This article was downloaded by:[ITC]

[ITC]
On: 16 May 2007
Access Details: [subscription number 769788768]
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Geographical

Information Science
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713599799

Raster-network regionalization for watershed data

processing

To cite this Article: Whiteaker, T. L., Maidment, D. R., Gopalan, H., Patino, C. and
Mckinney, D. C. , 'Raster-network regionalization for watershed data processing',
International Journal of Geographical Information Science, 21:3, 341 - 353
To link to this article: DOI: 10.1080/13658810600965255
URL: http://dx.doi.org/10.1080/13658810600965255

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction,
re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly
forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents will be
complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be
independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or
arising out of the use of this material.
Taylor and Francis 2007

Downloaded By: [ITC] At: 10:13 16 May 2007

International Journal of Geographical Information Science

Vol. 21, No. 3, March 2007, 341353

Research Article
Raster-network regionalization for watershed data processing
T. L. WHITEAKER*{, D. R. MAIDMENT{, H. GOPALAN{, C. PATINO{ and
D. C. MCKINNEY{
{Pickle Research Campus, Bldg 119, MC R8000, University of Texas, Austin, TX 78712,
USA
{Florida Department of Environmental Protection, 3900 Commonwealth Boulevard MS
49, Tallahassee, FL 32399, USA
(Received 11 February 2005; in final form 3 August 2006 )
Difficulties exist in calculating watershed parameters from raster datasets over
large regions because of the excessive computation time involved. A technique is
presented which divides a large region into hydrologically distinct subregions, in
each of which raster analyses are performed, in order to efficiently process large
raster datasets. The results of raster analyses are stored as attributes on the
resulting vector data. The vector data are then merged, and appropriate values
accumulated to obtain regional parameter values for points of interest along the
stream network. The technique, called Raster-Network Regionalization, uses the
vector stream network as the backbone for the integration of subregions into a
single region. A case study is presented which utilizes the technique to prepare
geospatial inputs for the Water Rights Analysis Package simulation model.
Keywords: Regionalization; ArcGIS Hydro data model (Arc Hydro); WRAP
Hydro; Water Supply Modelling

1.

Introduction

The State of Texas manages the supply and demand of surface water through a
priority-based system of water rights. Water users must apply for permits and
adhere to rules in the Texas Water Code regarding surface-water usage. In 1997, the
Texas legislature passed Senate Bill 1, mandating improved management and
modelling capabilities for surface-water resources in Texas. As part of the bill, the
Texas Commission on Environmental Quality was given the task of developing
water availability models for the major river basins in Texas. These models help
water-resources planners understand how much water is in the system, how that
water is being allocated, and the impacts of additions or changes in water-rights
permits (Hudgens and Maidment 1999). However, previous methods used to
support water-availability modelling with Geographic Information Systems (GIS)
were cumbersome, requiring the processing of large raster datasets. The motivation
for the research presented in this paper is to overcome the limitations of raster
processing in support of water-availability modelling.
This paper discusses the concept of Raster-Network Regionalization, which refers
to the integration of subregional datasets into regional datasets for analysis
purposes. First, problems with raster-based analysis of large datasets are outlined,
*Corresponding author. Email: twhit@mail.utexas.edu
International Journal of Geographical Information Science
ISSN 1365-8816 print/ISSN 1362-3087 online # 2007 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/13658810600965255

Downloaded By: [ITC] At: 10:13 16 May 2007

342

T. L. Whiteaker et al.

followed by a discussion of how network-based accumulation improves upon raster

techniques. With this background, the Raster-Network Regionalization Technique
is then described. Finally, a case study utilizing Raster-Network Regionalization to
prepare GIS data for use in the Water Rights Analysis Package (WRAP) simulation
model is presented.
2.

Previous work

Raster-based analysis of GIS data provides useful parameters for hydrologic

modelling, water-quality assessments, and other water-resources applications
(Mason 2000, Osborne 2000). Distributed parameters can be summarized and
accumulated using a variety of raster tools available through the GIS. However,
problems arise as the size of the raster datasets increase, as when working with very
large areas or with very-high-resolution data. These problems include:
1. The time required to process rasters becomes unreasonable. Figurski (2001)
reported processing times on the order of 10 days for calculating watershed
parameters for the Trinity Basin at 30-m raster resolution (over 100 million
cells).
2. Data storage becomes cumbersome. The calculation of multiple grids covering
a vast number of cells leads to a large amount of data that must be stored on
disk and makes it difficult to transfer data to other parties.
3. The required grid size exceeds limits. In ArcGIS version 8, grids are limited to
a maximum size of 2 GB. If a raster analysis required a grid with a resolution
sufficient to exceed the 2 GB limit, then that raster analysis cannot be
performed. Note that this limitation has been removed for ArcGIS version 9.
Despite these drawbacks, the distributed analysis capabilities of raster processing
are too valuable to discard. Figurski (2001) developed a technique of cascading
parameters, so that rasters could be split into parts. However, the technique still
relies on rasters for accumulation routines and is prone to error due to the large
number of hand manipulations of data involved.
Streit and Kleeberg (1996) encountered the problem of transferring hydrologic
information from small catchment scales to basin scales and used generalization to
address the issue. However, Streit and Kleeberg also recognized that generalization
may introduce errors into the data. Schumann and Funke (1996) addressed this
problem by creating components for rainfall-runoff modelling which utilize new
scale-independent parameters. However, those applications were limited to a
maximum basin size of roughly 10 000 km2.
Abrahart et al (1996) developed an application which divided a study basin into
sub-basins for use in MEDRUSH, a GIS combined with distributed process model
designed to simulate hydrologic and vegetative processes important in studying
desertification. Each sub-basin contains a set of flow-strips, in which hydrologic
simulations occur. The total output from each flow-strip in a basin is converted to
water and sediment output for that basin. Each basin feeds its total output to the
stream network, where information is accumulated in the downstream direction.
The application may operate on basins up to 5000 km2 in area.
The research presented in this paper introduces a Raster-Network
Regionalization Technique, in which large basins are subdivided into subregions
suitable for the calculation of distributed watershed parameters. The subregional
data are then merged to provide an integrated regional dataset attributed with the

Downloaded By: [ITC] At: 10:13 16 May 2007

Raster-network regionalization for watershed data processing

343

necessary hydrologic parameters. The technique has been successfully applied to the
Rio Grande basin in Texas, which has a contributing area of over 550 000 km2.
3.

ArcGIS Hydro data model

The Raster-Network Regionalization technique utilizes the data structure and tools
provided by Arc Hydro. The ArcGIS Hydro data model (Arc Hydro) is a waterresources data model that uses GIS to capture the essence of surface water features.
The goal of Arc Hydro is to represent data in a manner that supports hydrologic
simulation modelling (Davis 1999). Arc Hydro defines a set of water-resources
feature classes, such as watersheds, cross-sections, and monitoring points. In
addition to prescribing an attribute structure and organization, Arc Hydro also
defines relationships between features, so that a watershed may know which point
represents its outlet (Maidment 2002). An Arc Hydro toolset populates the
attributes of Arc Hydro and performs operations using the Arc Hydro data
structure. Although Arc Hydro and the Arc Hydro tools were designed to work in
the ArcGIS environment, the data structure of Arc Hydro could in principle be
applied to any GIS environment.
4.

Methodology

The Raster-Network Regionalization Technique computes watershed parameters

from raster sources by (1) using zonal statistics to summarize raster values for a set
of watersheds and (2) using Consolidation and Accumulation techniques to pass
values from the watersheds to the stream network, and subsequently downstream
through the stream network. The technique is valid for hydrologic parameters that
may be summed or spatially averaged for watersheds such as drainage area, average
precipitation, and average curve number. Before proceeding further, a definition of
terms is useful:

N
N

Consolidation: the process of summing values from a set of features onto a

related feature participating in the stream network;
Accumulation: the process of summing values downstream through the stream
network.

Whiteaker (2001) developed Accumulation and Consolidation GIS tools that use
relationships between features to accumulate information from one feature to
another. The Consolidation tool transfers values from features in the landscape to
related features along the stream network. The relationships between features are
specified in attributes. For example, drainage area from watersheds can be
consolidated to the junction feature on the stream network that serves as the outlet
for each watershed (figure 1), if each watershed contains an attribute storing the
identifier for its outlet junction.
The Accumulation tool accumulates attributes downstream among junction
features in a stream network. Downstream navigation is performed by reading a
next downstream ID attribute on a given junction, which stores the identifier of the
next downstream junction in the network. Each junctions value (e.g. drainage area)
is added to the next downstream junctions value, so that the final value for a given
junction contains the sum of values from all upstream junctions plus its own value
(figure 1). The Accumulation tool is more complicated than the Consolidation tool,
because the Accumulation tool must navigate through a series of attribute

T. L. Whiteaker et al.

Downloaded By: [ITC] At: 10:13 16 May 2007

344

Figure 1. (a) Area from Upstream Watershed Consolidated in Outlet Junction. (b)
Accumulated value for the outlet junction: the sum of all upstream values plus the value at
the outlet.

associations until all upstream junctions have been accounted for, for a given
junction, whereas the Consolidation tool only navigates a single attribute
association for each feature. Also, the Accumulation tool is designed to work
within the same feature class (e.g. stream junctions), while the Consolidation tool is
designed to work with two different feature classes.
An important caveat for the Accumulation tool is that the tool is designed to
work with a dendritic stream network. If branching occurs, each side of the branch
may not be accounted for correctly, as only one value for the next downstream ID
can be stored for a given junction, even if that junction is upstream of two other
junctions on either side of a branch. Therefore, when branching occurs, the user
must decide to which branch the values should be accumulated, and assign the next
downstream ID accordingly.
The Accumulation and Consolidation tools allow for rapid summation of
attributes in the vector domain. Not only is vector accumulation typically faster
than computing large raster datasets, but the size on disk of vector data required to
describe the entire basin is typically much less than the amount of space required to
store flow-accumulated rasters covering the study area.
For example, consider the Guadalupe River Basin in Texas, which covers
15 462 km2, or roughly 17 million raster cells at a 30 m630 m cell size. As a test case,
this basin was divided into 2035 watersheds, with an outlet junction on the stream
network for each watershed. The accumulation and consolidation tools were used to

Downloaded By: [ITC] At: 10:13 16 May 2007

Raster-network regionalization for watershed data processing

345

calculate total upstream drainage area for each junction. On a 2-GHz computer with
512 MB of RAM, the process takes 22 min and requires an additional 0.04 MB of
disk space to compute the result. Computing the same values using traditional raster
methods such as those found in Figurski (2001) requires 1 h 43 min and 281 MB of
additional disk space.
Yet, despite the efficiency in using network-based analysis techniques, some raster
analyses must still be performed. So, how can the Accumulation and Consolidation
routines be applied to geospatial processing for water-resources applications?
The key is in the watersheds. Often, hydrologic parameters are defined and
calculated for watersheds. For example, in the traditional method of calculating
watershed parameters in a GIS, control points (points of interest) are overlain on a
flow-accumulated raster. A grid cell underneath a given control point is read to
determine the value for that control point. The cell represents a weighted flow
accumulation for a certain value, such as average curve number, which includes the
influence of all grid cells which flow (hydrologically) to that grid cell. Together, all
of those cells form the watershed for that point of interest. Therefore, by reading the
value of the cell underneath a given control point, the average value for the
watershed draining to that control point is determined.
The Raster-Network Regionalization Technique uses a different approach. A
summarization of raster values over watersheds is determined by using the
watersheds as distinct zones which define the area of analysis for a zonal statistics
tool in GIS. This tool calculates statistics such as mean, sum, max, and min for each
zone by reading the values of cells within each zone and performing the necessary
statistical operations. Thus, with this approach, accumulated grids whose cell values
are influenced by all upstream cells are no longer needed. The only cells that a
watershed is interested in are the cells that lie directly over that watershed.
Once attribute values have been determined for watersheds, these values are
consolidated to outlet junctions and then accumulated throughout the stream
network in the vector domain. The watersheds become the basic processing unit
with basin-wide coverage, while raster coverage is reduced to each individual
watersheds extent. Thus, watersheds effectively replace grid cells as the units of
analysis. (Note that if a vector watershed dataset is not available, the vector
watersheds can be determined from elevation data using standard GIS tools.
Techniques for watershed delineation are beyond the scope of this paper.)
This allows a basin or region to be divided into hydrologically distinct subregions,
in which the necessary raster analyses take place. The smaller size of the subregions
permits faster raster processing, with results from raster analyses stored on vector
watersheds. The Consolidation and Accumulation techniques described above are
then used to accumulate watershed parameters across the entire basin.
The Raster-Network Regionalization Technique involves the following steps
(figure 2):
1. Clip raster and vector data to hydrologically distinct subregions. Subregions
may be defined by using established watershed boundaries such as USGS
HUC boundaries.
2. Perform raster processing on each subregion, to obtain necessary values
summarized onto vector watersheds.
3. Merge the subregional vector data to a regional vector dataset, and update
connectivity for the outlet junction of each subregion to point to the nearest
junction in the next downstream subregion.

T. L. Whiteaker et al.

Downloaded By: [ITC] At: 10:13 16 May 2007

346

Figure 2. Steps in the Raster Network Regionalization Technique: (a) defining subregions,
(b) performing subregional raster analyses, (c) merging vector subregions, and (d)
accumulating vector attributes using the stream network.

4. Use the Accumulation and Consolidation routines to compute final values for
each point of interest on the stream network.
The benefits of the Raster-Network Regionalization approach are (1) processing
time is reduced, since much of the processing occurs in the vector domain rather
than in the raster domain; (2) data storage requirements are reduced, since
accumulation grids no longer need to be created; and (3) the remaining grids can be
split into hydrologically distinct regions defined by one or more watersheds,
resulting in faster raster processing, more modular data storage, and less of a raster
reprocessing effort if data in a given watershed change (since only those cells within
that watershed matter to that watershed). With Raster-Network Regionalization,
the weight of processing is shifted from the raster side to the vector side. Even the
largest basins can now be processed with high-resolution raster data.
5.

Case study: WRAP Hydro

WRAP Hydro is a preprocessing data model created to provide the Water Rights
Analysis Package (WRAP) with geospatial inputs regarding watershed parameters
and water-rights connectivity. In addition to the data model, WRAP Hydro also
consists of a toolset called the WRAP Hydro tools, which calculates those geospatial
parameters for WRAP. WRAP Hydro utilizes the Raster-Network Regionalization
Technique to handle large river basins.

Downloaded By: [ITC] At: 10:13 16 May 2007

Raster-network regionalization for watershed data processing

5.1

347

Water Rights Analysis Package

The Water Rights Analysis Package was chosen by the Texas Commission on
Environmental Quality (TCEQ) as the model to simulate water-allocation in
support of water-availability modelling. WRAP uses a record of monthly
naturalized streamflows (naturalized streamflows are those that would exist
naturally without the influence of man or man-made structures and diversions),
reservoir evaporation rates, basin characteristics (such as curve number and
drainage area), and mean annual precipitation rates to simulate the hydrology for a
basin. WRAP uses water-right priorities, rules, and target relationships to allocate
water during the simulation period using a monthly time step. Simulation results
include such information as reservoir storage levels and reliability indices for
meeting water-use requirements (Wurbs 2001).
To facilitate the calculation of control point and watershed parameters (basin
characteristics) for WRAP, Hudgens and Maidment (1999) created a set of tools
encapsulated in an ArcView 3 project file called WRAP 1117, which computes
average curve number, mean annual precipitation, drainage area, and connectivity
for the watershed draining to each control point on the stream network. The first
three parameters are calculated using raster analyses. The procedure involves
burning streams (reducing the elevation values along the stream so that once water
enters the stream, it does not jump back onto the floodplain) into a DEM, filling
sinks, calculating a flow direction grid, and then calculating flow accumulation
grids. The drainage area value over a given point is the number of cells accumulated
to that point, times the area of each cell. A weighted flow accumulation is used to
determine the average curve number and mean annual precipitation over a given
point. Thus, those three parameters are calculated for every single cell over the
extent of the DEM, and the value for a given control point is found by picking up
the grid cell value for the cell underneath that control point. The connectivity of
control points is determined by identifying the next downstream control point in the
stream network. Once the parameters are calculated for all control points, the results
are then input to a WRAP model (Hudgens and Maidment 1999).
WRAP 1117 encounters scalability problems when applied to large river basins.
In basins large enough to cover more than 100 million cells in a raster dataset,
processing demands may exceed a high-end desktop computers processing
capabilities (Figurski 2001). The Raster-Network Regionalization technique
presented in this paper overcomes the problems encountered by WRAP 1117 when
processing large regions.
5.2

WRAP Hydro Data model

WRAP Hydro utilizes a GIS to store geospatial inputs required to calculate

parameters for WRAP. WRAP Hydro, developed primarily by Gopalan (2003),
divides data into separate geodatabases and folder locations for each region in the
analysis. Raster data for each region are processed separately, with resulting
watershed parameters attributed to vector watersheds. The vector data for each
region are then merged to produce so that the Consolidation and Accumulation
tools can be run over the entire region.
The actual processing of WRAP Hydro data is facilitated by the Arc Hydro and
WRAP Hydro tools. The resulting data contain all of the geospatial inputs required
by the WRAP simulation model.

Downloaded By: [ITC] At: 10:13 16 May 2007

348
5.3

T. L. Whiteaker et al.
WRAP Hydro tools

The WRAP Hydro tools consist of a set of public domain utilities developed on top
of the Arc Hydro data model. These tools operate in the ArcGIS ArcMap
environment, with some of the functions requiring the Spatial Analyst extension.
The tools provide functionality in addition to the Arc Hydro tools in order to
calculate drainage area, average curve number, average precipitation, and the next
downstream identifier for controls points in WRAP. The WRAP Hydro tools use
Raster-Network Regionalization, with additional refinements to better insure
proper parameter calculation, such as the delineation of drainage areas to the stream
network lines rather than the junctions. The general process flow utilized by the
WRAP Hydro tools is shown in figure 3.
This section describes some of the core functionality of the WRAP Hydro tools
and how the tools are used to calculate WRAP parameters. Full help documentation
for the tools, as well as the tools themselves, may be downloaded from http://
www.ce.utexas.edu/prof/maidment/grad/whiteaker/hydrotools.html.
The WRAP Hydro tools include utilities for delineating watersheds, which allow
the user to choose a feature class of point, line, or polygon geometry to serve as the
outlet zones for the watersheds. Polygon and line geometries are more secure than
points for delineating watersheds, because if a point is not over a cell within a
natural channel in the digital elevation model, then the watershed that is delineated
for that point will not define the intended drainage boundary. With points or lines

Figure 3.

WRAP Hydro processing flow chart.

Downloaded By: [ITC] At: 10:13 16 May 2007

Raster-network regionalization for watershed data processing

349

Figure 4. Use of different features as outlet zones for watershed delineation yielding
different results (only a portion of the watershed formed from using edges as outlets is
shown).

serving as the watershed outlets, there is a much greater chance that the geometries
will intersect a cell in the channel of the DEM. This approach becomes particularly
useful in flat areas, where the flow direction is more ambiguous (figure 4).
Once watersheds have been delineated, the WRAP parameters of average curve
number, average precipitation, and drainage area may be calculated using the
WRAP Hydro tools. The drainage area is simply read from the shape area of the
watershed feature, which is calculated automatically by ArcGIS, and then converted
to desired units of analysis. The average curve number for a given watershed is
found by laying a curve number grid over a watershed and then using zonal statistics
to computer the average curve number value for all cells over that watershed
(figure 5). The same process is used to calculate average precipitation for watersheds.
Once parameter values have been calculated for each watershed, they can be
accumulated onto control points using the consolidation and accumulation
techniques described above. The WRAP Hydro tools use those routines without
modification to accumulate drainage area values. For average curve number, each
watersheds curve number value is multiplied by that watersheds area. This
weighted curve number is then accumulated for the control points. The final
accumulated value is divided by the total combined area of upstream watersheds to
provide an average curve number for the entire drainage area of a given control

Figure 5.

Zonal statistics providing an average curve number for each watershed.

Downloaded By: [ITC] At: 10:13 16 May 2007

350

T. L. Whiteaker et al.

point, as shown in equation (1).

n
P

Ai CNi
CNj ~ i~1 n
,
P
Ai

i~1

where: CNj5average curve number for control point j; Ai5drainage area for the ith
watershed draining to control point j; CNi5curve number for the ith watershed
draining to control point j; n5number of watersheds draining to control point j.
The average precipitation for each control point is found in the same manner.
Once these attributes have been calculated, the results may be exported for use in the
WRAP simulation model.
With its well-organized data model and custom toolset, WRAP Hydro provides a
useful mechanism for managing Water Availability Modelling (WAM) data, and
calculating parameters required to run WRAP simulations. The Texas Commission
on Environmental Quality has adopted WRAP Hydro for its WAM projects. The
Raster-Network Regionalization Technique is vital to that effort, as the Rio Grande
comprises a contributing area over 550 000 km2, or well over 500 million grid cells at
30-m resolution (figure 6).
This case study has illustrated an example of an Arc Hydro-based Preprocessing
Data Model and application of the Raster-Network Regionalization Technique.
Both of these methods of geospatial integration facilitate the use of a GIS to prepare
input parameters for simulation models. Full details of WRAP Hydro, including the
role of each feature class in the data model and further elaboration on the use of the
WRAP Hydro tools, may be found in Gopalan (2003).
6.

Conclusions

This paper has illustrated a Raster-Network Regionalization Technique which

utilizes raster-based analysis at the subregional scale and network-based attribute
accumulation at the regional scale in order to process large regions in an efficient
manner. The benefits of raster-based analysis are preserved at the subregional level,
so long as the analyses are independent of other subregions, such as with zonal
statistics operations. These types of operations are useful in obtaining hydrologic
information pertaining to each watershed.
Subregions are integrated into regions through the vector datasets and relationships. The stream network provides the backbone for connectivity between
subregions, with watersheds being related to the network through their outlet
junction on the network. The Raster-Network Regionalization Technique has been
applied to the analysis of large river basins in Texas. The technique could also be
applied at a local level when high-resolution data, such as LIDAR data, are available.
An implication of the Raster-Network Regionalization Technique is the
categorization of datasets into one of three scales: Local, Subregional, and
Regional. Local datasets contain the most detailed information about an area,
such as detailed channel geometry or LIDAR data, and are useful for supplying
information to models with a local scope, such as a hydraulic model. These datasets
may be maintained by local entities and tend to vary greatly between locations.
Certain information from several Local datasets may be generalized and used to
create a Subregional dataset, which covers a much larger area than a single local

Downloaded By: [ITC] At: 10:13 16 May 2007

Raster-network regionalization for watershed data processing

Figure 6.

Rio Grande Basin.

351

Downloaded By: [ITC] At: 10:13 16 May 2007

352

T. L. Whiteaker et al.

dataset. These datasets may be maintained by state or federal agencies and provide
the most commonly used types of subregional information, such as the stream
network or digital elevation models at a 30-m resolution. Subregional datasets may
be integrated using the Raster-Network Regionalization Technique to form
Regional datasets, which allow analysis over very large areas.
A powerful conclusion from this research is that next downstream ID assignment
is critical to the success of regionalization. The ID enables the connection between
features in the landscape, including:

N
N
N

the connection of watersheds to outlet junctions;

the connection of junctions with next downstream junctions;
the integration of subregions into regions, through the update of the next
downstream ID in the most downstream junction in each region

The limits of network-based analysis were not reached in this research as far as
computing time and processing power are concerned, even when analysing river
basins such as the Brazos River basin, with an eastwest expanse larger than the
state of Texas. Thus, with Raster-Network Regionalization and Arc Hydro, the
same types of hydrologic analyses may be performed in the GIS, independent of
scale.
Acknowledgements
This research was financially supported by the GIS in Water Resources Consortium,
the National Council of Science and Technology (CONACYT), the Texas
Commission on Environmental Quality, the US Army Corps of Engineers, and
NAD Bank.
References
ABRAHART, R.J., KIRKBY, M.J., MCMAHON, M.L., BATHURST, J.C., EWEN, J., KILSBY, C.G.,
WHITE, S.M., DIAMOND, S., WOODWARD, I., HAWKES, J.C., SHAO, J. and
THORNES, J.B., 1996, MEDRUSHspatial and temporal river-basin modelling at
scales commensurate with global environmental change. In Application of Geographic
Information Systems in Hydrology and Water Resources Management, K. Kovar
and H.P. Nachtnebel (Eds), pp. 4754 (The Hague, Netherlands: Krips Repro,
Meppel).
DAVIS, K., 1999, Object-oriented modeling of rivers and watersheds in geographic
information systems. Thesis, The University of Texas at Austin. Available online
at: www.crwr.utexas.edu/reports/2000/rpt00-7.shtml (accessed 6 June 2004).
FIGURSKI, M.J., 2001, GIS algorithms for large watersheds with non-contributing areas.
Thesis, The University of Texas at Austin. Available online at: www.crwr.utexas.edu/
reports/2001/rpt01-7.shtml (accessed 6 June 2004).
GOPALAN, H., 2003, WRAPHydro data model: finding input parameters for the water rights
analysis package. Thesis, The University of Texas at Austin. Available online at:
www.crwr.utexas.edu/reports/2003/rpt03-3.shtml (accessed 6 June 2004).
HUDGENS, B.T. and MAIDMENT, D.R., 1999, Geospatial data in water availability modeling.
Thesis, The University of Texas at Austin. Available online at: www.crwr.utexas.edu/
reports/1999/rpt99-4.shtml (accessed 6 June 2004).
MAIDMENT, D.R. (Ed.), 2002, Arc Hydro: GIS for Water Resources (Redlands, CA: ESRI
Press, 2002).
MASON, D., 2000, An analysis of a methodology for generating watershed parameters using
GIS. Thesis, The University of Texas at Austin. Available online at: www.crwr.utexas.edu/reports/2000/rpt00-3.shtml (accessed 6 June 2004).

Downloaded By: [ITC] At: 10:13 16 May 2007

Raster-network regionalization for watershed data processing

353

OSBORNE, K.G., 2000, A water quality GIS tool for the City of Austin incorporating nonpoint
sources and best management practices. Thesis, The University of Texas at Austin.
Available online at: www.crwr.utexas.edu/reports/2000/rpt00-10.shtml (accessed 6
June 2004).
SCHUMANN, A.H. and FUNKE, R., 1996, GIS-based components for rainfall-runoff models. In
Application of Geographic Information Systems in Hydrology and Water Resources
Management, K. Kovar and H.P. Nachtnebel (Eds), pp. 477484 (The Hague,
Netherlands: Krips Repro).
STREIT, U. and KLEEBERG, H., 1996, GIS-based regionalization in hydrology: German
priority programme on spatial transfer of hydrological information. In Application of
Geographic Information Systems in Hydrology and Water Resources Management, K.
Kovar and H.P. Nachtnebel (Eds), pp. 485491 (The Hague, Netherlands: Krips
Repro).
WHITEAKER, T., 2001, A prototype toolset for the ArcGIS Hydro data model. Thesis, The
University of Texas at Austin. Available online at: www.crwr.utexas.edu/reports/
2001/rpt01-6.shtml (accessed 6 June 2004).
WURBS, R.A., 2001, Reference and Users Manual for the Water Rights Analysis Package
(WRAP) (College Station, TX: Texas Water Resources Institute, 2001).