estimates at each of the toadaig targeis specified by the user. The vertical doited grey lines denote the specified target
toads a; each loading targe*. Probability densities to the left of the grey Sre are ;oad estimates that meet the loading target, and probability densities to the right of the grey line are toad
estimates that exceed the loading target.

6 200 000

6.300.000

Total Nitrogen Loading
to CT River at MA border
[kg N / yr)

Total Phosphorus Loading
to Back Lake (NHLAK801010203-01-01)
(kg P/yr)

Figure 4.23: Viewing results of uncertainty analysis in total annual loading. The leading
sentence describes the lowest likelihood of successfully meeting the loading targets described out
of all the targets included. The table shows all of the included loading targets, their specified
reductions, the 95 percent confidence intervals of bootstrapped estimates of total annual loading,
and the likelihood of meeting loading targets. The likelihood of meeting loading targets refers to
the proportion of the probability density distribution that falls below the specified target-
reduction. Below the table are plots of the probability density distributions of estimated annual
total load at each specified loading target. Vertical grey dotted lines indicate the target-
reduction specified. Probability density to the left of this line meets the target, and probability
density to the right of this line exceeds the target.

-------
4.4.4 Download Detailed Results

COMID-specific BMP implementation decisions can be downloaded from the Shiny App for any
single scenario (Figure 4.24). Point-source decisions are not available for download and are
displayed within the Shiny interface. Additional contextual information is included in the
downloadable csv files including the HUC12 code and state for each COMID, the size of the
catchment, and the total area of agricultural land for ag BMPs, or urban land for urban BMPs, or
the total river reach length for riparian buffer BMPs.

The model chose to implement WWTP retrofits at the following locations.

Plant Name
Ludlow
Springfield
Clare mont
Hanover

Waterway Only (ac)

75 1

The model chose to implement the total area of the following urban BMPs.

Gravel Wetland (ac) Infiltration Basin (ac)

136.1 106,172.3

Download Urban BMPs by COMID

The model chose to implement the total length of the following riparian buffer BMPs.

Forested Buffer (ft of bank) Grassed Buffer (ft of bank)

2,184.2 60,348.7

Download Riparian Buffer BMPs by COMID

Figure 4.24: DownloadingRBEROSTresults.

The model chose to implement the total area of the following agricultural BMPs.

Fert 20 (ac) Filterstrip (ac) MIN TILL (ac) Ponds (ac) Terrace Waterway (ac)

2,889.2 74,529.2 117.1 2,435.6 20.9

Download Agricultural BMPs by COMID

-------
5 Model Sensitivity

RBEROST results are sensitive to provided data and user inputs. Model results are most
sensitive to the user-defined loading TN target in the terminal reach of the watershed, the
estimates of "other" TN loads to that terminal reach, and the TN removal efficiencies of several
BMPs that are implemented extensively (specifically Filterstrip, Ponds, and Infiltration Basins).
"Most sensitive" means that a 1% change or less in both the positive and negative direction cause
either total costs or spending on individual BMPs to change by more than 10%. Other parameters
may have been more sensitive in one direction than the other, or generally less sensitive in both
directions. RBEROST sensitivity to each individual user input is given in Table 5.1, and
sensitivity to each individual data parameter provided in the RBEROST preprocessing input csv
files is given in Table 5.2. The Parameter columns in Tables 5.1 and 5.2 are the names of the
parameters as they appear in the AMPL model and data scripts. The Set member columns
generally describe the specific BMP being defined but may also reflect other groupings, such as
capital or operations costs, urban or agricultural area, or the source of nutrient loading baselines.

Tolerances are defined as the relative amount of variation in the estimate of each parameter that
produces RBEROST results with costs that are within 10% of the reference cost, and that
produces results with relative amounts of spending on each BMP that are within 10% of the
amount spent on that BMP in the reference result. The reference result is the RBEROST solution
to a problem with all BMPs included and no limits on implementation (sometimes referred to as
the 'unconstrained' model. Parameters are considered sensitive if a change of less than 10%
produces results that are more than 10% different from the reference. Parameters are considered
highly sensitive if a change of less than 1% produces results that are more than 10% different
from the reference. The last two columns in Tables 5.1 and 5.2 describe the sensitivity to
parameters as they are decreased from their default value and as they are increased from their
default value. Parameters with tolerances marked as "not applicable" either have default values
of 0, or already have default values that are at the minimum or maximum of their permitted
range.

The user specified loading target at the terminal reach of the watershed was the single biggest
predictor of whether RBEROST could succeed in solving the optimization problem. Figure 5.1
shows the model's success (1 on the y-axis) or failure (0 on the y-axis) as a function of terminal
reach target loads for 1000 iterations of the same scenario. Each iteration differed in the exact
estimate of each parameter. The estimate of each parameter was sampled from a distribution that
described the likely values that the parameter may take. Most parameters were sampled from the
normal distributions used for uncertainty analysis. User specified parameters were mostly drawn
from uniform distributions encompassing the allowed range of that parameter. Maximum
implementations were sampled between 70%-100% and loading targets were sampled between
2% and 20% reductions. Minimum implementations were not varied, and were kept at 0%,
because it is unlikely that managers would wish to impose minimum implementations of
multiple BMPs at once. Additionally, minimum implementations impose more constraints on the
model than maximum implementations and more frequently prevent the optimizer from finding
feasible solutions. The image in Figure 5.1 is from an optimization of the Upper Connecticut
River watershed, which drains to the Massachusetts boarder. Occasionally optimization of
scenarios fail even when target loading is less restrictive. This is a function of an overly

-------
constrained problem and may be alleviated by relaxing maximum restrictions on BMPs, by
including more BMPs, or by relaxing intermediate loading targets.

1 ¦

0.75

0.50

0.25

0.00- ¦¦¦¦¦§

5500000 6000000 6500000

Target Loading at the MA Border (kg N/yr)

Figure 5.1: Sensitivity ofRBEROST success in finding an optimized result as a function of
loading target.

RBEROST predictions of annualized total cost increase exponentially as loading targets become
more restrictive (Figure 5.2). Figures 5.3, 5.4, 5.5, and 5.6 show the annualized spending on each
specific BMP. BMPs are differentiated by color. The clouds of points represent 635 of the 1000
scenarios described above which were optimized with feasible solutions. The solid lines show
the LOESS smoothed curve of the data. RBEROST consistently implements the same WWTP
upgrades (Figure 5.3) and reliably chooses the same agricultural BMPs dependent on target
loading (Figure 5.4). RBEROST less clearly distinguishes between urban BMPs (Figure 5.5) and
riparian zone BMPs (Figure 5.6). This means that the choice of which urban or riparian zone
BMPs to implement depends on the exact estimates of efficiency and cost. To test if this affects a
certain model scenario, users may wish to run RBEROST several times changing the costs in
01_UsenSpecs_BMPs. csv slightly for each run. If RBEROST subsequently chooses different
urban BMPs, additional modeling for urban BMPs may be advantageous. Generally, users will
benefit by having the most accurate estimates of cost, efficiency, suitability of sites for each
BMP, and the expected difficulty of retrofitting existing urban infrastructure with urban BMPs.

-------
$90,000,000-

5500000 6000000 6500000

Target Loading at the MA Border (kg N/yr)

Figure 5.2: Predicted annual total cost as a function of loading target.

Figure 5.3: Annualized spending on point source BMPs as a function of loading target

-------
BMP

Conservation

Contour_Farming

Fert_20

Filterstrip

Manure_lnjection

MIN_TILL

Ponds

Terrace_Waterway
Waterway_Only

5500000 6000000 6500000

Target Loading at the MA Border (kg N/yr)

Figure 5.4: Annualized spending on agricultural BMPs as a function of loading target

$60,000,000-

c $45,000,000

-------

c $3,000,000

C
-------
Table 5.1: RBEROST tolerance to variation in User Specifications
Parameter Set member (BMP or other)

agtrac

max

agfrac

max

agfrac

max

agfrac

max

agfrac

max

agfrac

max

agfrac

max

agfrac

max

agfrac

max

agfrac

max

agfrac

min

agfrac

min

agfrac

min

agfrac

min

agfrac

min

agfrac

min

agfrac

min

agfrac

min

agfrac

min

agfrac

min

loads_lim_N 1
loadslimPl
loads lim P2

Conservation

ContourFarming

Fert_20

Filterstrip

Manurelnjection

MINTILL

Ponds

TerraceOnly

T errace_W aterway

Waterway _Only

Conservation

ContourFarming

Fert_20

Filterstrip

Manurelnjection

MINTILL

Ponds

TerraceOnly

T errace_W aterway

Waterway _Only

Tolerance to decreased

values

not sensitive

highly sensitive

not sensitive

highly sensitive

not sensitive

highly sensitive

not sens

tive

not appl

cable

not appl

cable

not appl

cable

not appl

cable

not appl

cable

not appl

cable

not appl

cable

not appl

cable

not appl

cable

not appl

cable

highly sensitive

not sensitive

highly sensitive

Tolerance to increased

values

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

not

appl

cable

highly sensitive
sensitive
not sensitive

-------
Parameter

ripbuffracmax

ripbuffracmin

urbandesigndepth

urbanfracmax

urban frac max

Set member (BMP or other)

ForestedBuffer

GrassedBuffer

ForestedBuffer

GrassedBuffer

B i ofi 1 trati onwUnderdrai n

B i oretenti on_B asi n

En h an ced_B i ofi 1 trati on_w_I S R

ExtendedDryDetenti onBasi n

GrassS wal ewdetenti on

GravelWetland

InfiltrationBasin

I n fi 1 trati on_C h am b er

Infi 1 trati on_T rench

PorousPavementwsubsurfaceinfiltration

PorousPavementwunderdrain

SandFilterwunderdrain

WetPond

B i ofi 1 trati on w Underdrai n

B i oretenti on_B asi n

En h an ced_B i ofi 1 trati on_w_I S R

Extended Dry Detenti on Basi n

GrassS wal ewdetenti on

GravelWetland

Infiltration Basin

Tolerance to decreased

Tolerance to increased

values

sensitive

not applicable

highly sensitive

not applicable

not sensitive

sensitive

not sensitive

highly sensitive

sensitive

highly sensitive

not sensitive

sensitive

not sensitive

not applicable

not sensitive

not applicable

not sensitive

not applicable

not sensitive

not applicable

not sensitive

not applicable

highly sensitive

not applicable

sensitive

not applicable

-------
Parameter

Set member (BMP or other)

urban

frac

max

Infiltration Chamber

urban

frac

max

Infiltration Trench

urban

frac

max

PorousPavementwsubsurface

infiltration

urban

frac

max

PorousPavementwunderdrain

urban

frac

max

SandFilterwunderdrain

urban

frac

max

WetPond

urban

frac

min

Biofiltration w Underdrain

urban

frac

min

Bioretention Basin

urban

frac

min

Enhanced Biofiltration w ISR

urban

frac

min

Extended Dry Detention Basin

urban

frac

min

GrassS wal ewdetenti on

urban

frac

min

GravelWetland

urban

frac

min

Infiltration Basin

urban

frac

min

Infiltration Chamber

urban

frac

min

Infiltration Trench

urban

frac

min

PorousPavementwsubsurface

infiltration

urban

frac

min

PorousPavementwunderdrain

urban

frac

min

SandFilterwunderdrain

urban

frac

min

Wet Pond

Tolerance to decreased Tolerance to increased

values

not sensitive

not appl

cable

not sensitive

not appl

cable

sensitive

not appl

cable

not sensitive

not appl

cable

not sensitive

not appl

cable

not sensitive

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

not applicable

not appl

cable

-------
Table 5.2: RBEROST tolerance to variation in parameters in RBEROST data sets

Parameter

Set member (BMP or other)

Tolerance to
decreased values

Tolerance to
increased values

agcostscapital

Conservation

not sensitive

agcostscapital

Contour Fanning

not sensitive

agcostscapital

Fert 20

not sensitive

agcostscapital

Filterstrip

not sensitive

agcostscapital

Manure Injection

not sensitive

agcostscapital

MINTILL

not sensitive

agcostscapital

Ponds

sensitive

agcostscapital

Terrace Only

not sensitive

agcostscapital

T errace_W aterway

sensitive

not sensitive

agcostscapital

Waterway _Only

sensitive

not sensitive

agcostsoperations

Conservation

not sensitive

agcostsoperations

Contour Fanning

not sensitive

agcostsoperations

Fert 20

not sensitive

agcostsoperations

Filterstrip

sensitive

agcostsoperations

Manure Injection

not sensitive

agcostsoperations

MINTILL

not sensitive

agcostsoperations

Ponds

not sensitive

agcostsoperations

Terrace Only

not sensitive

agcostsoperations

T errace_W aterway

not sensitive

agcostsoperations

Waterway _Only

not sensitive

ag_effic_N

Conservation

not sensitive

ag_effic_N

Contour Fanning

not sensitive

ag_effic_N

Fert 20

not sensitive

-------
Parameter

Set member (BMP or other)

ag_effic_N

Filterstrip

ag_effic_N

Manure Injection

ag_effic_N

MINTILL

ag_effic_N

Ponds

ag_effic_N

Terrace Only

ag_effic_N

T errace_W aterway

ag_effic_N

Waterway _Only

agefficP

Conservation

agefficP

Contour Farming

agefficP

Fert 20

agefficP

Filterstrip

agefficP

Manure Injection

agefficP

MINTILL

agefficP

Ponds

agefficP

Terrace Only

agefficP

T errace_W aterway

agefficP

Waterway _Only

agcostfrac

: NA

area

urban

baseloads_N 1

point

baseloads_N 1

urban

baseloadsPl

Tolerance to
decreased values

highly sensitive

not sensitive

highly sensitive

not sensitive

highly sensitive

not sensitive

sensitive

not sensitive

Tolerance to
increased values

highly sensitive

not sensitive

highly sensitive

not sensitive

sensitive

not sensitive

sensitive

highly sensitive

not sensitive

-------
Parameter

Set member (BMP or other)

baseloadsPl

point

baseloadsPl

urban

baseloads_P2

point

baseloads_P2

urban

other_loads_N 1

; NA

otherloadsPl

other_loads_P2

| NA

pointcosts

capital

pointcosts

operations

point effic N

point

pointefficP

point

riparianloadN 1

j NA

riparianloadPl

| NA

riparianload_P2

riparianremoval N 1

Forested

Buffer

riparianremoval N 1

Grassed_

Buffer

riparianremoval PI

Forested

Buffer

riparianremoval PI

Grassed_

Buffer

riparianremoval P2

Forested

Buffer

riparianremoval P2

Grassed_

Buffer

ri pbufcostscapi tal

Forested

Buffer

ri pbufcostscapi tal

Grassed_

Buffer

ri pbufcostsoperati on s

Forested

Buffer

Tolerance to
decreased values

not sensitive

highly sensitive

sensitive

not sensitive

sensitive

not applicable

not sensitive

highly sensitive

sensitive

not sensitive

sensitive

not sensitive

Tolerance to
increased values

not sensitive

highly sensitive

not sensitive

highly sensitive

not sensitive

sensitive

not sensitive

not applicable

not sensitive

sensitive

highly sensitive

not sensitive

sensitive

not sensitive

-------
Parameter

ri pbufcostsoperati on s

runoffcoeffurban

totalbanklength

unbufferedbanklength

urbanbm p i m pi em entati onpotenti al

urban bm p i m pi em entati onpotenti al

urbancostadj ustmentcoef

urbancosts

urban costs

Set member (BMP or other)

Tolerance to

decreased values

increased values

GrassedBuffer

not sensitive

urban

sensitive

not sensitive

ForestedBuffer

sensitive

highly sensitive

GrassedBuffer

highly sensitive

sensitive

Biofiltration w Underdrain

not sensitive

Bioretention Basin

not sensitive

Enhanced Biofiltration w 1SR

not sensitive

Extended Dry Detention Basin

not sensitive

GrassS wal ewdetenti on

not sensitive

GravelWetland

highly sensitive

not sensitive

Infiltration Basin

sensitive

not sensitive

Infiltration Chamber

not sensitive

Infiltration Trench

not sensitive

PorousPavementwsubsurfaceinfiltration

sensitive

PorousPavementwunderdrain

not sensitive

SandFilterwunderdrain

not sensitive

WetPond

not sensitive

sensitive

Biofiltration w Underdrain

not sensitive

Bioretention Basin

not sensitive

capital

sensitive

Enhanced Biofiltration w 1SR

sensitive

not sensitive

Extended Dry Detention Basin

not sensitive

-------
Parameter

Set member (BMP or other)

Tolerance to

decreased values

increased values

urban

costs

GrassS wal ewdetenti on

not sensitive

urban

costs

GravelWetland

highly sensitive

sensitive

urban

costs

Infiltration Basin

sensitive

highly sensitive

urban

costs

Infiltration Chamber

not sensitive

urban

costs

Infiltration Trench

not sensitive

urban

costs

operations

sensitive

urban

costs

PorousPavementwsubsurface

infiltration

not sensitive

sensitive

urban

costs

PorousPavementwunderdrain

not sensitive

urban

costs

SandFilterwunderdrain

not sensitive

urban

costs

WetPond

not sensitive

urban

_effic_

Biofiltration w Underdrain

not sensitive

urban

_effic_

Bioretention Basin

not sensitive

urban

_effic_

Enhanced Biofiltration w ISR

not sensitive

sensitive

urban

_effic_

Extended Dry Detention Basin

not sensitive

urban

_effic_

GrassS wal ewdetenti on

not sensitive

urban

_effic_

GravelWetland

sensitive

highly sensitive

urban

_effic_

Infiltration Basin

highly sensitive

urban

_efflC_

Infiltration Chamber

not sensitive

urban

_efflC_

Infiltration Trench

not sensitive

sensitive

urban

_efflC_

PorousPavementwsubsurface

infiltration

sensitive

urban

_efflC_

PorousPavementwunderdrain

not sensitive

urban

_efflC_

SandFilterwunderdrain

not sensitive

urban

_efflC_

WetPond

not sensitive

urban

effic

Biofiltration w Underdrain

not sensitive

-------
Parameter

Set member (BMP or other)

Tolerance to
decreased values

Tolerance to
increased values

urbanefflcP

Bioretention Basin

not sensitive

urbanefflcP

Enhanced Biofiltration w ISR

not sensitive

urbanefflcP

Extended Dry Detention Basin

not sensitive

urbanefflcP

GrassS wal ewdetenti on

not sensitive

urbanefflcP

GravelWetland

not sensitive

urbanefflcP

Infiltration Basin

not sensitive

urbanefflcP

Infiltration Chamber

not sensitive

urbanefflcP

Infiltration Trench

not sensitive

urbanefflcP

PorousPavementwsubsurfaceinfiltration

not sensitive

urbanefflcP

PorousPavementwunderdrain

not sensitive

urbanefflcP

SandFilterwunderdrain

not sensitive

urban effic P

Wet Pond

not sensitive

6 Data Dictionary

Table 6.1 lists the input files that are read into the preprocessing and postprocessing R code, describes the contents of the
inputs and specific data fields used in the model, units of data, and data sources. Table 6.2 lists the output files that are created
by the postprocessing R code, the data units, and the sources of associated data.

-------
Table 6.1: Data Dictionary of Input Files
Input File Name

01_UserSpecs_BMPs.csv

Input File Description

User specification file for selection of
BMPs to be included in optimization
model, BMP capital and operations &
maintenance costs, and urban BMP
runoff depth specification.

File includes default capital and
operations & maintenance costs as well
as default runoff depths. Minimum and
maximum runoff depth fields
(Min_RD_in and Max_RD_in) indicate the
range between which the user can
specify urban BMP runoff depths.
Min_RD_in and Max_RD_in fields are not
used in scaled-up optimization model
code. Separate columns must be
included for the capital and operations
& maintenance costs for each state
included in the geographic range of
optimization. Column titles must follow
the naming convention "capital_[two
letter state code]" and "operations_[two
letter state code]".

Data Units (If

Source(s)

Applicable)

Per-area and

BMP cost

per-volume

data sources

BMP cost

described in

units

Table 2.

described in

Minimum

file. Runoff

and

depth in

maximum

inches. Buffer

runoff depths

width in feet.

for Urban

BMPs from

I New

Hampshire

MS4 permit

BMP

performance

curves (New

Hampshire

Department

Environment

al Services,

2020b).

Minimum

and

maximum

buffer widths

from Houle et

al. (2019).

-------
Input File Name

OlJJserSpecsJoadingtargets.csv

ACRE_H UC 12_HRU_Summary_compare Baseline, c
sv

Input File Description

User specification file for providing
loading targets to include in RBEROST.
Fields include Waterbody name, NHD+
C0M1D, Percent Reduction, choice of TN
or TP, flags for out of network, and flags
for terminal reaches. If the user has
loading targets on a mass basis, they
should input the percent reduction from
current loadings that the mass target
represents.

Agricultural BMP efficiencies
summarized by HUC12 (or if
unavailable, HUC10 and/or HUC8)
based on ACRE database. "HUC" fields
represent the hydrologic unit code
(HUC12/HUC10/HUC8). "Scenario" field
represents BMP scenarios, modeled by
White et al. (2019) for the ACRE
database. "MeanTN_Effic" field reflects
average nitrogen removal efficiency per
scenario, with respect to "Baseline"
conditions.

Data Units (If Source(s)
Applicable)

Percent Not

Applicable

Percent

Original
ACRE
database
from White
et al. (2019).
Data

summarized
at HUC-levels
by RBEROST
associated R
scripts.

-------
Input File Name

ACRE_HUC12_HRU_Summary_compareNoPractic
e.csv

AgBMPEffic_FertManure.csv

Input File Description

Agricultural BMP nitrogen removal
efficiencies for 20% fertilizer reduction
and manure injection (do not vary by
location). "Category" field reflects the
agricultural BMP category within
RBEROST, "BMP" field reflects the best
management practice considered, and
"N_Efficiency" field reflects the nitrogen
removal efficiency.

Data Units (If Source(s)

Applicable)

Percent Original

ACRE
database
from White
et al. (2019).
Data

summarized
at HUC-levels
by RBEROST
associated R
scripts.

Percent EPA, NRCS

-------
Input File Name

EQIPcosts_overyears.csv

Input File Description

EQIP costs for agricultural and riparian
buffer BMPs for years 2018-2021.
"BMP_Category" field reflects whether a
BMP is implemented on agricultural
(row crop) land or in riparian buffers.
"BMP" field reflects the best
management practice considered,
"capital_units" reflect the cost units for
capital costs as cost / unit area,
"operations_units" reflect the cost units
for operation costs as cost / unit area.
The remaining fields describe the costs
from EQIP payment schedules follow
the naming convention "[capital or
operations costs]_[state]_[year]". At
least one year of data must be provided
for each state included in the
geographic range of optimization.

Data Units (If Source(s)
Applicable)

Dollars per U.S.
unit area Department

Agriculture
Staff. 2021

-------
Input File Name

LengthinBuffer_2016.csv

Input File Description

The amount of each stream reach that is
in riparian buffers of certain widths, as
of 2016. Fields include "comid,"
"totalbanklength_ft," which is two times
the stream reach,

"totalbanklength_ft_se" describing the
standard error associated with the total
bank length, and fields describing the
length of stream bank in buffer and the
standard error around those values.
Naming conventions for the rest of the
fields are [Riparian buffer
BMP]_["buffer" for estimates,

"buffer_se" for standard
errors]_[minimum buffer width to be
considered "in buffer"]_ft.

Data Units (If Source(s)

Applicable)

Feet Stream

reaches from

McKay et al.

2012,

landcover

from Yang et

al. 2018.

Summaries

provided

following R

scripts and

ArcPython

model

builder

scripts

associated

with

RBEROST.

-------
Input File Name

NdepChange_2012_2019.csv

ne_sparrow_modelJnput.csv

Input File Description

Changes in TN deposition. Fields
include "comid," "TDEP_TN_2012"
(describing annual mass deposition in
2012), "TDEP_TN_2019 (describing
annual mass deposition in 2019)," and
"Change_2012_2019 (the percent
change between years)"

Northeastern Regional SPARROW
Model input dataset. Fields used in
scaled-up optimization model include
"HUC_12," "comid," "lncAreaKM2"
(incremental area per catchment), and
"urban_km2" (urban area per
catchment).

Data Units (If
Applicable)

kg-N/ha, kg-
N/ha, percent

Incremental
and urban
areas in km2

Source(s)

National
Atmospheric
Deposition
Program
(2021);
COMID
shapefiles
from McKay
et al. 2012;
summarized
with ArcMap
and R scripts
for RBEROST
Ator (2019)

-------
Input File Name

ne_sparrow_model_output_tn.csv

Input File Description

Northeastern Regional SPARROW Total
Nitrogen Model output dataset. Fields
used in RBEROST include "comid,"
"in_poin" (incremental point source
load per catchment), "in_urb"
(incremental urban source load),
"in_fert" (incremental fertilizer load),
"in_fix" (incremental load from direct
fixation by crops), "injnanu"
(incremental load from manure
applications), "in_sept" (incremental
load from septic systems), "in_atmo"
(incremental load from atmospheric
sources), and "DEL_FRAC" (fraction of
incremental load delivered to terminal
reach). Fields "sin_poin," "sin_urb,"
"sin_fert," "sin_fix," "sinjnanu,"
"sin_sept," "sin_atmo," and
"SE_DEL_FRAC" describe the standard
errors around each field that is used in
RBEROST.

Data Units (If Source(s)
Applicable)

Incremental Ator(2019)

loadings in

kg/yr

-------
Input File Name

ne_sparrow_model_output_tp.csv

Input File Description

Northeastern Regional SPARROW Total
Phosphorus Model output dataset.

Fields used in RBEROST include
"comid," "ip_poin" (incremental point
source load per catchment), "ip_urb"
(incremental urban source load),
"ip_fert" (incremental fertilizer load),
"ip_manu" (incremental load from
manure applications), "ip_rock"
(incremental load from bedrock
weathering), and "DEL_FRAC" (fraction
of incremental load delivered to
terminal reach). Fields "sip_poin,"
"sip_urb," "sip_fert," "sipjnanu,"
"sip_rock," and "SE_DEL_FRAC" describe
the standard errors around each field
that is used in RBEROST.

Data Units (If Source(s)
Applicable)

Incremental Ator(2019)

loadings in

kg/yr

-------
Input File Name

NH_streamcat_2011_cropland.csv

Input File Description Data Units (If Source(s)

Applicable)

New Hampshire percent cropland, Open Percent Hill et

development, low intensity al. (2015)

development, medium intensity

development, and high intensity

development data from National Land

Cover Database, 2011 (catchment-

specific data downloaded from

StreamCat). Fields used in scaled-up

optimization model include "comid,"

"PctCrop2011Cat," "PctOpUrb2011Cat,"

"PctLowllrb2011Cat,"

"PctMedUrb2011Cat," and

"PctHiUrb2011Cat."

*Note: if optimizing a geographic extent
that includes states beyond those used
in the Upper Connecticut Case study,
additional state files may be included.

Locating these files in the
Preprocessor/Inputs folder and
following the naming convention of
"[two letter state

code]_streamcat_2011_[cropland or
imperv].csv" will allow them to be used
automatically by RBEROST.

-------
Input File Name

NH_streamcat_2011_imperv.

NHD+infiltrationrates.csv

Input File Description

New Hampshire percent
imperviousness data from National
Land Cover Database (catchment-
specific data downloaded from
StreamCat). Fields used in scaled-up
optimization model include "comid"and
"PctImp2011Cat."

*Note: if optimizing a geographic extent
that includes states beyond those used
in the Upper Connecticut Case study,
additional state files may be included.
Locating these files in the
Preprocessor/Inputs folder and
following the naming convention of
"[two letter state

code]_streamcat_2011_[cropland or
imperv].csv" will allow them to be used
automatically by RBEROST.

Data Units (If
Applicable)

Percent

Source(s)

Hill et
al. (2015)

The average infiltration rate in each
COMID. Fields include "comid,"
"infiltrationratejnperhr" (the average
infiltration rate in inches per hour) and
"infiltrationrate_inperhr_dist" (a
resampling of infiltration rates used for
RBEROST uncertainty analysis. Values
are a character string of 10 samples that
are parsed within RBEROST)

inches per

Hydrologic

hour, list of

soil group

values in

values from

inches per

Soil Survey

hour

Staff (2020a,

2020b).

Infiltration

rates for

hydrologic

soil groups

from USDA

NRCS2009

-------
Riparian Efficiencies.csv

Performance curves describing nutrient
removal efficiency in riparian buffers.
Fields include "comid," and fields with
character entries of functions that
RBEROST will parse to calculate
efficiency. All functions are functions of
buffer width, and the coefficients
depend on slope and hydrologic soil
group. Naming convention follows [N or
P, designating which nutrient removal
efficiency is being described]_[type of
riparian buffer BMP]_[approximate
width of buffer].

Though all equations are a function of
buffer width, the slope and hydrologic
soil group may differ based on the
buffer width which may change the
coefficients. Columns that describe
uncertainty are designated with
"_uncertainty" at the end. Uncertainty
columns contain a character string of 10
different equations produced by
sampling the slope and hydrologic soil
groups of each buffer width. These
character strings are parsed within
RBEROST.

Percent

Methodology:
Houle etal.
2019. River
reach shape
files: McKay
et al. 2012.
Hydrologic
soil group:
gNATSGO
(Soil Survey
Staff, 2020a;
Soil Survey
Staff,

2020b;).
Slope data:
Verdin, 2017

-------
Riparian Loadings.csv

UrbanBMPPerformanceCurves.csv

Nutrient loading from riparian zones.
Fields include "comid,"
"N_riparian_kgyr" "P_riparian_kgyr,"
"N_riparian_kgyr_se," and
"P_riparian_kgyr_se," where N or P
designates the nutrient being described,
and _se designates standard errors.

kg-N/yr; kg-
P/yr; kg-N/yr;
kg-P/yr

Performance curves that describe the
nutrient removal efficiency of urban
BMPs. Fields include: "BMP" (the best
management practice under
consideration), "Pollutant" (designates
if the performance curve relates to N or
P), "InfiltrationRateJnperhr" (the
infiltration rate that the performance
curve assumes. BMPs that do not
depend on infiltration rate have "NA."),
"Best.Fit.Curve" (a character string that
lists the form of the performance
curve), and "Coef.l," "Coef.2" and
"Coef.3." The "Coef" columns list the
coefficients that correspond with the
function in "Best.Fit.Curve." Standard
error around these coefficients are
given in "Coef.l_se," "Coef.2_se" and
"Coef.3 se."

y in

"Best.Fit.Curv
e" are in units
percent,
where x is the
rating depth
of the BMP in
inches.

Methodology
for

calculating
loading:

Houle et al.
2019. Land
cover data:
Jin et al.
2019, Yang et
al. 2018.

River reach
shape files:
McKay et al.
2012
New

Hampshire

Department

Environment
al Service
(2020b)

-------
VT_streamcat_2011_cropland.csv

Vermont percent cropland, Open
development, low intensity
development, medium intensity
development, and high intensity
development data from National Land
Cover Database, 2011 (catchment-
specific data downloaded from
StreamCat). Fields used in scaled-up
optimization model include "comid,"
"PctCrop2011Cat," "PctOpUrb2011Cat,"
" P ctL o wU rb 2 011C at,"
"PctMedUrb2011Cat," and
"PctHiUrb2011Cat".

code]_streamcat_2011_[cropland or
imperv].csv" will allow them to be used
automatically by RBEROST.

Percent Hill et

, al. (2015)

-------
VT_streamcat_2011_imperv.csv

WWTP_BaselineRemoval_Finergrain.csv

Vermont percent imperviousness data
from National Land Cover Database
(catchment-specific data downloaded
from StreamCat). Fields used in scaled-
up optimization model include "comid"
and "PctImp2011Cat."

code]_streamcat_2011_[cropland or
imperv].csv" will allow them to be used
automatically by RBEROST.
An extension of

WWTP_COMIDs_BslnRemoval.csv that
includes loading for each year available
in New Hampshire Department of
Environmental Services 2020a.
RBEROST uses the

"load_lbday_2014/2015/2016/2017/2
018" columns to calculate the standard
error of change over time.

Percent

kg/yr or
lb/day as
designated in
column
headings

Hill et
al. (2015)

New

Hampshire

Department

Environment
al Services
(2020a)

-------
WWTP_COM I Ds_Bs 1 n Re m ova 1.

WWTP_COMIDs.csv

WWTP_RemovalEffic.csv

Percentage change in WWTP effluent Percent
loadings from 2014-2018 based on
Newport Wastewater Treatment
Facility NPDES Permit No. NH0100200.

Data used to scale point source loading
estimates from the Northeastern Total
Nitrogen SPARROW Model
(representing 2012 loadings) to current
(as of 2018) loadings.

"Rem_2 014_2 018_load_ch" field
represents percent loading change from
2014-2018 per WWTP.

Crosswalk between the WWTPs
included in the scaled-up optimization
model example application and their
locations along the NHDPlus Version 2
flow network.

Estimated nitrogen removal efficiencies Percent
based on BioWin modeling. "Category"
field reflects the point source BMP
category within the scaled-up
optimization model code, "BMP" field
reflects the WWTP considering
implementation of low-cost retrofits,
and "N_Efficiency" field reflects the
nitrogen removal efficiency.

New

Hampshire

Department

Environment
al Services
(2020a)

Not applicable

McKay et
al. (2012),
Ator (2019)

Environment
al (2015)

-------
Table 6.2: Data Dictionary of Output Files

Input File Name

AgBMP_by_ComID_Scenariol.csv (or
other scenarios, if not renamed by user)

Input File Description

Urban_by_ComID_Scenariol.csv (or other
scenarios, if not renamed by user)

RipBufBMP_by_ComID_Scenariol.csv (or
other scenarios, if not renamed by user)

Agricultural (row crop) BMP implementation by COMID.
Fields include "COMID," "HUC_12" (12 digit hydrologic
unit code), "IncAreaKm" (the area of the catchment
defined by COMID), "ag_km2" (the area within that COMID
defined as ag), "PctCrop2011Cat" (the percent of NLCD
2011 pixels categorized as row crop in the COMID as
summarized by StreamCat) and "StateAbbrev" (VT or NH).
Any additional fields follow the naming convention of
[BMP name]_[either Fraclmplement (the fraction of ag
area treated with this BMP) or areaac (the total acreage of
area treated with this BMP)]

Urban BMP implementation by COMID. Fields include
"COMID," "HUC_12" (12 digit hydrologic unit code),
"IncAreaKm" (the area of the catchment defined by
COMID), "urban_km2" (the area within that COMID
defined as urban), and "StateAbbrev" (VT or NH). Any
additional fields follow the naming convention of [BMP
name]_[either Fraclmplement (the fraction of urban area
treated with this BMP) or areaac (the total acreage of area
treated with this BMP)]

Riparian buffer BMP implementation by COMID.

Fields include "COMID," "HUC_12" (12 digit hydrologic
unit code), "IncAreaKm" (the area of the catchment
defined by COMID), "totalbanklength_ft" (the distance of
bank within the COMID, or 2 times "riverreachlength_ft"),
"StateAbbrev" (VT or NH), and "riverreachlength_ft" (the
length of river).

Additional fields follow the naming convention of [BMP
name]_LengthImplemented_ft.

Data Units (If
Applicable)

_FracImplement
fields as fractions,
_areaac fields as
acres

Length, in ft

-------
References

Ator, S. W. (2019J. Spatially referenced models ofstreamflow and nitrogen, phosphorus, and
suspended-sediment loads in streams of the Northeastern United States. U.S. Geological
Survey Scientific Investigations Report 2019-5118, 57 p. Retrieved from

Dell, C., Allen, A., Dostie, D., Meinen, R., & Maguire, R. (2016). Manure Incorporation and
Injection practices for Use in Phase 6.0 of the Chesapeake Bay Program Watershed Model.
Chesapeake Bay Program. Retrieved from

~ ~• eak r_ - T _ "-_r"_r

Detenbeck, N., ten Brink, M., Piscopo, A., Morrison, A., Stagnitta, T., Abele, R., et al. (2018).
Watershed Management Optimization Support Tool (WMOST) v3: Theoretical
Documentation. U.S. Environmental Protection Agency, Washington, DC EPA/600/R-
17/220. Retrieved from - ~ - - -m/w inner' ~n

Detenbeck, N., Piscopo, A., ten Brink, M., Weaver, C., Morrison, A., Stagnitta, T., et al. (2018).
Watershed Management Optimization Support Tool (WMOST) v3: User Guide. U.S.
Environmental Protection Agency, Washington, DC EPA/600/R-17/255. Retrieved from
~' i/wn -:' - -r - yrnen— ~ i

Heris, M. P., Foks, N. L., Bagstad, K. J., Troy, A., & Ancona, Z. H. (2020). A rasterized building
footprint dataset for the United States. Scientific Data, 7(207).

Hill, R. A., Weber, M. H., Leibowitz, S. G., Olsen, A. R., & Thornbrugh, D. J. (2015).
The Stream-Catchment (StreamCat) Dataset: A Database of Watershed Metrics for the
Conterminous United States. Journal of the American Water Resources Association (JAWRA),
52,120-128.

Houle, J., Riley, C., & Leonard, D. (2019). Pollutant Removal Credits for Buffer Restoration in
MS4 Permits: Final Panel Report. National Estuarine Research Reserve System Science
Collaborative. Retrieved from

Jin, S., Homer, C. G., Yang, L., Danielson, P., Dewitz, J., Li, C., et al. (2019). Overall
methodology design for the United States National Land Cover Database 2016 products.
Remote Sensing, ii (24).

JJ Environmental. (2015). Low Cost Retrofits for Nitrogen Removal at Wastewater Treatment
Plants in the Upper Long Island Sound Watershed. N-2012-047. Retrieved from

. . . itwpet.org/ttip-content/uploc ' " "18/ — .V. -r •port.pdf

McKay, L., Bondelid, L., Dewald, T., Johnston, J., Moore, R., & Rea, A. (2012). NHDPlus Version
2: User Guide. U.S. Environmental Protection Agency, Washington, DC. Retrieved from

; • " _" v " "r' " re# ' "_v J' "V

National Atmospheric Deposition Program. (2021). Total Deposition Maps. v2018.01.
Retrieved from p/tdepmaps/

New Hampshire Department of Environmental Services. (2020a). Authorization to
Discharge Under the National Pollutant Discharge Elimination System; NPDES Permit
No. NH0100200. New Hampshire Department of Environmental Services. Retrieved from

-------
New Hampshire Department of Environmental Services. (2020b). General Permits for
Stormwater Discharges from Small Municipal Storm Sewer Systems in New Hampshire. New
Hampshire Department of Environmental Services. Retrieved from
" "V :ionl/npdf" :"'rmw:

Schueler, T. R. (1987). Controlling Urban Runoff: A Practical Manual for Planning and
Designing Urban BMPs. The Metropolitan Washington Council of Governments, Washington
Metropolitan Water Resources Planning Board.

Soil Survey Staff. (2020a). The Gridded National Soil Survey Geographic (gNATSGO)
Database for New Hampshire. United States Department of Agriculture, Natural resources
Conservation Service. Retrieved from ; * ' "" "''P.box.coi _ : ~ils

Soil Survey Staff. (2020b). The Gridded National Soil Survey Geographic (gNATSGO)
Database for Vermont. United States Department of Agriculture, Natural resources
Conservation Service. Retrieved from , , - , : _ - :

U.S. Department of Agriculture, Natural Resources Conservation Service, (n.d.). Hydrologic
Soil Group. Retrieved from

http- :tives .v. ~~ov.usda.eov/QpenNonWebContent,aspx?content=22526.wba

U.S. Department of Agriculture Staff. (2021). Environmental Quality Incentives Program
Fiscal Years 2018-2021. United States Department of Agriculture. Retrieved from

U.S. Geological Survey, & U.S. Department of Agriculture, Natural Resources Conservation
Service. (2013). Federal Standards and Procedures for the National Watershed Boundary
Dataset (WBD) (4 ed.). Retrieved from

U.S. Geological Survey, National Geospatial Program. (2020). USGS National Hydrography
Dataset Best Resolution (NHD) for Hydrologic Unit (NH) 4 - 0410: Metadata. U.S. Geological
Survey. Retrieved from

lr~' \ /./_ — :v r\30br" 8754

University of Wisconsin in Madison. (2021). NEOS Server: State-of-the-Art Solvers for
Numerical Optimization. Wisconsin Institute for Discovery. Retrieved from I
server.ong/neos/

UVM Spatial Analysis Lab. (2019). Vermont High-Resolution Land Cover: Final Report. State
of Vermont. Retrieved from, iv/pagc

c" _ "r" locumei _ " n

Verdin, K. L. (2017). Hydrologic Derivatives for Modeling and Analysis - A new global high-
resolution database. U.S. Geological Survey Data Series 1053,16 p. Retrieved from
http" ~ " ". ""

Vermont Agency of Natural Resources. (2017). 2017 Vermont Stormwater Management
Manual Rule and Design Guidance. State of Vermont. Retrieved from

-------
Voorhees, M. (2016). Opti-Tool: Stormwater Nutrient Management Optimization Tool. EPA.
Retrieved from

Walker, J. T., Bell, M. D., Schwede, D., Cole, A., Beachley, G., Lear, G., & Wu, Z. (2019). Aspects
of uncertainty in total reactive nitrogen deposition estimates for North American critical
load applications. Science of the Total Environment, 690,1005-1018.

httds * / /doj opp/10 1016/i scitotenv 201. Q 06 3'^"7

White, M., DiLuzio, M., Gambone, M., Smith, D., McLellan, E., Bieger, K., et al. (2019).
Development of Agricultural Conservation Reduction Estimator (ACRE), a simple field-scale
conservation planning and evaluation tool. Journal of Soil and Water Conservation, 74(6).

Wickham, J., Stehman, S. V., Gass, L., Dewitz, J. A., Sorenson, D. G., Granneman, B. J., et al.
(2017). Thematic accuracy assessment of the 2011 National Land Cover Database (NLCD).
Remote Sensing of the Environment, 191, 328-341.

Yang, L., Jin, S., Danielson, P., Homer, C., Gass, L., Bender, S., et al. (2018). A new generation
of the United States National Land Cover Database: Requirements, research priorities,
design, and implementation strategies. ISPRSJournal ofPhotogrammetry and Remote
Sensing, 146,108-123. ^

-------