Exponential Growth Of Software And Hardware Department

Published Date: 02 Nov 2017

1 Introduction

The exponential growth of software and hardware developments and data availability have greatly influenced information systems in the last two decades and Geographic Information Systems(GIS) like other information systems have also been shaped by these developments. Today, GIS based methods are increasingly being used in hydrological modelling, geomorphology and cartography for terrain analysis. This is due to the availability of quality digital elevation model (DEM) coupled with the increasing computer power to process DEM data and extract topographic attributes. An example of topographic attribute commonly extracted from DEM data is the channel networks.

Extraction of channel networks is important for calculation of hydrologic parameters such as length, area, drainage density and shape. For hydrological modelling purposes, it is important that the extracted hydrological parameters from DEM matches with the property of channel network from field or high resolution topographic map. Therefore, there is the need for channel networks to be extracted from DEMs at an appropriate resolution or scale.

The simplest and most common method for extracting channel network was proposed by (O’Callaghan and Mark 1984). The proposed procedure is fill pits in DEM, calculation of flow direction, calculation of the accumulation drainage area and finally defining channels as cells exceeding accumulated support area threshold (Tarboton 1991). This method treats all cells with a drainage area greater than the support area threshold as streams and all others as non – stream cells. The selection of the support area threshold value has a significant influence on the resultant stream network. A lower support area threshold value will produce a larger and dense stream network whereas a higher support area threshold value will produce a small and less dense stream network. Therefore the resolution of the extracted stream network is control by the support area threshold. The selection of the justifiable support area threshold is one of the most important factors in the extraction of the stream networks from DEMs

A significant challenge is to select that appropriate threshold value that will result in delineating stream network from DEM whose properties corresponds to properties of networks obtained from traditional method, such as from high resolution topographic map (Tarboton et al. 1991).Towards this end, (Tarboton et al. 1991) have suggested a criterion for determining the appropriate resolution (drainage density) at which to extract stream channel networks from digital elevation data. The proposed method is to extract the highest drainage density network that satisfies constant drop property, which has been proved to hold for channel networks from traditional methods.

Stream drop is defined in literature as the difference in elevation between the beginning and end of Strahler streams. Constant drop property is therefore defined as the smallest support area threshold that produces a channel network where the mean stream drop in first order streams is not statistically different from the mean stream drop in higher order streams. The constant drop property has been proposed to be an empirical geomorphological attribute of properly graded channel networks that has a physical basis in terms of geomorphological laws governing drainage network evolution (Tarboton et al., 1992).

This study focuses on developing an ArcGIS toolbar (add-in) extension for the extraction of stream network using python programming. It provides a comprehensive user interface for the extraction of stream networks. The add-in is based on the constant stream drop law which was proposed by Tarboton et al. 1991 for the extraction of the channel networks, providing objectivity in the selection of the threshold for the drainage density.

In this case, existing desktop geoprocessing tools from ArcGIS Toolbox are extended with python to provide the required customize individual tools. ArcGIS script wizard is use to build a complete script tool combining all the individual tools developed. ArcGIS Python Add-In Wizard model is then used to build the final toolbar extension (plug-in) by integrating the script tool within the python add-in model for the extraction of the stream network.

Problem statement

Field mapping and topographic maps are still in practice the most precise methods in delineating channels and drainage networks because of their reliabilities of field observation. It is impractical to use these methods on a large scale because it is time consuming, labour intensive and requires more expertise in digital conversion. The easy availability of DEM has further promoted the use of various GIS software’s for the extraction and delineation of drainage network. Even though DEM approach is the current major approaches for terrain analysis and extraction of drainage networks, there is limitation in comparison with the real drainage network as far as some of its geomorphological properties are concern. The commercial GIS products put restriction for further research for the enhancement of their tools due to high cost and license for specific tools, therefore users are provided with general purpose tools rather than what are required by specific users.

Open source software (tools) provides solution to the scientific accountability such as geomorphological properties and also provides users with specific tools required for their needs. Unfortunately, many of the open source software packages lack visualization and general spatial analysis capabilities of a GIS. Therefore these open source models are developed as extensions to existing commercial software. The extensions are used to provide the flexibility and advanced functionality needed by the users. For example, in the case of stream network extraction, it is better to use ESRI ArcGIS extensions (add-in) for the extraction of the stream networks such that the properties of the resulted stream network extracted from DEM corresponds with the properties of stream networks obtained from traditional method.

Significance of the Research

Different algorithms exist for automatically extracting channel networks and delineating watersheds from DEM, therefore resulting in the development of different models (tools). The most common method of extracting channel networks from DEM is to specify a critical constant (arbitrary) support area threshold that defines the minimum drainage area required to initiate a channel. Most commercial GIS software’s like ESRI ArcGIS use this method to follow a particular workflow to delineate stream networks .This method is limited by the fact that the arbitrary stream network delineated does not always satisfy any geomorphological properties as compared to that of the traditional methods. Furthermore, the continuous changing of support area threshold on a trial and error bases until the resulted stream network properties corresponds to the properties of the network from filled leads to wasting of time. An alternative to the arbitrary threshold selection method describe above is implemented in Terrain Analysis Using Digital Elevation Models (TauDEM). TauDEM is a set of Digital Elevation Model (DEM) tools for the extraction of stream networks and analysis of hydrologic information from DEM. It is implemented as an ArcGIS toolbar extension using Visual Basic, C++ and the ESRI ArcObjects library. It extract stream network that corresponds with the stream network from traditional method using constant drop property (Tarboton 2003).

However, with the increasing computer processing power becoming cheaper and the easily availability of cross platform languages like python, users are demanding user friendly tools with no difficulties like COM registration and execution of process through command windows. This has necessitated the development of an ArcGIS toolbar extension in line with TauDEM by extending existing desktop geoprocessing tools from ArcGIS Toolbox with python to provide comprehensive user friendly toolbar for the extraction of stream networks and watershed from DEM. The significance of python ArcGIS extension for stream network delineation is that, it increases the potential of ArcGIS as a means for stream network extraction that fulfils specific scientific requirement, in this case geomorphological laws.

To achieve this goal a number of challenges need to be overcome. First, the existing TauDEM contains individual geoprocessing tools for the extraction of the stream network from DEM such as pit removal, flow direction, move outlet to stream among others. However, they totally depend on third party software (MPICH2 library) for them to function. There is a challenge in the installation of MPICH2, the procedure follows series of steps which is difficult for inexperienced people and the installation determines the efficiency of TauDEM. Further, there is also an additional challenge in the execution of the tools due to the fact that it is executed through command line which is inconvenient to new users. Python as a cross platform language has been identified as a programming language for the development of an ArcGIS extension due to the fact that ArcGIS support python as its main scripting language. Python add-in does not require compiling DLLs to build. Python is a cross platform that does not depend on any third party software .The creation of user friendly interface is easier using python to compensate for the execution of tools through command line.

Secondly, it is important to examine the processes that went into the development of TauDEM. TauDEM is based on D8 flow direction model, which is the traditional method of directing flow from one grid cell or pixel into one of its eight neighbours. However, the D8 approach has disadvantages arising from the discretization of flow into only one of eight possible directions (Tarboton, 1997).A possible solution to this challenge is suggested by Tarboton (1997). He suggested D infinity procedure which partitions flow from a cell into two adjacent cells. Another toolbar examines the impact of using Dinfinity flow direction method in delineating watershed and stream network extraction.

In terms of significance to the scientific community, TauDEM uses T statistics for the comparison of means of different populations to determine whether constant drop property is achieved because stream drops are highly variable. The validity of constant drop property determines whether the extracted stream network is corresponding with that from traditional method. However normal distribution is preferred over the T distribution when sample contain more than 30 values (Paul and Charles, 1997), therefore it necessitate the use of different statistics to verify the validity of the constant drop property. Furthermore, with any open source tool, the scientific impact which is easily appreciated is that of providing a basis for future research. Python add-in is easy to understand and the toolbar is an extension of an existing geoprocessing tools, therefore making it easier for modification by any interested person.

The significance in business perspective especially for multipurpose companies is also high as modern hydrologic research and water resource companies rely heavily on computer-based tools that allow users to manage and analyze data in an efficient ways that would not have been possible without the aid of software capabilities. It is important that such tools contain easily understandable functionality and be accessible to a broad spectrum of users. Python add-ins can be installed on a general server and be available to all ArcGIS users, unlike TauDEM which is available to an individual computer. They are also easier to share, install, delete and require no com registration, therefore improves flexibility and efficiency.

In the view of the foregoing, this study aims to answer the following questions:

What is the feasibility of extending existing ArcGIS geoprocessing tools available using python for developing a comprehensive user friendly ArcGIS plug-in for automatic stream network extraction.

How effective is to extract stream networks objectively using constant drop properties of the stream

Stream drop are highly variable, how effective is to verify the validity of constant drop property using different statistical methods.

How applicable is using multiple flow direction method in delineating stream networks.

How applicable is initiating watershed delineation at points of specific interest other than junction.

1.3 Objectives.

The overall aim of this study is to develop and evaluate a set of ArcGIS add-ins that takes input from digital elevation model (DEM) to objectively extract drainage networks that is compliant to TauDEM using python. The specific objectives of this study are:

Develop an ArcGIS add-ins for delineating watersheds at a point of specific interest using Python.

Extract channel network based on objective procedure for drainage density estimation in compliant with TauDEM.

Evaluate ArcGIS add-ins for channel network extraction using D infinity flow direction method rather than D8 flow direction.

Evaluate ArcGIS add-ins using F test to validate the constant drop properties of a channel.

1.4 Area of Study

The study area for the test runs of the software is Linach Dam with the purpose of extracting Linach Valley. Linach Dam is located in a small valley of the Black Forest in South-West Germany.

linach

Figure 1-1: Location of Study Area.

http://images1.everysingleplace.com/features/1324145-linach-germany-information

1.5 Data and Software

The data sources used for this project consist of digitized contour lines, digitized stream lines and digitized singular height points from topographic map 1:25 000 for the Linach valley. The datasets were used to provide detailed DEM used for the project. These datasets were acquired from Stuttgart University of Applied Sciences (HFT) winter semester (2012) project.

In terms of software, for creation of Digital Elevation Model (DEM), ArcGIS 10.1 geoprocessing framework comprising spatial analyst in Arc Toolbox is used. Python 2.7 through ArcPy site package is used for the extension of geoprocessing tools of ArcGIS. Further, the ArcGIS Python Add-in Wizard is used for the creation of ArcGIS add-ins

Table 1-1: Software used in the study

Software

Purpose

ArcGIS 10.1

DEM creation

Python 2.7

Extension of existing geoprocessing tools

ArcGIS 10.1 Python Add-in Wizard

ArcGIS 10.1 add-in creation

1.6 Thesis Structure

This study is organized into seven main chapters as follows.

In chapter one, a general overview of the study is provided beginning with an introduction, problem statement, elaborations of significance of the research and the total objectives of the study. A description of the study area for testing the software follows as well as the data and software used in the research.

Chapter two focuses on the background of the study including, fundamental techniques for automatic delineate watersheds, a review of geomorphological properties and statistical methods, ArcGIS Add-Ins and review of related research in stream network extraction.

In chapter three, a review of existing method and tools for watershed and stream network delineation is presented

The methodology adopted for this research is described in chapter four including creation of individual tools and development of ArcGIS add-in using python.

Chapter 5 evaluates the python add-in created from the previous chapter with respect to tools for extraction of stream network in the literature.

Chapter 6 provides an analysis of the overall implementation of the python add-in workflows and result.

Conclusions and recommendations of the study are presented in chapter seven while chapter eight contains the references and bibliography used in the research.

2 Background

This chapter begins by providing background information regarding watershed and the fundamental techniques used in automatic delineating watersheds. Further, it provides a review of few geomorphological properties of streams and statistical methods. The chapter then concludes with a presentation of related work.

2.1 Watershed

The integral component to most scientific research is that of a study area, the location where the research is focus. Watershed is one of the most common study areas in hydrology. Watersheds, also known as contributing area or sub-basin is normally defined as the total area of land ,defined by topographic divides , that drains surface water to a given outlet point or pour point(kang-Tsung,p298).Mapping of watersheds depends on the catchment drainage pattern of the watershed. Hence, it depends on the relief of the area. Traditional method of delineating watershed from topographic maps has proved to be successful. However, today different computer programs are used to delineate watershed in a fraction of the time needed for the traditional method.

Watershed analysis is referred to in literature as the process of using DEMs and the power of computer technology to automatically delineate watersheds and to obtain topographic features like stream networks. The automatic watershed delineation is either area – base or a point based method. An area-base method produces a series of sub-basins, one for each stream section. A point base method, on the hand produces a watershed for each used point. The point can be location of an outlet of the network or a gauge station. The automatic method for delineating watersheds from DEMs follows a series of steps which is used for data manipulation. The type of DEM use and the fundamental processes are described below.

2.2 Fundamentals in watershed delineation.

The automatic method for delineating watersheds from DEMs follows a series of steps which is used for data manipulation. In what follow, the standard methodology for working with DEMs to automatically delineate watershed and methods for channel network extraction is reviewed.

2.2.1 Digital elevation model (DEM)

The grid base DEM is a cell- based data representation of continuous topographic surface of the earth. Grid-base DEMs have seen widespread application in the automatic watershed delineation than any of the other DEMs types. This is due to the fact that it is readily available and simple to use (Moore et al., 1991). However, there are open concerns which need to be considered in using grid-base DEM for hydrological analysis like extraction of channel networks. The resolution of the DEM has major influence on the watershed analysis. The finer-resolution DEM data will not always give improved spatial accuracy result. There is always a trade off which need to be considered when using DEM to delineate watershed. The knowledge of the study area is very important before using the DEM for the extraction of the stream network (Michael N.DeMers, 2002).A stream network extracted from 10-meter DEM for example will produce more detailed stream network than 30-meter DEM((kang-Tsung,p304).However, higher – resolution DEM like 10-meter also tends to produce smaller watershed area than a lower resolution DEM like 30-meter does (Usery et al.2004).

2.2.2 Fill Pit

Filling pits in DEM is one of the important steps needed to be carried out as a first step in automatic extraction of drainage networks and delineation of watersheds. A filled DEM is free from sinks. Pit is a cell or cells surrounded by cells with higher elevation values, thus creating an area of internal drainage. The presence of pits in DEM create sinks which prevents the smooth flowing of water and result in creating errors in the delineation of watershed. Some pits are real, however many are imperfections in the DEM.The most common method for removing pit is to increase its cell value to the lowest overflow point out of the pit (Jenson and Domingue 1988). This method results in flat surface which needs to be considered in delineating watershed. An approach in compensating the flat surface proposed by (Garbrecht and Martz 2000) is to impose two shallow gradients and to force flow away from higher terrain surrounding the flat surface toward the edge bordering lower terrain. Other researchers like Band, 1986 have also proposed different methods in filling pits in DEM. The method used by ArcGIS which is also implemented in the study is the Jenson and Domingue method which raises the average elevation of the terrain but creates artificial flat areas. For many applications Fill tool can be used but some precise analysis can be negatively affected.

2.2.3 Flow Direction

One of the most fundamental steps in watershed delineation process is the Flow Direction. A flow direction raster shows the direction water will flow out of each cell of a filled DEM using the elevations in the filled DEM. The accuracy of this step will influence the streams flow paths, subwatersheds boundaries and the outlets of the stream. A widely used method for deriving flow direction is the Deterministic (D8) method. The Deterministic (D8) method assigns a cell’s flow direction to one of its eight surrounding cells that have the steepest distance weighted gradient. In ArcGIS, it is encoded to correspond to the orientation of one of the eight cells that surround the cell under consideration using 1, 2,4,8,16,32,128. However there exist some problems with D8 method as recognized by various authors. Tarboton (1997) stated that the D8 approach has disadvantages arising from the discretization of flow into only one of eight possible directions, separated by 45.

Multiple flow direction methods have also been suggested. These proportion flows from a grid cell between more than one downstream grid cells, and by doing this avoid some of the biases associated with D8 flow directions, resulting in better estimates of drainage area, especially on hill slopes (Tarboton and Ames, 2001).An example of the multiple flow direction is the D Infinity method which partitions flow from a cell into two adjacent cells (Tarboton, 1997). Figure 1 below illustrates Dinfinity (D∞) flow direction method algorithm. The D∞ multiple flow direction models is proposed to be useful for the calculation of specific catchment area where flow is dispersed on a hillslope and can be readily extended to include weighted accumulation. This study will make use of both D8 and D¥ flow direction methods to investigate the extraction of the drainage networks.

Figure 2-1: The DÂ¥ Algorithm (Source: Tarboton, 1997)

2.2.4 Flow Accumulation

The third fundamental principle makes use of the flow direction raster data set to create the flow accumulation raster data set, where each cell is assigned a value equal to the number of cells that flow to it. There are different types of flow accumulation (contributing area), because contributing area is dependent on the flow direction. This study will focus on flow direction due to D8 and DÂ¥ flow direction methods respectively. It is stated by Tarboton (2003) that the contributing area calculated using the DÂ¥ approach is smoother and avoids bias as compared to the D8 approach.

A flow accumulation is interpreted by Chang (2010) in two ways. First, stream channels are considered as cells having high accumulation values, whereas cells having an accumulation value of zero are proposed to correspond to ridge lines. Second, if flow accumulation is multiplied by the cell size, the accumulation value equals the drainage area.

2.2.5. Stream Network

A variety of methods have been developed to process raster DEMs automatically to extract stream network. Using a constant threshold value to extract stream networks automatically from flow accumulation raster is the most popular method. Extracting a stream network with a threshold value of 100, for example, means that each cell of the stream network has a minimum of 100 contributing cells. Given the same flow accumulation raster, a higher threshold value will result in a less dense stream network and fewer internal watersheds than a lower threshold value. Therefore the threshold value is a major requirement to watershed delineation.

2.2.6. Stream Link

After the overall stream network is derived from a flow accumulation raster, they need to be broken up into individual segments. A stream link raster is the result of the broken stream network into segment. Each section of the stream link raster is assigned a unique value and is associated with a flow direction. Researchers have describe stream link raster as topology-based stream layer where the intersections are like nodes and the stream segments between nodes are like reaches or arcs.

In ArcGIS, the output of the stream link tool can be used as data input for watershed tool to create watershed based stream junctions (ArcGIS resource centre).

2.2.7 Area – based watershed

Area – based watershed is a process where the watershed for each stream section is delineated. This is done by using the flow direction grid and the stream link grid above to delineate the sub-watersheds. Therefore the accuracy of delineating sub-watersheds using stream link depends on the accuracy of the flow direction and the stream link (Chang, 2010).The above description of Area – based watershed is similarly implemented in ArcGIS.

2.2.8 Point – Based Watersheds

Pour points (Outlets) are normally defined in literature as the points of interest in watershed analysis. The points of interest can be stream gauge locations or dams. Point-based watershed is delineated based on the points of interest or pour points. However, the procedure for delineation of watershed based on pour points follow the same procedure as in delineating watershed based on stream link. The only difference is the replacement of stream link with pour points. For most systems like ArcGIS, the pour point must be located over a cell that is part of the stream network. If the pour point is not located directly over a stream, it will result in a small, incomplete watershed. The point based watershed delineation relies on the accuracy of the location of the pour point. For this reason, various systems have developed an algorithm to ensure that the pour point is located over a cell on the stream network. For example, ArcGIS has a command that can snap a pour point to a stream cell with the highest accumulation value within a user – defined search radius.

Furthermore, the algorithm for delineating point –based watershed varies according to the location of the point of interest. If the point of interest is located at a junction, then the watersheds upstream from the junction are merged to form the watershed for the point of interest. Also, if the point of interest is located between two junctions, then the watershed assigned to the stream segments between the two junctions is divided into two, one upstream from the pour point and the other downstream. The upstream portion of the watershed is then merged with watersheds further upstream to form the watershed for the pour point (Chang, 2010).

2.2.9 Vector Conversion

The calculation of the terrain characteristics like drainage density, shape and area are based on vector data. Once the stream segment grid and watershed grid are formed, it is a relatively simple process to convert the grids into a vector format. In ArcGIS for example, streams are converted to polyline shapefiles using stream to feature tools and also watershed grid turned to watershed polygon using raster to polygon tool in Arc toolbox. Most systems calculate various important properties for each of vector stream segments and the watershed polygons. These parameters include length, area, average elevation etc. All these are important for the study of watershed analysis.

2.3 Review of stream networks properties

This chapter review some of the basic quantitative description of stream networks as were defined in various literatures. The stream network is describe like a tree whose trunk is the outlet point or lowest point and sources are points farthest upstream. The point of intersection of stream channels is termed junction or node. Exterior links also known as outside links, are the segments of channel between a source and the first junction downstream ,whereas inside links are the segments of channel between two successive nodes or a node and the outlet(Tarboton et al,1991).Further, each stream segment or link also has certain properties. Elevation differences (drop), average slope, stream length, local area, the area draining directly into segments are examples of properties of stream segment.

The most common approach for classifying stream networks is the stream ordering system. Stream ordering is a method of assigning numerical values to streams based on the number of tributaries branching from each trunk stream. For example, first-order streams are streams with no tributaries. There are different methods use for stream ordering, each of which ranks the branching streams differently. The ordering systems implemented in ArcGIS are proposed by Strahler (1957) and Shreve (1966).

The two methods assign an order of 1 to all streams with no tributeries; however they differ in the classification of the subsequent orders. Strahler method increases the stream order as each link is encountered. For example, the intersection of two first-order streams result in a second-order link. The order increases only at the intersection of streams of the same order. Unlike Strahler method, Shreve method increase the stream order at the intersection of two streams irrespective of whether they are of the same order or not. For example, the intersection of a first order stream and second order stream creates a third order stream. Strahler order is described in various literatures as a topological, dimensionless measure of size or scale of a channel segment or network.

The drainage density is defined by Horton (1945) as the ratio of the total length of all streams within a watershed to the total contributing area of the watershed. Further, drainage density is proposed as a physical scale associated with the dissection of the landscape by a river network.

2.4. STATISTICAL TESTING

The statistical test provides a mechanism for making quantitative analysis about processes. It is a statement that determines whether an estimated value matches with a given theoretical value or another independent estimated value within a given probability. It provides the bases to determine whether to reject or accept a hypothesis about a process. The test statistic is computed from the sample data and is the value used to determine whether the null hypothesis should be accepted or rejected. When the null hypothesis is rejected, it is said that the sample statistic computed is not consistent with what is expected from the population. For example, suppose there is interest in selecting a threshold for delineating stream networks where the difference in drops of first order and higher order is not greater than 2. The null hypothesis in this case is that the difference in drops of first order and higher order is not greater than 2.

On the other hand, an alternative hypothesis is what is accepted when a decision is made to reject the outcome of a test. For example the alternative hypothesis that is being guarded against in the example above, is the selection of a threshold where the difference in drop is higher than 2. Therefore, a statistical test requires a pair of hypotheses; namely, a null hypothesis and an alternative hypothesis.

When a decision is made concerning the acceptance of a test, there is a probability of making a wrong decision since one can never be 100% certain about a statistic or a test. For example, using a confidence interval of 95% means there is a 5% chance that the decision is wrong. The 95% is termed the significance level of the test (www.itl.nist.gov/div898/handbook/prc/section1/prc13.htm,).

2.4.1 Difference of two arithmetic means test (Two-Sample t-Test for Equal Means)

At times it may be desirable to test a sample mean against a known value. The two-sample t-test is used to determine if two population means are equal. A common application is to test if process 1 is equivalent to process 2 (www.itl.nist.gov/div898/handbook/prc/section1/prc13.htm,). In this case, we have two sets of observations. And we have two alternatives: Either the variances of the two sets of observations are equal or they differ from each other.

In this study, it is assumed that the variances of the two sets of the observations are equal. Therefore the following test scheme below is used.

Null hypothesis: µ1 = µ2 and alternative hypothesis: µ1 ≠ µ2

Test Statistic:

Where

D is the t-distribution value calculated

n1 and n2 are the sample size, and are the sample means and s1 and s2 are the sample standard deviations

Critical Region: Reject the null hypothesis that the two means are equal if:

T1 – α, n1 + n2 – 2 out of the statistical table of Fs ( t, f)

Where f = n1 + n2 – 2 degree of freedom

(GSM lecture note,HFT)

2.4.2 F-Test for Equality of Two Variances

Another statistical procedure is used to compare the ratio of two population variances. An F-test is used to test if the variances of two populations are equal. The test is either a two-tailed test or a one-tailed test. The two-tailed test is use to test against the alternative that the variances are not equal. Whiles the one-tailed test is use to test whether the variance from the first population is either greater than or less than the second population variance. Therefore, there are two sets of data under consideration. Out of these data a decision regarding variance 1(σ12 ) and variance 2 can be made (σ22 ).

The F hypothesis test is defined as:

Null hypothesis :

σ12 = σ22

Alternative: σ12 ≠ σ22     for a two-tailed test

Test Statistic: F = s1**2/s2**2 (GSM lecture note, HFT)

F is fisher distributed value with f1 = n1-1 and f2 = n2 – 1 degrees of freedom

Where s1**2 and s2**2 are the sample variances. The more this ratio deviates from 1, the stronger the evidence for unequal population variances.

Critical Region: Reject the null hypothesis that the two variances are equal if:

2.6 Related Research

The advent of Digital Elevation Model (DEM) have spurred growth in research and developments of automatic, computer-based channel network tools that use digital elevation data as their primary input for hydrological analysis (Moore,et al.,1991). Today, tools used for delineating watershed and other hydrological analysis can include anything from websites to open source programs for standalone computers. United States Geological Survey for instance has developed a tool for web based geographical information systems known as StreamStats. The primary functionality of StreamStats allows users to delineate drainage networks and also obtain watershed characteristics (Kernell et al., 2009).

Several implementations of the research have been released by the open source community and closed source for hydrological analysis. For example, GRASS and ArcGIS contain tools which are providing frameworks for drainage networks extraction and delineation of watersheds. Some of the open source tools are stand alone tools, whiles others are add-ins to commercial GIS software.

2.6.1. DEM corrections

Development of tools for optimum removing spurious pits in digital elevation models has also received some attention in the research recently. Jackson (2012) has for instance investigated the implementation of hybrid approach proposed by (Soille 2004) for pit removals as ArcGIS add-in. The hybrid approach is defined as the absolute total elevation change to the DEM, by utilizing two existing approaches, carving and filling methods. This ArcGIS add-in is develop using c++ programming language and python programming to provide interface and connection with ArcGIS.

2.6.2 Stream network and watershed delineation

In terms of tools for the extraction of stream network and watershed delineation, investigations have been conducted using various methods and different programming languages.

Peucker and Douglas 1975 (cited by Kikuchi et al.1982) use DEM to identify potential streams. They employed the ‘upward concave’ and ‘convex’ procedure. This algorithm flags the pixel of highest elevation from each possible square of four adjacent grid cells. After one sweep of the matrix the unflagged grid cells represent drainage courses. This method eliminates the use of arbitrary threshold. However, it needs further thinning process to be able to delineate the stream networks. Kikuchi, et al. (1982) developed a FORTRAN program called High and Low Point Detection (HILO), which uses the Peucker and Douglas algorithm to delineate stream networks (Naveen 2005).

Jasiewicz and Metz (2011) conducted a study of developing a new GRASS GIS toolset designed for Hortonian analysis of drainage networks. The new toolset called r.stream uses a multiple flow direction method for stream network extraction as well as for calculating other geomorphological features in the catchment’s area. The package is now free and open-source software, available for GRASS version 6.4 and later

Tarboton (2000) conducted a study to develop a toolset for delineating stream network and watershed. The toolset is a command line programs that run from a DOS prompt of a windows or shell on UNIX. The toolset is based on the relationship between slope and contributing area, and the constant stream drop property to objectively decide upon a support area threshold for the delineation of stream network. This procedure brings objectivity to the procedure of delineating channel networks with spatially uniform drainage density.

Alternative methods to uniform drainage density tool has been explored by Tarboton(2003) where he researched on the use of C++ as a library of functions compiled into a component object model (COM) dynamic link library that is callable from other COM compliant systems such as Visual Basic and ESRI ArcGIS for the delineation of channel network. The software is based on the weighted accumulation of upwards curved grid cells. This method is adaptive to spatial variability in drainage density. A t test is used to select a weighted support area threshold that conforms to geomorphological laws, therefore bringing objectivity in the selection of the threshold.

Maidment (2002) presented an Arc Hydro as a tool to assist in hydrological analysis. Arc Hydro is an Add-on tool to ArcGIS software use to build hydrologic information systems which synthesize geospatial and temporal water resources data for hydrologic analysis and modelling. ArcHydro is use to comprehensively delineate drainage networks and watersheds.

3 Review of Existing methods and tools for stream network extraction and watershed delineation.

3.1 Introduction

This chapter focuses on an evaluation of some of the existing approaches in the extraction of stream network and delineating of watersheds. The reason for conducting this evaluation is based on the fact that there are different approaches and tools in the extraction of the stream network of a given area. Four (4) different general approaches in total have been researched about with three (3) of the approaches using Digital Elevation Model (DEM) and the remaining method being a field based approach. The evaluation plan of the procedure is based on a data input, method for the mapping of stream, the scale, and the software used being open source or commercial and how the calculation of various hydrological parameters is achieved.

In the following text the review of the approaches is presented starting with the field based approach.

3.2 Field Based procedure

The existing channel network representations in most countries are the blue lines stream networks that are drawn on 1:25000 scale topographic maps. This method is proposed by most researchers as the best estimate of the actual drainage network. The field based procedure of extracting stream network parameters is based on the mapped major channel sections of the drainage network since the blue line stream network comprises mostly only permanent and major channels. Traditionally, the outlines of the channel network are mapped by land surveying techniques. Despite high accuracy of field measurements, this method is time consuming and also labour intensive. Furthermore, is prone to unavoidable human and technical errors.

Photogrammetry supplements the land surveying techniques in the production of the modern topographic maps. The use of photogrammetry however requires experts’ intervention and also the scale of what can be seen on the photos also need to be considered. Aerial photos are normally use in the mapping of major channels, therefore the land surveying techniques are required to compensate for the identification of the other seasonal and minor channel stream segments. The basic hydrological parameters like stream lengths, slope, etc of the drainage networks and its watershed are calculated after digitizing the stream network from the topographic maps. There are various modern machines like Co-ordinate digitizer which are employed to convert images into digital data .Despite the availability of high precision digitizers, their use still introduces many errors in the data due to the involvement of intensive labour in controlling the digitizing machines.

Montgomery and Foufoula-Georgiou (1993) have proposed that the printed blue lines do not represent a viable data source for many applications. They attributed the inaccuracies in the present published maps to the fact that, stream channels are difficult to detect and cartographic generalization and decision rules also contribute to the inaccuracies in published drainage networks.

3.3.

The use of computer programmes for stream network delineation has provided an alternative to field base method described above. The software’s are coming from closed source community, open source community and the integration of open source in closed source.

3.3.1 R.STREAM PACKAGE

The r.stream toolkit package is part of the Geographic Resources Analysis Support System (GRASS GIS) which comprised of several modules use for extracting stream network and delineation of watershed. The r.stream package is an ANSI C application that utilizes ready to use GRASS API to provide modes for calculation various stream parameters. The package can be used as a stand-alone GRASS GIS package or cooperation GRASS GIS tools and external software (Jasiewicz and Metz).

The GRASS GIS is develop by a team of multinational consisting of developers around the world. It is part of the official project of the Open Source Geospatial Foundation. It is available as open source on a GNU General Public License (GPL) >= V2.

The previous version of GRASS GIS is available as command based only but it is now available with good graphical user interface (GUI), with options in Tc and python. An advance GUI based modeller tool has also been incorporated with GRASS GIS for user defined workflow (Singhai and Saxena).

In terms of approach for extracting stream networks from DEM, r.stream package of GRASS GIS has the capability of extracting stream networks based on accumulation map created with different flow direction method. The methods include D8, D-infinity, DEMON, and FD8 methods. Another advantage in the use of r.stream of GRASS GIS is the availability of different ordering systems including Horton, Hack and topological diameter approaches which are not available in any other software (Jaroslaw and Markus 2011).

The r.stream package employs the constant threshold procedure in the initiation of the stream network but the user has to choose it arbitrary. Therefore, the extracted stream network does not fulfil geomorphological law which is the bases of this research.

3.3.2 ESRI ArcGIS Hydrology Tools

The ESRI ArcGIS hydrology tool set is the leading model use in the extraction of stream network. It is part of the ESRI ArcGIS spatial analyst toolbox .The tool set include tools like Fill, Flow Direction, Stream Link and Flow Accumulation. The spatial analyst toolbox functionality can be access through different ways. Geoprocessing operations can be performed either by the use of Tool dialog box present in the ArcGIS toolbox , by using python to access geoprocessing functionality or by the use of ArcGIS Model builder.

The ESRI ArcGIS hydrology tools set is a proprietary software for spatial analysis which is available on a commercial license. The development of the ArcGIS hydrology tools set is carried out by ESRI Inc. Based in the USA and is part of the ArcGIS suit of geospatial software for the desktop, server and mobile platforms.

The greatest advantage in the use of ArcGIS hydrology tool is the availability of better graphical user interface, making it easier to use by both experience and non experience users. ArcGIS supports many DEM formats and also the ability to produce a DEM from different data source is also an added advantage.

Regarding the procedure use in the extracting stream networks from DEM, the hydrology tools set has the ability of extracting stream networks based on accumulation DEM created with D8 Flow direction method. In this tool set, there is only D8 flow direction tool unlike the r.stream of GRASS. Furthermore, there are two types of stream ordering tools available in the hydrology tool set, which are strahler and shrieve ordering tools.

The main disadvantage of the Hydrology tools set as far as this research is concern is the procedure in the extraction of the stream network. It is also based on the constant arbitrary threshold approach, which is applied on accumulation map created with D8 flow direction method.

However there are possibilities of using other geospatial libraries with the hydrology tool set with python or ArcObjects.(ESRI Resource center)

3.3.3 TOPAZ

TOPAZ (TOpographic PArameteriZation) is a raster based topographic analysis tool use for processing Digital Elevation Models (DEM) to delineate drainage networks, extracts watersheds, sub catchment identification and the hydrological and topological parameters of stream networks. TOPAZ is written in ANSI standard FORTRAN 90 and comes with six (6) different programs that must be executed systematically (USDA).

TOPAZ application is produced cooperatively by Dr Jurgen Garbrecht of US Department of Agriculture (USDA) and Prof Lawrence Martz of the University of Saskatchewan, Canada. As of my review, the current available version of TOPAZ is 3.12 which were released in 1999.The source code of the software and the manuals are available free of charge upon written request to the authors.

TOPAZ has the capability to distinguish between sink – depressions and impoundment –depression, which differentiate it from ESRI ArcGIS that identifies only sink- depression (Martz and Garbrecht, 1998).TOPAZ has an added advantage of generating spatially varying drainage densities of watersheds and subwatersheds. The calculation of many geomorphologic statistics and Strahler ordering of stream are also some features of the model.

But the model is based on D8 flow direction method for the creation of flow accumulation raster, which is use in the extraction of stream networks. It is also based on user choosing arbitrary stream area threshold using the constant threshold area method for the identification and delineation of stream networks.

TOPAZ is an example of open source hydrological systems that lacks its own graphical user interface. The model depends on other commercial systems like ESRI ArcGIS for its visualisation.