Methods for Gully Characterization in Agricultural Croplands Using Ground-Based Light Detection and Ranging

The purpose of this book is to put together recent developments on sediment transport and morphological processes. There are twelve chapters in this book contributed by different authors who are currently involved in relevant research. First three chapters provide information on basic and advanced flow mechanisms including turbulence and movement of particles in water. Examples of computational procedures for sediment transport and morphological changes are given in the next five chapters. These include empirical predictions and numerical computations. Chapters nine and ten present some insights on environmental concerns with sediment transport. Last two contributions deal with two large-scale case studies related to changes in the transport and provenance of glacial marine sediments, and processes involving land slides.

resolution for gully investigations at individual plot scale. In order to capture the microtopography impacting gully formation and development over time it is necessary to work with DEMs with spatial resolution ranging between 5mm to 15cm (Schmid et al., 2004). Additionally, ephemeral and classical gullies are affected by precipitation events, management and conservation practices, crop phenology, and crop canopy development. In ephemeral gully investigations, timing becomes significant as farming practices can remodel the field's topography by filling the gully channel. Classic gullies are also dynamic, responding to condition changes through headcut migration, deepening, and channel widening processes. Standard DEM datasets produced either by government agencies on a national scale or by private companies on a regional scale can only capture a snap shot in time of gully development. Recent developments in laser scanner technology provide new opportunities for scientific investigation of gullies located in croplands. Although, laser scanners have been previously used in similar investigations such as large-scale classical gullies in different locations such as mountain-side sites (Perroy et al., 2010), forest sites (James et al., 2007), desert sites (Park, 2008), and landslides (Hsiao et al., 2003 andRoering et al., 2009), its application to ephemeral and classical gully investigation is in the early stages of development. Groundbased laser scanners provide the tools for detailed multi-temporal analysis of microtopography of gullies. With this detailed information it is possible to obtain digital terrain models (catchment area, slope, planform curvature, thalweg, and cross sections), estimate soil roughness, perform volume calculations, and others. This is especially important in research setups where the same location needs to be surveyed multiple times over extensive periods of time as conditions change due to rain events, field management changes, and/or different conservation practices investigations. Ground-based laser systems are capable of providing high levels of topographic and morphologic details. Ground-based light detection and ranging (LiDAR) systems differ from airborne systems. During data collection, the scanner (LiDAR sensor) is located on a tripod at ground level ( Figure 1). Proximity to the ground surface causes laser pulses to be sent with larger range of vertical incidence angles than the airborne system. Ground sampling density is influenced by operator controlled factors such as investigation area selection (vertical and horizontal incidence angles), instrument resolution (average point density at mid-range), and degree of overlap between scans. The large volume of points combined with irregular point distribution of ground-based LiDAR data sets makes production of digital elevation models computationally challenging. Additionally, the presence of vegetation, standing residues, and shadows (regions with no data) represents an extra level of complexity that often cannot be addressed by standard off-the-shelf commercial geospatial software packages. In this chapter, we describe new and enhanced techniques to pre-process and evaluate detailed topographic information acquired with ground-based LiDAR, which were specifically devised to quantitatively and morphologically characterize gullies located in agricultural fields. In the following section (Section 2) background information about ground-based LiDAR systems is provided. In section 3, multiple techniques to pre-process LiDAR data in the original point cloud format are depicted, such as cross-validation between scans (geo-referencing accuracy investigation), point sampling pattern investigation in terms of density (quadrant methods) and spatial distribution (distance to nearest neighbours), removal of unwanted points (thinning and outlier removal), and surface reconstruction (surface smoothing). Morphological analysis through cross-section www.intechopen.com investigation follows in Section 4. An alternative method to overcome the high computational cost involved in the conversion of large point cloud into raster grid is addressed in Section 5. We provide a brief description of the implementation details as well as the computer programming language libraries utilized in Section 6. A study case of a gully located in a field of the Cheney Lake Reservoir Watershed (in South Central Kansas, USA) is included in Section 7 to illustrate some the techniques previously described. The chapter ends with discussion of the importance of this work and points to future research needs (Section 8). The material in this chapter can be used as a basis to develop guidelines on the use of ground-based LiDAR technology to characterize topographic conditions associated with the formation and evolution of agricultural gullies at field scale. This is especially important in research sites where the same geographical location needs to be surveyed multiple times over lengthy periods of time as conditions change due to precipitation and runoff events, field management, and implementation of conservation practices.

Light detection and ranging technology
LiDAR technology, measures the laser pulse travel time from the transmitter to the target and back to the receiver allowing the distance to be calculated (Wehr and Lohr, 1999). In this process, a very accurate timing system is needed to guarantee an accurate resolution since laser pulses are sent at the rate of thousand times per second. Additionally, the transmitter and the receiver must be located at the same physical location, which is a single-ended system because the phase the incoming signal is compared to the phase of an internal clock at the same location (Measures R. M., 1984). This technology is often classified into air-borne and ground-based systems. In airborne LiDAR systems, the aircraft position is recorded using accurate positional systems such as inertial navigation unit, GPS receivers located in the aircraft, and GPS base locations. With these combined equipment, it is possible to determine three-dimensional geo-referenced coordinates for each pulse and then correct for the aircraft positioning in terms of roll, pitch, and heading, thus improving the accuracy of the system.

www.intechopen.com
Ground-based LiDAR systems, are simplified because the sender/receiver's location is known. They operate in a similar fashion to the reflectorless total stations. For each pulse, these systems can measure: (1) x, y, z coordinate values with respect to the position of the laser sensor; (2) the intensity of the returned signal as reflectance; and (3) RGB values from an integrated digital camera within the instrument. These systems are capable of collecting information with a wide range of ground sampling densities as a result of operator controlled factors such as scan angle (area covered by individual scans), average point density, and degree of overlap between scans. Higher point density can be achieved by higher sensor resolution, smaller vertical field of view angles, and multiple scans of the same ground location. The instrument resolution is often controlled by an imaginary plane located at the middle range of the vertical field of view and is orthogonal to the sensor's normal sight. The vertical field of view angles also influence the point density as point sampling becomes sparser as vertical scan angles tends to the horizon. Multiple scans can be used to collect data over the same geographical location resulting in increased sampling density and overcomes problems such as shadowing and limited coverage due to vegetation.

Techniques to pre-process topographic information
Pre-processing is an important step introduced in a topographic data analysis scheme to minimize the influence of unwanted points and to maximize the influence of the desired ones, thus simplifying the computational work needed in the later stages of the analysis process in order to reach the investigation goals. Pre-processing is composed of operations such as filtering, enhancing, transformation, or combination of these processes. The preprocessing techniques described herein include: cross-validation to assess geo-referencing accuracy between multiple scans of the same geographical location, investigation of point sampling patterns through quadrant method, evaluating the internal distribution of points using the average distance to the nearest neighbour, unwanted point removal, and surface reconstruction through surface smoothing.

Cross-validation
During field surveys, such as the one conducted a field in the Cheney Lake watershed in Kansas, the same geographical location can be covered by different scans. This technique is often used to increase sampling density and to avoid problems such as shadowing and limited coverage due to vegetation. The overlap in LiDAR point sampling can also be used to evaluate survey final geo-referencing accuracy. Since each individual scan is collected using local coordinates, geo-referencing is the procedure to convert from local coordinate system into global coordinate system. Given that the overlapping scans collected have sufficient point density, neighbouring points can be identified, and elevation differences between these points can be determined. Similar approach is used to evaluate the accuracy of airborne LiDAR sensors by comparing elevation values of nearby points collected by different flight paths in overlapping areas. For each combination of two scans, points were analysed to form a list of neighbouring points. For illustration, consider two scans i s and j s where: ij  . A pair of points i xyz P and j xyz P are recorded into the list when two conditions are met: a) the points are from different scans ij  and b) the horizontal distance between the two laser points is smaller than the distance threshold   threshold dd  .

www.intechopen.com
For example, in Figure 2, scans 4, and 12, 13, 14, and 15 overlap. Scans 12, 13, 14, and 15 were designed to increase point sampling at the main gully channel. Each of these scans is geo-referenced using a set of four pre-registered targets (the global coordinates of the targets are known prior to the geo-referencing procedure).
Using the cross-validation method for each pair of overlapping scans produces the tabular information presented in Figure 3. These values represent the upper quartile of the elevation difference between pairs of near points from different scans, multiplied by 1000. The values highlighted in Figure 3 exemplify geo-referencing problems between scans 4 and 12, 13, 14, and 15. Upon verification of the coordinates of the four targets used to geo-reference scans 12, 13, 14, and 15; the target with the largest RMS error was removed. A new crossvalidation assessment between scan 4 and 12, 13, 14, and 15 was performed yielding new values of 17, 13, 16, and 17 respectively.

Quadrant method
A simplified and efficient way of exploring point pattern (spatial distribution and density) is to partition the location surveyed into sub-regions of equal area, or quadrants and then sum the number of laser points in each of the quadrants (Bailey and Gatrell, 1995). The division of the number of points by the area of the quadrant grid generates a measure of sampling intensity or laser points sampling density. The density map represents a two-dimensional frequency distribution (Bailey and Gatrell, 1995). This can be used to identify locations with high and low density, providing guidance to evaluate over and under-sampled areas. The density information can be correlated with topographic and land cover information for identification of areas with no laser point sampling due to shadowing and the influence of vegetation cover in the density information. The quadrant method is influenced by the quadrant grid size. Larger quadrant grid sizes mask small density variations while small quadrant grid sizes create a larger range of density values with many quadrant grids containing no points. Using two different values of virtual quadrant grid, the gully represented by lighter colour in Figure 4A, is depicted by laser point sampling density associated with a virtual grid of 1.0 meter ( Figure 4B) of 0.25 meter ( Figure 4C). Both density maps reveal under-sampled locations.

Average distance to nearest neighbor
The use of the quadrant method provides the means to evaluate the spatial variation of the laser point sampling density. Density is often referred to as first order property of point pattern (Bailey and Gatrell, 1995). The use of the average distance to the nearest neighbours within each quadrant grid depicts an indication of the internal distribution of the laser points ( Figure 5A). Based on this information binary maps can be generated of particular average distances ( Figure 5B) to investigate sampling coverage. Although the average distance to the nearest neighbour provides an indication of how clustered or distributed laser points are in each quadrant grid, care should be used on analysis of large quadrant grids. In those analyses, two quadrants can produce the same average distance to nearest neighbour values, but one quadrant can contain a larger quantity of points uniformly distributed while the other quadrant contains fewer clustered points. To minimize this problem, a simple alternative is to combine the density map with the average distance to nearest neighbour map ( Figure 5C). www.intechopen.com

Thinning
Similarly to the quadrant method, thinning generates a two-dimensional virtual quadrant grid covering the entire extent of the point cloud, identifies the points within each quadrant, and keeps only the point with the lowest elevation in each quadrant. This technique, originally devised to filter trees out in order to obtain bare Earth data sets, is based on the www.intechopen.com assumption that in surveys with enough sampling density, vegetated areas will have points that can penetrate the tree canopies and reach the ground. The same concept can be adopted when working with terrestrial laser scanners. Vegetation foliage and residue stubs can be partially removed by thinning procedure (Figure 6).

Outlier removal
Gully investigations in cropland require artificial removal of outlier laser points to generate the true bare Earth representation. These points can constitute standing crop residue, vegetation canopy, or even individual incorrect registered scan lines. Two different approaches were considered here: square average distance (SAD) to the K-nearest neighbours (KNN) and distance to the fitted surface (DFS) to the KNN (Bäsken, 2010, Delage, 2010, Cazals and Pouget 2008. Unlike the previously described quadrant methods that compute two-dimensional distances, these two approaches classify a point as an outlier or not based on three-dimensional Euclidean distance. In the SAD algorithm, for each point the K nearest neighbours are selected and the square average distance between these points is computed. If this value is larger than a user provided threshold the point is assigned to be removed. A subset of a gully channel with different outliers, standing crop residue (two sharp peaks) and small lines of elevated points were evaluated ( Figure 7A). An initial attempt to remove these outliers used a square average distance threshold of 5 cm for the 50 nearest neighbours resulting in the removal of one point ( Figure 7B). A second attempt using 10 nearest neighbours and a square average distance threshold of 1 cm removed an additional 14 laser points. The DFS algorithm, identifies the K nearest neighbours for each laser point, fits a polynomial surface to these points, and computes the three-dimensional distance of the point being evaluated to the fitted surface. If the distance is greater than the user-specified threshold the point is then marked for removal. The polynomial order is also a parameter defined by the user. The same information shown in Figure 7A is filtered using a first order polynomial surface, planar surface, removing a total of 4312 points ( Figure 7D). Visual comparison to the original dataset ( Figure 7A) indicates the removal of most outlier points.

Smoothing
The presence of vegetation and/or standing crop residue can yield point clouds, and subsequently digital terrain representations, with large elevation variations (quantified by large standard deviations). In such noisy datasets, techniques can be devised for surface reconstruction (herein referred to as smoothing). The objective is to attenuate large local relief variations while maintaining the global relief of the terrain. Unlike the filtering procedures, no points are removed but rather three-dimensionally moved to a fitted surface to the KNNs of each point (Delage, 2010, Cazals andPouget 2008). The output point cloud contains the same number of points than the original one, however, the coordinate values of each point is slightly modified. The parameters required are the k number of neighbours and the degree of the polynomial to generate the fitted surface. The lower the degree of the polynomial the smoother the reconstructed surface will be. The original surface ( Figure 7A) can be reconstructed using a third order polynomial ( Figure  7E) while a surface using a first order polynomial (plane) can also be constructed ( Figure  7F). Visual inspection indicates a more realistic representation using the third order polynomial over than the first order.

Summary
Although the selection of the most appropriate algorithm along with their parameters varies according to the terrain (topography and land cover) and scale (flume, plot, or field), best results can be obtained by combining these procedures (filtering and smoothing). A list of the tools described in this section with the necessary inputs, the generated output, and the main purpose are listed in Table 1.

Morphological analysis
In scientific settings, different gullies are surveyed multiple times with the objective to capture a channel's dynamic changes. The objective is to quickly assess channel morphological changes such as headcut migration, widening, deepening, and deposition, as well as estimation of the volume of soil eroded/deposited. As the number of surveys increases and consequently the volume of data, an automated and standard procedure to extract key parameters is necessary.

Automated cross-section generation
Channel geometry can be quantified through a series of cross-sections. Although crosssection methods exist in commercial off-the-shelf Geographic Information Systems (GIS) software packages, the generation of a series of closely spaced cross-sections can become a time consuming process due to the necessary file format transformations (conversion of point cloud into either raster or TIN format) and the lack of automation (user has to define the individual location of each cross-section). The procedure developed requires the user to define a centreline approximating the gully thalweg in the ESRI Shapefile file format and input parameters describing how the crosssections will be generated ( Figure 8A). The cross-section geometric parameters are: length of the cross-section, distance between cross-sections (b in Figure 8A), distance between points within the cross-section (c in Figure 8A), and search radius (a in Figure 8A). The final elevation for each point of the cross-section is defined by the Inverse Weighted Distance (IDW) algorithm (Equation 1).  Fig. 7. Point cloud pre-processing algorithms: A -TIN representation of original point cloud; B -TIN representation of point cloud after outlier removal using a SAD of 5 cm and 50 KNN (k) points; C -TIN representation of point cloud after outlier removal using a SAD of 1 cm and 10 KNN points; D -TIN representation after outlier removal using distance to fitted surface (DFS) to KNN points (4312 points removed); E -TIN surfaces were reconstructed by three-dimensionally moving points to the surfaces fitted to the KNN points using a third order polynomial surface; and F -TIN surfaces were reconstructed by threedimensionally moving points to the surfaces fitted to the KNN points using a first order polynomial surface.

www.intechopen.com
The IDW algorithm is a local interpolation algorithm that assigns different weights to laser points based on their proximity to the unknown location (z 0 ). In Equation 1, the z i is the elevation at the known point i, d i is the distance between point i and the point being estimated (unkown location), and k is the power controling the degree of local influence (closer laser points are assigned higher weight) and is also an input parameter. Based on using the same centerline and input parameters utilized in different surveys, gully channel geometrical changes between surveys can be quantified (Figure 9). The crosssectional area difference between two cross-sections is computed by defining a reference elevation and computing the area of the closed polygon (red polygon in Figure 9). The change in cross-sectional area is determined by computing the difference between the individual cross-sectional areas . The thalweg is generated by simply connecting the point with the lowest elevation in each cross-section ( Figure 10).

Cross-section modeling
Quantification of gully channel geometrical properties is often performed by simplifying and fitting measured cross-sections through trapezoidal cross-section shapes. However, as the number of cross-sections grow as a result of gully channel extension, small cross-section interval, and high survey revisit time, identification of top-of-the-bank and channel toes www.intechopen.com manually can become tedious and time consuming. Alternatively, a simplified optimization algorithm can be used to fit a trapezoidal cross-section to the LiDAR derived cross-section in a semi-automated fashion. Fig. 10. Three-dimensional representation of the thalweg generated by connecting the crosssection point with the lowest elevation.
The algorithm iteratively fits two lines to each of the cross-section sides individually (left hand-side is shown in Figure 11). A series of scenarios are considered. Initially, the algorithm fits one line to the three points furthest from bank toe (blue points in Figure 11A) and another line to the remaining points (red points in Figure 11A). The root mean squared deviation (RMSD) of each of the fitted lines is computed and then the two combined (Equation 2). The process is repeated by increasing the number of points considered in the first line while reducing the points to the second line ( Figure 11B), until only three points are left for the second line ( Figure 11C). The optimum point is selected by identifying the minimum combined RMSD value trough minimization of Equation 2 ( Figure 11D).
Where, a i Y is the elevation of the first fitted line a at point i, o i Y is the elevation of the second fitted line b at point i, j is the last point in the first set of points, and i Y is the measured elevation at location i. A line is fitted to the gully bottom through least square error and the intercept of this line with the others generated with the optimization method (left and right banks) for a trapezoidal cross-section shape (Figure 12).

Summary
The main objective of the cross-section processes is to standardize and to expedite the crosssection generation and analysis. These procedures play an important role when conducting multi-temporal investigations devised to quantitatively assess gully changes over time (formation and evolution). The main components described herein are listed in Table 2, along with their required input, output, and the main purpose.

Inputs Outputs Purpose
Crosssection generation Systematic extraction of geometric parameters from cross-section. Table 2. Summary of tools implemented for gully morphological analysis in agricultural fields using ground-based LiDAR.

Interpolation using local binning algorithms
Point clouds generated using Ground-based LiDAR systems at field-scale and even at flume-scale can yield large datasets (millions of points) with varying sampling densities and uneven point distributions posing a challenge to interpolation techniques based on global estimates such as a regularized spline with tension and ordinary Kriging algorithms. The algorithm proposed by Kim (2006) was adopted and expanded to generate grids of intensity, red, green, and blue values in addition to elevation values. This algorithm produces a tree data-structure (Quadtree) point cloud designed to recursively partition the two-dimensional space (X and Y coordinates) to expedite queries of laser points in the desired neighbourhood (Bäsken, 2010). www.intechopen.com

Implementation
The described algorithms were implemented in a combination of C++ and Python computer programming languages as a result of the need to have customizable tools that were not previously available in GIS software packages. The point clouds were stored and manipulated using the American Society of Photogrammetry and Remote Sensing (ASPRS) LASer file format (LAS) through the libLAS library (http://liblas.org/). To overcome the large point cloud size (millions of laser points) the Computational Geometry Algorithms Library (CGAL) was selected to perform efficient nearest neighbour queries (Delage, 2010), compute distances, outlier removal (Allez et al., 2010), normal estimation, and fitting of three-dimensional polynomial surfaces (Cazals and Pouget, 2008). The cross-section modelling algorithm was developed using the Scientific Tools for Python (SciPy) due to the plotting capabilities offered. Additionally, the ESRI Shapefile file format was used when generating/manipulating cross-sections and thalweg files.   Fig. 11. Optimization procedures to fit a theoretical trapezoidal cross-section to the measured cross-section obtained with ground-based LiDAR points from Kansas field surveys. In the first scenario two lines are fitted, one to the first three points and the second to the remaining points (A). The procedure continues until the second line has only three points (C). The optimal fit is obtained by the minimum RMSD value (B) after all the scenarios are evaluated (D).

Study case
A gully located in South Central Kansas was selected for analysis. The gully is within the Cheney Lake Reservoir watershed and is approximately 96 metres long, 1.3 metres wide at the widest location, from 10 to 50 centimetres deep, and oriented North-South. The channel is free of vegetation and crop residues, while the surrounding field is covered by crop residues resulting from no-till management used in winter wheat followed by sorghum (milo) in the 2010 crop rotation. Historical cultivation practice records indicate that initially this gully did not disrupt farming operations; however, as no-tillage practices were adopted in 2005, the channel grew wider and deeper to the point that the farming equipment could not be used to travel across the gully and the ensuing cropping activity was performed around the main channel . The site was visited and surveyed in March and November of 2010. Measured precipitation records indicate two significant precipitation events occurred in June and July, 2010 ( Figure 13). www.intechopen.com The equipment used was a TOPCON GLS-1000 series that has a 4 mm single point accuracy distance (Table 3). The surveys are composed of 11 and 7 scans resulting in a total of 945,668 and 3,673,167 laser points for March and November, 2010, respectively. The surveys presented different laser point sampling densities (Figure 14). The survey conducted in March yielded higher variation in the sampling rate (density) in the gully channel with two clusters of high density. Conversely, the survey conducted in November generated a more uniform sampling rate specifically on the gully channel.  Both surveys were subset to the same extent (gully only) to remove unwanted laser points regarding agricultural furrows and standing crop residues. Additional outliers were removed using the square average distance to KNN algorithm with threshold of 5 cm and k of 50 followed by a thinning algorithm using a virtual grid of 1.5 cm.
Using the local binning algorithm with a raster grid size of 2.5 cm, search radius of 5.0 cm and IDW power 2, two raster grids were generated ( Figure 15). Areas were identified with elevation changes signifying different physical processes based on computing the difference www.intechopen.com between the two grids. The active headcut migration path is represented by the darker blue colour indicating the occurrence of incision processes between cross-section 50 and 90 ( Figure 15). Further downstream, between cross-sections 99 and 145 (Figure 15), the predominant physical process is widening of the gully channel as result of encountering a less erodible layer. Fig. 15. Raster grid generated using local binning interpolation algorithm. Differencing of these grids illustrate elevation changes that occurred during the time period investigated. The red dots mark selected cross-sections.
Using the same user-provided centreline for both surveys a set of cross-sections were automatically generated and compared. Cross-sections at different locations of the gully were computed (Figure 16) based on various parameters (Table 4). A total of 448 crosssections were generated for each survey varying in the extent and number of points as result of the non-uniform distribution of the laser points.

Parameter
Value Cross-section size 3.0 meters Distance between points in the crosssection 2.5 centimeters Distance between cross-sections 10.0 centimeters Power used in IDW 2.0 Search radius 5.0 centimeters Table 4. Parameters considered in the generation of cross-sections for multi-temporal morphological analysis of gully evolution in croplands. The profiles, one for each of the surveys, were generated by deriving elevation information from the intersection of the user-defined centreline and each measured cross-section ( Figure  17). Comparison between the two profiles indicates the rate of headcut migration (between distances 5 and 10) as well as the amount of incision in that section ( Figure 15). Beyond 10 meters, the November profile (blue line in Figure 17) points to a situation of stable slope or a less erodible layer. The deposited material in March was likely removed as result of the large storm events. Differences in cross-sectional area are plotted in Figure 18. Visual comparisons between the cross-section difference plot and the series of cross-sections identify erosional patterns. Between cross-section 70 and 90 the difference in cross-section area is attributed mainly by headcut migration and incision processes. From station 90 to 100 a combination of widening of the channel and incision are responsible for the differences in cross-section area. Beyond station 110 the change in cross-sectional area is mainly due to widening of the channel, 110-130 side walls only, 130-175 side walls and bed, and 175-250 widening of the channel bed.

Summary and closing remarks
The successful monitoring of gully evolution (ephemeral and classical) at individual gully scales in agricultural fields depends on accurate and detailed topographic information. The utilization of new technologies, such as ground-based LiDAR, provides new venues for scientific investigation of physical processes responsible for gully formation and evolution. Ground-based LiDAR can generate point clouds containing millions of points with point intervals varying from 1 mm to 200 mm depending on operator-defined parameters. These systems generate non-uniform spatial distributions of laser points (varying sampling density) because of the sensor's proximity to the target (equipment setup on tripods). The non-uniform distribution of points is influenced by relief, standing obstructions such as vegetation and crop residue, and vertical incidence angle. The format and massive amount of measured data represents a challenge for existing watershed modelling software applications that lack the necessary tools to pre-process the data and require the point clouds to be downgraded or interpolated in order to be used (Luccio, 2011). Watershed management planning will need these tools to implement economical and environmentally effective conservation practices.

www.intechopen.com
Cross-Section Area Difference, meterssquare This chapter described a set of tools, devised primarily for gully investigation, to preprocess and extract morphological information from gully channels at different scales. These tools work with the point clouds in industry standard file format (LAS files) to minimize the need of file format conversions as well as to facilitate sharing between different systems. The tools also utilize advanced data structures and fitting algorithms to efficiently perform queries, compute distances between nearby laser points, and fit surfaces to KNNs, facilitating the removal of outlier points and the smoothing of the overall surface. Additionally, the extraction of morphological information in a semi-automated fashion was implemented through the analysis of cross-sections derived straight from the point clouds. This is an important feature providing a standard procedure to analyse multi-temporal data sets. Future work includes the extension of the described system to include enhanced features such as classification of laser points based on elevation and intensity values, including combinations of each. In addition, correction of intensity values based on range and incidence angle, edge detection, and extraction of gully-specific geometric parameters are needed.

Acknowledgments
support. Thanks are also due to Don Seale and Antonia Smith for their indispensible assistance during data collection. Finally, the authors would like to acknowledge the technical suggestions and discussions offered by Dr. Langendoen. The purpose of this book is to put together recent developments on sediment transport and morphological processes. There are twelve chapters in this book contributed by different authors who are currently involved in relevant research. First three chapters provide information on basic and advanced flow mechanisms including turbulence and movement of particles in water. Examples of computational procedures for sediment transport and morphological changes are given in the next five chapters. These include empirical predictions and numerical computations. Chapters nine and ten present some insights on environmental concerns with sediment transport. Last two contributions deal with two large-scale case studies related to changes in the transport and provenance of glacial marine sediments, and processes involving land slides.