1

Gsflow-Arcpy - Python Toolkit for GSFLOW Model Development

Introduction

This manuscript describes a toolkit of ArcGIS Python codes, known as Gsflow-Arcpy, that can
be used to develop input data required for the U.S. Geological Survey (USGS) coupled
Groundwater and Surface-Water Flow (GSFLOW) numerical, hydrologic simulation code,
which is an integration of the USGS Precipitation-Runoff Modeling System (PRMS) and
Modular Groundwater-Flow (MODFLOW) simulation codes (Markstrom et al., 2008;
Markstrom et al., 2015; Niswonger et al., 2011). This toolkit does not produce all the input data
required for GSFLOW; however, it produces all the data requirements for constant-area square
HRU PRMS models, along with several input datasets required for MODFLOW simulations.
Although PRMS has several modules for distributing precipitation, temperature, and solar
radiation, and for simulating evapotranspiration, runoff, and streamflow, Gsflow-Arcpy
generates input datasets for PRMS modules precip_1sta, precip_laps, temp_1sta, temp_laps,
ddsolrad_hru_prms, potet_jh, srunoff_smidx, and strmflow, respectively. Datasets required for
alternative PRMS modules are not supported in the first release of the toolkit.

Gsflow-Arcpy supports development of PRMS models, and the MODFLOW model top for
constant area square MODFLOW cells for the Discretization (DIS) Package, the model exterior
boundary array for the Basic (BAS) Package, and the stream network input for the Streamflow-
Routing (SFR2) Package (Niswonger et al., 2005; Niswonger et al., 2011). Gsflow-Arcpy
developes GSFLOW models assuming MODFLOW grid cells and PRMS HRUs are equivalent.
The toolkit works in tandem with the Cascade Routing Tool (CRT) to develop the input required
for the PRMS Cascade Routing Module (Henson et al., 2013). The toolkit consists of a series of
python scripts that, when run in succession will generate model ready PRMS Parameter file(s),
and a parameter shapefile that can be used to visualize results throughout the process and extract
the required MODFLOW datasets. The PRMS Data File and Control File are not created by the
toolkit and must be created separately, but templates are provided with the toolkit, along with
example models.

The toolkit was developed to provide automation and reproducibility for creating GSFLOW
models within the CONUS. The toolkit could be used to develop models outside the CONUS;
however, required data input formats were developed using products developed for the CONUS,
and data products developed outside the CONUS may have different formats that require
modification. Some steps of the development process, especially steps required to create the DIS
Package, SFR2 Package, and Cascade Routing Module are difficult to fully automate in some
cases. It may be necessary for the user to intervene within the automated procedure, and to
manually modify the elevation of grid cells within the ArcGIS environment. Additionally, it is
recommended that the flow lines for the stream network provided by the NHD dataset be used as
a guide for creating the SFR2 Package stream network. NHD stream network is not used directly
to create SFR2 input due to inconsistencies between the resampled DEM used for creating model
grid and the NHD stream network. Where stream lines generated by the toolkit deviate
significantly from the NHD streams, the user may need to modify the altitude of the model top
data to create drainage directions that follow the NHD streams. Thus, especially in complex
 2

terrain, developing a stream network and cascades may be an iterative process that includes: 1)
filling closed depressions in the model top array using the CRT program; 2) generating the
stream network and comparing to the NHD stream network; and 3) making local adjustments to
the elevation of individual grid cells, and re-generating the stream network. This process may
need to be repeated 2-3 times to generate a consistent model top, SFR2 stream network, and set
of cascade parameters. However, for many applications these manual interventions are not
required, and model input generation is fully automated.

Gsflow-Arcpy consists of a set of Python scripts that use tools supplied by ArcMap and
associated add-in features, including ArcHydro. However, the toolkit scripts should be run while
the ArcMap software is not running. If ArcMap is running in the background, the scripts may not
run correctly, as ArcMap locks files that must be edited by toolkit. Thus, it is recommended that
the toolkit be run to generate datasets, followed by loading the parameter shapefile into ArcMap.
Datasets contained within the parameter shapefile can be displayed within ArcMap to evaluate
progress and accuracy, as well as to make small modifications to HRU elevations to improve the
generated stream network.

Previous work

Watershed modeling encompasses a vast number of processes and specifications, variable in

time and space, and development of hydrologic models can be arduous. Despite the abilities of a
process-based model, some may resist the technique due to practical developmental matters
(Fatichi et al., 2016). Both research and management can benefit from hydrologic models, but
their development and application is a complex process. A wider acceptance and distribution of
process-based approaches will require improved model visualization tools and a streamlined
approach for model setup, execution, and analysis (Fatichi et al, 2016). Integrated Hydrologic
Models (IHMs) can provide important information about water resources and are often used as
decision support tools for resource management (Laniak et al., 2013). Construction of IHM
inputs is not well formalized or automated, making reproducibility very difficult. Various
techniques have been developed to create drainage networks, but none have the benefit of a fully
automated and topologically consistent approach (Turcotte et al., 2001). The progression of
IHMs has spurred research and development for improving data processing and graphical user
interfaces (Band, 1986; Gruber and Peckham, 2009; Metz et al., 2011; Wilson, 2012; Tian et al.,
2016).

Advancements in GIS technology and remote sensing have provided the ability to highly
parameterize systems of interest, but not without complex data processing and utilization. For
example, PRMS requires upwards of 90 input parameters. Creating input files requires complex
GIS processing of raster and vector datasets from various sources. For example, developing a
stream network employs several ArcHydro tools, while importation of soil and vegetation data
calls for high level raster processing and time-consuming GIS reformatting. Research has only
recently begun to focus on GIS processing paired with model input construction. The GIS
Weasel, described by Viger and Leavesley in 2007, provided basic steps for developing input
parameters for a watershed model, but required extensive data management and lengthy GIS
processing. Alongside the GIS Weasel, PRMS had commonly been run through the USGS
modular modeling system (MMS) that provided a platform upon which certain modules could be
 3

coupled to simulate a variety of processes (Leavesley et al, 1996). The GIS Weasel explains the
tools to delineate, characterize, and parameterize features for distributed modeling. The MMS
(no longer in existence) applied a model builder GUI that integrated process modules stored in a
library. These primitive tools were valuable for combining the multiple components that
contribute to the dynamics of a watershed, but required a high level of process knowledge and
serious time investment on the front end of a modeling project. Together these platforms
required the user to weave through a seemingly endless series of steps, tools, and phases.

Here we present a series of python scripts that provide a formalized technique for the model
development of the IHM GSFLOW. This python toolkit, called Gsflow-Arcpy, automates many
of the necessary but laborious processes of parameterization, including stream network
development, land coverage and meteorological distribution, and Parameter File construction.
Stream network development based on digital elevation models (DEMs) are needed in hydrology
to determine the paths of water, sediment, and contamination movement (Tarboton, 1997).
Similar techniques to those developed in older models such as TOPMODEL (Beven and Kirkby,
1979) and HYDROTEL (Fortin et al., 1995, 2001) that utilize a direction matrix and a compute
topographic index are applied automatically in this toolkit through the use of the integrated
ArcHydro tools (Maidment, 2002). The Cascade Routing Tool (CRT) is utilized by Gsflow-
Arcpy to provide cascade parameters for the streamflow routing packages (SFR, SFR2) used by
GSFLOW.

Program and computing requirements

• ArcGIS 10.1 or newer (will include Python 2.7)

• Cascade Routing Tool (CRT) executable (version 1.3.1) http://water.usgs.gov/ogw/CRT/
• GSFLOW 1.2.2 http://water.usgs.gov/ogw/gsflow/#downloads
• 4GB or more of RAM, depending on the size of the model.
• USDA Soil Data Viewer and Microsoft Access
http://www.nrcs.usda.gov/wps/portal/nrcs/detail/soils/survey/geo/?cid=nrcs142p2_05361
4
• A GitHub repository https://github.com/gsflow gives users the ability to download the
latest versions of the python scripts and templates.

Geospatial and other ancillary data requirements

Input data required for GSFLOW models are derived from diverse geospatial data that are
available for the continental United States and other places across the globe. These datasets are
typically available through centralized data portals and can be downloaded freely to local
computers for selected basins of interest, or by county. Gsflow-Arcpy does not access these
geospatial datasets through web portals, so they must be stored locally on the same computer as
 4

the toolkit. The toolkit develops the GSFLOW inputs by making calculations for PRMS
parameters and MODFLOW datasets according to methods described by Viger and Leavesley
(2007), Markstrom et al. (2008), Markstrom et al. (2015), Harbaugh (2005), Niswonger et al.
(2011), and Henson et al. (2013), Regan et al. (2017). Almost all the data required by Gsflow-
Arcpy can be downloaded for basins in the United States using the USDA Geospatial Data
Gateway https://gdg.sc.egov.usda.gov/. Each particular dataset is explained individually below.
In some cases, users may want to apply local datasets which may come from various sources.
Datasets of this nature can be applied in the same manner as those obtained from USDA or
USGS.

Elevation

Elevation data is typically in the form of 10m (1/3 arc-second) or 30m (1 arc-second) National
Elevation Dataset (NED) rasters, which can be downloaded through Geospatial Data Gateway, or
through the USGS National Map launch site (http://viewer.nationalmap.gov/launch/) under
‘Download GIS Data’. The user will find several choices for elevation products, alongside a map
to visualize and select tiles of these data. A file transfer site, also under the ‘GIS Data’ section of
the launch page, provides access to elevation data but without the clarity of the first option. If
multiple digital elevation models (DEMs) are downloaded to cover the study area, they will need
to be combined into a single DEM for Gsflow-Arcpy processing (we recommend the mosaic tool
in ArcGIS). The elevation datasets will be resampled to the grid size resolution of the model
based on user preferred statistics within the area of each model grid cell. LIDAR can be used if
available. DEMs are generally available for the United States and across the globe, however, the
availability and the resolution of these DEMs varies with location.

Study Area

The study area is a land polygon that defines the portion of an area of interest where hydrologic
calculations will be made by GSFLOW. The boundary polygon will dictate the spatial projection
of the model grid and the flow direction for each cell within the polygon. GSFLOW study areas
may or may not correspond to the watershed boundaries, but all flow will be directed to the given
model outlet points within the study area boundary. When the study area is the same or nearly
the same as the watershed boundaries, the study area boundary can be downloaded from the
Geospatial Data Gateway. Watershed boundaries can be retrieved by selecting the “Hydrologic
Units” option within the Geospatial Data Gateway website. Sub-basins contained within the
study area can be developed from point shapefiles of downstream exit points, or pour points, in
ArcGIS. If the study area is not available as existing watershed boundaries, ArcGIS tools can be
applied to develop a polygon shapefile of the model domain using a DEM, or a polygon
shapefile can be created manually. Any land polygon dataset that defines the study area
boundary can be used with the toolkit.

Vegetation

Spatially varying vegetation type and coverage must be defined within the study area for
GSFLOW models. These data are provided by the LANDFIRE dataset, which can be
downloaded as tiles or for the CONUS (http://www.landfire.gov/vegetation.php). Remap files
 5

are used to categorize the diverse sets of vegetation defined in the LANDFIRE vegetation
coverage datasets to a subset of vegetation types that are supported by PRMS (Viger and
Leavesley, 2007). Existing vegetation type (EVT) and existing vegetation cover (EVC) datasets
can be downloaded for a study area as a single tile or as multiple tiles that must be aggregated to
a single image before processing.

Soil

PRMS requires several parameters that are calculated from soil properties. Important soil
properties that are used for parameterizing PRMS include the available water capacity (AWC),
percent sand, silt, and clay, and saturated hydraulic conductivity (Ksat). These soil properties are
provided for the CONUS by the STATSGO or SSURGO datasets, and can be downloaded
through the USDA Geospatial Data Gateway. Typically, SSURGO data is used first, and if gaps
exist in these data, STATSGO can be used to fill the gaps. Shapefiles of these soil properties can
be extracted using the NRCS Soil Data Viewer tool and then converted to raster formats at the
same resolution as the DEM for the study area. The Soil Data Viewer tool can be downloaded
and installed as an ArcMap add-in:
http://www.nrcs.usda.gov/wps/portal/nrcs/detail/soils/survey/geo/?cid=nrcs142p2_053614

Remap files

Text files with lists of the LANDFIRE vegetation codes and corresponding PRMS codes are
necessary to re-assign LANDFIRE key codes to the vegetation characterization supported by
PRMS. PRMS datasets derived from vegetation data include coverage type, coverage density
(summer and winter), snow interception, rain interception (summer and winter), root depth, and
optionally soil type. Example remap files are provided with the toolkit, and may need to be
modified to include new or missing LANDFIRE vegetation types. Remap files will vary slightly
depending on the version of ArcGIS being used, for example, Arc 10.1 allows the remap files to
have strings after the values, but Arc 10.2 does not accept the strings (examples of both remap
formats are provided).

Geology

To realistically distribute the saturated hydraulic conductivity (Ksat) of the soil, surficial geology
maps can be applied. State maps are available for download here:
https://mrdata.usgs.gov/geology/state/
The rock type can be used to assign a multiplier to the Ksat values in the equations concerning
soil moisture, specifically regarding gravity drainage.

Impervious Cover

Impervious cover represents locations in the study area where the land surface is impervious to
water, including, for example, paved areas or impervious bedrock outcrops. Impervious cover is
used in PRMS to calculate runoff and infiltration. The impervious coverage data can be
downloaded from the National Land Cover Database (NLCD
 6

http://www.mrlc.gov/nlcd11_data.php) for 2011. This dataset is intersected with the HRU grid to
calculate the percent of impervious area per HRU.

Climate

The python toolkit produces the required PRMS parameters for use by the precip_1sta or
precip_laps and temp_1sta or temp_laps modules. These parameters are used to distribute
precipitation, and minimum and maximum temperature to all HRUs in the study area using daily
climate data provided within the PRMS Data File. PRMS calculates daily precipitation at each
HRU using daily values measured at one or more climate stations set in the PRMS Data File and
average monthly precipitation adjustment factors (PRMS parameters rain_adj and snow_adj).
The python toolkit calculates precipitation adjustment factors using the Precipitation Runoff on
Independent Slopes Model (PRISM) data (Daly and others, 1994). Typically, 30-year monthly
average precipitation at an 800m resolution is used, but longer periods of data from the PRISM
group are available for a cost. Temperature adjustment factors, based on HRU aspect, (PRMS
parameters tmin_adj and tmax_adj) used to calculate daily maximum and minimum temperature
at each HRU are calculated by the Gsflow-Arcpy. PRISM data can be downloaded from the
PRISM climate portal (http://www.prism.oregonstate.edu/normals/). Observed climate station
data can be downloaded from several sources, depending on the study area. Typical station data
for the Western U.S. are NRCS SNOTEL stations
(http://www.wcc.nrcs.usda.gov/snow/snow_map.html) and WRCC NOAA coop stations
(http://www.wrcc.dri.edu/climatedata/climsum/). Daily climate data across the CONUS can be
found via the NOAA daily summaries map
(https://gis.ncdc.noaa.gov/maps/ncei/summaries/daily). Local stations can be found to contain
good data as well. If two climate stations at different elevations are available, precip_laps and
temp_laps modules may be used, and PRMS will calculate daily lapse rates based on data at the
two stations. If only a single station is used, monthly temperature lapse rates must be provided by
the user, while precipitation distribution will be based off the relationship between station data
and PRISM data.

National Hydrography Dataset (NHD)

A stream network provided by the NHD should be used as a guide when developing a stream
network with Gsflow-Arcpy. The python toolkit uses the ArcHydro flow accumulation and flow
direction tools to develop a stream network that is consistent with the model top dataset. For this
approach, the user specifies minimum flow accumulation and stream length thresholds to create a
stream network at a resolution appropriate for the modeling study. The flow direction is
determined using the DEM resampled to the GSFLOW model grid. The DEM must have
continuous downward sloping paths along the streams, such that every DEM cell in the study
area is adjacent to a cell with lower altitude in 1 of the 8 adjacent cells, referred to as d-8 filling.
The Cascade Routing Tool (CRT) provides the capability to d-8 fill the DEM. Additional to the
NHD stream network, aerial imaging of the study area is also helpful for verifying the location of
stream lines in flat areas, or for canals that do not follow the natural drainage defined by
topography. Aerial images are available through ArcGIS or can be downloaded through the
Geospatial Data Gateway.
 7

Configuration File (.ini) and Field List

The Configuration File for the python toolkit is a text file (.ini) that provides site specific
parameters and folder specific information that is read by each python script. The location of this
file is considered the working directory, which will house the folders and files created by
Gsflow-Arcpy. The Configuration File holds input and output filenames and locations, spatial
projections, stream network development thresholds, mean monthly precipitation data, monthly
lapse rates, CRT settings, and many other project specific settings related to parameter
development and ArcMap operations. Several flags can be turned on or off in the Configuration
File, depending on preferences of the user and project specifics. Folder locations and names of
input and output rasters, shapefiles, and text files are designated throughout the Configuration
File, allowing the python scripts to call and read or operate on these files. The purpose of the
Configuration File is to eliminate changing the python scripts, requiring only changes to be made
in the Configuration File for each new project. More detail on the Configuration File is explained
throughout the tutorial section, and a template Configuration File is provided with the toolkit.

Another text file, field_list.ini, is located in the python script folder. This file contains a list of
the parameter names (or fields) that will make up the attribute table of the parameter shapefile.
Some of these field names differ from PRMS parameter names, but are typically very close to
the actual parameter names. This list does not need to be changed, but must be present, and is
provided with the toolkit. If the user wants to change the name of a field, they may, but it is not
necessary, and would generally require changing the name in certain scripts as well. A table is
provided with the toolkit that holds explanations of each field name in the parameter shapefile.

Python Scripts

The python scripts make up the bulk of the Gsflow-Arcpy toolkit. The scripts can be stored in a
folder separate from the working directory. Along with the scripts, a template Configuration File,
field list file, and two CSV files that hold dimension and parameter settings are provided. For
each new project, it is recommended that the user copies the Configuration File and the CSV
files to a separate directory, where the files will be edited specifically to each project. The
Configuration File, stored in the working directory, will guide the scripts to the correct folder
locations, and the CSV files will provide specifics for the PRMS Parameter Files. More detail is
provided in the tutorial section about both the Configuration File and the parameter and
dimension CSV files. As the scripts are run, folders and files will appear in the working
directory. The parameter shapefile holds the model attributes and can be examined in ArcMap
after each script to check progress. The following section gives a brief description of each python
script. More detailed explanations and functionalities of the scripts are provided in the appendix.

fishnet_generator.py

The fishnet generator uses the fishnet tool in Arc to build the model top dataset. This
model top grid is used for defining HRU elevations, areas, stream network, soil
 8

properties, MODFLOW layer bottom elevations, etc. This script reads cell size and study
area boundary settings from the Configuration File. The fishnet will adapt the spatial
projection of the boundary shapefile. The fishnet script is needed if no fishnet shapefile
exists, therefore it is recommended but not mandatory. In some cases, the user may want
to use a pre-existing grid, or a section of an existing grid. In these cases, the script does
not need to be run, and the pre-existing grid will determine the projection and cell size.
To snap to a pre-existing grid, the user must input X and Y snap points in the
configuration file.

hru_parameters.py

The hru_parameters.py script constructs the initial fields for the attribute table in the
parameter shapefile, based on the field_list.ini file. The attribute table will display each of
these field names as column headers. These columns start out blank, and fill with data as
the scripts are run. This script also designates HRU_TYPE (0 = inactive, 1 = active, 2 =
lake, 3 = swale) based on the study area boundary and lake and sink shapefile, if lakes or
sinks are present. Swale is the PRMS term for sink, referring to a cell that does not have
a cascade outlet. Endorheic basins must contain a designated swale or CRT will attempt
to fill cells in order to route water out of the basin. The lake shapefile is typically
extracted from an NHD dataset, or can be manually created. If swales exist, or known
closed basins are being modeled, a shapefile must be provided with the sink points. The
parameter shapefile attribute table can be edited by the user.

dem_parameters.py

This script populates the elevation fields based on the original DEM (DEM_MEAN,
DEM_MIN, DEM_MAX, DEM_ADJ, DEM_FLOWAC, DEM_SUM, DEM_COUNT,
and HRU_ELEV). HRU_ELEV is altitude in feet. DEM_ADJ values represent the
surface elevation resampled to the model grid resolution, and are the elevation data that
the stream network and cascade parameters are developed from. They are eventually
written to a PRMS Parameter File as the hru_elev parameter. As these elevation-based
parameters are being developed, dem_parameters.py runs a flow accumulation function
to weight the elevation of streams heavier than the surrounding elevations, in order to
better represent stream altitudes at the model grid scale. If the original DEM is a smaller
resolution than the model grid, these values get written to DEM_FLOWAC. The default
setting in the Configuration File is to set these values as the original DEM_ADJ. The user
can change this setting to give DEM_ADJ values from any of the DEM parameters, but
the DEM_FLOWAC values are recommended, especially in complex terrain. This also
provides an original model scale DEM column in the parameter shapefile that can be used
as a fall back if mistakes are made while editing the DEM_ADJ values.

dem_2_streams.py

This script uses DEM_ADJ values to develop the gridded stream network. This is the
first step in the iterative process of stream network development. The script operates Arc
Hydro tools (flow accumulation and direction), based on thresholds set in the
 9

Configuration File, to build the initial stream network. The thresholds allow the user to
obtain a desired level of detail in the stream network (figures 10 & 11). Once again, flow
direction from all cells within the boundary shapefile will be forced to the given model
points (outlet, sub-basin, or swale). Common issues with the stream network can include
dead ends, misalignment relative to the NHD lines or aerial photos/basemaps, incorrectly
located confluences, or unnecessary parallelization of streams (typically in flatter area).
These issues can be fixed by either manually adjusting the elevation (DEM_ADJ), or by
running crt_fill_parameters.py and using the CRT fill values to adjust DEM_ADJ. A
combination of these methods can be used to route the streams correctly.

crt_fill_parameters.py

This script runs the Cascade Routing Tool (CRT) without streams and will identify
undeclared swales and honor declared swales. The script will write CRT fill values to the
fishnet (CRT_FILL), but won’t modify the DEM_ADJ values unless the fill flag is TRUE
in the Configuration File. This step can be left for the user to do manually, allowing for
more user intervention. In the case that the user wants to use some of the CRT fill values,
selected sections of the CRT_DEM column can be copied directly to the DEM_ADJ
column. It is recommended for to users evaluate the swale areas and make manual
adjustments, especially if the fill values are relatively large. After editing the DEM_ADJ
values, the dem_2_streams.py script must be re-run to re-build streams based on the
updated DEM_ADJ values, followed by re-running this script (crt_fill_parameters.py). A
few iterations may be necessary to eliminate swales and route all streams correctly. Since
the CRT does not fill cells that have a stream segment in them, running CRT without
streams first ensures that there are no swales along the streams and that there is a
continuous downward slope from cell to cell in the downstream direction.

stream_parameters.py

The stream parameter script runs the CRT with streams in order to generate the cascade
routing parameters for PRMS. The script also generates the streamflow routing (SFR)
input values for MODFLOW, including flow direction, reach (KRCH, IRCH, JRCH,
IREACH), reach length (RCHLEN, MAXRCH), segment (ISEG, IRUNBOUND,
OUTSEG, IUPSEG), and slope (STRM_SLOPE). These data are all added to the
parameter shapefile and can be seen in the attribute table. If any edits to DEM_ADJ take
place after this script has run, all three stream building scripts should be re-run, starting
with dem_2_streams.py.

veg_parameters.py

The vegetation parameters script generates all PRMS parameters derived from the
LANDFIRE vegetation dataset, including coverage type, winter and summer coverage
density, winter and summer interception (snow and rain), and root depth. These
parameters are written to the parameter shapefile. This script must be run before the soil
scripts in order to establish root depths.
 10

soil_raster_prep.py

The soil preparation script projects and clips the soil rasters to the model grid, and fills no
data cells in the input rasters in order for soil_parameters.py to run smoothly. It is
recommended that STATSGO data is used to cover any areas where SSURGO data is
unavailable, but in the case that no data areas exist, this script will use neighboring soil
types to assign values to the no data areas. This script must be run previous to
soil_parapeters.py.

soil_parameters.py

The soil parameter script uses the prepared soil rasters to generate percentages of sand,
silt, and clay, as well as available water content (AWC), saturated hydraulic conductivity
(Ksat), and soil type. Soil depth will also be derived if a raster is provided, but is not
required, as the greater of root depth and soil depth will be used for soil calculations. The
script writes these values to the parameter shapefile and uses them to calculate PRMS
soilzone parameters soil_moist_init, soil_moist_max, soil_rechr_init, soil_rechr_max,
ssr2gw_rate, slowcoef_lin, slowcoef_sq.

prism_800m_normals.py or prism_4km_normals.py

The PRISM parameter script determines a maximum and minimum temperature along
with a precipitation value for each model grid cell for each month. Values are typically
based on PRISM 30-year monthly normals (4km or 800m), but could be from another
gridded climate dataset if available, and would have to be formatted to match that of the
PRISM data. The script operates ArcMap’s zonal statistics tool to resample the PRISM
grid data to the model grid cells. This script must be run before the precipitation ratio
script (ppt_ratio_parameters.py) is run.

ppt_ratio_parameters.py

The precipitation ratio script calculates values assigned to each grid cell for each month
that acts as a multiplier based on the ratio between the PRISM gridded data and observed
monthly mean precipitation values from a designated climate station (or multiple
stations). These ratios (labeled as PPT_RT_01, PPT_RT_02, etc. in the parameter
shapefile) are the monthly rain and snow adjustment parameters (rain_adj, snow_adj) that
PRMS applies to each HRU to account for elevation differences throughout the model
domain.

impervious.py

The impervious script uses impervious cover data from the National Land Cover
Database to derive the percentage of impervious surface within each HRU grid cell. The
script writes the impervious percentage values to the IMPERV_PCT column in the
parameter shapefile, which provides the values for the hru_percent_imperv parameter in
PRMS.
 11

prms_template_fill.py

The final script in the toolkit, template fill, writes the complete PRMS Parameter Files.
Values written to the Parameter Files are directly read from the parameter shapefile that
has been developed through the Gsflow-Arcpy toolkit. The PRMS dimensions and
parameters CSV files are also read by this script. The dimensions CSV file holds basic
settings, several of which the user can adjust based on the project, such as number of
climate stations (ntemp), for example. The parameter CSV holds the name of all
parameters and designates which Parameter File they will be written to. Templates of
these files are provided with the toolkit, and further explanation is provided in the tutorial
section.

Notes on scripts

Scripts are generally run in the order that they are listed above; however, the scripts may be run
in slightly different sequences. When the scripts are being used solely for generating a stream
network for a MODFLOW model, for instance, the user may not need to prepare any files or run
scripts related to vegetation or soil. Or, if the user is interested only in developing precipitation
ratios, the PRISM scripts (prism_parameters.py and ppt_ratio_parameters.py) can be run before a
stream network or land use parameters are developed. The fishnet_generator.py script must be
run first for applications (assuming no pre-existing grid), followed by the hru_parameters.py and
then dem_parameter.py scripts, as these define the model top HRU grid, parameter fields, and
elevations.

Another python file, support_functions.py, exists within the toolkit. This script contains
functions that are used by multiple scripts. This file must be present in the directory with the rest
of the python files, but never needs to be explicitly run.

Tutorial Section
Example watershed – Russian River, Mendocino and Sonoma Counties, California, USA

This tutorial will guide the user through the process of applying the Gsflow-Arcpy toolkit to help
develop a GSFLOW model of the Russian River basin, located in northern California, USA
(figure 1). The user is encouraged to replicate the example model, or follow along using another
area of interest. The Russian River basin was chosen to showcase the Gsflow-Arcpy capabilities
within a large, dynamic watershed, but smaller examples are available with the toolkit as well.
These small examples are tailored to specific model scenarios, such as a closed basin or a basin
with a lake. These examples will run quickly and the user can compare results easily. The toolkit
requires some amount of input file development in ArcGIS, which is briefly reviewed but not
explained in any great detail, keeping the focus on the python tools. Plenty of outside resources
exist for ArcMap operations.
 12

In the case where only a stream network is desired, the user can develop the necessary fishnet
parameters and build the stream network without running through the entire toolkit. These steps
are explained in section A. Section B explains the remaining scripts that allow for complete
development of PRMS Parameter Files, along with several MODFLOW input parameters. The
tutorial describes the process of developing the model top grid (HRUs), grid cell altitudes, stream
network, land use and coverage, soil characteristics, and climate data. When complete, the user
will have a parameter shapefile containing the data necessary to create PRMS and MODFLOW
input files that can be then applied as a GSFLOW model. Gsflow-Arcpy saves the user an
incredible amount of time developing these data, allowing for more time to be spent on more
significant tasks, such as model calibration, application, and analysis.

Figure 1: Russian River watershed, geographic location and hydrographic border

A. Develop fishnet grid, DEM parameters, and build stream network

I. Fishnet grid development using fishnet_generator.py

Step 1. Download a digital elevation model (DEM): In this tutorial, the National
Elevation Dataset (NED) 10m (1/3 ac-second) is being used. Elevation data was
 13

downloaded in an IMG format through the USGS National Map Viewer launch site
(https://viewer.nationalmap.gov/launch/) in 4 tiles that were aggregated in ArcMap using
the mosaic tool to become one single raster image (figure 2). The four 1x1 degree IMG
files that cover the Russian River basin are:
USGS NED n39w123 1/3 arc-second 2013
USGS NED n39w124 1/3 arc-second 2013
USGS NED n40w123 1/3 arc-second 2013
USGS NED n40w124 1/3 arc-second 2013
Download tiles to cover the area of interest, import into ArcGIS and use the mosaic tool
to patch the tiles together, if necessary. This process generates a single DEM covering the
study area. A study area boundary is not needed for this step, and it is recommended to
download tiles to cover the entire area of interest and beyond to ensure that the entire
study area is completely encompassed within the DEM. Save the DEM image file and set
the location of the file to ‘dem_orig_path’ in the Configuration File in the ‘DEM
Parameters’ section. The DEM file for this tutorial is named ‘ned10m_nad83z10.img’.

Step 2. Download study area boundary: There are several ways to do this, depending
on the study area. Some watersheds may need to be extracted from larger datasets, while
other sources provide datasets for individual watersheds. The Russian River watershed
boundaries were downloaded from the Geospatial Data Gateway
(https://gdg.sc.egov.usda.gov/GDGOrder.aspx), which contains country wide datasets
that can be ordered by county. Boundary datasets can be found under ‘Hydrologic Units’
in the Geospatial Data Gateway, after the county or counties are entered. The datasets
named “8 digit watershed boundary” (NRCS version), for both counties within the
Russian River watershed (Mendocino and Sonoma), contained boundaries for the study
area. The datasets named “10 and 12 digit” contained sub-basin scale watersheds, and
were not used for this tutorial, but would be acceptable for models focused on local, sub-
basin hydrology. The watershed boundary polygon encompassing the Russian River
basin was selected and extracted from the Mendocino County dataset in ArcMap (figure
2). The shapefile file path and file name were set in the Configuration File (figure 3).

Download watershed boundary file(s), and export the study area boundary to an
individual shapefile in ArcMap. Enter the location of this file in the Configuration File
under the ‘Study Area’ section (study_area_path). If no boundary is available through
online resources, a watershed boundary can be created in ArcMap using the DEM and an
outlet (pour) point. ArcMap applies the flow direction and accumulation tools in the
Hydrology toolbox, under Spatial Analyst Tools. A watershed boundary gets created
above the designated outlet point, which can be manually created, or can be a location of
an actual streamgage. More information on this process can be found through ESRI help
online. http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/how-
watershed-works.htm). Be aware that the spatial projection of the boundary shapefile
will dictate the projection of the model fishnet and therefore all resulting rasters and data.
The boundary will also control all flow directions, meaning that if the boundary is drawn
outside the watershed, the elevations of those cells will be raised so that flow can be
directed to the model outlet point.
 14

Figure 2: The combined DEM and Russian River watershed boundary (8-digit),
visualized in ArcMap. These two datasets are the building blocks for Gsflow-Arcpy.

Step 3: Setup Configuration File: The Configuration File contains several settings that
apply to the scripts, and also guides each script to specific files it needs to read. At the
very least, the section related to the script being currently run must be filled out prior to
running that script. For the fishnet section, set the desired grid cell and buffer cell size,
and designate the folder path of the DEM and study area boundary. Buffer cells size
designates how many layers of cells will be constructed outside of the model boundary. It
is recommended to use at least 2 buffer cells. Smaller cell sizes cause longer runtimes for
the scripts, so a balance must be found between model discretization and efficiency. This
Russian River model was created with a 300 m x 300 m cell size. For the DEM to be re-
sampled efficiently to the fishnet grid size, it is recommended that the cell size be set as a
multiple of the DEM resolution. The boundary shapefile will provide a spatial projection
and location for the fishnet. This will ensure spatial congruency for following operations.
If a part of or entire pre-existing grid is to be used, the user may include a reference X-Y
point in the Configuration File so the grid will snap to a specific location. The
fishnet_generator.py script is not necessary in this case, and the spatial projection will be
based off the grid provided. If no reference point is given, the lower left corner of the
study area, plus buffer cells, will be used. Be sure to enter the folder and file names that
will be given to the parameter shapefiles. The hru_fishnet_path will be the feature class
 15

shapefile that holds all model attributes, while the hru_centroid_path will be the point
shapefile that holds centroid locations of each HRU for zonal statistics operations. An
example section of the Configuration File and the settings used for the Russian River can
be seen in figure 3. Several sections of the Configuration File will be shown throughout
this tutorial, and an example template is provided with the toolkit.

Figure 3: A portion of the Configuration File that directs fishnet characteristics and
provides the study area path.

To run any script in the python toolkit, the DOS command prompt is typically used. In
the command prompt, the directory needs to be changed to the folder containing the
Configuration File (the working directory). The folder containing the scripts must then be
entered after the working directory, along with the name of the script being run, followed
by a ‘-i’ and the Configuration File name (figure 4). A ‘-d’ typed after the Configuration
File name will run the script in debug mode which provides more detail on the activity as
the script runs.
 16

Figure 4: Command prompt for running the python scripts. Note the folder path for the
script is followed by a ‘-i’ and then the Configuration Filename, as seen on the line next
to the blue arrow. A ‘-d’ can be added at the end of the command to run the script in
debug mode in order to see more details on the script operation.

Step 4: Run fishnet_generator.py: One this script has run, the fishnet grid will appear
in the folder designated in the Configuration File, and the shapefile can be opened in
ArcMap (named ‘hru_params.shp’ in this tutorial). The initial shapefile will contain only
three columns at this point: FID, SHAPE, and ORIG_FID. This fishnet shapefile is the
model top, each grid cell representing a PRMS hydrologic response unit (HRU). The
attribute table will build upon itself as each script is run, holding parameter values for
each cell.

After running fishnet_generator.py, open the newly created shapefiles in ArcMap, along
with the boundary shapefile, and observe the results of the script (figure 5). Check that
buffer cells sufficiently provide extra space outside the watershed border, so that no
active cell is the outermost cell of the fishnet. This will help when developing the stream
network. Open the point shapefile (named ‘hru_params_label.shp’ for this tutorial) and
check that each point represents the centroid of one cell. This file is used for several steps
throughout the process regarding zonal statistic operations, such as re-projecting the
DEM to the fishnet. Close the parameter shapefiles in Arc before running the next script.
(Note – closing the shapefiles doesn’t guarantee the next script will be able to acquire a
lock. It is recommended to either close ArcMap completely, or open a new blank map in
Arc where the shapefile can be re-opened once the next script is done running.)
 17

Figure 5: Fishnet shapefile open in ArcMap overlain by the Russian River watershed
boundary polygon shapefile. Gsflow-Arcpy writes input data for PRMS and
MODFLOW to the attribute table of the fishnet shapefile for each HRU grid cell.

II. Creation of HRU attribute table using hru_parameters.py

Step 1: Run hru_parameters.py: This script builds the attribute table for the HRU
fishnet with field names for each parameter, as specified in the field_list.ini file provided
with the toolkit. The script also designates HRU_TYPE: 0=inactive, 1=active, 2=lake,
and 3 = swale. If lakes exist in the study area, a shapefile containing the lake polygons
will be needed, with a filename and path designated in the Configuration File for
lake_path. Water body data can be obtained from the National Hydrography Dataset
(NHD) through the Geospatial Data Gateway. For the Russian River model, two major
lakes were selected from the NHD and extracted, in ArcMap, to create a shapefile
containing just the water bodies significant to the model. This file (rr_lakes.shp) is
specified in the Configuration File as the lake_path. If no lakes are being modeled, set
the set_lake_flag to false. To designate model outlet, sub-basin, and sink points, a point
shapefile must be provided that contains a field with the point type designations
OUTLET, SUBBASIN, or SWALE (figure 6).

All cells with an area >50% outside the study area boundary are declared inactive, and all
cells with a greater area within the boundary are designated as active, or lake if more than
half the cell area resides inside a lake boundary. This lake designation setting can be
changed in the Configuration File (lake_area_pct), to account for a narrow section of
lake.
 18

Figure 6: Configuration File clip regarding model points and lakes shapefiles. The field
that contains the point type must be designated and sub-basin numbering will be based on
the values in the field provided for model_points_zone_field.

After running hru_parameters.py, check the attribute table in the parameter shapefile to
ensure the fields have been created and values have been generated for HRU_TYPE. Use
the Categories display option to view HRU_TYPE and make sure the lakes and swales (if
any) are in the correct location (figure 7).

Figure 7: HRU fishnet symbolized with HRU_TYPE = 0 for inactive cells (blank),
HRU_TYPE = 1 for active cells (green), and HRU_TYPE = 2 for lakes (blue). No sub-
basins within the Russian River watershed are endorheic, therefore no swale points were
provided. If so, these would appear in a different color and be HRU_TYPE = 3.
 19

III. Write DEM-based parameters to attribute table using dem_parameters.py

Step 1: Designate DEM settings and file paths in Configuration File: The original
DEM file path and associated settings need to be designated in the Configuration File,
including dem_orig_path and dem_cellsize (figure 8).

Figure 8: Clip of the ‘DEM parameters’ section of the Configuration File.

As can be seen in figure 8, dem_adj_copy_field (line 55 in figure 8) is set to

DEM_FLOWAC. As explained in the introduction, this setting designates which
elevation based field will set the initial values for DEM_ADJ. Of all the elevation
parameters, DEM_FLOWAC is the preferred option for most cases, especially in steep
topography, as it provides better altitude values along streams during the resampling of
the original scale DEM. However, the user may choose DEM_MEAN, for example, to
provide initial values for DEM_ADJ.

Along with the elevation parameters, dem_parameters.py also calculates the aspect,
slope, and relative temperature adjustment for each grid cell. The script reads aspect and
temperature remap text files, the file paths of which are set under the ‘Remap files’
section of the Configuration File (figure 9). The temperature adjust remap file used for
this example is “temp_adj_x10.rmp”. The format of the remap file is for ArcMap 10.2
and 10.3. The remap values are multiplied by 10 in order to convert the floating point
values to integers because the latest versions of ArcMap require integers when
calculating the aspect and temperature adjustment. The script divides the values by 10
internally before generating the results. Remap templates are provided with the toolkit.
The remap values should be checked by the user prior to running the script, and may be
changed based on the study area.
 20

Figure 9: Remap settings in the Configuration File. The aspect and temperature adjust
remap files are used by dem_parameters.py, while the rest are used in later scripts.

Step 2: Run dem_parameters.py: This script will re-sample the original DEM to the
model grid cell size, and create all elevation, slope, and aspect related parameters. Once
the script has finished running, the parameter shapefile can be opened in ArcMap to
check and visualize the data that have been written to the attribute table (figure 10).
Because this script will overwrite any existing DEM values, the user will see a message
warning of this and will be prompted to press ENTER to continue running the script. This
is a safeguard to avoid an unintentional overwrite of manual adjustments. If the user
wishes to refresh the DEM and start anew, the script can easily be re-run, or the
dem_adj_copy_field can be copied into the DEM_ADJ column.

Figure 10: A) Original 10m DEM, clipped to the Russian River watershed, B) inset at the
original DEM scale, and C) inset of the 300m DEM re-sampled for the model grid.
 21

IV. Development of the stream network and cascade numbering using

dem_2_streams.py, crt_fill_parameters.py, and stream_parameters.py

Development of the stream network is typically an iterative process. Aerial imagery and
the NHD stream lines can be used to evaluate the accuracy of the stream network
generated by Gsflow-Arcpy. The NHD is not used directly because it is not consistent
with the DEM_ADJ data, and this causes unstable solutions for the SFR2 Package in
MODFLOW. Filling the DEM using 8-way fill in ArcMap and the CRT modifies the
DEM_ADJ dataset to provide 1) continuous slopes in the direction of streamflow for all
grid cells that contain stream reaches, and 2) continuous drainage pathways for all HRUs
within the study area domain, such that all cascade termination points end at a stream,
lake, or specified swale. The change in elevation applied to each cell during the fill
process should be evaluated by examining the fill rasters created in the ‘flow_rasters’
folder in the working directory. These are created by dem_2_streams.py.

Solely applying 8-way fill does not always result in a satisfactory DEM_ADJ dataset for
creating the stream network. In some cases, the altitude of cells in DEM_ADJ can be
better modified by hand. Adjustments are made by hand when the stream network is not
continuous or if stream lines do not follow closely enough the NHD stream lines or aerial
imagery. After changes (fills) to DEM_ADJ are adopted from CRT results or made by
hand, CRT should be run again to verify that no additional fill is required and to generate
the cascades. If the fill process does not work well or the user is unhappy with the stream
network, DEM_ADJ can be reset to the dem_adj_copy_field (DEM_FLOWAC in this
tutorial), which holds the original DEM_ADJ values. The user would then apply different
adjustments. Details on this process are described in the following steps.

Step 1: Run dem_2_streams.py: This script develops the initial stream network, based
on DEM_ADJ. It also builds PRMS sub-basins if any model points are designated as
SUBBASIN (figure 11). Sub-basins are defined using outlet points, or “pour points” as
they are referred to in the ArcMap literature. These points can be actual streamgage
locations or manually inserted points to designate the bottom (outlet) of a sub-basin. The
script uses the model points shapefile and the watershed tool in ArcMap to delineate sub-
basins. The script creates a stream network within the constraints of the flow length and
flow accumulation thresholds set by the user in the Configuration File, in the DEM 2
Streams section (figure 12). The flow accumulation and flow length thresholds can be
adjusted to attain a stream network with the desired level of detail (figure 13).

Figure 11: Example of a model points shapefile attribute table, created by the user to
designate the model outlet point, sub-basin points, and swales. The ZONE_VALUE
determines the basin numbering.
 22

Figure 12: Configuration File settings for the stream network. The flow_acc_threshold
(line 59) and flow_length_threshold (line 61) can be adjusted to construct the desired
level of detail of the stream network. The calc_flow_dir_points_flag (line 63) will build
flow direction arrows if set to True. This is a nice visualization tool for the user to
observe flow direction results.

Figure 13: Stream networks generated with different flow accumulation and flow length
threshold settings; A) Russian River watershed, and B) close up of a sub-basin showing
different stream resolutions. The darkest blue lines (streams_100_10) show a network
with a flow accumulation threshold of 100 and flow length threshold of 10. HRUs with
flow accumulation and flow lengths smaller than the specified threshold will not contain
streams. The dark and light blue lines show the levels of detail that can be obtained by
decreasing the flow accumulation and stream length thresholds (streams_50_5 and
 23

streams 30_3, respectively). These thresholds are site specific and depend on the
resolution and purpose of the model.

In the case of an endorheic basin that does not have any outflow, a cell within the basin
must be specified as a swale (HRU_TYPE = 3). This will have been assigned in
hru_parameters.py. This will create a sub-basin with no outlet, where local streams will
flow to the sink point. If a sink point is not designated, ArcMap and CRT fill algorithms
will attempt to fill the local depression until an outflow exists.

Step 2: Observe initial stream network and modify DEM_ADJ: The

dem_2_sreams.py script creates a stream shapefile (streams.shp) located in the
‘flow_rasters’ folder that gets created in the working directory. Import streams.shp into
the ArcMap workspace, overlaying it on the HRU grid and basemap imagery, along with
NHD streams for comparison. A hillshade map also can aid in evaluating the generated
stream network. Evaluate the headwater drainages and tributaries to assess the effect of
the flow accumulation and length thresholds on the level of detail in the stream network.
Several threshold re-adjustments may be necessary, re-running dem_2_streams.py each
time, to arrive at the desired stream detail. Zooming and panning throughout the model
domain will reveal areas in the network that may require changes to the DEM_ADJ in
order to clean up, smooth, or re-route sections of the network.

Image rasters that display the characteristics of the stream network are also created in the
‘flow_rasters’ folder with this script. Undeclared swales in the active model domain,
named ‘dem_sink.img’, shows all cells that ArcMap fill algorithms found to need filling
to create outflow. Other rasters created by this script include flow direction points, flow
accumulation, stream links, stream order, stream count, sub-basins, and watersheds (per
segment). The ‘dem_fill.img’ raster shows how much the DEM_ADJ values will have to
be raised in each cell. These values are written to the DEM_SINK column in the
parameter shapefile attribute table. In cases where fill values are relatively low (<5m), the
user may choose to use computed fill values, therefore continuing on to run
crt_fill_parameters.py and using the CRT_ELEV field to directly replace DEM_ADJ
values in the swale locations. (The ArcMap fill and the CRT fill values should be very
similar). However, in some cases, there will be areas within the model domain that can
benefit from manual adjustments before continuing. Adjusting DEM_ADJ during this
step should be done to fix issues with sinks or with the stream network before running the
following stream scripts. An example of a manual edit of DEM_ADJ to alter the stream
network can be seen in the following figures.
 24

Figure 14: A clip of the initial stream network (streams.shp), shown in light blue lines,
overlaying the HRU grid (thin black lines), NHD streams (thin yellow lines), and
basemap imagery. The teal outline of a single grid cell shows where the elevation
adjustment should take place to guide the stream back to its actual path.

It can be seen in figure 14 that the tributaries coming in from the top-right are not well
represented by the initial stream shapefile. It is clear they diverge from the stream paths
visible in the imagery and the NHD lines, and there is an unnecessary parallelization
where the tributary meets the main stem. In this case, the user may choose to manually
edit the DEM_ADJ values of some surrounding cell(s) in order to direct the streams more
realistically to the main channel (running from top-left to bottom right in the figure).
Users must be sure to avoid creating any swales while manually editing. Regarding the
tributary approaching from the left side of the main channel, the user must decide if an
adjustment here is necessary to follow the NHD stream path or not. This depends on each
individual model and purpose.

Figure 15 shows the HRU grid cells labeled with the DEM_ADJ altitudes, and the cells
with light green borders have been slightly adjusted (by hand) in order to re-route the
stream to better follow the true streams seen in the imagery and NHD lines, and to
eliminate the parallelization of steam segments seen in figure 14.
 25

Figure 15: DEM_ADJ values labeled in each cell, and selected cells that have been
slightly adjusted in order to re-route the stream to follow the actual stream path. The
amount of manual editing of the stream network is left up to the user.

Once edits have been made and the user would like to see the results, save and close the
project in ArcMap, and re-run dem_2_streams.py. Re-open the parameter and stream
shapefiles and observe the changes in the stream network and any remaining sinks. The
user may choose to edit several of the largest sink areas and perform some manual stream
routing to clean the network up, and then rely on the CRT_ELEV values (created in the
next step) to fill the rest of the swales throughout the model domain. Figure 16 shows the
result for this example section of stream network editing.
 26

Figure 16: The resulting flow paths after manual adjustments. Note the cleaner, more
realistic flow lines, resulting from a few small changes to select cells.

Step 3: Finalize stream network by running crt_fill_parameters.py: Once satisfied

with the stream network and the magnitudes of the sinks, close the ArcMap project and
run crt_fill_parameters.py. This step (and step 4) require the CRT executable to be
located by the scripts, which must be downloaded by the user (link provided in
introduction). Set a file path in the Configuration File under the “CRT parameters”
section (figure 17). A folder will be created named ‘fill_work’ in the working directory,
which will store the CRT input and output files, compete with swales and associated
HRU cell ids. The parameter shapefile attribute table will contain fields named
CRT_FILL and CRT_ELEV. CRT_FILL is the amount of elevation change needed to
fill a swale, and CRT_ELEV is the elevation of the cell with that fill amount applied.
Remember, in the case that a model domain contains actual swales that the user wants to
model, HRU_TYPE can be designated as a swale if set to a value of 3 (Henson et al.,
2013). This will keep the CRT fill from operating on that swale area. After observing the
potential CRT fill amounts, the user can choose here to step back and perform manual
edits to eliminate these swales by hand, or to simply copy the CRT_ELEV values to the
DEM_ADJ column where they differ, and re-run both dem_2_streams.py and
crt_fill_parameters.py. If the use_crt_fill_flag is True, the values from the CRT_ELEV
will automatically be copied to DEM_ADJ. These steps will result in a stream network
free of undeclared swales and ready for CRT to run with streams to build cascades (figure
18). Keep in mind that if any editing of the DEM_ADJ takes place, dem_2_stream.py
must be run again, followed by crt_fill_parameters.py, so that stream_parameters.py will
build cascades correctly.
 27

Figure 17: CRT settings in the Configuration File. The scripts run CRT internally, but
need to locate the executable. Folders named ‘fill_work’ and ‘cascade_work’ contain
CRT results that will be created in the working directory. The CRT parameters can be
adjusted for someone familiar with CRT, but the template Configuration File will be
preset to the recommended CRT settings. More information regarding the CRT settings
can be found in Henson et. al., 2013.

Figure 18: Stream network (blue lines) overlying NHD flowlines (purple), with the inset
displaying grid cells (gray lines).
 28

Step 4: Develop the PRMS and MODFLOW cascade parameters by running

stream_parameters.py: The stream_parameter.py script will run CRT with streams and
assign the final cascade numbering to each cell. This is the final script in the stream
building process, and will result in the final cascade parameters for each HRU. It is not
unusual iterate through these stream development scripts to eliminate all swales and
develop the final stream network. For large watersheds, this process may take place in
sections as the user moves through the watershed. For smaller models, only one or two
iterations may be needed. In some cases, a simple model may not require any manual
adjustments. In this case, the use_crt_fill_flag would be TRUE and the stream building
steps would interate a single time, only repeating dem_2_streams.py and
crt_fill_parameters.py once, after the initial sinks were filled. Results of
stream_parameters.py can be observed by checking HRU_CASC.DAT,
OUTFLOW_HRU.DAT, LAND_ELEV.DAT, XY.DAT and the ‘outputstat.txt’ file in
the ‘cascade_work’ folder, and visualizing parameters in ArcMap (ISEG, IRUNBOUND,
FLOW_DIR, KRCH, IRCH, JRCH, ISEG, REACH, RCHLEN, MAXREACH,
OUTSEG, IUPSEG, SUBBASIN, SEGBASIN, STRM_TOP, STRM_SLOPE). The
cascade parameters are used for input to the Streamflow Routing (SFR) package in
MODFLOW.

B. Prepare remap files, soil, vegetation, and climate parameters, and create the PRMS
Parameter File

I. Preparing remap files, vegetation data, and running veg_parameters.py

Step 1: Download vegetation type and cover: For the Russian River watershed,
existing vegetation type (EVT) and cover (EVC) from the LANDFIRE data distribution
site was used (http://landfire.cr.usgs.gov/viewer/). These datasets were stored to a local
folder and the file paths were set in the Configuration File (figure 19). Attribute data
dictionaries can also be downloaded (http://www.landfire.gov/
NationalProductDescriptions21.php for vegetation type and http://www.landfire.gov/
NationalProductDescriptions23.php for vegetation cover). The attribute data dictionaries
help update the vegetation remap files for the study area.

Figure 19: Vegetation type and cover sections of the Configuration File. Cell size is the
size at which the data will be re-sampled to the model grid, so a factor of the model grid
cell size is recommended to ensure best alignment.
 29

Step 2: Check/update remap files: It is recommended that the user scans through the
remap files provided with the toolkit to ensure that all vegetation coverage types,
densities, root depths, and interceptions are acceptable, referring to LANDFIRE
vegetation data files if there are questionable vegetation types assigned in the study area.
If the study area contains vegetation types that are not included in the remap files, they
will need to be added to the remap list. Save these remap files to a folder and set the file
paths in the Configuration File under ‘Remap files’.

Step 3: Generate PRMS vegetation parameters from veg_parameters.py: The

veg_parameters.py script will use the given LANDFIRE datasets and the associated
remap files to write vegetation codes acceptable to PRMS to the parameter shapefile
attribute table (figure 20). The cell size setting designates the resolution at which the data
will be re-projected, so a factor of the fishnet grid is the most acceptable for best
alignment (10m is default). Choosing a projection cell size that is not a multiple of the
fishnet cell size could cause unwanted misalignment of the dataset. The script will scan
the vegetation rasters to check that all values are in the remap files, and will alert the user
of missing vegetation codes, if any. The remapped files will be stored in a folder named
‘veg_rasters’ in the working directory that will appear once the script has run. To
evaluate these files, open the rasters and their attribute tables in ArcMap. Also check the
HRU parameter shapefile to ensure the vegetation related fields were populated for all
active cells. Vegetation fields include COV_TYPE, COVDEN_SUM, COVDEN_WIN,
RAD_TRNCF, SNOW_INTCP, SRAIN_INTC, and WRAIN_INTC.

A B

PRMS
Bare
LANDFIRE US120EVT Grasses
Shrubs
Deciduous
Conifer
veg_parameters.py

Figure 20: Vegetation type remapped from A) LANDFIRE existing vegetation type to
B) PRMS values.
 30

II. Generate PRMS soilzone parameters using Soil Survey Spatial and Tabular Data
(SSURGO), and/or the U.S. General Soil Map (STATSGO2).

Some preparation of the soil data is necessary for Gsflow-Arcpy. The required soil data
must be imported through the ArcMap Soil Data Viewer add-in, then formatted as raster
images. The reason the Soil Data Viewer is listed as a recommended software is because
the user could choose to simply use PRMS default values for soil parameters. This is not
recommended, but is an option for a quick model build. Detailed instructions for
installing and using the Soil Data Viewer are available on the USDA web page
(www.nrcs.usda.gov/wps/portal/nrcs/detail/soils/survey/geo/?cid=nrcseprd337066).
The following steps take the user through the process of downloading, importing,
mapping, and formatting soil data.

Step 1: Prepare soil data for ArcMap: Soil data can be downloaded through the
Geospatial Data Gateway. For the Russian River watershed, SSURGO data was used for
Mendocino and Sonoma counties, as it completely covered the study area. In cases where
SSURGO data does not completely cover the model domain, STATSGO data must be
downloaded and can be used in conjunction with the SSURGO data. The data for
Mendocino County downloaded in three separate folders, while Sonoma county data was
contained in a single folder. Zipped folders must be extracted and the Microsoft Access
files that come with the data must be opened in order to import the soil data needed for
PRMS. Upon opening the Microsoft Access file for each database, a window labeled
“Macro Single Step” may appear. Select ‘Stop All Macros’ and then Enable Content. The
user will then be prompted to enter the directory (file pathname) where the soil data is
stored. Paste the directory for the tabular data into the SSURGO Import window and
select OK to import the data (figure 21). Once imported, the data will be available for the
NRCS Soil Data Viewer Tool add-in in ArcMap.

Figure 21: Importing SSURGO data into Microsoft Access. Once hitting OK, the data
can take several minutes to import.
 31

Figure 22: A visualization of how soil data is transferred from A) the mosaic of the
SSURGO map unit keys for the Russian River watershed and resulting rasters of soil
properties; B) MOIST_MAX (PRMS parameter soil_moist_max); C) SSR2GW_RATE
(PRMS parameter ssr2gw_rate); and D) SLOWCOEF_L (PRMS parameter
slow_coef_lin).

Step 2: Load soil property maps into ArcMap: With the SSURGO/STATSGO soil
data shapefiles (located in the spatial folder) open in ArcMap, open the Soil Data Viewer
tool (an icon should appear on the ArcMap toolbar after importing the add-in). A box will
appear for the user to choose which soil dataset will be used, and then the main Soil Data
 32

Viewer window will open. Paste the Microsoft Access pathname of the matching file in
the Database box (figure 23). These two files (Map Layer and Database) should
synchronize, giving the green light to download specific attributes to ArcMap. Once the
database file path has been entered into the Soil Data Viewer, the user will see all related
attribute folders associated with the database. PRMS is interested in Available Water
Capacity (AWC), Percent Clay, Percent Sand, Percent Silt, and Saturated Hydraulic
Conductivity (Ksat), and soil depth. These parameters are in the Soil Physical Properties
folder when the ‘Advanced Mode’ option is selected, except soil depth, which is located
in the Soil Qualities and Features folder. Each parameter must be selected and rating
options chosen before mapping the data to ArcMap. Soil depth is optional since Gsflow-
Arcpy will apply the greatest value of root depth vs soil depth.

Figure 23: The Soil Data Viewer add-in open in ArcMap. Map Layer and Database file
pathnames at the bottom of the window are synchronized. The drop-down menu below
the Soil Physical Properties attribute contains the soil properties needed for PRMS.
Rating options must be selected prior to clicking the ‘Map’ button.

Regarding the rating options, the ‘Weighted Average’ aggregation method is typically
used, and ‘Higher’ is selected for the tie break rule (‘Fastest’ for Ksat). For the Layer
Option, a depth range can be determined or ‘All Layers’ may be chosen. These settings
do not seem to pose significant differences in the mapping of the soil properties relating
to PRMS, so the user may feel free to experiment here and perform outside research to
come up with the best options for a particular property. After filling out these options,
 33

press the ‘Map’ button to create a shapefile containing the selected soil property. This
step needs to be done for each map layer (soil shapefile) and each soil property.

Step 3: Create raster layers for each soil property: Shapefiles containing soil
properties may only cover partial sections of the model domain, especially if the study
area crosses county lines. Thus, the shapefiles must be joined to form one single shapefile
(per soil property) covering the entire study area. These shapefiles must then be
converted to raster images in order to be read correctly by Gsflow-Arcpy. The final result
should be raster image files covering the study area for each soil property. For more
information on these ArcGIS methods, refer to ESRI online resources:
http://resources.arcgis.com/en/help/.

A surficial geology raster can also be created, in order to apply a more realistic saturated
hydraulic conductivity (Ksat) to the soil gravity drainage calculations. This raster is
optional, and a default multiplier value can be set in the Configuration File if the user
does not choose to use a geology raster. The rock type from a surficial geology map is
used to apply a multiplier to the Ksat (acquired from the SSURGO data). It has been
observed that often these conductivity values are too large and cause parameters such as
ssr2gw_rate to be much too high initially. State geology maps that are GIS ready can be
downloaded at https://mrdata.usgs.gov/geology/state/. These provide general geologic
maps that are usually discrete enough for regional scale models, though other geologic or
hydrogeologic data may be applied for a finer or coarser resolution. The user must
convert the polygons covering the model domain to a raster, just like the soil data. A
VALUE field must be added to the attribute table, where a multiplier for each rock type
is added. This optional step can help designate recharge zones, for example, that may not
have been as well defined by the soil properties alone.

Step 4: Run soil_raster_prep.py and soil_parameters.py: Ensure that the correct file-
names for the soil rasters have been entered in the Configuration File under the ‘Soil
parameters’ section (figure 24), and run soil_raster_prep.py. This script evaluates each
soil raster, projecting and clipping to the HRU parameter grid and filling ‘no data’ areas
using Nibble techniques (spatial analyst tool in Arc). Once the soil prep script has
finished, run soil_parameters.py. This will populate the soil property fields in the
parameter shapefile attribute table (AWC, CLAY_PCT, SAND_PCT, KSAT,
ROOT_DEPTH, SOIL_TYPE, MOIST_INIT, MOIST_MAX, RECHR_INIT,
RECHR_MAX, SSR2GW_RATE, SLOWCOEF_L, and SLOWCOEF_S). The soil files
will be stored in the ‘soil_rasters’ folder in the working directory, allowing for the user to
observe the image files in ArcMap.
 34

Figure 24: Soil parameters section of the Configuration File. The folder path containing
the soil data is designated, and each raster image filename is set for each soil property.

Step 5: Run impervious_parameters.py: The impervious cover data from the National
Land Cover Database is applied to the parameter shapefile in this script, as an impervious
percentage within each HRU grid cell. No formatting or altering is necessary for this
data. Ensure that the folder in which the impervious database is stored is entered in the
Configuration File under ‘Impervious cover parameters’ (figure 25) and run
impervious_parameters.py. Open the parameter shapefile once the script finishes and
observe the results by checking the attribute table and/or using ‘Symbology’ properties to
visualize results.

Figure 25: Impervious coverage parameters in the Configuration File. Cell size is the re-
projection size, once again best if a multiple of the grid cell size is used.
 35

III. Set temperature and precipitation parameters using observed station data and PRISM
climate data.

PRMS parameters used for distributing temperature and precipitation to HRUs are
generated using data collected at select climate station(s) within or near the study area
along with PRISM gridded climate data. Specific data is needed depending on the climate
modules being applied to the model. See the PRMS documentation for more information
on the specific parameters needed for each climate module. The following steps explain
the process and calculations that result in each HRU receiving a unique monthly
maximum and minimum temperature and a monthly precipitation amount along with a
precipitation factor. These data will appear in the parameter shapefile attribute table and
are written to the PRMS Parameter File to be applied with the several possible modules
available. As PRMS is driven by daily climate data from one or two climate stations
(point data), the precipitation multipliers, or ratios, derived through the following steps
allow for a more realistic distribution of climate data over the model domain.

Step 1: Prepare PRISM and climate station data: PRISM data can be downloaded
through the link provided at the beginning of this manuscript. The options available for
data download include different scales and time periods. Typically, and for this tutorial,
the 30-year 800m monthly normals are used; however, the user may choose to use
different PRISM data. Longer periods are available for purchase. The downloaded
climate data must be saved into separate folders containing the precipitation (ppt),
maximum temperature (tmax), and minimum temperature (tmin). In the Configuration
File, enter the folder path for the PRISM data in the ‘PRISM parameters’ section (figure
25).

Figure 25: PRISM parameter settings in the Configuration File. PRISM cell size is the
output cell size desired (cell size, or multiple, of the model grid).

To distribute precipitation data over the model domain, mean monthly precipitation
values need to be determined from observed data. These are values calculated from a
selected climate station or stations, typically the same station(s) that will be driving the
model with daily values. If a single station is being used, the observed mean monthly
values are entered in a comma separated row in the ‘PPT ratios’ section of the
Configuration File for ppt_obs_list (figure 26).
 36

Figure 26: For use of a single climate station, set_ppt_zones_flag is set to false, and
ppt_obs_list is set equal to monthly mean precipitation values from a climate station.
Two climate stations are shown in this example, but only one is read by the script
(Healdsburg), while the other (Cloverdale) is commented out by the ‘##’. Units of the
observed data are also set here.

A single climate station can be used for simplicity or for smaller models, but an option
exists for more than one climate station to be used. The Russian River is a large
watershed and contains several climate stations with decent records. A handful NWS-
COOP stations were chosen to distribute precipitation and temperature over the model
domain (figure 27). Data were obtained through the Western Regional Climate Center
(WRCC) climate summaries: http://www.wrcc.dri.edu/climate-summaries/. From this
link, the NOAA cooperative stations interactive map can be accessed to locate each
station and its data.

network sta_id sta_name record_start record_end Lat lon elev_ft

COOP 049684 WILLITS 1NE 3/6/1902 9/27/2012 39.41944 -123.34 1353

COOP 047109 POTTER VALLEY 9/1/1937 6/9/2016 39.36194 -123.12 1018
COOP 049122 UKIAH 1/1/1893 5/24/2013 39.14667 -123.21 636
COOP 049173 UPPER LAKE 1/1/1893 11/12/2006 39.19056 -122.89 1335
COOP 044701 LAKEPORT 1/1/1920 6/20/2002 39.03333 -122.91 1315
COOP 041838 CLOVERDALE 7/22/1950 5/31/2016 38.8 -123.01 330
WARM
COOP 049440 SPRINGS DAM 6/1/1973 1/3/2006 38.71611 -122.99 224
COOP 043875 HEALDSBURG 2/1/1893 8/31/2012 38.6175 -122.87 108
COOP 041312 CALISTOGA 3/1/1906 5/20/2016 38.59611 -122.60 400
SANTA ROSA
SONOMA CO
COOP 047971 AP 6/8/1998 6/9/2016 38.50389 -122.81 114
COOP 043578 GRATON 1/1/1926 4/30/2016 38.43056 -122.86 200
COOP 046370 OCCIDENTAL 5/1/1943 5/26/2016 38.38583 -122.96 865
COOP 047965 SANTA ROSA 6/1/1902 1/31/2013 38.43806 -122.69 174
Figure 27: NWS-COOP stations chosen to represent climate over the Russian River
model domain

In the case that multiple stations will be used to distribute climate data over a large model
domain, zones must be designated to assign the HRUs to a specified climate station. To
do this, a shapefile needs to be created that contains the chosen precipitation zone
polygons and holds the mean monthly precipitation values for each station (PPT_01,
 37

PPT_02 etc.) associated with a precipitation zone. The user must set set_ppt_zones_flag
to True in the Configuration File, and provide the location and name of the precipitation
zone shapefile, along with the field name of the zone field (figure 28).

Figure 28: PPT Ratios section when using more than one climate station. Zone flag is set
to True, observed data is commented out, and ppt_zone_path, ppt_zone_field, and
ppt_hru_id_field are specified. Observed units are still designated here.

Figure 29: Precipitation zone polygons and NWS-COOP climate stations. Precipitation
distribution will be based on the climate station assigned to each zone, and the
relationship of the mean monthly precipitation of that station to the PRISM data.
 38

To assign the precipitation zones for the Russian River model, HUC 12 polygons were
merged together to designate 8 zones (figure 29). A climate station was chosen to
represent each zone and the mean monthly precipitation was input to the attribute table of
the precipitation zone shapefile (figure 30). The HRU_ID where each station is located
can be entered into the PPT_HRU_ID field to ensure the precipitation ratio at that HRU
is equal to 1. If the station is outside the zone or the model domain, a zero entered will
keep the script from attempting to adjust the precipitation ratio to 1 at the station location.
The final PPT zone shapefile should contain only as many rows as PPT zones. It can help
to create the zone shapefile directly from the HRU parameter shapefile, by extracting the
data to a new shapefile and editing the attribute table. This will ensure no gaps are in the
precipitation zones and they will snap to the model grid. When selecting the PPT zones,
be sure to make the zone shapefile the only selectable layer so that the selected cells can
be merged into one polygon. Also make sure that the twelve (monthly) PPT columns are
type float. The final requirement for the precipitation zone shapefile is a column
containing HRU_PSTA numbering, which designates the precipitation station number
that is associated with each zone. This should mimic the order that the stations are listed
in the data file.

Figure 30: Example of a PPT zone shapefile attribute table. Eight zones were designated
for the Russian River tutorial, and mean monthly observed precipitation values from each
associated station were entered. The ratio of the observed station values to the PRISM
data will determine the precipitation distribution within each precipitation zone. This
option allows for a more realistic precipitation distribution over a larger model domain.

Step 2: Run prism_800m_normals.py and ppt_ratio_parameters.py: These scripts

will populate the temperature and precipitation fields in the parameter shapefile attribute
table. The first script, prism_800m_normals.py (or prism_4km_normals, if the user is
applying 4km PRISM data), reads the PRISM data from the three folders (ppt, tmax,
tmin) and re-samples the temperature and precipitation data to the HRU grid, for each
month of the year (figure 31). The second script, ppt_ratio_parameters.py, uses the
observed mean monthly precipitation values to calculate a ratio between the PRISM
precipitation and the observed data, and assigns each HRU a monthly precipitation factor
that acts as a multiplier to account for changes in elevation (and effects on precipitation)
throughout the model domain. Observe the results in the attribute table in ArcMap, and
make sure the precipitation factors make sense (figure 32). For example, in a
mountainous region, the ratios would mostly be <1 if a high elevation climate station was
used, and >1 if a low elevation station was used. These factors influence the rain and
snow adjust parameters in the PRMS Parameter File.
 39

Figure 31: Precipitation and maximum temperature distribution for the month of January,
based on observed climate data and the PRISM network. (note: these distributions are
based off a single climate dataset)

Figure 32: Precipitation factors for January are calculated from January PRISM values
and daily precipitation values measured at climate stations located within the watershed.
Nine precipitation stations and respective zones were used in this example. This process
is repeated for each month.
 40

IV. Finalize and build PRMS Parameter File

At this point, all PRMS related scripts have been run and the attribute table in the
parameter shapefile will be fully populated and prepared to develop the Parameter File
needed to run PRMS, along with several input files for MODFLOW. A couple final steps
will ensure successful development of the Parameter File.

Step 1: Designate Parameter File name, and CSV file locations: Review the PRMS
Parameters section of the Configuration File, and make sure to have the file path names
entered correctly, along with several model specific settings for the parameter and
dimension CSV files (figure 33). The parameter CSV file holds default values for some
parameters as well parameters that will be designated from the HRU shapefile and the
Configuration File. Look over these settings and compare the default values to those
given in the PRMS documentation. The user may want to adjust the defaults within the
possible range of values depending on the study area. Typically, however, the default
values are left alone until the calibration period, where they may be changed in the actual
Parameter File. The dimension CSV file contains counts of certain parameters, some of
which are specific to a study area, such as number of precipitation stations (nrain),
temperature stations (ntemp), and streamgage stations (nobs). For a model without lakes,
make sure nlake and nlake_hrus are set to zero. If lakes are being modeled, typing
‘calculated’ for these dimensions will derive the values from the parameter shapefile.
These two CSV files must be present in the working directory for the final
prms_template_fill.py script to run. Temperature station settings in this section of the
Configuration File must reflect the temperature module that will be used. These
parameters are explained in the PRMS documentation. Depending on the version of
PRMS being used, the user may choose to have Gsflow-Arcpy make a single Parameter
File with parameters listed in a single column (applicable to versions through PRMS 4) or
multiple Parameter Files with parameters in 2D arrays (applicable to PRMS 5). These
settings are easily designated by the flags in the Configuration File under the ‘PRMS
Parameters’ section (figure 33).
 41

Figure 33: Folder paths and file names for the PRMS Parameter File and accompanying
CSV files in the Configuration File. Temperature and elevation settings must be set based
on the module being used (temp_1sta vs temp_laps). This will provide PRMS with
adequate information to distribute temperature over the model domain.

Step 2. Run prms_tmeplate_fill.py: This script will create the initial Parameter File(s)
that PRMS will read. The Parameter File option (lines 193-195 in figure 33) directs the
script to create a single PRMS Parameter File or multiple Parameter Files (categorized in
the parameters.csv file), and the option to create parameter values in arrays or columns.
Default (false) settings will create multiple Parameter Files with arrays. After choosing
the settings and running the script, review the PRMS Parameter File(s) and ensure that
the initial build contains reasonable parameter values, and correct units are assigned to
match the data. It is easy to forget about maintaining consistent units while developing
the model data. Units used in the PRMS Data File to drive the model must match those of
the Parameter File. These include temperature, precipitation, and runoff. For example, the
units for the precipitation factors used in this tutorial are millimeters. This means that the
precipitation data in the Data File must also be in millimeters. This is the final step for
this version of Gsflow-Arcpy. To run PRMS, the user must also create a Data File and a
Control File. Guidelines for making these files can be found in the PRMS documentation.

For MODFLOW models, several columns in the HRU shapefile attribute table are ready
to be formatted into MODFLOW input files. Resources are available that can automate
this procedure somewhat, such as the MODFLOW FlowPy toolkit
(https://github.com/modflowpy/flopy). Layer bottom elevations can quickly be calculated
based on DEM_ADJ, and SFR package inputs are in the attribute table. A GSFLOW
model can be created once both PRMS and MODFLOW models have been separately
constructed and calibrated.
 42

References

Band, L. E., 1986. Topographic partition of watersheds with digital elevation models. Water
resources research. 22(1), 15-24.
Beven, K. J., Kirkby, M. J., 1979. A physically based, variable contributing area model of basin
hydrology/Un modèle à base physique de zone d'appel variable de l'hydrologie du bassin
versant. Hydrological Sciences Journal. 24(1), 43-69.
Daly, C., Neilson, R.P., Phillips, D.L., 1994. A statistical-topographic model for mapping
climatological precipitation over mountainous terrain. J. Appl. Meteor., 33, 140-158.
Fatichi, S., Vivoni, E., Ogden F., Ivanov, Z., Mirus, B., Gochis, D., Downer, C., Camporese, M.,
Davison, J., Ebel, B., Jonesk, N., Kim, J., Mascaro, G., Niswonger, R., Restrepoo, P.,
Rigon, R., Shen, C., Sulis, M., Tarboton, D., 2016. An overview of current applications,
challenges, and future trends in distributed process-based models in hydrology. Journal of
Hydrology. 537, 45-60
Fortin, J. P., Turcotte, R., Massicotte, S., Moussa, R., Fitzback, J., Villeneuve, J.-P., 2001. A
distributed watershed model compatible with remote sensing and GIS data. Part 1:
Description of the model. Journal of Hydrologic Engineering. 6(2), 91–99.
Gruber, S., Peckham, S., 2009. Land-surface parameters and objects in hydrology. Developments
in Soil Science. 33. 171-194.
Harbaugh, A.W., 2005, MODFLOW-2005, The U.S. Geological Survey modular ground-water
model—the Ground-Water Flow Process: U.S. Geological Survey Techniques and
Methods 6-A16, variously p.
Henson, W.R., Medina, R.L., Mayers, C.J., Niswonger, R.G., and Regan, R.S., 2013, CRT—
Cascade Routing Tool to define and visualize flow paths for grid-based watershed
models: U.S. Geological Survey Techniques and Methods 6-D2, 28 p.

Laniak, G.F., Olchin, G., Goodall, J., Voinov, A., Hill, M., Glynn, P., Whelan, G., Geller, G.,
Quinn, N., Blind, M., Peckham, S., 2013. Integrated environmental modeling: a vision
and roadmap for the future. Environmental Modelling & Software. 39, 3-23.
Leavesley, G.H., Restrepo, P.J., Markstrom, S.L., Dixon, M., Stannard, L.G., 1996. The Modular
Modeling System (MMS): User’s Manual: U.S. Geological Survey Open File Report 96-
151, 142 p.
Markstrom, S.L., Niswonger, R.G., Regan, R.S., Prudic, D.E., Barlow, P.M., 2008. GSFLOW-
Coupled groundwater and surface-water flow model based on the integration of the
Precipitation-Runoff Modeling System (PRMS) and the Modular Ground-Water Flow
Model (MODFLOW-2005): U.S. Geological Survey Techniques and Methods. 6-D1, 240
p.
 43

Markstrom, S.L., Regan, R.S., Hay, L.E., Viger, R.J., Webb, R.M.T., Payn, R.A., and
LaFontaine, J.H., 2015, PRMS-IV, the precipitation-runoff modeling system, version 4:
U.S. Geological Survey Techniques and Methods, book 6, chap. B7, 158 p.,
http://dx.doi.org/10.3133/tm6B7

Maidment, D.R., 2002, Arc Hydro: GIS for Water Resources, vol 1, ESRI, Inc. 203 p.

Metz, M., Mitasova, H., Harmon, R. S., 2011. Efficient extraction of drainage networks from
massive, radar-based elevation models with least cost path search. Hydrology and Earth
System Sciences. 15(2), 667.

Niswonger, R.G., Panday, S., Ibaraki, M., 2011. MODFLOW-NWT, A Newton formulation for
MODFLOW-2005: U.S. Geological Survey Techniques and Methods. 6-A37, 44 p.

Niswonger, R.G., Prudic, D.E., 2005. Documentation of the Streamflow-Routing (SFR2)

Package to include unsaturated flow beneath streams - A modification to SFR1: U.S.
Geological Survey Techniques and Methods. 6-A13, 50 p.

Regan, R.S., Markstrom, S.L., Hay, L.E., Viger, R.J., Norton, P.A., Driscoll, J.M., LaFontaine,
J.H., 2017, Description of the National Hydrologic Model for use with the Precipitation-
Runoff Modeling System (PRMS): U.S. Geological Survey Techniques and Methods,
book 6, chap B9, xxx p. DOI 10.3133

Tarboton, D. G., 1997. A new method for the determination of flow directions and upslope areas
in grid digital elevation models. Water Resources Research. 33(2). 309-319.
Tian, Y., Zheng, Y., Zheng, C., 2016. Development of a visualization tool for integrated surface
water–groundwater modeling. Computers & Geosciences 86. 1-14.

Viger, R.J., and Leavesley, G.H., 2007, The GIS Weasel User's Manual: U.S. Geological Survey
Techniques and Methods, book 6, chap. B4, 201 p; http://pubs.usgs.gov/tm/2007/06B04/

Wilson, J. P., 2012. Digital terrain modeling. Geomorphology. 137(1). 107-121.