Technical Manual For The Geospatial Stream Flow Model

Prepared in cooperation with the U.S.
Agency for International Development, Office of

Foreign Disaster Assistance (USAID/OFDA)
Technical Manual for the Geospatial Stream Flow Model

(GeoSFM)
By Kwabena O. Asante, Guleid A. Artan, Shahriar Pervez, Christina Bandaragoda, and James P. Verdin
Open-File Report 20071441
U.S. Department of the Interior

U.S. Geological Survey
U.S. Department of the Interior

DIRK KEMPTHORNE, Secretary
Mark D. Myers, Director
U.S. Geological Survey, Reston, Virginia 2008
For product and ordering information:

World Wide Web: http://www.usgs.gov/pubprod
Telephone: 1-888-ASK-USGS
For more information on the USGSthe Federal source for science about the Earth,
its natural and living resources, natural hazards, and the environment:
World Wide Web: http://www.usgs.gov
Telephone: 1-888-ASK-USGS
Suggested citation:
Asante, K.O., Artan, G.A., Pervez, S., Bandaragoda, C. and Verdin, J.P., 2008, Technical Manual for the
Geospatial Stream Flow Model (GeoSFM): U.S. Geological Survey Open-File Report 20071441, 65 p.
Any use of trade, product, or firm names is for descriptive purposes only and does not imply
endorsement by the U.S. Government.
Although this report is in the public domain, permission must be secured from the individual
copyright owners to reproduce any copyrighted material contained within this report.
ii
Contents
Abstract................................................................................................................................................................................1
Introduction .........................................................................................................................................................................1
GeoSFM Architecture........................................................................................................................................................2
Preprocessing Module ......................................................................................................................................................4
Terrain Analysis Routine ...............................................................................................................................................4
Basin Characterization Routine ...................................................................................................................................5
Unit Hydrograph Generation Routine..........................................................................................................................9
Weather Data Processing Routine............................................................................................................................10
Soil Moisture Accounting Module ................................................................................................................................11
Linear Soil Moisture Accounting (LSMA) Routine..................................................................................................12
Nonlinear Soil Moisture Accounting (NSMA) Routine ..........................................................................................16
River Transport Module...................................................................................................................................................21
Pure Lag Flow Routine .................................................................................................................................................21
Diffusion Analog Routine.............................................................................................................................................23
Muskingum Cunge Routine .........................................................................................................................................26
Calibration Module...........................................................................................................................................................30
One-At-a-Time (OAT) Sensitivity Analysis ...............................................................................................................33
The Multi-Objective Shuffled Complex Evolution Metropolis (MOSCEM) Algorithm.......................................34
Postprocessing of Calibration Results......................................................................................................................35
Postprocessing Utilities...................................................................................................................................................35
Conclusions .......................................................................................................................................................................36
References Cited ..............................................................................................................................................................36
Appendix 1: GeoSFM Data Specifications ...................................................................................................................38
Figures
Figure 1. Process map and system diagram for the Geospatial Stream Flow Model............................................3
Figure 2. Calibration module relationships between the Geospatial Stream Flow Model and the MultiObjective Shuffled Complex Evolution Metropolis algorithm...................................................................................31
Figure 3. Flowchart of the calibration process. ..........................................................................................................32
Tables
Table 1. SCS runoff curve numbers used for various soil and land cover classes in GeoSFM...........................6
Table 2. Mannings roughness values used for various land cover classes in GeoSFM....................................10
Table 3. List of key variables for the Linear Soil Moisture Accounting routine....................................................12
Table 4. List of key variables for the Nonlinear Soil Moisture Accounting routine. ............................................16
Table 5. List of key variables for the pure lag flow routine.......................................................................................21
Table 6. List of key variables for the diffusion analog routine. ................................................................................24
Table 7. List of key variables for the Muskingum Cunge routine.............................................................................27
iii
Conversion Factors
SI to Inch/Pound
Multiply
By
To obtain
Length
centimeter (cm)
0.3937
inch (in.)
millimeter (mm)
0.03937
inch (in.)
meter (m)
3.281
foot (ft)
kilometer (km)
0.6214
mile (mi)
meter (m)
1.094
yard (yd)
0.0002471
acre
Area
square meter (m2)
2
square kilometer (km )
247.1
square meter (m )
10.76
square kilometer (km2)
0.3861
acre
square foot (ft2)
square mile (mi2)
Flow rate
cubic meter per second (m3/s)
meter per second (m/s)
70.07
3.281
acre-foot per day (acre-ft/d)

foot per second (ft/s)
cubic meter per second (m /s)
35.31
cubic foot per second (ft3/s)
cubic meter per second (m3/s)
22.83
million gallons per day (Mgal/d)
iv
Technical Manual for the Geospatial Stream Flow

Model (GeoSFM)
By Kwabena O. Asante1, Guleid A. Artan1, Shahriar Pervez1, Christina Bandaragoda2, and James P.
Verdin3
1
Science Applications International Corporation, Contractor to the U.S. Geological Survey (USGS)
Earth Resources Observation and Science (EROS) Center, Sioux Falls, SD, work performed under
USGS contract 03CRCN0001
2
Water Resource Research Institute, Utah State University, Logan, UT
Abstract
The monitoring of wide-area hydrologic events requires the use of geospatial and time
series data available in near-real time. These data sets must be manipulated into information
products that speak to the location and magnitude of the event. Scientists at the U.S. Geological
Survey Earth Resources Observation and Science (USGS EROS) Center have implemented a
hydrologic modeling system which consists of an operational data processing system and the
Geospatial Stream Flow Model (GeoSFM). The data processing system generates daily forcing
evapotranspiration and precipitation data from various remotely sensed and ground-based data
sources. To allow for rapid implementation in data scarce environments, widely available terrain,
soil, and land cover data sets are used for model setup and initial parameter estimation. GeoSFM
performs geospatial preprocessing and postprocessing tasks as well as hydrologic modeling tasks
within an ArcView GIS environment. The integration of GIS routines and time series processing
routines is achieved seamlessly through the use of dynamically linked libraries (DLLs) embedded
within Avenue scripts. GeoSFM is run operationally to identify and map wide-area streamflow
anomalies. Daily model results including daily streamflow and soil water maps are disseminated
through Internet map servers, flood hazard bulletins and other media.
Introduction
During the past two decades, a number of major flooding events, such as the Midwest
floods of 1993 (Kunkel and others, 1994) and the southern African floods of 2000 (Christie and
Hanlon, 2001), have drawn attention to the need for wide-area monitoring systems for hydrologic
hazards. The goal of such monitoring is to provide early warning to at-risk populations and decision
makers, and to direct the efforts of relief agencies involved in humanitarian assistance and
sustainable development activities. In many of these areas, poor ground-based monitoring or data
transmission systems hamper the efforts of local water authorities to issue spatially specific flood
warnings in a timely fashion. Consequently, hydrologic modeling has been adopted as the main
approach for estimating flows for warning generation. While many hydrologic models are in
common use today, most of these models are unsuitable for wide-area implementation in data
scarce environments. The models may not be designed to ingest geospatial data sets, an important
requirement for wide-area models. The estimation of model parameters often also requires
observed streamflow data, which is often not available in many parts of the world. Another
common problem is the absence of a steady and reliable stream of daily forcing data sets to
facilitate near-real time operation.
To help address these information needs, scientists at the U.S. Geological Survey Earth
Resources Observation and Science (USGS EROS) Center, with the cooperation of the U.S.
Agency for International Development, Office of Foreign Disaster Assistance (USAID/OFDA),
have developed a system for operational monitoring of wide-area hydrologic hazards in data scarce
environments. The system consists of a Geospatial Stream Flow Model (GeoSFM) for flow
simulation and a supporting data processing system for generating daily forcing data (Artan and
others, 2004). GeoSFM uses spatially distributed terrain, soil, and land cover data sets for model
parameterization. It also includes hydrologic routine algorithms for data preprocessing,
postprocessing, and time series manipulation. The data processing system handles the acquisition,
transformation, and staging of data sets required for initial model parameterization and for daily
model simulations. The integration of these spatially distributed time series of evapotranspiration
and precipitation and static grids of surface elevation, soil, and land cover characteristics into an
operational monitoring system is described in this document and in Asante and others (2007). This
technical manual also describes the various algorithms and routines which GeoSFM uses for runoff
estimation and transport.
GeoSFM Architecture
The development of the model was driven by the need to establish a common visual
environment for the topographic analysis, geospatial data processing, time series manipulation, and
results presentation activities that go into the monitoring of hydrologic conditions over large areas.
The spatially distributed nature of the raster grids used in many of these processes pointed to the
adoption of a customizable geographic information system with excellent raster functionality. The
ArcView GIS software was adopted for the implementation because it provided a visual,
customizable development environment with excellent support of raster operations. Figure 1
presents an overview of GeoSFM processes showing the relationship among model inputs and
outputs.
Figure 1. Process map and system diagram for the Geospatial Stream Flow Model.
Geospatial
Data Inputs
GeoSFM Process Map
Precip itation
S ource: NO AA
Hydrographs
Color-coded
maps
ArcView GIS
Model Interface
Land Cover
S ource: ED C
Preprocessing
Run Initiation
Postprocessing
Hydrologic Algorithm Services

Subbasin
Soil Water Balance
Linea r
Nonlinear
S oils
S ource: FA O
Optional:
User supplied
data inputs
(v ari ab le su b je ct t o
av ai la b il it y )
Balparam.txt
Balfiles.txt
Routparam.txt
Routfiles.txt
Order.txt
1D ASCII
Files
Dynamic data sou rce
Static data source
Formatted Model Inputs

Response.txt
Basin.txt
Ra in.txt
River.txt
Times Series
ASCII
Fo rmat :
B asi n ID
P arame te rs
Tim e
Baseflow
Groundwater
loss
Interflow
Ba si n ID
Streamfl ow
Forecast
Time Series
ASCII
Massbalance
TimeSeries
ASCII
Riverdepth
LocalFlow
E xcessflow actualE T
Soilwater
Fo rmat :
Route
Between Basins
Diffusion Analog,
Musk-Cunge
L ag
Formatted Model Outputs

Basinrunoff
yield
PET .txt
2D ASCII
Files
S
Fo rmat :
Parame te rs
Graphical Model Outputs
Data Processing Services
PET
Sou rce: NOA A + EDC
Visualization Services
Forma t :
Ti me
B asi n ID
An ArcView extension was developed in Avenue (ArcViews in-built customization

language) for the geospatial processing operations. GeoSFM contains preprocessing and
postprocessing modules to perform such tasks. It also includes a hydrologic algorithm module for
the initiation of time series analysis tasks and a calibration module which assesses the outputs from
model runs for proximity to observations, modifies input files, and initiates new model runs in an
iterative effort to reduce difference between model results and observations. All of these modules
are initiated within ArcView to take advantage of its interface, associated tools, and input/output
forms. However, the manipulation of rainfall and evapotranspiration time series to generate
streamflow is better handled with compiled programming languages than with interpreted
languages that are available in the GIS environment. For this reason, routines for performing the
hydrologic computations involved in flow mass balance and routing operations were developed in a
mixed programming environment (C/C++ and FORTRAN) and compiled as dynamically linked
libraries (DLLs) using Microsofts Visual Development Studio. The calibration module also
includes both executable (*.exe) programs and DLLs.
The Visual Studio environment allows codes created by various developers in programming
languages of their preference to be compiled together into a single DLL file. Procedures for calling
up DLL routines were implemented from within Avenue scripts, allowing the ArcView extension
to serve as a common model interface for both GIS and time series analysis. These procedures
eliminate the need to develop a separate interface for hydrologic routing operations. The DLLs and
3
ArcView scripts work together seamlessly; the user cannot tell that a different program has been
initiated. The only major problem encountered in this implementation is that errors generated
within the DLL will cause the parent program (ArcView in this case) to shut down. Three measures
were implemented in GeoSFM to overcome this problem. First, all data files to be used in the DLL
program are screened for completeness using a series of validation checks in ArcView before a
routine call is initiated to the DLL. Second, any errors occurring within the DLL are reported to test
files instead of through error windows. Third, explicit exit routines are implemented within the
DLL to close any open data files prior to returning program control to ArcView.
Preprocessing Module
The GeoSFM preprocessing module contains several routines for deriving flat ASCII files
required by other modules from geospatial data sets available for most regions of the world.
Detailed procedures for downloading and processing the input geospatial data sets from their
original sources are presented in the Users Manual, which accompanies this publication. The
routines in this section extract modeling units and parameters from the geospatial data sets and
transcribe the results into text files for use in hydrologic computations.
Terrain Analysis Routine

The analysis of topographic data for hydrologic modeling applications relies on the simple
principle that water flows in the direction of steepest descent. Hence, by comparing the elevation of
a given cell with that of the eight surrounding cells, it is possible to determine which direction
incident drops of water would flow. Flow direction in GeoSFM is assigned using the eight direction
pour point model (Jenson and Dominque, 1988) in which each grid cell is assigned one of eight
compass directions depending on which of its eight neighboring cells its discharges flow to. The
computation of a flow direction grid paves the way for the determination of other parameters of
hydrologic interest such as upstream contributing area, distance to the basin outlet, and the slope of
the land surface.
The division of river basins into catchments with associated river reaches, and the
parameterization of these modeling units from Digital Elevation Model (DEM) data have been
automated in many hydrologic models including GeoSFM. While the model supports the use of
DEMs of any resolution, the 1-kilometer resolution HYDRO1k DEM produced at USGS EROS is
used in GeoSFM because of its global coverage. HYDRO1k is a hydrologically corrected DEM,
which implies that it is devoid of spurious pits that interrupt hydraulic connectivity over the land
surface. These pits are artifacts of the interpolation procedure used in the creation of DEMs, and
they result in breaks in the flow network unless they are removed. HYDRO1k data in which pits
have been removed using the ArcInfo Fill command are available for download at
http://edcdaac.usgs.gov/gtopo30/hydro. The terrain analysis proceeds from the hydrologically
corrected DEM to produce the following grids:
Flow Direction Grid (FLOWDIR): A measure of direction of steepest descent from each cell
location, computed using ESRIs flowdirection function with no special treatments of edges.
Flow Accumulation Grid (FLOWACC): A count of cells upstream of a given cell location,
computed using ESRIs flowaccumulation function with no weights.
Flow Length Grid (FLOWLEN): A measure of distance from a given cell location to the nearest
sink cell along the path defined by the flow direction grid, computed using ESRIs flowlength
function with no weights.
Stream Grid (STREAMS): A grid identifying the location of river cells, computed by assigning
a value of 1 to river cells with flow accumulation grid values greater than a user-defined
threshold and NODATA to all other cells.
Stream Link Grid (STRLINKS): A grid grouping cells belonging to the same river reach with a
unique value, computed using ESRIs streamlink function with the flow direction and stream
grids as inputs.
Outlet Grid (OUTLETS): A grid identifying the most downstream cell (a cell with the highest
flow accumulation value) within each river reach in the stream link grid, computed using
ESRIs zonalmax and equality test (==) grid functions.
Basin Grid (BASINS): A grid identifying the subbasins associated with each river reach in the
stream link grid, computed using ESRIs watershed function with the flow direction and outlet
grids as inputs.
Hill Length Grid (HILLLENGTH): A measure of distance from a given cell location to the
nearest river cell along the path defined by the flow direction grid, computed using ESRIs
flowlength function with the flow direction grid modified by assigning NODATA values to
river cells.
Slope Grid (SLOPE): A measure of the maximum rate of change of elevation around a given
cell location, computed using ESRIs slope function and the input elevation grid with results
expressed as percent rise.
Downstream Grid (DOWNSTREAM): A grid identifying the subbasin immediately

downstream of each river reach, computed using ESRIs grid neighborhood notation to query
the cell next to each outlet grid cell in the direction defined by the flow direction grid.
The results of the terrain analysis are stored in ArcInfo grid format for access by other
routines.
Basin Characterization Routine

In addition to the elevation derivatives, GeoSFM contains algorithms for estimating flow
simulation parameters for each catchment and river reach from land cover and soil data sets. Land
cover influences both the rate of runoff generation and the rate of overland flow transport. The
USGS Global Land Cover Characterization (GLCC) database (Loveland and others, 2000) is used
in GeoSFM because of its global coverage. The GLCC data were derived from 1-km AVHRR data
and are available in the Interrupted Goode Homolosine and Lambert Equal-Area Azimuthal
projections. The data are available for download at http://edcdaac.usgs.gov/glcc/. Land cover data
are used to compute an impervious area grid (MAXCOVER) to account for the presence of water
bodies in a subbasin. Land cover is also used in assigning vegetation roughness, which is
subsequently used in conjunction with land surface slope to estimate overland velocity.
For estimating soil parameters such as water holding capacity (WHC), hydrological active
depth (DEPTH), texture (TEXTURE), and saturated hydraulic conductivity (KC), GeoSFM uses
data from the Digital Soil Map of the World (DSMW) jointly produced by the United Nations Food
and Agriculture Organization (FAO) and the United Nations Educational, Scientific and Cultural
Organization (UNESCO) (FAO, 19711981) because of its global coverage. The DSMW has 10
map sheets covering Africa, North America, Central America, South America, Europe, central
Asia, northeast Asia, the Far East, Southeast Asia, and Oceania with data sets available in ArcInfo
Vector Export, ERDAS Raster, and IDRISI Raster formats. All the data are in geographic
5
projection and are produced from original map sheets with a scale of 1:5,000,000. The Zobler data
set (Zobler, 1986) is used for characterizing soil type and texture. For soil depth characterization,
the Global Data Set of Soil Particle Size Properties produced by Webb and others (1993) was used.
The data set specifies the top and bottom depths and the percentage of the texture type (sand, silt,
and clay) of soil horizons in 106 different soil types cataloged for nine regions. Detailed procedures
for extracting the required grids from these data sets are described in the User Manual, which
accompanies this document, because the procedures are performed outside of GeoSFM. The
expectation is that, where possible, individual users can use more accurate or alternate data sets to
generate the input grids for use in GeoSFM.
The soil and land cover data sets are also used together to determine Soil Conservation
Service (SCS) runoff curve numbers in GeoSFM. The curve numbers are used by some routines in
GeoSFM to determine the amount of incident precipitation that becomes surface runoff. As shown
in table 1, SCS curve numbers are assigned based on land cover type and the soil hydraulic classes
(A, B, C, or D) of the U.S. Department of Agricultures (USDA) soil classification system. The
coarse and medium/coarse textures in the FAO data set are matched with the loam and sandy loam
(USDA group A) classes. Likewise, medium and medium/fine FAO classes are matched to the
USDA soil classes B and C respectively while FAO fine class is matched to the USDA class D.
The intersection of the land cover and soil texture grids yields a runoff curve number (RCN) grid,
which is placed in the SOILDATA directory of GeoSFM for user access.
Table 1. SCS runoff curve numbers used for various soil and land cover classes in GeoSFM.
Anderson
Code
0
100
211
212
213
280
290
311
321
330
332
411
412
421
422
430
500
620
610
770
820
810
850
830
900
Land Cover Description

Unclassified
Urban and Built-Up Land
Dryland Cropland and Pasture
Irrigated Cropland and Pasture
Mixed Cropland and Pasture
Cropland/Grassland Mosaic
Cropland/Woodland Mosaic
Grassland
Shrubland
Mixed Shrubland/Grassland
Savanna
Deciduous Broadleaf Forest
Deciduous Needleleaf Forest
Evergreen Broadleaf Forest
Evergreen Needleleaf Forest
Mixed Forest
Water Bodies
Herbaceous Wetland
Wooded Wetland
Barren or Sparsely Vegetated
Herbaceous Tundra
Wooded Tundra
Mixed Tundra
Bare Ground Tundra
Snow or Ice
Soil Class A
54.0
81.0
68.0
62.0
65.0
65.0
45.0
54.0
45.0
49.5
57.0
45.0
45.0
25.0
25.0
35.0
98.0
30.0
25.0
68.0
98.0
98.0
98.0
98.0
98.0
Soil Class B
70.0
88.0
79.0
71.0
75.0
75.0
66.0
70.0
66.0
68.0
73.0
66.0
66.0
55.0
55.0
60.5
98.0
58.0
55.0
79.0
98.0
98.0
98.0
98.0
98.0
Soil Class C
80.0
91.0
86.0
78.0
82.0
82.0
77.0
80.0
77.0
78.5
82.0
77.0
77.0
70.0
70.0
73.5
98.0
71.0
70.0
86.0
98.0
98.0
98.0
98.0
98.0
Soil Class D
85.0
93.0
89.0
81.0
85.0
85.0
83.0
85.0
83.0
84.0
86.0
83.0
83.0
77.0
77.0
80.0
98.0
78.0
77.0
89.0
98.0
98.0
98.0
98.0
98.0
The basin characterization routine summarizes the elevation derivatives and soil and land
cover parameters over each subbasin and exports the results into ASCII files formatted to GeoSFM
specifications. For any river basin application, the basin characterization is only performed once
during the model creation phase, and the resulting files are stored for access by subsequent
simulation routines. The routine produces two basin characteristics files that can be manually
altered as necessary to enhance model performance using any text editor. The first of these files is
the basin file (basin.txt). This file contains the following characteristics required for soil moisture
accounting in each subbasin:
Soil Water Holding Capacity Parameter (SOILWHC): Contains a value in mm/m of the mean
soil water holding capacity for each subbasin, computed using ESRIs zonalmean function with
the WHCGRID and BASINS grids as inputs.
Soil Depth Parameter (SOILDEPTH): Contains a value in cm of the mean soil depth for each
subbasin, computed using ESRIs zonalmean function with the DEPTH and BASINS grids as
inputs.
Texture Parameter (TEXTURE): Contains a value indicating the predominant soil type (1 =
sandy soil, 2 = loam, and 3 = clay) or water body (5 = water) in each subbasin, computed using
ESRIs zonalmajority function with the TEXTURE and BASINS grids as inputs.
Saturated Hydraulic Conductivity Parameter (KS): Contains a value of saturated hydraulic

conductivity in cm/hr for the predominant soil type in each subbasin, computed using ESRIs
zonalmajority function with the KS and the BASINS grids as inputs.
Area of the Subbasin (AREA): Contains the area of each subbasin in km2, computed from the
BASINS grid using the count and unit area of cells within each subbasin.
Residence Time for the Interflow Reservoir (INTERFLOWLAG): Contains an interflow

residence time in days for each subbasin, computed as the time it would take water to drain
from the subbasin AREA to the river if drainage occurred along the entire length of the river at
the KS rate. Limits of between 2 and 120 days are imposed on the interflow residence times.
Hill Slope Parameter (HSLOPE): Contains the average subbasin slope for each subbasin,
computed as the average change in elevation between subbasin cells and their associated outlet
divided by the average flow length between the same cells and outlets.
Residence Time for the Baseflow Reservoir (BASEFLOWLAG): Contains a baseflow

residence time in days for each subbasin, initially estimated as three times the residence time of
the baseflow reservoir.
Runoff Curve Number (RCNUMBER): Contains a SCS runoff curve number for each subbasin,
computed using ESRIs zonalmean function with the RCN and the BASINS grids as inputs.
Percent Impervious Cover (MAXCOVER): Contains a fraction of subbasin covered by a water

body or other impervious area, computed by applying ESRIs zonalsum function with BASINS
grid as the zonal grid over the STREAMS grid and the LANDCOV grid and dividing the result
by the total count of cells of each subbasin.
Local Moisture Loss Adjustment (BASINLOSS): Contains a fraction applied to subbasin soil
moisture to account for local losses to regional groundwater, assigned based on predominant
soil type as 0.95 for sandy soil, 0.97 for loam, 0.99 for clay, 0.98 for water bodies, and 0.99 for
other undefined classes.
Local Evapotranspiration Adjustment (PANCOEFF): Contains a coefficient applied to

reference evapotranspiration to obtain potential evapotranspiration; a default value of 0.95 is
assumed for all subbasins.
Topsoil Fraction (TOPSOIL): Contains a fraction of the total soil layer that is classified as
topsoil; a default value of 0.1 is assumed for all subbasins.
Aridity Index (ARIDITY): Contains an index which is used by the Nonlinear Soil Moisture
Accounting model for selecting among surface runoff mechanisms; values can range from 1 to
3 (1 = dry, 2 = normal, 3 = dry); a default value of 2 is assumed for all subbasins during initial
parameterization.
The basin characterization routine also produces the river file (river.txt), which contains the
following characteristics required for river flow transport in each river reach:
2
Area of the River Subbasin (AREA): Contains the area in km of the subbasin around each river
reach, computed from the BASINS grid using the count and unit area of cells within each
subbasin.
Drainage Area of River Reach Outlet (UPAREA): Contains the drainage area in km2 upstream
of each river reach including the local subbasin, computed using ESRIs zonalmax function
with the FLOWACC and BASINS grids as inputs.
River Slope Parameter (RIVSLOPE): Contains the slope of the river channel, computed with
ESRIs zonalrange function as the difference between the maximum and minimum elevation
within each reach in the DEM grid divided by the difference between the maximum and
minimum flow length in the FLOWLEN grid with STRLINKS defining the zones for both
computations.
River Length Parameter (RIVLENGTH): Contains the length of the river channel, computed
with ESRIs zonalrange function as the difference between the maximum and minimum value
in the FLOWLEN grid with zones defined by the STRLINKS grid.
Downstream Reach Identifier (DOWNID): Contains the identity of the downstream river reach,
computed with ESRIs zonalmajority function from the DOWNSTREAM grid with zones
defined by the STRLINKS grid.
Mannings Roughness Coefficient (RIVMANNING): Contains a value of Mannings roughness

coefficient for the reach; a default value of 0.035 is assumed for all reaches.
River Loss Factor (RIVLOSS): Contains a channel infiltration loss factor for each river; a
default value of 1.0 (implying no loss) is assumed for all reaches.
River Floodplain Loss Factor (RIVPLOSS): Contains a floodplain infiltration loss factor to be
used for each river for flows exceeding bankfull flow; a default value of 1.0 (implying no loss)
is assumed for all reaches.
Dam Indicator (HASDAM): Contains a value that indicates the absence (0) or presence (1) of a
dam within the river reach; a default value of 0 (implying no dam) is assumed for all reaches.
Rating Curve Indicator (HASRATING): Contains a value which indicates the absence (0) or
presence (1) of a flow rating curve for any location within the river reach; a default value of 0
(implying no rating curve) is assumed for all reaches.
Flow Data Indicator (HASFLOWDATA): Contains a value which indicates the absence (0) or
presence (1) of observed flow data for any location within the river reach; a default value of 0
(implying no flow data) is assumed for all reaches.
Kinematic Wave Celerity (CELERITY): Contains a value of celerity (in m/s), which is the rate
at which a flood wave is propagated through the reach; values between 0.3 m/s and 5 m/s are
assigned based on RIVSLOPE and UPSTREAM (used as a surrogate for hydraulic radius).
Dispersion Coefficient of the Diffusion Analog Equation (DIFFUSION): Contains a value of

the dispersion coefficient (in m2/s), which accounts for flow attenuation in the diffusion analog
2
equation; values between 100 and 10,000 m /s are computed based on reach length and celerity
by equating the second moments of the Muskingum equation (with X = 0.35) and the diffusion
analog equation (Dooge, 1973).
River Width (RIVWIDTH): Contains a value of river channel width in m, computed based on
drainage area using the empirical equation of Dunne and Leopold (1978).
Reference Flow (FLOWREF): Contains a value of reference flow for the Muskingum Cunge
equation, computed based on upstream drainage area using the empirical equations of Dunne
and Leopold (1978).
Run Type Indicator (RUNTYPE): Indicates whether the simulation should proceed as a new
run (0) or as a continuation of an old run (1); a default value of 0 (implying new run) is
assumed for all reaches.
Unit Hydrograph Generation Routine

As a semidistributed hydrologic model, GeoSFM requires a single input value of
precipitation and other forcing data for each catchment during each modeling time step. The
distribution of water within the catchment is not modeled explicitly during each time step. Instead,
a unit hydrograph is developed to simulate the typical response of the catchment to a uniformly
distributed water input event. The unit hydrograph is developed for each catchment during the
model preprocessing phase. GeoSFM contains automated GIS-based algorithms for generating
these responses. The default procedure involves computing a uniform overland velocity for each
catchment based on the mean slope of the catchment and dominant land cover type present. The
algorithm also supports the use of nonuniform velocity grids or predetermined uniform velocity
values supplied by the user. The distance along the flow path from each grid cell in the catchment
to the catchment outlet is also computed. The travel time from each grid cell to the catchment outlet
is also computed using ArcViews flowlength function with the FLOWDIR grid defining flow
connectivity and the inverse of the VELOCITY grid serving as a weight grid. The default approach
for estimating overland velocity from land cover uses Mannings equation with values of hydraulic
radius assigned to each cell based on drainage area, as shown in equation 1.
VELOCITY
2
1
RH 3 HILLSLOPE
MANNINGN
(1)
where RH is the hydraulic radius

HILLSLOPE is the average elevation change divided by the average flow length from each
cell to the catchment outlet
MANNINGN is the Manning roughness for the dominant land cover in the subbasin
VELOCITY is the average overland velocity in the subbasin
The hydraulic radius values are 0.002 m, 0.005 m, 0.01 m, 0.02 m, and 0.05 m for drainage
areas up to 1,000 km2, 2,000 km2, 3,000 km2, 4,000 km2, and 5,000 km2, respectively. Cells with
drainage areas greater than 5,000 km2 are assumed to be river cells, and velocities ranging from 0.3
m/s to 1.5 m/s are directly assigned based on drainage and slope. For nonriver cells, Mannings
roughness values are estimated based on the land cover type as shown in table 2.
Table 2. Mannings roughness values used for various land cover classes in GeoSFM.
Anderson Code
Description
Manning Roughness
100
Urban and Built-Up Land
0.03
211
Dryland Cropland and Pasture
0.03
212
Irrigated Cropland and Pasture
0.035
213
Mixed Dryland/Irrigated Cropland and Pasture
0.033
280
Cropland/Grassland Mosaic
0.035
290
Cropland/Woodland Mosaic
0.04
311
Grassland
0.05
321
Shrubland
0.05
330
Mixed Shrubland/Grassland
0.05
332
Savanna
0.06
411
Deciduous Broadleaf Forest
0.1
412
Deciduous Needleleaf Forest
0.1
421
Evergreen Broadleaf Forest
0.12
422
Evergreen Needleleaf Forest
0.12
430
Mixed Forest
0.1
500
Water Bodies
0.035
620
Herbaceous Wetland
0.05
610
Wooded Wetland
0.05
770
Barren or Sparsely Vegetated
0.03
>800
Tundra, Snow or Ice
0.05
The distribution of discharge at the catchment outlet is given by the probability density
function (PDF) of travel times in the catchment. The PDF is analogous to the geomorphologic
instantaneous unit hydrograph (GIUH) of Rodriguez-Iturbe and Valdes (1979). The unit
hydrograph which is the probability mass function of flow travel times is obtained by discretizing
the PDF over the routing interval. The resulting unit hydrograph is only applicable for the particular
routing interval for which it was defined. For example, a change from a daily routing interval to a
six hourly interval would require the definition of a new unit hydrograph response. The unit
hydrograph response is computed for each catchment during preprocessing and stored in an ASCII
file for subsequent use.
Weather Data Processing Routine

Precipitation data is the most essential input to any hydrologic model, and in the case of the
GeoSFM, it was the driving factor in the design of the model. While ground-based gauges are
10
widely considered to provide the most accurate measure of precipitation at any location, the spatial
coverage of such gauges is very poor in many regions of the world, and access to data from those
gauges is very limited and rarely timely. Consequently, a number of satellite precipitation products
are generated by meteorological agencies around the world to meet user needs for timely data. The
primary rainfall data used for daily simulations in GeoSFM are satellite-derived rainfall estimates
(RFE) produced by the Climate Prediction Center (CPC) of the National Oceanic and Atmospheric
Administration (NOAA) (Xie and Arkin, 1997).
The NOAA RFE product uses satellite images from a variety of sources in conjunction with
ground-based gauges to process rainfall estimates daily. The theoretical basis for satellite rainfall
estimation derives from the fact that rainfall forms when moisture in the atmosphere is cooled to
condensation. In the absence of condensation nuclei, moisture condenses at temperatures of 235 K
and below. By monitoring the temperature of cloud tops from infrared imagery from geostationary
satellites such as METEOSAT 7, scientists can determine areas where moisture condensation, and
consequently rainfall, is occurring. Gauge measurements at experimental sites are used to obtain a
first estimate of rainfall rates. Microwave imagery from sensors, such as the Special Sensor
Microwave/Imager (SSM/I) of the Defense Meteorological Satellite Program and the Advanced
Microwave Sounding Unit (AMSU) from NOAA, are also incorporated to estimate nonconvective
rainfall. The observed rainfall data used in the NOAA RFE comes from the Global
Telecommunications System (GTS) operated by the World Meteorological Organization (WMO).
The system aggregates ground-based data collected by national meteorological agencies around the
world. While there are not enough stations in the GTS system to support hydrologic modeling on
their own, they facilitate daily ground truthing and adjustment of the satellite-derived products.
Another important model input is the daily potential evapotranspiration (PET) data
produced by the Early Warning group at the USGS EROS (Verdin and Klaver, 2002). Data from
the Global Data Assimilation System (GDAS) (Kanamitsu, 1989) is used to solve the PenmanMonteith equation to generate grids of PET at a daily time step. GeoSFM contains procedures for
ingesting the resulting PET grids and computing actual daily evapotranspiration based on
antecedent soil moisture conditions. The Penman-Monteith equation takes into consideration
weather conditions including air temperature, fluxes of long and short wave radiation, actual and
saturation vapor pressure, relative humidity, and wind speed at 2 m above the ground. A reference
crop height of 0.12 m, a fixed surface resistance of 70 s m-1, and an albedo of 0.23 are assumed. A
daily PET grid with a spatial resolution of 100 by 100 km and global coverage is produced and
distributed via the USGS EROS FTP site at ftp://edcftp.cr.usgs.gov/pub/edcuser/fewsips/global.
GeoSFM contains algorithms for computing mean areal precipitation (MAR) and mean
areal evapotranspiration (MAE) values for each catchment from these grids. To correctly associate
a given grid with the time interval for which it was computed, the algorithms rely on a grid file
naming convention which includes the year and Julian day. GIS routines are used to identify and
average the grid cells within each catchment, and the results are stored in ASCII files. For areas
over which spatially distributed precipitation data is not available, data interpolation routines are
provided for converting station readings into a continuous surface. The routines include inverse
distance weighting, spline, and kriging interpolation methods. A generation of evapotranspiration
time series in ASCII format is the final step before the initiation of hydrologic model runs.
Soil Moisture Accounting Module

As a continuous simulation model, GeoSFM contains routines for computing runoff and soil
moisture conditions daily. The two options provided for performing this accounting in GeoSFM are
the Linear Soil Moisture Accounting (LSMA) and the Nonlinear Soil Moisture Accounting
11
(NSMA) modules. The two modules have been coded in a mixed programming environment and
compiled as dynamically linked libraries (DLLs). Avenue code has been developed to call up and
execute the DLL routines from within ArcView. Consequently, the user is able to perform either of
the SMA runs from within the GIS environment. Irrespective of the choice of soil moisture
accounting model, GeoSFM generates a series of ASCII files containing the surface, interflow,
baseflow, and percolation fluxes as well as the soil moisture storage. The output files are formatted
for ingestion by the channel routing routines in GeoSFM. The features of the models are described
below.
Linear Soil Moisture Accounting (LSMA) Routine

The LSMA routine is a bucket model in which surface runoff is generated with a partial
contributing area formulation, interflow is generated from the bucket, and baseflow is generated
from an unbounded storage below the bucket. The soil layer acts as a single control volume with
rainfall as the only input, and evapotranspiration, surface runoff, subsurface runoff, and deep
groundwater percolation as outputs. The relative simplicity of this model makes it easy to
implement and parameterize in a wide variety of settings with little or no calibration. However, this
model is less well suited to high spatial or temporal resolution applications or to settings where a
model well calibrated with observed data is required. In such settings, the nonlinear module may be
a better option.
Table 3. List of key variables for the Linear Soil Moisture Accounting routine.
Item
B
T
R
STARTDAY
STARTYR
LDAY
LYEAR
INIFRACTION
BALMODE
INTERVAL
INHOURS
Type
Integer
Integer
Integer
Real
Real
Integer
Integer
Real
Real
Real
Real
HSLOPE
KS
TOPSOIL
Real
Real
Real
PANCOEF
Real
IMPRATIO
SOILLAG
GWLOSSTOTAL
GWLOSSFACTOR
GWLAG
Real
Real
Real
Real
Real
STMAX
STTEMP
AEVAL
REVAL
Real
Real
Real
Real
Variables
Description
Number of subbasins
Simulation time steps
Response time steps
Simulation start day
Simulation start year
Simulation end day
Simulation end year
Infiltration fraction
Run initialization mode
Interval
Simulation time step duration in
hours
Hill slope
Hydraulic conductivity
Topsoil as fraction of full soil
layer
Evapotranspiration Pan
coefficient
Impervious cover ratio
Interflow residence time in days
Total groundwater loss
Groundwater loss factor
Groundwater residence time in
days
Maximum storage
Temporary storage
Evapotranspiration
Residual evapotranspiration
12
Comments
Source = balparam.txt
Source = basin.txt
Source = basin.txt
Source = basin.txt
Source = basin.txt
Source = basin.txt
Source = basin.txt
Source = basin.txt
Source = basin.txt
Source = basin.txt
Source = Computations
RESPLIST
BASLIST
TIMELIST
WORKDIR
Character
Integer
Integer
Character
RAIN(B,T)
EVAP(B,T)
BASIN(B,12)
RESPONSE(B,R)
LOCALFLOW(B,T)
AEVAP(B,T)
GWLOSS(B,T)
SURP(B,T)
EXCESSFLOW(B,T)
BASEFLOW(B,T)
INTFLOW(B,T)
STORE(B,T)
GWSTORE(B,T)
REVAP(B,T)
Real
Real
Real
Real
Real
Real
Real
Real
Real
Real
Real
Real
Real
Real
RAINFN
EVAPFN
BASINFN
RESPFN
SURPFN
Input
Input
Input
Input
Output
Response file header

Basin file header
Time step listing
Working directory
Arrays
Mean Areal Precipitation
Mean Areal Evapotranspiration
Subbasin Characteristics
Unit Hydrograph File
Local contribution to flow
Actual evapotranspiration
Water loss to groundwater
Total runoff generated daily
Surface runoff contribution
Baseflow runoff contribution
Interflow runoff contribution
Total soil water storage
Groundwater store
Residual Evapotranspiration
Input-Output Files
Subbasin mean areal rainfall file
Mean areal evapotranspiration file
Subbasin characteristics file
Subbasin unit hydrograph file
Total runoff yield file
STOREFN
AEVAPFN
Output
Output
Soil water storage file

Actual evapotranspiration file
GWLOSSFN
OUTSWFN
LOGFN
Output
Output
Output
Regional groundwater loss file

Final soil moisture file
Runtime messages and errors file
EXCESSFN
Output
Excess precipitation file
BASEFN
INTFN
PARAMFN
BALFILESFN
DEFAULTFN
Output
Output
Input
Input
Output
Baseflow file
Interflow file
Listing of simulation parameters
Listing of simulation files
Default methods file
Source = response.txt
Source = basin.txt
Source = rain.txt
Source = balfiles.txt
Input from rain.txt
Input from evap.txt
Input from basin.txt
Input from response.txt
Output to localflow.txt
Output to aevap.txt
Output to gwloss.txt
Output to basinrunoffyield.txt
Output to excess.txt
Output to baseflow.txt
Output to interflow.txt
Output to soilwater.txt
Temporary Variable
Temporary Variable
Default = Rain.txt, Source = balfiles.txt
Default = Evap.txt, Source = balfiles.txt
Default = Basin.txt, Source = balfiles.txt
Default = Response.txt, Source = balfiles.txt
Default = Basinrunoffyied.txt, Source =
balfiles.txt
Default = Soilwater.txt, Source = balfiles.txt
Default = Actualevap.txt, Source =
balfiles.txt
Default = Gwloss.txt, Source = balfiles.txt
Default = Cswater.txt, Source = balfiles.txt
Default = Logfilesoil.txt, Source =
balfiles.txt
Default = Excessflow.txt, Source =
balfiles.txt
Default = Baseflow.txt, Source = balfiles.txt
Default = Interflow.txt, Source = balfiles.txt
Default = Balparam.txt, Fixed File Name
Default = Balfiles.txt, Fixed File Name
Default = Default.txt, Fixed File Name
The FORTRAN language script, 1DBalance.for, computes the LSMA as documented

below. The listing of variables and input and output files used in this script are presented in table 3.
At the start of a soil moisture accounting simulation, initiate values of storage state variables are
required for each subbasin. The maximum storage of each subbasin, STMAX(B), is computed by
multiplying its soil water holding capacity, WHC(B), by the depth of the soil column,
SOILDEPTH(B), as shown in equation 2. A conversion factor of 0.01 is applied to obtain a result
in mm because WHC(B) is in mm/m and SOILDEPTH(B) is in cm in the basin file (basin.txt).
STMAX(B) WHC(B) SOILDEPTH( B) 0.01
(2)
An initial estimate of moisture storage at the beginning of the simulation run is required. In
GeoSFM, a user-supplied estimate of initial soil moisture fraction is used to make this
determination. A software default of 0.1 (implying 10 percent moisture content) is provided, but the
13
user may change this estimate by providing a value between 0 and 1 prior to initiation of the
simulation run. Given this fraction, the initial soil moisture storage is computed by multiplying
maximum storage by the infiltration fraction (equation 3).
STORE(B,0) STMAX(B) INIFRACTIO N
(3)
Note that a single value of initial fraction is currently applied throughout the simulation
area. Initial storage in the groundwater reservoir, GWSTORE(B,0), is assumed to be zero. Some
units are converted during model initiation to ensure consistency of computations. For example, the
saturated hydraulic conductivity of the soil in each subbasin, KS(B), is provided in the basin file,
basin.txt, in cm/h. It is converted to m/day by multiplying by 0.024. Likewise, the Hill Slope
parameter provided in the basin file is converted from a percent slope to a fraction by dividing by
100.
A fraction of incident rainfall becomes surface runoff before entering the control volume.
The magnitude of this fraction is directly proportional to the ratio of the fraction of the subbasin
that is impervious. The partial contributing area fraction (PARATIO) is updated at every simulation
time step based on total soil moisture storage at the beginning of the time step by assuming a linear
relationship between the depth of water in subsurface storage exceeding the soil depth and the
partial contributing area, subject to a maximum of 1.0, as shown in equation 4.
GWSTORE(B, T - 1) - SOILDEPTH
Max
HILLSLOPE HILLLENGTH
PARATIO(B, T) = MIN
(1 - IMPRATIO)

, 0

(4)
This relationship is based on two assumptions. The first assumption that the land surface is
V-shaped allows the hill slope and hill length to be used in determining the average elevation of the
watershed divide above the river. The second assumption is that the groundwater table is directly
below the soil layer used for moisture accounting. Hence, by comparing the average elevation with
the subsurface storage less than the soil depth, it is possible to determine what portion of the
subbasin surface area is being saturated by subsurface storage.
Once impervious area ratio has been updated, the excess precipitation (EXCESSRAIN) is
computed as the fraction of rainfall (RAIN) landing on the permanent impervious area and the
partial contributing area created by the saturation from subsurface storage. The remainder of the
rainfall is allowed to enter the soil layer where it either meets the evapotranspiration requirements
or augments the subsurface reservoir, STORE(B,T), from which interflow is generated. Additional
excess precipitation is generated when the soil storage capacity, STMAX(B), is exceeded, as
presented in equation 5.
STORE(B, T - 1) RAIN(B, T) - STMAX(B)
EXCESSRAIN(B, T) MAX
RAIN(B, T) PARATIO( B, T) IMPRATIO
(5)
Potential evapotranspiration is first adjusted by a pan coefficient (PANCOEF) to account

for known biases in satellite or pan-based estimates supplied in the evap.txt file. Where no known
biases exist, a pan coefficient of 1 is used. Values of pan coefficients for each subbasin are
14
specified in the basin characteristic file, basin.txt. The conversion of potential to actual
evapotranspiration is based on the availability of moisture from rainfall, soil moisture, or runoff
sources to satisfy this demand (equation 6).
EVAP(B, T)
AEVAP(B, T) MIN
STORE(B, T - 1) RAIN(B, T) - EXCESSRAIN(B, T)
(6)
To ensure conservation of mass is maintained throughout the computation, a temporary

storage term, STORETEMP, is used to track the moisture remaining in storage after each flux
extraction or addition during the simulation time step, as shown in equation 7.
STORETEMP STORE(B, T - 1) RAIN(B, T) - EXCESSRAIN (B, T) - AEVAP(B, T) (7)
A first-order exponential loss function is used to compute the amount of water percolating
from the bounded soil bucket, BFLOSS(B,T). The loss rate is computed based on storage, soil
depth, and the saturated hydraulic conductivity (equation 8).
STORETEMP KS(B)
KS(B)
BFLOSS(B, T)
exp
SOILDEPTH(B)
SOILDEPTH(B)
(8)
The water enters the groundwater storage from which baseflow is generated. After this
extraction, the residual storage within the soil layer is updated to reflect current moisture conditions
(equation 9).
STORE(B, T - 1) RAIN(B, T) - EXCESSRAIN (B, T)
STORETEMP
- AEVAP(B, T) - BFLOSS(B, T)
(9)
Percolation to regional groundwater is similarly computed using a linear loss function but
with the groundwater residence time supplied in the basin file (equation 10).
GWSTORE(B, T - 1) BFLOSS(B,T)
1
GWLOSS(B,T)
exp
(10)
GWLAG(B)
GWLAG(B)
A linear reservoir formulation is used to compute the amount of water from the bounded
soil bucket that is released to the surface water network of rivers and lakes as interflow,
INTERFLOW(B,T). The residence time required for this computation is supplied in the basin file
as the interflow lag time (equation 11).
STORETEMP
1
INTERFLOW(B, T)
exp
INTLAG(B)
INTLAG(B)
(11)
After estimating interflow, residual moisture in the soil layer is computed for use in the next
simulation time step, using equation 12:
15
STORE(B, T - 1) RAIN(B, T) - EXCESSRAIN(B, T)

STORE(B, T)
- AEVAP(B, T) - BFLOSS(B, T) - INTERFLOW(B, T)
(12)
Baseflow contribution to surface runoff is also estimated with a linear reservoir function. As
with the interflow, the residence time is supplied in the basin file as the groundwater lag time
(equation 13).
GWSTORE(B,T - 1) BFLOSS(B,T) - GWLOSS(B,T)
(13)
1
BASEFLOW(B, T)
exp
GWLAG(B)
GWLAG(B)
After estimating interflow, residual moisture in the groundwater reservoir is computed for
use in the next simulation time step (equation 14).
GWSTORE(B,T - 1) BFLOSS(B, T)
GWSTORE(B,T)
- BASEFLOW(B, T) - GWLOSS(B,T)
(14)
The moisture accounting for the time step ends with the fluxes and storage computed. The
entire processing sequence is repeated for subsequent simulation time steps until the final rainfall
input has been processed. An additional 100 time steps of simulations is performed assuming no
additional rainfall input. The additional simulations allow runoff to be generated from the moisture
in soil storage for flow forecasting applications. The resulting moisture fluxes and storage are
written to ASCII files for postprocessing or ingestion into the flow routing modules.
Nonlinear Soil Moisture Accounting (NSMA) Routine

The NSMA module provides a more complete representation of subsurface processes by
creating separate soil layers within which interflow and baseflow processes occur. The SCS runoff
curve number method, which takes into account vegetation, soil type, and antecedent moisture, is
used for separating runoff into surface and interflow components, while the Green-Ampt equation
(Green and Ampt, 1911) is used for extracting water from the interflow soil layer to feed the
baseflow soil layer. In addition to the parameters from the single-layer model, parameters
describing the fraction of the soil column that is attributable to the respective interflow and
baseflow layers, SCS runoff curve numbers, and Green-Ampt parameters are required for the twolayer model. The superior representation of land surface and soil moisture processes makes the
two-layer model better suited for application in areas where finer resolution data is available to
support its parameterization. The variables used in this script can be grouped under three broad
categories: input variables, model state variables, and model output variables, as presented in table
4.
Table 4. List of key variables for the Nonlinear Soil Moisture Accounting routine.
Variable Group
PERIOD
Variable Name
DAY
DOY
YEAR
MONTH
HOUR
Type
Input
Integer
Integer
Integer
Integer
Integer
Description
Day of the month
Julian day of the year
Year of the dam release
Month of the dam release
Hour of the release
16
SOIL
LANDFORM
LANDUSE
METEOROLOGY
MODELPARAM
PROGNOSTIC
RUNOFF
SWCX
SWCM
SWCF
WHC
USD
LSD
TSD
Ks
Bp
GAMMA
TEXTURE
HLSLP
AREA
UPAREA
CN
XIA
MIA
PRECP
Float
Float
Float
Float
Float
Float
Float
Float
Float
Float
Integer
Float
Float
Float
Float
Float
Float
Float
PET
Float
Kc
LAGSTR
LAGGWT
LAGIRF
LGRW
Float
Float
Float
Float
Float
GUH
MOISTDISTR
Float
Float
RMODE
Integer
NDAY
USWC
dUSWC
LSWC
dLSWC
X
FGWS
Integer
State
Float
Float
Float
Float
Float
Float
SGWS
Float
INTERFLOWSTOR1
INTERFLOWSTOR2
Float
Float
BASINIMPERVAREA
SURFACEWATER
EXPRP
DRNFF
Float
Float
Output
Float
Float
IFLW
GWL
FBS
SBS
Float
Float
Float
Float
Saturation soil water content (mm)

Minimum soil water content (mm)
Soil water content at field capacity (mm)
Water holding capacity (mm)
Upper soil layer depth, percent of the total active soil
Lower soil layer depth, percent of the total active soil
Total soil, upper and lower, layer depth (cm)
Saturation soil hydraulic conductivity (cm)
Bubbling pressure (m)
Pore-size index
Soil texture; 1 = sandy soil, 2 = loam, and 3 = clay
Mean slope of the basin
Basin area (km)
Basin outlet upstream watershed area (km)
Basin mean SCS runoff curve number
Maximum fraction of the basin area that can be impervious
Fraction of the basin area that is permanently impervious
Time series of observed basin daily mean areal rainfall
(mm)
Time series of observed basin daily evapotranspiration
(mm)
Crop coefficient, between 0 and 1
Linear reservoir routing constant for overland runoff
Linear reservoir routing constant for groundwater runoff
Linear reservoir routing constant for interflow runoff
Fraction of the baseflow loss to the regional groundwater
flow
Unit hydrograph parameters
Represents the degree of areal nonuniform distribution of
tension water capacity, Xinanjiang model
Excess rainfall calculation mode, 1 = Philips equation, 2 =
SCS CN method, and 3 = excess moisture to soil water
deficit
Maximum convolution days for the unit response function
Upper soil water content (mm)
Change in upper soil water content (mm)
Lower soil water content (mm)
Change in lower soil water content (mm)
Retention rainfall (mm)
Conceptual reservoir for 1st groundwater reservoir level
(mm)
Conceptual reservoir for 2nd groundwater reservoir level
(mm)
1st conceptual reservoir for soil moisture storage (mm)
2nd conceptual reservoir to describe soil moisture storage
(mm)
Percent of basin area impervious at any time
Excess precipitation of the past days (mm)
Excess rainfall (mm)
Surface runoff from excess rainfall and impervious area
(mm)
Interflow (mm)
Losses to the regional groundwater flow (mm)
Fast baseflow flux (mm)
Slow baseflow flux (mm)
17
STATISTICS
TOTALFLUXES
TRFLW
AET
FLOW [12] [500]
DAYS [12]
SUMET
SUMRAIN
SUMRUNOFF
SUMBASEFLOW
SUMLOSSGROUND
SUMINTERFLOW
Float
Float
Float
Integer
Float
Float
Float
Float
Float
Float
Total local runoff at the basin outlet (mm)

Actual evapotranspiration (mm)
Streamflow sorted by month (m3/sec)
Number of days in each month of the runs
Total Evapotranspiration
Total precipitation
Total runoff
Total baseflow
Total groundwater loss
Total interflow
Excess precipitation (EXPRP) is computed based on precipitation input (PRECP) using one
of three infiltration equations (Philips infiltration equation, the SCS curve number method, and a
simple bucket model) depending on what is specified in the basin file. The default is the SCS curve
number method. At each time step, the default curve number is adjusted based on the percentage
saturation of the upper soil layer ( SWCX UpperLayer ), as shown in equation 15 (Woolhizer, 1976).
1.95 CN 0 2.718282 (-0.00663 CN 0 )
CN
CN 0
0.39 CN 2.718282 (0.009 CN 0 )
0
(0.9 SWCX UpperLayer 1)
(0.6 SWCX UpperLayer 0.9)

(0 SWCX UpperLayer 0.6)
(15)
The adjusted curve number is used to generate excess precipitation when the daily rainfall
(PRECP) is higher than 20 percent of SCS runoff, using equation 16.
2
1000
PRECIP 0.2
10
CN
RUNOFF.EXPRP =
1000
PRECIP 0.8
10
CN
(16)
To simulate variable source area for surface runoff abstraction, the impervious area is
allowed to vary during each time step between a minimum that equals the stream and open water
areas and a predetermined maximum. Impervious runoff area (BASINIMPERVAREA) is
computed as a function of the upper zone water content (USWC). If the USWC is less then 90
percent of the field capacity of the topsoil ( SWCFUpperLayer ), then the impervious runoff generating
area is equal to the minimum impervious area in the basin (MIA). Otherwise, the impervious runoff
area is a function of the maximum impervious area (XIA), the saturation capacity of the topsoil
( SWCX UpperLayer ), and the USWC, as shown in equation 17.
SWCUpperLayer - (0.9 SWCUpperLayer USD)

(17)
BASINIMPERVAREA = MIA + ( XIA - MIA)
SWCX
(0.9
SWCF
)
UpperLayer
UpperLayer
Evapotranspiration from the upper soil layer is limited to 20 percent or less of the potential
evapotranspiration. When the water content in the upper soil layer is less than 70 percent of field
capacity, the evapotranspiration is further reduced by a ratio of the current and maximum residual
water holding capacity (equation 18).
18
AETUpperSoil
(USWC - (SWCM USD)) PET 0.2

(0.7 SWCF USD) - (SWCM USD)
=
PET 0.2
USWC 0.7
(18)
USWC 0.7
After the actual evapotranspiration is removed from the upper layer, the soil water content
in the layer is updated by adding nonexcess precipitation and subtracting evapotranspiration
(equation 19).
USWC = USWC PRECP - AETUpperSoil - EXPRP
(19)
Validation checks are included to ensure that USWC is greater than or equal to the upper
soil layer wilting point but less than its field capacity. Any excess moisture above the upper soil
layer field capacity becomes available for infiltration or percolation using equation 20.
USWC (WHC USD )
ESWUpperLayer =
0
(USWC ( SWHC USD ))
(20)
(USWC (WHC USD ))
Water in the topsoil layer exceeding the upper soil layer field capacity is available for either
interflow or percolation. The rate at which water moves out of the upper soil layer is characterized
by the sorptivity (SORP), which is computed as a function of soil properties and water content
(equation 21) (Brooks and Corey, 1964).
2 3 GAMMA
SORPUpperLayer = 2 K s (SWCX UpperLayer - USWC)
(21)
(1 3 GAMMA) Bp 0.5
The fractions that actually will be interflow (equation 22) and percolation (equation 23) are
computed based on sorptivity SORP (from Green-Ampt), hydraulic conductivity KS, and basin
slope HlSlp. In general, basins with little slope have only a small fraction of interflow and a high
fraction of percolation. In most cases, the fraction for percolation will be larger than the fraction for
interflow because slopes are typically much smaller than 45o (both fractions will be 0.5 for a slope
of 45).
sin(HlSlp)
INTERFLOWUpperLayer =
ESWUpperLayer
( sin(HlSlp) ) + ( cos(HlSlp) )
cos(HlSlp)
PERCOLATION UpperLayer =
ESWUpperLayer
( sin(HlSlp) ) + ( cos(HlSlp) )
(22)
(23)
Transpiration from the lower soil layer amounts to 80 percent or less of the potential
evapotranspiration (PET). The potential evapotranspiration is further reduced by the ratio of current
to maximum available water content in the lower layer if the current lower soil water content
(LSWC) is less than 70 percent of field capacity (equation 24).
19
(LSWC - (WHC LSD) PET 0.8

(0.7 SWCF LSD) - (SWCM LSD)
AETLowerSoil =
PET 0.8
LSWC 0.7
(24)
LSWC 0.7
The lower soil moisture content is updated by adding percolating flows from the upper layer
and subtracting transpiration from the lower soil layer, with equation 25.
LSWC = LSWC PERCOLATIO N UpperLayer - AETLowerSoil
(25)
As with the upper layer, validation checks are used to ensure that lower soil layer moisture
content does not fall below its soil wilting point but stays below field capacity. Any moisture in
excess of field capacity of the lower soil layer goes into one of three storage components:
groundwater, fast interflow storage, and slow interflow storage. A simple groundwater loss fraction
(LGRW) is used to determine how much water of the excess is released to increase regional
groundwater flow (equation 26). The loss only occurs when water content in the lower soil layer
exceeds its water holding capacity. For time periods when no excess is available, the groundwater
storage remains unchanged.
((LSWC - (WHC LSD)) LGRW)
GWSTORE GWSTORE MAX
(26)
Of the remainder, 25 percent enters a slow response interflow storage

( INTERFLOWSTOR Slow
LowerLayer ), while the remaining 75 percent enters a fast response interflow
storage ( INTERFLOWSTOR Fast
LowerLayer ), as shown in equations 27 and 28.
Fast
INTERFLOWSTOR Fast
LowerLayer INTERFLOWSTOR LowerLayer (EWCLowerLayer - GWLOSS) 0.75 (27)
Slow
INTERFLOWSTOR Slow
LowerLayer INTERFLOWSTOR LowerLayer (EWCLowerLayer - GWLOSS) 0.25 (28)
Water is gradually released from the two interflow and baseflow reservoirs for transfer to
the nearest stream, downstream through the stream network, and eventually to a terminal point in
an ocean or inland lake. The rate of release of water from each of these reservoirs is governed by a
linear response function with residence times obtained from the input river characteristics file. The
responses for the fast and slow interflow and baseflow reservoirs are given in equations 29, 30, and
31.
Fast
INTERFLOWLowerLayer
(INTERFLOWSTOR Fast
LowerLayer / LAGIRF)
(29)
Slow
INTERFLOWLowerLayer
(INTERFLOWSTOR Slow
LowerLayer / (0.5 LAGIRF) )
(30)
BASEFLOW ( GWSTORE / LAGGW T )
(31)
20
Descriptions of the routines available in GeoSFM for river transport are provided in the
following sections.
River Transport Module

After runoff generated within a given catchment travels to the catchment outlet, it enters the
river network and works its way downstream to the basin outlet. GeoSFM supports two linear
routines (pure lag and the diffusion analog) and one nonlinear method, (the Muskingum Cunge) for
simulating the in-channel phase of flow.
Pure Lag Flow Routine

Lag routing is the simplest of the river transport approaches. It accounts for the advection of
flow but does not include any attenuation or deformation of the input. Consequently, the input flow
remains unaltered in magnitude at the discharge point, and a single flow parameter, travel time,
between the input and discharge locations is required for the computation. The FORTRAN script,
Lagroute.for, retrieves runoff generated by the soil water balance in each subbasin and routes it
through the river network by pure translation as documented below. The variables, input files, and
output files used in this program are presented in table 5.
Table 5. List of key variables for the pure lag flow routine.
Name
Type
B
T
R
F
STARTDAY
STARTYR
LDAY
LYEAR
INHOURS
Integer
Integer
Integer
Integer
Real
Real
Integer
Integer
Real
AREA
UPAREA
Real
Real
DOWNID
Integer
RIVLENGTH
RIVMANNING
Real
Real
RIVERLOSS
RIVERPLOSS
HASDAM
CELERITY
Real
Real
Integer
Real
SURP(B, T)
RIVER(B, 12)
RESPONSE(B, R)
STREAMFLOW(B,
T+F)
LOCALFLOW(B,T+F)
Real
Real
Real
Real
Description
Variables
Number of subbasins/rivers
Response time steps
Forecast time steps
Simulation end day
Simulation end year
Simulation time step duration in
hours
River reach subbasin area
Total drainage area at river reach
outlet
Subbasin downstream of river
reach
Length of the river reach
Manning roughness coefficient for
reach
In-Streamflow loss coefficient
Overbank flow loss coefficient
River reach dam flag
Flow celerity for river reach
Arrays
River Characteristics File
Streamflow File
Real
21
Comments
Source = routparam.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Input from basinrunoffyield.txt
Input from river.txt
Output to streamflow.txt
INFLOW(B, T+F)
FORECAST(B, F)
Real
Real
SURPFN
Output
UpStreamflow into a subbasin

Streamflow Forecast File
Input and Output Files
RIVERFN
RESPFN
Input
Input
River characteristics file

RESERVOIRFN
Output
File containing reservoir listing
STOREFN
Output
AEVAPFN
Output
GWLOSSFN
OUTSWFN
Output
Output

LOGFN
Output
Runtime messages and errors file
EXCESSFN
Output
BASEFN
Output
Baseflow file
INTFN
Output
Interflow file
PARAMFN
BALFILESFN
DEFAULTFN
Input
Input
Output

Output to inflow.txt
Output to forecast.txt
balfiles.txt
Default = River.txt, Source = balfiles.txt
Default = Response.txt, Source =
balfiles.txt
balfiles.txt
Default = Soilwater.txt, Source =
balfiles.txt
balfiles.txt
Default = Cswater.txt, Source =
balfiles.txt
balfiles.txt
balfiles.txt
Default = Baseflow.txt, Source =
balfiles.txt
Default = Interflow.txt, Source =
balfiles.txt
In lag routing, inflow into the upstream end of a river reach is translated with a time delay
but without change in amplitude to the downstream end of the same reach. The flow is then
transferred to the next downstream river reach. This routing approach is easily implemented in data
sparse settings because it only requires a single parameter, the lag time. Mathematically, lag routing
can be expressed as equation 32.
Q (t ) I (t t ' )
(32)
Where Q(t) is the discharge at time, t

t is the travel time between input and discharge locations
I(tt) is the input at time (tt)
The routing process begins by reading in the runoff generated within the subbasin
associated with each river reach from the runoff yield file (SURPFN) into the runoff yield array
(SURP(B, T)). A set of parameters associated with each river reach is also read from the river
characteristics file (RIVERFN) into the river characteristics array (RIVER(B, 12)). The
characteristics contained in this array include RIVERID of the next downstream reach, indicators of
the presence of dams along the reach. If a dam is present in the basin, additional filesa dam link
file and a reservoir characteristics file, or a file containing prespecified discharge valuesare
opened for each dam. The surplus runoff from the various sources, including excess precipitation,
interflow, and baseflow, are aggregated (equation 33).
SURP(B, T) EXCESSRAI N(B, T) INTERFLOW( B, T) BASEFLOW(B , T)
22
(33)
The surplus runoff is converted from depth units such as millimeters to flow units such as
cubic meters per second using equation 34.
INRUNOFF(B, T)
SURP (B, T) BAREA POLYLOSS

1000 3600 STEPSIZE
(34)
The result is convolved with the basin response function and transferred to the subbasin
outlet as the local contribution to river flow (equation 35).
LOCALFLOW (B, T R) INRUNOFF(B, T) * RESPONSE(B, R)
(35)
where * is the convolution integral

UpStreamflow contribution entering a river reach must also be translated to the downstream
end of the reach and added to the local flow contribution, allowing for translational losses along the
river reach. For reaches without dams, this translation can be represented using equation 36.
OUTFLOW (B, T) LOCALFLOW(B, T) (INFLOW(B, T - T' ) RIVLOSSCOEFF) (36)
If the river reach contains a DAM, the preexisting reservoir water storage is updated by
adding the upStreamflow. Reservoir discharge is then computed from the updated storage by a
user-defined storage function (equation 37).
OUTFLOW (B, T) LOCALFLOW(B, T)
DAMSTORE(B, T - 1 INFLOW(B,T - T' ) ) (37)

T
The discharge from the river reach is transferred to the next downstream reach, D, as
presented in equation 38.
INFLOW (D, T) OUTFLOW(B,T)
(38)
where n is the number of reaches immediately upstream of reach D.

The discharge from each river reach is then written to the streamflow file.
Diffusion Analog Routine

The diffusion analog method is another linear transport routine available in GeoSFM. This
method goes beyond the lag routing by accounting for both flow advection (using flow time or
celerity) and attenuation (using a flow dispersion coefficient). In fact, the diffusion analog equation
is the linear solution of the advection-dispersion equation (also known as the Navier-Stokes
equation) for a plane rectangular source (Maidment, 1993). Mathematically, the diffusion analog
equation can be expressed as equation 39.
x
Q (t ) I (to )
4 D (t t ) 3
o
exp (V (t to ) x )
4 D (t to )
23
(39)
where
D is the dispersion coefficient in m2/s
V is the flow celerity in m/s
x is the length of the river reach in m
to is the time of the input event in seconds
i is the present time in seconds
Q(t) is the discharge at the downstream end of the river reach
I(to) is the inflow at the upstream end of the river reach
In its implementation in GeoSFM, the diffusion analog routine is essentially the lag routing
with additional functionality to account for the deformation or attenuation of the input. The
attenuation is controlled by a time invariant parameter called the dispersion coefficient.
Consequently, one parameter controls how quickly water travels through the river reach while a
second parameter controls the spreading that occurs within the reach. The FORTRAN program,
diffroute.for, retrieves runoff generated by the soil water balance in each subbasin and routes it
through the river network by diffusion analog as documented below. The variables, arrays, input
files, and output files used in this program are presented in table 6.
Table 6. List of key variables for the diffusion analog routine.
Name
Type
B
T
R
F
D
Integer
Integer
Integer
Integer
Integer
STARTDAY
STARTYR
LDAY
LYEAR
INHOURS
AREA
UPAREA
Real
Real
Integer
Integer
Real
Real
Real
DOWNID
Integer
RIVLENGTH
RIVMANNING
RIVERLOSS
RIVERPLOSS
HASDAM
CELERITY
Real
Real
Real
Real
Integer
Real
SURP(B, T)
RIVER(B, 12)
RESPONSE(B, R)
STREAMFLOW(B,T+F)
LOCALFLOW(B, T+F)
INFLOW(B, T+F)
FORECAST(B, F)
DIFFRESP(B, D)
Real
Real
Real
Real
Real
Real
Real
Real
Description
Variables
Number of subbasins/rivers
Response time steps
Forecast time steps
Number of river response
ordinates
Simulation end day
Simulation end year
Simulation time step in hours
River reach subbasin area
Total drainage area at reach
outlet
Subbasin downstream of river
reach
Length of the river reach
Manning roughness coefficient
In-Streamflow loss coefficient
Overbank flow loss coefficient
River reach dam flag
Flow celerity for river reach
Arrays
River Characteristics File
Streamflow File
UpStreamflow into a subbasin
Streamflow Forecast File
River Response Ordinates
24
Comments
Internal computation
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Source = river.txt
Input from basinrunoffyield.txt
Input from river.txt
Output to streamflow.txt
Output to inflow.txt
Output to forecast.txt
Internally Computed
SURPFN
Output
Input and Output Files

RIVERFN
RESPFN
RESERVOIRFN
STOREFN
AEVAPFN
Input
Input
Output
Output
Output
River characteristics file

File containing reservoir listing
GWLOSSFN
OUTSWFN
LOGFN
Output
Output
Output
EXCESSFN
Output

Runtime messages and errors
file
BASEFN
INTFN
PARAMFN
BALFILESFN
DEFAULTFN
Output
Output
Input
Input
Output
Baseflow file
Interflow file

balfiles.txt
Default = River.txt, Source = balfiles.txt
Default = Response.txt, Source = balfiles.txt
Default = Reservoir.txt, Source = balfiles.txt
Default = Soilwater.txt, Source = balfiles.txt
balfiles.txt
Default = Cswater.txt, Source = balfiles.txt
balfiles.txt
balfiles.txt
Default = Baseflow.txt, Source = balfiles.txt
Default = Interflow.txt, Source = balfiles.txt
The surplus runoff from excess precipitation, interflow, and baseflow are aggregated
(equation 40).
SURP(B, T) EXCESSRAI N(B, T) INTERFLOW( B, T) BASEFLOW(B , T)
(40)
The surplus runoff is converted from depth units such as millimeters to flow units such as
cubic meters per second using equation 41.
INRUNOFF(B, T)
SURP (B, T) BAREA POLYLOSS

1000 3600 STEPSIZE
(41)
The result is convolved with the basin response function and transferred to the subbasin
outlet as the local contribution to river flow, represented with equation 42.
LOCALFLOW (B, T R) INRUNOFF(B, T) * RESPONSE(B, R)
(42)

UpStreamflow contribution entering a river reach must also be transferred to the
downstream end of the reach with a convolution integral to account for the spreading out of flow
along the reach. Since the dispersion coefficient in the diffusion analog equation is time invariant,
the ordinates associated with its response function along any particular river reach are also time
invariant. Consequently, the ordinates are estimated once at the beginning of the simulation and
applied to the flow entering the upstream end of the reach at all subsequent simulation time steps.
The resulting discharge is added to the local flow contribution after allowing for translational losses
along the river reach. For reaches without dams, this translation can be presented by equation 43.
OUTFLOW(B, T) LOCALFLOW(B, T) (INFLOW(B,T - T' ) * DIFFRESP(B, D)) RIVLOSSCOEFF (43)
25

D is the number of ordinates of the river response function
DIFFRESP(B, D) is the matrix containing the river response function ordinates
If the river reach contains a dam, the preexisting reservoir water storage is updated by
adding the upStreamflow. Reservoir discharge is computed as the updated storage by a user-defined
storage function represented by the differential term in equation 37. The discharge from the river
reach is also transferred to the next downstream reach, D, using equation 38, and the discharge
from each river reach is then written to the streamflow file.
Muskingum Cunge Routine

The Muskingum Cunge is a nonlinear, variable parameter routing method. Like the linear
Muskingum routing method, it uses the continuity equation and an empirical storage equation. It
relies on the Muskingum K coefficient (analogous to flow time) to control the rate of advection and
the Muskingum X coefficient to control the rate of attenuation, or spreading out. Cunge proposed to
amend the method by allowing the Muskingum X coefficient to vary during each time step based
on the flow at the previous time step. In effect, the rate of attenuation of flow is dependent on the
condition of flow. Mathematically, Muskingum Cunge routing can be expressed using equations 44
and 45.
dS
I Q
(44)
dt
S K XI 1 X Q
(45)
The Muskingum K and X coefficients are estimated using equations 46 and 47,
respectively.
x
c
1
Q
X
2 2c B Se x
(46)
(47)
where
K is the Muskingum K coefficient in seconds
X is the dimensionless Muskingum X coefficient
c is the average flood wave celerity in m/s
B is the average channel width in m
Se is the dimensionless friction slope
x is the length of the channel
S is the storage in the channel at time t in m/s
I is the inflow in m3/s
3
Q is the discharge in m /s
All catchments within the analysis area are parameterized for the three routing methods
during preprocessing such that the user can opt for any of the three methods just prior to execution
without additional preprocessing. The variables used in this script are described in table 7.
26
Table 7. List of key variables for the Muskingum Cunge routine.

Variable Group
PERIOD
CHANNEL
LANDFORM
RESERVOIRS
PROGNOSTIC
River Routing
STREAMFLOW
Variable Name
DAY
DOY
YEAR
MONTH
HOUR
ROUGH
RIVLENGTH
SLOPE
WIDTH
BANKFULLFLOW
LOSSFLDP
LOSSRIV
HASDAM
HASRATING
RATINGTYPE
HLLNG
AREA
UPAREA
ISREGULATE
RINFLOW
ROUTFLOW
RSTORAGE
RLEVEL
RAREA
LAGRSV
Type
Input
Integer
Integer
Integer
Integer
Integer
Float
Float
Float
Float
Float
Float
Float
Integer
Integer
DEPTH
ROUTSTOR
DSTORSTART
Integer
Float
Float
Float
Integer
Float
Float
Float
Float
Float
Float
State
Float
Float
Float
FLDSTOR1
Float
FLDSTOR2
Float
COEFF[4]
NSPACE
NTIME
Float
Integer
Integer
dx
dt
IFLOW
LAG
CELERITY
Float
Float
Float
Float
Float
Output
Float
Float
Float
Float
Float
Float
Float
Integer
OUTFLOW(T)
INFLOW(T)
LOCALFLOW
OBS
INFLOW(T-1)
OUTFLOW(T-1)
Qref
HASOBS
Description
Day of the month
Julian day of the year
Four-digit year
Two-digit month
Two-digit hour
Basin mainstream channel mean Manning n coefficient
Basin mainstream channel length (m)
Basin mainstream channel mean slope
Basin mainstream channel mean width (m)
Basin mainstream channel bankfull flow (m3/sec)
Floodplain to regional groundwater water loss fraction
In-stream infiltration and evaporation water loss fraction
Flag for presence of a dam in river reach, 1 = yes, 0 = no
Flag for presence of rating curve for river reach, 1 = yes, 0
= no
Type of rating curve, 1 = exponential, 2 = polynomial
Mean length of the mainstream tributaries (m)
Basin area (km2)
Basin outlet upstream watershed area (km2)
Flag indicating if the reservoir is regulated, 1 = yes, 0 = no
Reservoir inflow (m3/sec)
Reservoir outflow (m3/sec)
Amount of water storage in the reservoir (m3)
Reservoir water level (m)
Reservoir surface area (km2)
Reservoir flow lag time (sec)
Depth of the water in mainstream channel (m)
conceptual channel reservoir storage (m3/sec)
Initial conceptual channel reservoir storage capacity
(m/sec)
1st conceptual reservoir to describe flood plain storage
(mm)
2nd conceptual reservoir to describe flood plain storage
(mm)
Muskingum Cunge routing scheme coefficients
Number of Muskingum Cunge routing segments in a reach
Number of Muskingum Cunge routing interval per time
step
Routing segment length for Muskingum Cunge scheme
Routing interval for Muskingum Cunge flow segment
Initial streamflow for Muskingum Cunge flow segments
Travel for water passing through channel segment
celerity of the flow
present streamflow at the subbasin outlet (m3/sec)
present streamflow at the subbasin inlet (m3/sec)
Flow generated within the local subbasin (m3/sec)
observed streamflow (m3/sec)
past streamflow at the basin headwater inlet (m3/sec)
past streamflow at the basin headwater outlet (m3/sec)
Reference flow or theoretical bankfull flow (m3/sec)
flag that indicates if basin stream has observed data
27
RATING
RATPARAM[3]
HMAX
Float
Float
Parameters of the rating curve

Bankfull stage (m)
A theoretical bankfull flow is estimated from upstream contributing area for each river to
allow for the separation of overbank flow from channel flow. The relationship is presented as
equation 48.
BANKFULLFLOW = 36.0 (UPAREA / 2)0.68 0.02832
(48)
While the input runoff data have a predefined time interval, subscale timing and routing
intervals must be defined for the Muskingum Cunge routing scheme to ensure numerical stability of
computations. The space and time subinterval are based on reference flow and celerity (from the
river file) and fixed throughout the simulation, even though the actual flow celerity varies during
each simulation time step. The size of the time subinterval is computed as the smaller of the time it
takes for the kinematic wave to travel through the river reach or one-sixth of the input data time
interval (equation 49).
RIVLENGTH / WAVECELERITY
6.0
t = MIN
INTERVAL
6.0
(49)
The routing space subinterval is similarly estimated as equation 50.
(0.6 t WAVECELERITY)
MAX
Q ref
WIDTH
SLOPE
WAVECELERI
TY
(50)
x = MIN
Q ref
WIDTH SLOPE WAVECELERITY
During each time step, a new wave celerity is computed based on the water surface slope
and shape during the preceding routing time subinterval (equation 51).
0.5
Q d ROUGH
SLOPE
WAVECELERITY =
1.6666667
ROUGH
SLOPE WIDTH
28
0.4
(51)
Muskingum K and X parameters are also recomputed to reflect the new celerity using
equations 52 and 53, respectively.
MUSKK =
x
WAVECELERITY
MUSKX = 0.5
(52)
(1 - (Qd / WIDTH) )
WAVECELERITY SLOPE x
(53)
After calculating K and X parameters, Reynolds (D) and courant (C) numbers are also
calculated to verify numerical stability with equations 54 and 55, respectively.
REYNOLDSD = (1 - 2 MUSKX)
COURANTC =
(54)
WAVECELERITY t
x
(55)
If the sum of these two numbers (REYNOLDSD + COURANTC) is less than 1.05 or if
COURANTC is greater than 2, then the routing time and space subinterval (dt and dx) are adjusted
until the two conditions are met. For propagation of flow within the river channel, a set of
Muskingum coefficients are computed using the four coefficients in equations 55A, 55B, 55C, and
55D.
COEF0 =
t - (2.0 * MUSKK * MUSKX)

2 * MUSKK * (1 - MUSKX) t
(55A)
COEF1 =
t (2.0 * MUSKK * MUSKX)

(55B)
COEF2 =
(55C)
COEF3 =
( Qlocal / RIVLENGTH ) * x * t
(55D)
The coefficients are applied to the flow at the upstream section for the current and previous
time step and to the flow at the current section for the previous time step to estimate current flow at
the current section with equation 56.
COEF0 CFLOW(nx - 1, nt)
CFLOW(nx, nt) = COEF1 CFLOW(nx - 1, nt - 1)

COEF1 CFLOW(nx, nt - 1)
(56)
The flow in the floodplain outside the river channel is simulated using two conceptual linear
reservoirs in series, the lag time for each of which is computed using equation 57.
29
0.42 ( RIVLENGTH/0.3048) ROUGH 1.2

LAGFLD 0.5
p
SLOPE0.5
0.8
(57)
Since the two reservoirs have the same residence time, their net effect can be represented by
the 2 parameter gamma distribution. Discharge from river reaches containing reservoirs is handled
in one of two ways. Where discharges are supplied through an input file, releases are made
according to such specifications. Otherwise, reservoir releases are made using a conceptual levelpool reservoir. This involves computing and applying three coefficients to current and previous
reservoir inflows and discharge with equations 58, 59, and 60.
1
LAGRSV
RCOEF1 RCOEF2
1
LAGRSV
1
LAGRSV
RCOEF3
1
LAGRSV
RCOEF1 RINFLOW(T - 1)
ROUTFLOW(T ) RCOEF2 RINFLOW(T)
RCOEF3 ROUTFLOW(T - 1)
(58)
(59)
(60)
The reservoir discharge is then added to streamflow at the downstream end of the routing
reach and propagated downstream through the stream network.
Calibration Module
GeoSFM includes an automatic calibration method intended to improve the application of
the rainfall-runoff model. Because of uncertainties in the input and output data sets, model structure
error, and the quantity or quality of data, it is difficult to determine a best set of model parameters
(Duan and others, 1992). The calibration algorithm performs a global search of the specified
parameter space, which has been shown to be effective in dealing with difficulties in searching such
as multiple local optima and derivative discontinuities. The algorithm is also efficient because it
needs relatively few model runs to obtain the optimum parameter set.
The relationship between the streamflow model and the calibration algorithm is presented in
figure 2. By the end of a successful calibration, the streamflow model will have been run many
times (iterations) while tracking information from the parameter values input for each of the model
runs. The parameter values that give the best model performance become the optimal parameter set.
The GeoSFMCalib is linked to the spatial input data through basin and river watershed attribute
files that list parameter values for each subwatershed. These attribute files (basin.txt and river.txt)
are rewritten with new parameter values each time the model is run. The GeoSFM.dll is used to call
the streamflow model. The modeled output is compared to the observations using an aggregated
statistic over the time series. The user can choose one of six statistical tests to make this
30
comparison. The statistical test used to compare the model results to the observation can also be
called an objective function. At the time of writing this report, the software connection between the
calibration results and the ArcView interface had not been completed and is shown with a dashed
connector.
Figure 2. Calibration module relationships between the Geospatial Stream Flow Model and the
Multi-Objective Shuffled Complex Evolution Metropolis algorithm.
GeoSFM
Calibration
Automatic Optim ization with the
Multi-Objective S huffled Complex
Evolution Me tropolis algorithm
(MOSCEM )
ArcView
Interface
6. View calibration
results when
parameters have
converged or max
iterations run
5. Use comparison to inform

calibration algorithm in the
search for new parameter
values
Observed
Streamflow
2.
Calibration
Algorithm
Rewrite GeoSFM inputs in

search of optimal parameter
values
Basin.txt
River.txt
GeoSFM
Input files
iterations
4. Compare modeled
output to observed
streamflow
Modeled
Streamflow
Output
1. Link spatial data to calibration

setup and calibration outputs
3. Run hydrology model
GeoSFM
A flowchart of the calibration process is presented in figure 3. The calibration process

begins with the model setup. The terrain analysis extracts model parameters and creates basin.txt
and river.txt watershed attributes. Copies of basin.txt and river.txt are saved under different names
(basin_original.txt and river_original.txt) because the calibration process overwrites the former pair
of files with each program iteration. Next, the user must assess the uncalibrated model. The best
calibration results can only be achieved with realistic parameter inputs and an appropriate model
structure for the watershed that is being modeled. The uncalibrated model results should be
analyzed before beginning the calibration. The program geosfmcalib.exe consists of a caller
program (dllcaller.exe) which calls the GeoSFM and runs the model one time using the basin.txt
and river.txt inputs and a statistical analysis program (UncalibStats.exe) which reports basic model
performance by comparing model output with observed streamflow. The user should also plot
observed and modeled streamflow time series to visually check and ensure that the calibration
results are reasonable. While subjective, visual analysis may help to identify large errors in input
31
data, parameter ranges, or observed streamflow, which the calibration process cannot be expected
to overcome.
Figure 3. Flowchart of the calibration process.
1. Use Terrain Analysis to

extract Model Parameters
from land surface data
2. Test Uncalibrated Model
3. Perform SA
400 model runs
4. Calibrate sensitive parameters

5,000-10,000 model runs
5. Run Postprocessing
Returns uncertainty boundaries on streamflow
and model performance statistics
6.
Calibrated model parameters

Can be used for
operational model runs
Next, sensitivity analysis should be performed. Sensitivity analysis (SA) serves the dual
purpose of testing which sensitive parameters should be used for calibration and analyzing feasible
parameter ranges. Simply put, the goal of SA is to see if there is a change in model results when the
parameter values are changed. The program geosfmsarun.exe tests 20 input parameters over userspecified ranges. A total of 400 model runs are used to complete the SA test. The SA result can be
used to test the model structure; if parameters assumed to have a strong impact on model results do
not show any sensitivity, the model structure should be reassessed. Likewise, the SA can give a
clear understanding of the feasible parameter space. If the SA results show a parameter to be too
sensitive or not sensitive at all, the parameter range may need to be refined. Getting the minimum
and maximum parameter space refined as much as possible using realistic values for the specific
watersheds being modeled is imperative for obtaining realistic calibration results.
For calibration, a subset of the 20 parameters tested in SA should be selected. For the
parameters not selected for calibration, the uncalibrated values in the watershed attribute files are
used. The program geosfmcalib.exe is the main calibration program. Although many calibration
algorithms have been developed for hydrologic modeling over the years, the MOSCEM algorithm
has been chosen for its ability to consistently find optimum parameter sets with the least number of
model runs. The number of model runs necessary to converge to the optimum parameter values
32
depends on the length of streamflow record, the number of parameters being tested, and the
complexity of the relationship between model parameters and the watershed response being
modeled. The range will generally be 5,00010,000 model runs to converge to a global minimum.
One-At-a-Time (OAT) Sensitivity Analysis

The sensitivity analysis method is intended to give an efficient look at model parameter
sensitivity so that sensitive parameters can be chosen for calibration in the least number of model
runs. The One-At-a-Time (OAT) method (Morris, 1991) is a local test because in each model run,
only one parameter is changed while all other parameters are held constant. With this method,
changes in the output for each model run can be unambiguously attributed to the change of the
parameter. To ensure that parameters are tested over the full range, parameter values are taken at a
20 equal interval sample for each of the 20 parameters. This sampling results in a total of 400
model runs. This efficient method for this application ensures the full range of parameters is tested
while attributing parameter effects on simulations with relatively few model runs. However, this
approach does not examine interactions among parameters.
Because the GeoSFM has spatially distributed parameter information, care was taken to
retain the spatial pattern of parameter values derived from land surface data and terrain analysis.
The minimum and maximum ranges to test for each parameter (given in range.txt) and the basin
average value for each parameter (derived from subbasin parameter values in basin_original.txt and
river_original.txt) are used to derive minimum and maximum multiplier values. Values at equal
intervals between the minimum and maximum multiplier values for each parameter are used to
write model inputs (basin.txt and river.txt) one parameter by one interval at a time. Intervals are
calculated using a fraction (f) of the distance between the minimum and maximum parameter
multiplier value. The range of multiplier values used to define parameter search space is calculated
with equation 61.
OATmultiplier
f Max range.txt
Max watershed
Min range.txt
Min watershed
(61)
where Min range.txt and Maxrange.txt are user-specified feasible upper and lower bounds for the
parameter space given in range.txt, and where Min watershed and Maxwatershed are the upper and lower
bounds extracted from the subbasin parameter values given in the watershed attribute files
(basin.txt and river.txt). The parameters written for each subbasin in the watershed attribute files
are calculated using equation 62.
Subbasin parameterOAT OATmultipl ier Subbasin parameteroriginal
(62)
The OAT sensitivity test changes the parameter values at all subbasins for each OAT
parameter test. For the remaining subbasin parameter values, original subbasin parameter values are
used (from basin_original.txt and river_original.txt).
For any subbasin, if the result from equation 62 is greater than the maximum set in
range.txt, the OAT parameter is set to the maximum, calculated using equation 63.
OATmultipliermax
Max range.txt
(63)
Max watershed
33
When the basin characteristics files (basin.txt and river.txt) are written, the GeoSFM.dll is
used to run the model. A model configuration consisting of one soil moisture accounting routine
(linear or nonlinear) and one river transport routine (lag, diffusion analog, or Muskingum Cunge) is
specified in an input file (whichModel.txt). For each OAT test, the model results (streamflow.txt)
are analyzed using three statistical tests (stest): maximum, mean, and standard deviation. These
tests give different information on the effect that a parameter change may have on model results.
Generally, parameter changes have the greatest numerical effect on the peaks or the maximum
streamflow modeled. The sensitivity for each of 20 parameters is calculated for three statistical
tests by averaging the equal interval model results (found in intermediate file, SensAnalysis.txt), as
shown in equation 64.
20
Stest
i 1
Stest i 1
(64)
20
Results can be charted in order to choose which parameters are most important in the
calibration (see SarunOutput.txt). Sensitive parameters may be model dependent, watershed
dependent, or time period dependent. The sensitivity must be tested while preparing to calibrate so
that resources are not wasted optimizing parameters that have little or no effect on model output.
Sensitivity analysis measures the impact on model outputs due to changes in inputs.
Sensitivity analysis gives a picture of the importance of uncertainty in model inputs and can help
refine model structure. OAT is one of the simplest of many techniques for sensitivity analysis.
Other techniques include Latin-Hypercube sampling, Monte Carlo sampling, Generalized
Sensitivity Analysis (GSA), Multi-Objective GSA (MOGSA), and the Fourier Amplitude
Sensitivity Test (FAST).
The Multi-Objective Shuffled Complex Evolution Metropolis (MOSCEM) Algorithm

The purpose of calibration is to adjust model parameters so that model results closely match
real observations. Although many GeoSFM model parameters are derived from spatially distributed
observed data, uncertainties in initial parameter estimates due to differences input data scale, errors
in observed data, and inaccuracies in the structure of the model require that parameters be adjusted.
As with all hydrologic models, parameter adjustment allows the GeoSFM model to overcome what
is unknown as well as what cannot be measured about a watershed.
Much research has been done on automatic optimization strategies. These strategies take
advantage of computers to objectively find an optimal set of parameters that minimizes the
difference between the model-simulated output and the observed output values. The alternative is to
use manual trial-and-error calibration, which is time consuming, subjective, and limited in search
capabilities to a small parameter space. The research in automatic optimization strategies seeks to
create algorithms that can consistently find the optimum parameters (effective) in as few model
runs as possible (efficient). The shuffled complex evolution (SCE) (Duan and others, 1992) is the
most popular algorithm for searching a global parameter space without having to run a random
Monte Carlo search, which would make the problem several orders of magnitude larger. Vrugt and
others (2003a) improved on the SCE by replacing the downhill simplex search aspect of the
algorithm with a Metropolis-annealing scheme and incorporating a multiobjective framework
(Vrugt and others, 2003b). The resulting MOSCEM algorithm combines complex shuffling,
controlled random search, and probabilistic search to give an estimate of probabilistic uncertainty
34
for a set of parameters that minimizes the difference between modeled and observed values, given
multiple objectives.
An important aspect of modeling hydrologic systems is capturing multiple characteristics of
the watershed. For a given model, multiple characteristics may include 1) the driven versus
nondriven portions of the hydrograph, 2) surface water, groundwater, and evapotranspiration or
other modeled fluxes, 3) streamflow at multiple locations within the watershed, or 4) characteristics
representing model performance using different statistical tests or objective functions such as RMS,
bias, or a maximum likelihood estimator. To capture more than one characteristic in an automatic
optimization, multiobjective methods have been developed (Vrugt and others, 2003b). Using
multiple sources of observed data can improve the model calibration.
Postprocessing of Calibration Results

The program GeoSFMpostproc.exe is used to process the calibration results. The
postprocessing program reads in calibration output files. The calibration algorithm does not save
the streamflow results for each of the thousands of model runs due to computer memory and space
restrictions, but it does retain the parameter values that created the best model results. The
postprocessing runs the model using the best parameter values so that the streamflow results can be
plotted and analyzed.
The calibration algorithm works on the assumption that there is no best parameter set due
to the uncertainties in model structure and data inputs. The postprocessing program runs the model
for the entire set of multiple best parameter values and returns upper and lower bounds to enable
plotting of the uncertainty boundaries. The one parameter set which shows the most balanced
tradeoff between the minimized objectives is chosen as the singular best parameter set. Six
difference model performance statistics are reported for this model run. However, this parameter
set (the final basin.txt and river.txt in the working directory) should be used with the understanding
that this is only one of many equivalent solutions for the global optimum.
Postprocessing Utilities
A number of postprocessing utilities included in the GeoSFM system to facilitate ease of
use are described in this section. Because input precipitation data are spatially distributed, it is often
difficult to fully discern the evolution of rainfall events over time. Precipitation and
evapotranspiration data may also be required for longer accumulation periods, such as 10-day and
monthly totals. GeoSFM includes a routine for vertically integrating grid layers to perform pixel
level accumulations and other temporal statistics such as maxima, minima, standard deviations,
medians, modes, variety, range, and nonzero values for each pixel. The results of the analyses are
presented as grids. While not directly related to the computation of streamflow, the spatial statistics
routines are useful for running verification checks on input data and for studying sources of
anomalously high streamflow or soil water levels in simulation results.
At the end of a simulation run, a summary of the results can be computed. GeoSFM
includes a tool for computing a variety of flow statistics including the maximum, minimum, mean,
standard deviation, median, 25th, 33rd, 66th, and 75th percentile flows for each catchment. The
33rd and 66th percentile flows may be adapted as the respective low- and high-flow thresholds.
These thresholds can be updated in the catchment data layer and subsequently used in the
generation flow status maps. The results of the flow statistics computations are stored in the form
of tables linked to the catchment data layer. The user can access the values for each catchment by
clicking on the catchment in the visual interface or by modifying the legend of the associated data
layer in the visual interface to display the computed values.
35
Visual maps are considerably easier to interpret than tabular time series data, particularly
when dealing with large river basins. GeoSFM contains a tool for displaying the results of
simulations for any given date in a visual map. The streamflow values on a user-selected date are
presented in the form of indices which present the values in the context of predefined criteria. The
default criterion for differentiating between low and normal flow is the 33rd percentile flow for the
analysis period, while the 66th percentile flow is the minimum threshold separating normal flows
and high flows. However, the user can define other criteria such as return period flow or
predetermined drought and flood warning levels for the classification of flows. Each catchment is
assigned a flow status index of 1, 2, or 3 to signify the respective low, normal, and high flow
conditions. A color-coded map is then produced showing the flow status of each catchment for the
selected day.
GeoSFM contains a graphing tool for plotting hydrographs at the completion of a
simulation run. The tool can be activated from either a program menu or the tool menu bar. The
user can then visually select the catchment for which a hydrograph is required by clicking on the
catchment in the visual interface. The program automatically matches the spatial information with
the time series and produces a hydrograph. The flexibility of this tool is limited by ArcViews
rather limited charting capabilities. The user is consequently encouraged to import the ASCII files
from flow simulations into spreadsheet programs such as Microsofts Excel for more sophisticated
time series plotting capabilities.
Conclusions
The geospatial streamflow modeling system developed at the USGS EROS is a
semidistributed, catchment-based hydrologic model, complete with supporting parameterization
and forcing data for monitoring wide-area hydrologic events. It is designed to run operationally
using widely available remotely sensed data sets and ground observations. The preprocessing
modules used for processing the input data within a GIS environment have been briefly presented
in this manual. Next, the hydrologic analysis module, which consists of a linear soil moisture
accounting routine, a more complex nonlinear soil moisture accounting routine, and a choice of
three river flow transport routines, namely pure lag, diffusion analog, and Muskingum Cunge, is
presented. The hydrologic analysis routines are written as external DLLs, which are initiated
seamlessly from ArcView. A calibration module is also presented which links the hydrologic
analysis routines to a One-At-a-Time sensitivity analysis routine and a Multi-Objective Shuffled
Complex Evolution Metropolis algorithm for parameter calibration. Finally, a postprocessing
module, which contains various routines for analyzing and presenting model output in visual
formats, is also presented. Taken together, the modules allow users to undertake basic hydrologic
assessments in locations with limited in situ information. It is anticipated that GeoSFM will be used
to preserve life, property, and the environment through flood hazard and water resource monitoring
applications around the world, particularly in locations with limited in situ monitoring
infrastructure.
References Cited
Artan, G.A., Asante, K.O., and Verdin, J.P., 2004, A continental scale flood hazard monitoring
system, Eos Trans. AGU, v. 85, no. 17, Jt. Assem. Suppl.
Asante, K.O., Macuacua, R.D., Artan, G.A., Lietzow, R., and Verdin, J.P., 2007, Developing a
flood monitoring system from remotely sensed data for the Limpopo Basin, IEEE Transactions
on Geoscience and Remote Sensing, v. 45, no. 6, p. 17091714.
36
Brooks, R.H., and Corey, A.J., 1964, Hydraulic properties of porous media, Hydrology Paper 3,
Colorado State University, Fort Collins, CO.
Christie, F., and Hanlon, J., 2001, Mozambique and the Great Flood of 2000. Indiana University
Press, Bloomington, IN.
Dooge, J.C.I., 1973, Linear theory of hydrologic systems. Tech.Bull. no. 1468, US Agricultural
Research Service, Washington, D.C.
Duan, Q., Gupta, V.K., and Sorooshian, S., 1992, Effective and efficient global optimization for
conceptual rainfall-runoff models, Water Resources Research, v. 28, p. 10151031.
Dunne, T., and Leopold, L.B., 1978, Water in Environmental Planning. Freeman, San Francisco,
CA.
Food and Agriculture Organization (FAO), 19711981, Soil Map of the World. UNESCO, Paris,
vol. 1-10 (1:5M scale maps and accompanying texts).
Green W.H., and Ampt, G.A., 1911, Studies on soil physics, part I, the flow of air and water
through soils, Journal of Agricultural Sciences, vol. 4, no. 1, p. 124.
Jenson, S.K., and Dominque, J.O., 1988, Extracting topographic structure from digital elevation
data for geographic information system analysis, Photogrammetric Engineering and Remote
Sensing, vol. 54, no. 11, p. 15931600.
Kanamitsu, M., 1989, Description of the NMC global data assimilation and forecast system,
Weather Forecasting, vol. 4, p. 335342.
Kunkel, K.E., Changnon, S.A., and Angel, J.R., 1994, Climatic aspects of the 1993 Upper
Mississippi River Basin flood, Bulletin of American Meteorological Society, vol. 75, no. 5, p.
811974.
Loveland, T.R., Reed, B.C., Brown, J.F., Ohlen, D.O., Zhu, J., Yang, L., and Merchant, J.W., 2000,
Development of a Global Land Cover Characteristics Database and IGBP DISCover from 1-km
AVHRR Data, International Journal of Remote Sensing, vol. 21, no. 6, p. 13031330.
Maidment, D., ed., 1993, Handbook of Hydrology. McGraw Hill, New York, NY.
Morris, M.D., 1991, Factorial sampling plans for preliminary computational experiments,
Technometrics, vol. 33, no. 2, p. 161174.
Rodriguez-Iturbe, I., and Valdez, J.B., 1979, The geomorphological structure of unit response,
Water Resources Research, v. 15, p. 14901520.
Verdin, J.P., and Klaver, R.W., 2002, Grid cell based crop water accounting for the Famine Early
Warning System, Hydrologic Processes, vol. 16, p. 16171630.
Vrugt, J.A., Gupta, H.V., Bouten, W., and Sorooshian, S., 2003a, A Shuffled Complex Evolution
Metropolis algorithm for optimization and uncertainty assessment of hydrologic model
parameters, Water Resources Research, v. 39, no. 8, p. 1201.
Vrugt, J.A., Gupta, H.V., Bastidas, L.A., Bouten, W., and Sorooshian, S., 2003b, Effective and
efficient algorithm for multiobjective optimization of hydrologic models, Water Resources
Research, v. 39, no. 8, p. 1214.
Webb, R.S., Rosenzweig, C.E., and Levine, E.R., 1993, Specifying land surface characteristics in
general circulation models: soil profile data set and derived water-holding capacities, Global
Biogeochemical Cycles, vol. 7, p. 97108.
Xie, P., and Arkin, P.A., 1997, A 17-year monthly analysis based on gauge observations, satellite
estimates, and numerical model outputs, Bulletin of the American Meteorological Society, vol.
78, no. 11, p. 25392558.
Zobler, L., 1986, A world soil file for global climate modeling. NASA Tech. Memo. 87802, NASA
Goddard Institute for Space Studies, New York, NY.
37
Appendix 1: GeoSFM Data Specifications

Preprocessing Module: Performing Terrain Analysis
Input Grid
Description
Elevations
Hydrological corrected Digital Elevation Model
source HYDRO1k DEM.
Input Shapefile
Limpbas.shp
Input Parameters
Analysis Extent
Analysis Cell Size
Stream Delineation Threshold
Output Grids
FlowDir
Units
Values
Shapefile defining analysis extent.

Boundary defined for analysis area (same as
Limpbas.shp).
Boundary defined for cell area (same as elevations).

Minimum number of cells upstream of a given
location for initiating stream. Default = 1,000,
minimum = 500.
Flow direction grid consists of numeric values
assigned using the eight direction pour point model.
Each cell is assigned one of eight compass directions.
FlowAcc
Flow accumulation grid defines the number of

contributing cells draining into each cell. Values
range from 0 at topographic highs to large numbers at
the mouths of rivers.
FlowLen
DownStreamflow length grid defines the distance

from the cell to the basin outlet.
Streams
StrLinks
Stream grid defines stream cells.

Stream link grid assigns a unique value to each
stream ID.
Outlets
Outlet grid defines number of outlets; one outlet per

subbasin, each outlet assigned a unique value.
Basins
Subbasin grid assigns a unique ID value to each

subbasin.
HillLength
Hill length grid defines distance to stream, from cell

to nearest stream.
Slope
Hill slope grid describes the maximum change in

elevation between each cell and its eight neighbors.
Expressed in percentage.
Downstream
Downstream grid assigns a unique ID number of the

subbasin immediately downstream.
BigBasins
Basply.shp
Basin boundary polygon.

Shapefile defining all subbasins with unique ID and
gridcodes.
Rivline.shp
Shapefile defining streams with unique IDs,

gridcodes, and from_node and to_node.
Preprocessing Module: Generating Basin Characteristics Files

Input Grids
Description
38
m
1
Units
Values
Rcn
SCS runoff curve number grid uses hydraulic soil

classes together with the GLCC classes to derive an
RCN value.
Whc
Soil water holding capacity grid is calculated from

the DSMW soil class and the total available water
median soil moisture value.
mm/m
Soildepth
Hydrological active soil layer depth grid created from

DSMW soil code value and Webbs soil depth
values.
cm
Texture
Texture grid created from Zoblers seven classes

converted to three classescoarse, medium, and fine.
Ks
Saturated soil hydraulic conductivity grid created on

Ks values, which are based on Zoblers texture
classes.
Maxcover
Maximum impervious cover grid based on wetland

areas and water bodies found in the GLCC data set
and stream cells defined in the flowacc grid.
Output Files
Basin.txt
see .txt file detail below
File containing each subbasin characteristics. This

file is rewritten during the calibration process for
each model run.
Basin_original.txt
Copy of basin.txt. Retains original basin

characteristic data.
Order.txt
Computational orderused in subsequent program

operations.
river.txt
File containing river characteristics. This file is

rewritten during the calibration process for each
model run.
river_original.txt
Copy of river.txt. Retains original river characteristic

data.
describe.txt see .txt file detail

below
File descriptions for 12 different files generated in

modeling process.
Basin.txt
BasinID
SoilWHC
SoilDepth
cm/hr
Subbasin identification number.

Soil water holding capacity calculated from the
DSMW soil class and the total available water
median soil moisture value.
mm/m
Hydrological active soil layer depth created from

DSMW soil code value and Webbs soil depth
values.
Texure value created from Zoblers seven classes and
converted to three classescoarse, medium, and fine.
cm
Ks
Saturated soil hydraulic conductivity value created

based on Zoblers Ks and texture values.
cm/hr
Area
Interflowlag
Hslope
Baseflowlag
Drainage area of the subbasin.

Residence time for the interflow reservoir.
Average subbasin slope.
Residence time for the baseflow reservoir.
km
days
Texture
39
1,2,3
days
RCNumber
SCS runoff curve number uses hydraulic soil classes

together with GLCC classes to derive a RCN value.
0100
MaxCover
Maximum impervious cover value based on wetland

area and water bodies found in the GLCC data set a
long with the stream cells defined in the flow
accumulation data.
Fraction of the soil water infiltrating to groundwater.
Pan coefficient for correcting PET readings.
Fraction of soil layer classified as hydrologically
active topsoil layer.
01.0
BasinLoss
Pancoeff
Topsoil
Aridity
Order.txt
BasinID
01.0
01.0
01.0
Aridity index range (1 = arid, 2 = wet).
1,2
Most downstream subbasin Number.

2nd most downstream subbasin Number.
3rd most downstream subbasin Number.
Most upstream subbasin Number.
River.txt
RiverID
Area
UpArea
RivSlope
Identification number of the river reach.

Local drainage area of the river reach.
Total area upstream of the river reach.
Average slope of the river reach/basin mainstream
channel mean slope.
km2
km2
RivLength
DownID
RivManning
Length of the river reach.

Identification number of the downstream river reach.
Value of Mannings n for the river reach, Default =
0.035.
RiverLoss
RiverPloss
HasDam
Fraction of river water lost to infiltration.

Fraction of the river water lost to evaporation.
Indicates whether the river reach contains a dam (0 =
no, 1 = yes).
01.0
01.0
0,1
HasRating
Indicates whether the reach has a rating curve (0 =

no, 1 = yes).
0,1
HasFlowData
Indicates whether the river reach has observed flow

data (0 = no, 1 = yes).
0,1
Celerity
Diffusion
Velocity of the flood wave through the river reach.

Flow attenuation (or dispersion) coefficient of the
reach.
m/s
m2/s
RivWidth
FlowRef
RunType
Average channel width.

Reference flow for the section.
Indicates if simulation should proceed from an
existing or a new run (0 = new, 1 = existing).
m
m3/s
describe.txt
Basin Characteristics
List and description of the 12 different characteristics

generated in the basin.txt.
River Characteristics
List and description of the 17 different characteristics

generated in the river.txt.
Response Characteristics
Explanation of the time step runoff data generated in

the response.txt.
40
0,1
Computation Order
Explanation of the subbasin

downstream/upStreamflow in the order.txt.
Rainfall Characteristics
Explanation of the time step rainfall data generated in

the rain.txt.
Evapotranspiration
Characteristics
Explanation of the time step PET data generated in

the evap.txt.
Balance Parameters
List and description of the nine different

characteristics generated in the balparam.txt.
Balance File Listing
List and description of the 17 different input/output

files in the balfiles.txt.
Route Parameters
List and description of the nine different

characteristics generated in the routparam.txt.
Route File Listing
List and description of the 13 different input/output

files in the routfiles.txt.
Total Runoff from each

Subbasin
Explanation of the time step total runoff data

generated in the basinrunoffyield.txt.
Total Streamflow from each

Subbasin
Explanation of the time step total discharge generated

in the streamflow.txt.
Preprocessing Module: Basin Response Computation

Input Grids/Files
Description
Basins
Subbasin grid assigns a unique ID value to each
subbasin.
Flowdir
Flow direction grid consists of numeric values

assigned using the eight direction pour point model.
Each cell is assigned one of eight compass directions.
Flowlen
DownStreamflow length grid defines the distance

from the cell to the basin outlet.
Outlets
Outlet grid defines number of outlets, one outlet per

subbasin, each assigned a unique value.
Usgslandcov
Velocity
USGS Land Cover grid.

Overland flow velocity value or user-defined velocity
grid.
Elevations
Flowacc
Elevation grid.
Flow accumulation grid defines the number of
contributing cells draining into each cell. Values
range from 0 at topographic highs to large numbers at
the mouths of rivers.
Order.txt see .txt file detail -1.3
Computational order file.
Output Grids/Files
Traveltime
Velocity
Travel time from each grid cell to the catchment

outlet based on slope, watershed size, flow resistance
(Mannings roughness coefficient), and water input.
Option 1: Estimated uniform overland velocity for
each catchment based on the mean slope of the
catchment and dominant land cover.
41
Units
Values
Option 2: Velocity values for each catchment defined

by user.
Option 3: Velocity default value for each catchment
defined by user.
response.txt
response.txt
BasinID
Day 0
Day 1
Day 2
Day 20
File containing fraction of excess runoff arriving at

subbasin outlet in each subsequent time step, similar
to a unit hydrograph for each subbasin.
BasinID of subbasin beginning with the most
downstream.
Most upstream subbasin.

Fraction of excess runoff arriving at subbasin outlet
in each subsequent time step.
Preprocessing Module: Extracting Rain/Evap Data

Input Grid/File
Description
Basins
Basin Grid.
order.txt see .txt file detail -1.3
Computational order file.
Input Parameters
raindata
evapdata
Start Year
End Year
Start Day Number
End Day Number
Output Files
rain.txt see .txt file detail below
evap.txt see .txt file detail
below
rain.txt
Time
BasinID of subbasin beginning

with the most downstream
0.01.0
Rain data from directory.

Evap data from directory.
Units
mm
tenths of
mm
Start year of data.

End year of data.
Start day of data defaults 1 Julian Date.
End day of data number of days included for data
estimate defaults -240- entered 10 (for shorter
processing time) Julian Date.
File containing time step rainfall for each subbasin
(per day).
mm
File containing time step PET for each subbasin (per

day).
tenths of
mm
Date example 1999001 format yyyyddd
Date
1999010
time step rainfall for each subbasin.
mm
Most upstream subbasin
42
Values
evap.txt
Time

Date
1999010
time step PET for each subbasin.
tenths of
mm
Simulation Module: Computing Soil Water Balance

Input Files
Description
rain.txt see .txt file detail -2.1
File containing time step rainfall for each subbasin.
evap.txt see .txt file detail -2.1
File containing time step PET for each subbasin.
basin.txt see .txt file detail -1.3
File containing each subbasin characteristics.
response.txt see .txt file detail 1.4
File containing time step excess runoff arriving at

subbasin outlet.
Input Parameters
Computation Start Year
Year entered into field.
Computation Start Day

Number of Rain/Evap Days
Daydefault 001 Julian Date

Daysdefault 10
Number of Response Days
Daysdefault 21 (how long the hydrograph is

stretched)
Number of Subbasins
Initial Soil Moisture
Numberdefault 365
Assume soil is dry containing only 10% of its storage
capacitydefault 0.1 (decimal).
Data Format (l for Daily Data)
1 = Day (default), 0 = Hourly
New Run or Continue Previous

Run
0 = New run (default), 1 = Continue previous run
Basin Polygon Theme
Polygon of analysis area selected basply.shp for

exercise.
Key Field e.g. Grid Code
Selected gridcode for exercisenumber ID.
Mode
Soil Model
0 = Simulation, 1 = Calibration -defaults-0

Option of Single Layer Soil Model or Double Layer
Soil Modelselected Single Layer for exercise.
Output Files
balparam.txt see .txt file detail
below
File containing: Number of ordinates of unit

hydrograph response, number of simulation time
steps, simulation start year/day, and number of
catchments, simulation interval in hours, data format,
model initialization code, and initial soil fraction.
43
Units
mm
tenths of
mm
Values
balfiles.txt see .txt file detail

below
Output Files
basinrunoffyield.txt see .txt file
detail below
File listing input and output .txt files.
Basin runoff yieldrunoff for each subbasin.
soilwater.txt see .txt file detail

below
Output soil moisture storage.
actualevap.txt see .txt file detail

below
Actual evapotranspiration.
mm
gwloss.txt see .txt file detail
Output file water lost to the regional groundwater

flow.
mm
cswater.txt see .txt file detail

below
Output file containing last model time step simulated

soil water content.
mm
excessflow.txt see .txt file detail

below
Output file excess runoff fluxes.
mm
interflow.txt see .txt file detail

below
Output file interflow or fast base flow fluxes.
mm
baseflow.txt see .txt file detail

below
Output file base flow fluxes.
mm
logfilesoil.txt see .txt file detail

below
Process/Error Log File.
massbalance.txt see .txt file

detail below
Output file of sum of water cycle components for the

total simulation period only for Double Layer Soil
Model.
initial.txt see .txt file detail

below
Input/output file that contains initial conditions for

model state variablesonly if Double Layer Soil
Model is selected.
default.txt see .txt file detail

below
Soil model and routing configuration selected for

model run.
balparam.txt
No header
Number of ordinates of unit hydrograph response.

Number of simulation time steps.
Simulation start year.
Simulation start day.
Number of catchments.
Simulation interval in hours.
Data format indicator (1/0)defaults 1.
Model initialization mode (1/0)defaults 0 (i.e. no
mode initialization).
Initial soil fraction, Default = 0.1.
Model Calibration Mode (0/1)defaults 0 (i.e.
noncalibration mode).
balfiles.txt
No header
Input rainfall file (rain.txt).

Input potential evapotranspiration file (evap.txt).
Input basin characteristics file (basin.txt).
Input unit hydrograph response file (response.txt).
Output balance parameter file (balparam.txt).
44
24
01
01
01
01
Output runoff yield file (basinrunoffyield.txt).

Output soil moisture storage file (soilwater.txt).
Output actual evapotranspiration file (actualevap.txt).
Output groundwater loss file (gwloss.txt).
Output final soil water storage file (cswater.txt).
Output surface precipitation excess file (gwloss.txt).
Output interflow file (interflow.txt).
Output baseflow file (baseflow.txt).
Output mass balance file (massbalance.txt).
Output log file (logfilesoil.txt).
Output initialization file (initial.txt).
Output workdirectory (<full directory pathname>).
basinrunoffyield.txt
Time

Date
1999110
Time step runoff for each subbasin.
mm

soilwater.txt
Time
ID of subbasin beginning with

the most downstream
Date example 1999001 format yyyyddd (first

column)
Date
1999110
Time step soil moisture storage for each subbasin.

actualevap.txt
Time

the most downstream

column)
Date
1999110
Time step for actual evapotranspiration for each
subbasin.

gwloss.txt
Time

column)
Date
1999110
45
mm

the most downstream
Time step water lost to the regional groundwater flow

for each subbasin.

cswater.txt
Basin ID
CurrentSW
excessflow txt
Time

the most downstream
ID of subbasin beginning with the most downstream.

Contains last model time step simulated soil water
content.
mm

column)
Date
1999110
Time step of excess runoff fluxes.
mm

interflow.txt
Time

the most downstream

column)
Date
1999110
Time step of interflow or fast base flow fluxes.
mm

baseflow.txt
Time

the most downstream

column)
Date
1999110
Time step base flow fluxes.

logfilesoil.txt
No header
Open output files for calib/sens/standard run modes.

Read basin parameter file.
Open output files for standard runs.
46
mm
Initialized basin objects.

Read basins characteristics.
Calculated runoff contributing partial area model
parameters.
Opened met input rain and evap files.
Converted slope from percent to fraction.
default.txt
No header
massbalance.txt
basinId
precipitation
Soil configuration selected for model run.

Soil1 Single layer soil model.
Soil2 Double layer soil model.
Routing method selected for model run.
Rout1 Simple lag routing method.
Rout2 Diffusion analog routing method.
Rout3 Muskingum Cunge routing method.

Sum of water cycle precipitation for the total
simulation period.
mm
surfaceRunoff
Sum of water cycle surface runoff for the total

simulation period.
mm
interflow
Sum of water cycle fast base flow flux for the total
simulation period.
mm
baseflow
Sum of water cycle base flow flux for the total

simulation period.
mm
evaporation
Sum of water cycle evaporation for the total

simulation period.
mm
moisture_change
Difference of the water cycle value from start of

model to end of model for the total simulation period.
mm
residual
Difference of water cycle (rain in, evaporation/runoff

out) for the total simulation period.
mm
initial.txt
Basin ID
upperSoilWater
lowerSoilWater
fastBaseflow
slowBaseflow
interflowWater
upperSoilExcessWater
lowerSoilExcessWater
surfaceWaterAtTime0
surfaceWaterAtTime1

Upper soil water values for each subbasin.
Lower soil water values for each subbasin.
Fast base flow values for each subbasin.
Slow base flow values for each subbasin.
Interflow reservoir water values for each subbasin.
Upper soil excess water values for each subbasin.
Lower soil excess water values for each subbasin.
Surface water value at first time step of hydrograph
for each subbasin.
Surface water value at next time step of hydrograph
for each subbasin.
47
surfaceWaterAtTime2

for each subbasin.
surfaceWaterAtTime19

for each subbasin.
Simulation Module: River Flow Transport

Input Files
Description
basinrunoffyield.txt see .txt file Basin Runoff Yield File.
detail -2.2
response.txt see .txt file detail 1.4
File containing time step excess runoff arriving at

subbasin outlet.
river.txt see .txt file detail -1.3
File containing river characteristics.
reservoir.txt see .txt file detail

below
Reservoir Characteristic File.
rating.txt see .txt file detail

below
Stream gauge input containing BasinID, type,

parameters and maximum river depth.
Input Parameters
Basin Polygon Theme
Selected gridcode for exercise.
Number of Days of Forecast

Required
Default - 3
Mode
Routing Method
0 = Simulation, 1 = Calibration -default 0Needed for the computation: three optionssimple

lag routing method, diffusion analog routing method,
or Muskingum Cunge routing methodselected
simple lag routing method for exercise.
OpFilName
rating.txt
BasinID
Type
Values
Polygon of analysis area selected basply.shp for

exercise.
Key Field e.g. Grid Code
reservoir.txt
DamID
Storage
Residence Time
Operated
Units
BasinID where dam is located.

Storage in 1,000m3.
Defaults 30.
0 No user-defined file.
1 User-defined file used.
Path and file name if user-defined file is used.
Defaults None.
0, 1
BasinID of subbasin beginning with the most

downstream.

Two different types based on the following
equations:
Q = m3/sec
Type 1 Q = A(H+B)^C
Type 2 Q = (A*H^2) + BH + C
H=m
where: A = Par1
48
1,2
Par1
Par2
Par3
hmax
Output Files
routparam.txt see .txt file detail
below
routfiles.txt see .txt file detail

below
B = Par2
C = Par3
Equation Parameter for A
Equation Parameter for B
Equation Parameter for C
Maximum height where equation is valid.
File containing: number of simulation time steps,
simulation start year/day, number of catchments,
simulation interval in hours, model type, number of
forecast days, model initialization mode, number of
reservoirs.
File listing input and output .txt files.
Output Files
inflow.txt see .txt file detail
below
Flow values coming from upstream for each

subbasin.
streamflow.txt see .txt file detail

below
Streamflow file contains a time series of velocity

values in cubic meters per second for each subbasin.
localflow.txt see .txt file detail

below
Flow value generated in each subbasin.
riverdepth.txt see .txt file detail

below
River depth file populates when Muskingum Cunge

routing method is selected.
Reservoir status file, present only if there are dams in
the basin.
damstatus.txt see .txt file detail

below
damlink.txt see .txt file detail
below
Present only if there are dams in the basin.
forecast.txt see .txt file detail

below
Forecast file contains the streamflow forecast for the

number of days that were selected (3 days in
exercise).
times.txt see .txt file detail

below
Start/end time and date of computing streamflow

process.
logfileflow.txt see .txt file detail

below
Process/Error log file.
routparam.txt
No Header
routfiles.txt
No Header

Simulation start year.
Simulation start day.
Simulation interval.
Model type (0 = Musk-Cunge, 1 = Linear Reservoir).
Number of forecast day.
Model initialization mode (1/0).
Number of reservoirs.
Route parameter file (routparam.txt).
Input river characteristics file (river.txt).
Input river initialization file (initial.txt).
Input runoff yield file (basinrunoffyield.txt).
49
YYYY
JDAY
hours
24
0,1
0100
1,0
Input reservoir-river link file (damlink.txt).

Input observed flow file (obsflow.txt).
Input rating curve flow file (rating.txt).
Output streamflow file (streamflow.txt).
Output subbasin flow contribution file (localflow.txt).
Output average river depth file (riverdepth.txt).
Output subbasin upStreamflow contribution
(inflow.txt).
Output flow routing log file (logfileflow.txt).
Output work directory (<full directory pathname>).
inflow.txt
Time

Date
1999013
UpStreamflow values for each subbasin.

streamflow.txt
Time

Date
1999013
Time step total discharge for each subbasin.
m3/s

localflow.txt
Time


forecast.txt
Time

Date
1999013
Flow value generated in each subbasin.
Flow value generated within each subbasin.
m3/s
Date
1999013
Forecasted flow value for each subbasin.
m3/s
50
times.txt
No header
Starting time: Day of the week, month/day,

hour/minute/second, year.
Ending time: Day of the week, month/day,
hour/minute/second, year.
Example Fri Oct 28 16:49:59 2005
logfileflow.txt
No header
riverdepth.txt
Time

File history. Sample messages below.

***CREATED LOGFILEFLOW.TXT FILE***
****READ ROUTFILES.TXT ****
***OPENED ROUTPARAM.TXT FILE***
****READ ROUTPARAM.TXT ****
***OPENED RIVER.TXT FILE***
***OPENED BASINRUNOFFYIELD.TXT
FILE***
***CREATED INFLOW.TXT FILE***
***CREATED STREAMFLOW.TXT FILE***
***OPENED DAMLINK.TXT FILE***
***CREATED LOCALFLOW.TXT FILE***
****READ BASIN FILE HEADER****
****READ RUNOFF FILE HEADER****
****READ BASIN FILE****
****READ BASIN RUNOFF YIELD FILE****
****INITIATING ROUTING****
***WRITING OUTPUT FILES***
***CLOSING TIME SERIES FILES***
***RETURNING TO ARCVIEW***
Additional Output for Muskingum Cunge routing method.
column)
Date
1999013
Time step river depth for each subbasin.

damstatus.txt
damlink.txt
Present only if dams present.
Calibration Module: Performing Sensitivity Analysis

Input Parameters
Description
51
Units
Values
River Reach
Drop-down list of river reachselected 140 for

exercise 4.
Model Configuration
Drop-down list of all possible model configurations

soil/routing. Selected one soil layer, lag routing for
exercise 4.
Input Files
range.txt
see above table
balparam.txt
see section 2.2
basin_original.txt
basin.txt
Provides minimum and maximum values of a feasible

parameter range for testing model sensitivity.
Original basin characteristics.
file is rewritten during the sensitivity analysis
process.
river_original.txt
river.txt
Original river characteristics.

rewritten during the sensitivity analysis process.
moscem_param.txt
Gives location of outlet in model output file.
whichmodel.txt see .txt file

detail below
Sets choice of water balance and routing model

configurations.
Moscem_param.txt
No header
Nflux
Ntstep1
Nstep2
Obj_func,
Missing Value
Nflux_obs
Nflux_model
whichModel.txt
Model Index
Index Description
Number of output variables (fluxes).

Number of time steps of input variables.
Number of time steps of output variables.
Statistical test for optimization:
1 = RMSE root mean square error
2 = STD standard deviation
3 = HMLE maximum likelihood error function
4 = NSE Nash-Sutcliffe efficiency
5 = NSC Number of sign changes
6 = BIAS
Value of missing data in observed_streamflow.txt.
Column in observed_streamflow.txt to test.
Column in streamflow.txt to extract modeled data.
Models:
Water balance model
Routing model
Water balance model 1 = 1D balance, 2 =
2Dbalance
Routing model 1 = diffusion, 2 = Muskingum
Cunge, 3 = lag
Output Files
52
1,2,3,4,5,6
9999
oat.txt see .txt file detail below
One-At-a-Time parameter index and multiplier

values used in sensitivity analysis.
SensAnalysis.txt
Results of sensitivity analysis for each model run,

includes parameter index, multiplier value, mean,
maximum, and standard deviation of modeled
streamflow results.
Gives the mean absolute difference of test results
over the parameter range for each parameter. The
greater the differences, the more sensitive the
parameter.
SArunOutput.txt
oat.txt
Whichparam
OAT Parameter Value

SensAnalysis.txt
whichParam
OAT Parameter Value

Max_mod
Mean_mod
stddev
SArunOutput.txt
whichparm
max_mod
List of 1
through 20
For all 20 runs, total of 400 lines.
Multiplier values used in sensitivity analysis.
List of 1
through 20
For all 20 runs, total of 400 lines.
Multiplier values used in sensitivity analysis.
Maximum of modeled streamflow results.
Mean of modeled streamflow results.
Standard deviation of modeled streamflow results.
List of 1
through 20
The maximum difference of test results over the
parameter range for each parameter.
mean_mod
The mean difference of test results over the

stddev
The standard deviation of test results over the

Calibration Module: Performing Model Calibration

Input Parameters
Description
Observed Streamflow Stations
The streamflow station ID displayed in a drop-down
list. Selected 1 for exercise 4.
BasinID for Streamflow Station Drop-down list of basin IDs, select ID that
corresponds with the streamflow station selected.
Selected 140 for exercise 4.
Number of Runs
Drop-down list of choices. Selected 54 for exercise 4.

(Selected for faster processing time.)
53
Units
Values
The number of model runs depends on the length of

streamflow record, number of parameters being
tested, and the complexity of the parameter
dependent model response for the watershed being
modeled. The range will generally be 5,00010,000
model runs for convergence to a global minimum.
Objective Function Type
Input Files
balparam.txt
Convergence can be measured in terms of

Root Mean Square Error (RMSE)
Standard Deviation (STD)
Maximum Likelihood Error (MLE)
Nash-Sutcliffe Efficiency (NSE)
Number of Sign Changes (NSC)
BIAS
Selected RMSE for exercise 4.
basin_original.txt
basin.txt

process.
river_original.txt
river.txt

streamflow.txt see .txt detail

below

parameter.in
Range of multiplier values used to define parameter

search space:
see file detail below
Minparameter.in = Minrange.txt / (basin ave param)

Maxparameter.in = Maxrange.txt / (basin ave param)
Main calibration algorithm control file.
moscem.in see file detail below

MOSCEMx.txt see .txt file
detail below
File listing which parameters to calibrate.
moscem_param.txt see .txt

detail below
Calibration control file.
ObjOptFlag.in see file detail

below
Flag for number of fluxes to multi-calibrate.
whichModel.txt
observed_streamflow.txt
Sets choice of water balance and routing models.

Observations from which to compare model results.
parameter.in
No
Name
Index.
Parameter name see Parameters to be Calibrated.
54
Default
Lower
Upper
OptIdx
Description
moscem.in
No Header
nOptPar
nOptObj
nSamples
Dummy variable.
Minimum multiplier value.
Maximum multiplier value.
Optimization flag: 0 = do not optimize, 1 = optimize
Description of parameter see Parameters to be
Calibrated.
Number of parameters used for optimization.

Number of fluxes used for optimization.
Number of random samples from nOptPar
dimensioned parameter space per shuffle.
nComplex
nMaxDraw
Number of complexes.
Maximum number of function evaluations/model
runs.
ParamMult file
Name of the parameter file including path. Default =

parameter.in
whichFlux file
Name of the file specifying which fluxes to be

calibrated, including path. Default = objoptflag.in
Observed streamflow file
Name of the file providing evaluation data, including

path. Default = observed_streamflow.txt
ObjectiveResults output
Name of file storing the parameter values of the final

points, including path. Default = objectives.out
Parameter values output
Name of file storing the objective function values of

the final points, including path. Default =
parameter_values.out
Parameter convergence output
Name of the file keeping track of the convergences of

each of the parameters, including path. Default =
par_convergence.out
MOSCEMx.txt
Parameter Index
ObjOptFlag.in
No Header
One line for each parameter to be calibrated.
Last line setting to denote end of life.
999
Runoff_1
Runoff_2
1,0
1,0
observed_streamflow.txt
Time
Streamflow Stations
Last station
Output Files
Par_convergence.out see file
detail below
Objectives.out see file detail
below
Date
1999365
Streamflow values.
Parameter convergence for best nSamples.

Parameter_values for best nSamples.
55
Parameter_values.out see file

detail below
Objectives.out
nSampleIndex
Objective values for each flux
Objective function results for best nSamples.
Sample number starting with 1
Last sample 60
Parameter_values for best nSamples.
Fitness/Probability for each

sample
Calibration Module: Postprocessing Calibration Results
Input Files
Description
balparam.txt
basin_original.txt
basin.txt
process.
river_original.txt
river.txt

postproc.in see file detail

below
File containing nSamples, observed streamflow

values, Parameter_values.out, Objectives.out,
Timeseries.out, ObjectStats.out and
Trdoff_bounds.out.
postproc.in
No Header
nSamples
Same nSamples from moscem.in, equivalent to

number of parameter sets saved in
parameter_values.out.
Observed streamflow values file. Default =
Observed_streamflow.txt
Parameter value output file from calibration. Default
= Parameter_values.out
Objective function results output file from
calibration. Default = Objectives.out
Timestep, Observed Streamflow, Single timeseries of
best calibration run. Default = Timeseries.out
Objective Statistics for all six objective functions for
the streamflow runs presented in timeseries.out.
Default = ObjectStats.out
Timestep, Lower boundary, upper boundary, and
average streamflow over the nSample model runs.
Default = Trdoff_bounds.out
Output Files
56
Units
Values
timeseries.out see file detail

below
Objective function results output file from

calibration.
objectiveStats.out see file

detail below
All statistic tests from best calibration run in

timeseries.out.
Trdoff_bounds.out see file

detail below
Timestep, lower boundary, upper boundary, and

average streamflow over the nSample model runs.
timeseries.out
Timestep
Date
Date
365
Observed Streamflow values.
Single time series of best calibration run.
1365
Date
1365
Observed Station 1
Calibrated Station 1
trdoff_bounds.out
Timestep
LowerBound Station1
UppBound Station1
Average Station1
Date
365
Lower boundary streamflow values for station 1.
Upper boundary streamflow values for station 1.
Average streamflow values for station 1.
Postprocessing : Update Bankfull and Flow Statistics

Input Files
Description
below
Output Files
monthlyflow.txt see .txt detail
below
Streamflow file contains a monthly time series of

velocity values in cubic meters per second for each
subbasin.
annualflow.txt see .txt detail

below
Streamflow file contains an annual time series of

subbasin.
riverstats.txt see .txt detail

below
List of streamflow statistic types and corresponding

values.
statsparam.txt see .txt detail

below
File which contains the start year, start month, start

day, end year, number of simulation time steps,
number of catchments, statistical type, and a list of
input/output file names and paths.
logfilestats.txt see .txt detail

below
Log file of statistical parameters and output file

names and paths.
monthlyflow.txt
Time
Units
Date example 199901 format yyyymm (year,

month)
Streamflow file contains a monthly total time series
of velocity values in cubic meters per second for each
subbasin.

annualflow.txt
57
m3/s
Values
Time
Date example 1999 year

Streamflow file contains an annual total time series
subbasin.

riverstats.txt
ID
Gridcode
Soilwater
Currentsw
Sum
Count
Mean
Maximum
Minimum
Range
Variance
Stddev
Median
Quart25
Quart75
Quart33
Quart66
Highflow
Lowflow
Medflow
statsparam.txt
No Header
File record number.

Subbasin ID number.
Soil water value for subbasin.
Current soil water value (same as Soilwater) for
subbasin.
Total Streamflow.
Number of days of data.
Streamflow mean value for each catchment.
Streamflow maximum value for each catchment.
Streamflow minimum value for each catchment.
Streamflow range value for each catchment.
Streamflow variance for each catchment.
Streamflow standard variation for each catchment.
Streamflow median value for each catchment.
25 percentile flow for each catchment.
Streamflow high flow value for each catchment.
Streamflow low flow for each catchment.
Streamflow medium flow value for each catchment.
Start Year.
Start Month.
Start Day of the Month.
End Year.
Statistical Type.
Input file name and path: Default = streamflow.txt
Monthly Flow Output file and path: Default =
monthlyflow.txt
Annual Flow Output file and path: Default =
annualflow.txt
logfilestats.txt
No Header
sample messages below

Start Year.
Start Day.
Number of Records.
Number of Fields.
Operation (0-MEAN, 1-MEDIAN, 2-MAX).
58
m3/s
Input filename e.g.

C:\GeoSFM\workspace\streamflow.txt.
Monthly Output filename e.g.
C:\GeoSFM\workspace\monthlyflow.txt.
Annual Output filename e.g.
C:\GeoSFM\workspace\annualflow.txt.
Calculating monthly max for month 1.
Writing annual values.
Postprocessing: Display Flow Percentile Map
Input File
Description
streamflow.txt see .txt detail in Streamflow file contains a time series of velocity
section 3.2
Input Parameter
Date
Output
Map
Units
Values
Units
Values
Units
Values
Selected from drop-down list of dates for flow

percentile map yyyydddJulian day
Map showing daily variation of streamflow values
displayed as low, normal, and high conditions.
Postprocessing: Display Flow Hydrographs

Input Parameters
Description
Basin Theme
Drop-down list of choices basply.shp selected.
Subbasin ID
Drop-down list of choices Gridcode selected.
Output
Hydrograph
Graph showing daily variation of streamflow at the
outlet of the selected subbasin.
GeoSFM Utilities: Downloading Rain/Evap Grid
Input Parameters
Description
FTP Site Information
FTP Site Location default edcftp.cr.usgs.gov
Site Username default anonymous
Site Password default anonymous@edc
Specify Type of Data to
Options: Get Rainfall Grids, Get Forecast Rainfall
Download
Grids, Get Global Pet Grids, and Get Global
TRMM Rainfall Grids selections made for exercise
6 are in bold.
Region of Analysis
Options: Global, Africa, SE Asia, Afghanistan,

Australia, NAmerica, CAmerica, SAmerica, and
Europe selected Africa for exercise 6 data is not
available for all regions.
Select Grids to downloaded
A list of 16 days of data is presented 10 days were

selected to complete exercise 6.
rain_yyyyddd.tar.gz or evap_yyyyddd.tar.gz
Zip File
rain_yyyyddd.tar.gz
evap_yyyyddd.tar.gz
Zip rainfall data downloaded to local working

directory one file per day.
mm
Zip evap data downloaded to local working directory

one file per day.
tenths of
mm
59
Input Grid
rain_yyyyddd.tar.gz
GeoSFM Utilities: Unzip, Untar, Imagegrid

Description
Zip rainfall data downloaded to local working
directory one file per day.
evap_yyyyddd.tar.gz
Zip evap data downloaded to local working directory

one file per day.
rain_yyyyddd.bil
Unzipped rainfall image files processes in local

working directory.
evap_yyyyddd.bil
Unzipped evap image files processes in local working

directory.
Output Grid
rain_yyyydddg
evap_yyyydddg
Input Parameters
Select Continent for Analysis
Analysis Properties
Projection
Output Grid
rain_yyyyddd
Evap_yyyyddd
Values
Units
Values
Grid containing rainfall data for each day

geographic projection, added to ArcView table of
contents.
Grid containing evap data for each day geographic
projection added to ArcView table of contents.
GeoSFM Utilities: Projecting Geographic Grids to Lambert Azimuthal

Input Grid
Description
rain_yyyydddg
Grid containing rainfall data for each day
geographic projection, added to ArcView table of
contents.
evap_yyyydddg
Units
Grid containing evap data for each day geographic

projection added to ArcView table of contents.
Options: Global, Africa, SE Asia, Afghanistan,
Australia, NAmerica, CAmerica, SAmerica, and
Europe selected Africa for exercise 6.
AnalysisExtent: Current Value
Analysis Cell Size: Current Value
Use all default values.
Lambert Equal-Area Azimuthal.
Grid containing rainfall data for each day Lambert
Azimuthal projection added to ArcView table of
contents.
mm
Grid containing evap data for each day Lambert

Azimuthal projection added to ArcView table of
contents.
tenths of
mm
GeoSFM Utilities: Compute Rain/Evap Grid Statistics

Input Parameters
Description
raindata
evapdata
Start Year
Start year of data 1999.
End Year
End year of data 1999.
Start Day Number
Start day of data defaults 1-Julian Day
60
Units
Values
End Day Number
End day of data number of days included for

calculation of statistics defaults 365-entered 10
(shorter processing time) Julian Day
Input Parameter
GRID_STATYPE_SUM
Sum of values for date range.
GRID_STATYPE_MAX
Maximum value for date range.
GRID_STATYPE_MEAN
Average of values for date range.
GRID_STATYPE_STD
Standard Deviation value for date range.
GRID_STATYPE_MEDIAN
Median value for date range.
GRID_STATYPE_RANGE
Range of values for date range.
GRID_STATYPE_MIN
Minimum value for date range.
GRID_STATYPE_MAJORITY
Majority value for date range.
GRID_STATYPE_MINORITY
Minority value for date range.
GRID_STATYPE_VARIETY
Number of different values for date range.
GRID_NONZERO_VALUES
Nonzero values for date range.
Output Grid
R10SUM1-grid name depends
on type and number of days
selected
R = rain, 10 = number of days, SUM = statistical

type, 1 = each calculation run is incremented.
E10SUM1-grid name depends

on type and number of days
selected
Point theme
gauges2.shp
Input Parameters
raindata
evapdata
Start Year
End Year
Start Day Number
End Day Number
E = evap, 10 = number of days, SUM = statistical

type, 1 = each calculation run is incremented.
GeoSFM Utilities: Pick Grid Values at Point

Description
Site IDs, latitude and longitude coordinates in
decimal degrees.
Start year of data 1999.
End year of data 1999.
Start day of data defaults 1-Julian Day
End day of data number of days included for
calculation of statistics defaults 365-entered 10
(shorter processing time) Julian Day
61
Units
mm
mm*10
Values
Output File
rainpoint.txt see .txt detail
below
File containing time step rainfall (mm) for each

station ID (per day).
evappoint.txt see .txt detail

below
File containing time step PET (tenths of mm) for

each station ID (per day) File.
rainpoint.txt
Time
Station ID number starting with

the smallest number
Largest number
evappoint.txt
Time
Station ID number starting with

the smallest number
Date
1999010
Time step rain for each point (rain gauge station).
Date
1999010
Time step PET for each point (evap gauge station).
mm
Tenths of
mm
Largest number
GeoSFM Utilities: Interpolate Station Data to Grid
Input Shapefile
Description
gauges2.shp
Site IDs, latitude and longitude coordinates in
decimal degrees.
Limpbas.shp
Input Files
rainstations.txt see .txt detail
below
File containing time step rainfall (mm) for each

station ID.
evapstations.txt see .txt detail

below
File containing time step PET (tenths of mm) for

each station ID.
Input Parameters
Date
Interpolate Surface Method
Shapefile defining modeled basin region.
Data start date formatted m/d/yyyy

Two interpolation methods are displayed in the dropdown list:
Inverse Distance Weighted (IDW) interpolator
assumes each input has a local influence that
diminishes with distance; the points closer to the
processing cell are given a greater weight.
Spline interpolator is a general purpose method that
fits a minimum-curvature surface through the input
points.
Inverse Distance Weighted

(IDW) parameters
This is the default interpolation method.
Z Value Field
The field containing the values to be used in the

interpolation. Stations_i
62
Units
Values
For IDW specify either Nearest Neighbors or a

Fixed Radius in the interpolation. defaults Nearest
Neighbors.
Number of Neighbors / Fixed
Radius
Power
Number of Neighbors or input points defaults 12.

For Fixed Radius enter the radius to search for
points.
Controls the significance of the surrounding points on
the interpolated value the lower the number the
more influence from distance points. Defaults 2
For exercise 6 changed to 1.
A break that limits the search for input points.
Defaults No Barriers.
Barriers
Spline Parameters
Z Value Field
The field containing the values to be used in the

interpolation. Stations_i
Input Parameters cont.

Weight
No: of Points
Type
Output Grids
Rain_1999001
Rain_1999010
Evap_1999001
Evap_1999010
Input Grid
DEM
Output Grid
fill1
Weight factor.
Specifies the number of points to use per region.
Select from drop-down list:
Regularized a smooth surface.
Tension tunes the stiffness of the surface according
to the character of the modeled phenomenon.
Grid showing rainfall values interpolated across

subbasins originating from ground station data for
one day.
(10-day example)
Grid showing PET values interpolated across
subbasins originating from ground station data for
one day.
tenths of
mm
(10-day example)
GeoSFM Utilities: Sink Filling Menu
Description
Elevation data requiring fill sink process.
Description
Elevation data after fill sink process.
GeoSFM Utilities: Converting Time Series From Daily to Monthly and Annual
Input Files
Description
section 3.2
Input Parameter
Statistic to be computed
mm
List includes max and mean max chosen for

exercise.
63
Units
Values
Units
Values
Output Files
monthlyflow.txt see .txt file
detail below
Streamflow file contains a monthly time series of

subbasin.
annualflow.txt see .txt file detail

below
Streamflow file contains an annual time series of

subbasin.
statsparam.txt see .txt file detail

below
File containing the start year, start month, start day,

end year, number of simulation time steps, number of
catchments, statistical type, and a list of input/output
file names and paths.
monthlyflow.txt
Time
Date example 199901 format yyyymm (year,

month)
Streamflow file contains a monthly total time series
subbasin.
m3/s

annualflow.txt
Time
Date example 1999 year

Streamflow file contains an annual total time series
subbasin.
m3/s

Statsparm.txt
No Header
Start Year.
Start Month.
Start Day of the Month.
End Year.
Statistical Type.
Input file name and path:
C:\GeoSFM\workspace\streamflow.txt
Monthly Flow Output file and path:
C:\GeoSFM\workspace\monthlyflow.txt
Annual Flow Output file and path:
C:\GeoSFM\workspace\annualflow.txt
GeoSFM Utilities: Frequency Distribution Computation

Input File
Description
section 3.2
Input Parameters
64
Units
Values
Basin coverage/grid theme
Selected basply.shp for exercise.
Select statistics to be computed
Selected Total Flow, Mean Flow, Maximum Flow,

and Minimum Flow for exercise multiple choices
from a drop-down list.
Select field to display
Selected Mean for exercise multiple choices from a

drop-down list.
Output
Map
Map showing daily variation of streamflow values

displayed as selected statistic for the exercise Mean
was selected multiple choices are presented in
drop-down list.
65

Technical Manual For The Geospatial Stream Flow Model

Uploaded by

Copyright:

Available Formats

Technical Manual For The Geospatial Stream Flow Model

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Technical Manual For The Geospatial Stream Flow Model

Uploaded by

Copyright:

Available Formats

Prepared in cooperation with the U.S.

Agency for International Development, Office of

Technical Manual for the Geospatial Stream Flow Model

Open-File Report 20071441

U.S. Department of the Interior

U.S. Department of the Interior

For product and ordering information:

square kilometer (km )

square kilometer (km2)

acre-foot per day (acre-ft/d)

cubic meter per second (m /s)

cubic foot per second (ft3/s)

cubic meter per second (m3/s)

million gallons per day (Mgal/d)

Technical Manual for the Geospatial Stream Flow

Water Resource Research Institute, Utah State University, Logan, UT

U.S. Geological Survey

GeoSFM Process Map

Hydrologic Algorithm Services

Dynamic data sou rce

Static data source

Formatted Model Inputs

Formatted Model Outputs

Graphical Model Outputs

Data Processing Services

An ArcView extension was developed in Avenue (ArcViews in-built customization

Terrain Analysis Routine

Downstream Grid (DOWNSTREAM): A grid identifying the subbasin immediately

Basin Characterization Routine

Land Cover Description

Saturated Hydraulic Conductivity Parameter (KS): Contains a value of saturated hydraulic

Residence Time for the Interflow Reservoir (INTERFLOWLAG): Contains an interflow

Residence Time for the Baseflow Reservoir (BASEFLOWLAG): Contains a baseflow

Percent Impervious Cover (MAXCOVER): Contains a fraction of subbasin covered by a water

Local Evapotranspiration Adjustment (PANCOEFF): Contains a coefficient applied to

Mannings Roughness Coefficient (RIVMANNING): Contains a value of Mannings roughness

Dispersion Coefficient of the Diffusion Analog Equation (DIFFUSION): Contains a value of

Unit Hydrograph Generation Routine

where RH is the hydraulic radius

Urban and Built-Up Land

Dryland Cropland and Pasture

Irrigated Cropland and Pasture

Mixed Dryland/Irrigated Cropland and Pasture

Deciduous Broadleaf Forest

Deciduous Needleleaf Forest

Evergreen Broadleaf Forest

Evergreen Needleleaf Forest

Barren or Sparsely Vegetated

Tundra, Snow or Ice

Weather Data Processing Routine

Soil Moisture Accounting Module

Linear Soil Moisture Accounting (LSMA) Routine

Response file header

Soil water storage file

Regional groundwater loss file

Excess precipitation file

The FORTRAN language script, 1DBalance.for, computes the LSMA as documented

STORE(B, T - 1) RAIN(B, T) - STMAX(B)

RAIN(B, T) PARATIO( B, T) IMPRATIO