Spatiotemporal variations of soil surface roughness from in-situ laser scanning

Abstract

Microtopography and roughness are highly dynamic properties of the soil surface and important factors governing surface runoff and erosion processes. While various remote sensing technologies were successfully applied for topography measurements at different spatial scales, there is a lack of field studies that collected systematically microtopography data over long observation periods. In this paper an approach to measure and quantify surface roughness in the field based on laser scanning technologies is presented. Between June 2004 and November 2005 97 in-situ measurements were conducted in a test site with two different sandy substrates in vegetation-free conditions. Two-dimensional high-resolution (1 mm) datasets where generated for eight micro erosion plots of 0.25 to 2.9 m² in size. Dynamics and pattern formation were quantified for surface roughness and surface height changes. Roughness patterns at different scales were analyzed by local roughness indices using sliding windows of 3 to 55 mm in size. Results show strong spatial and temporal dynamics in surface roughness as well as substrate-specific variations. Temporal roughness variations could be detected and were linked to precipitation patterns. The methods presented in this paper are considered suitable to generate high-resolution datasets on spatiotemporal and multi-scale microtopography patterns and to advance the understanding of surface processes at small scales in natural environments.

Introduction

Soil surface roughness is a highly dynamic variable playing an important role for surface processes in natural environments (Kirkby, 2001). Under the influence of precipitation, discharge and wind erosion, substrate movements affect the structure of the soil surface and form spatial heterogeneities in microtopography. At the same time, resulting roughness patterns have an effect on infiltration and runoff rates by forming physical barriers or providing discharge channels for water (Solé-Benet et al., 1997, Govers et al., 2000, Römkens et al., 2001, Huang et al., 2002). Bull and Kirkby (2002) developed a conceptual model describing dominant geomorphological processes and their impacts in dryland environments. According to this model, high erosion rates cause a segregation of fine and coarse textured materials finally resulting in typical spatial distribution patterns of microtopography among hillslopes and channel beds. While the general sequences and interactions resulting in such patterns are well understood on a conceptual level, underlying processes need to be quantified at the microscale in the field to better assess their role in the context of soil erosion.

For the purpose of a better process understanding, methodologies for monitoring topography and roughness dynamics at the microscale are necessary. Among the published field studies dealing with the quantification of soil surface properties, several describe data collection methods based on mechanical devices. van Wesemael et al. (1996) used pin meters in 50 cm transects along contour lines to study the effect of rainfall and soil properties on surface roughness in Spain. Desir and Marin (2007) used the same methodology in addition to erosion pins to analyze erosion rates for a different test site in Spain. Their studies showed that this data collection method is easily and efficiently applied in the field. However, while pin meters or erosion pins are able to reflect surface topography in one-dimensional transects of arbitrary length, they do not provide data necessary for the generation of two-dimensional topography models in millimeter or sub-millimeter resolution.

Remote sensing methods have the advantage of measuring coherent areas without physical contact. Several technologies exist to generate three-dimensional surface models using remote sensing devices. While photogrammetry (Carbonneau and Lane, 2004), laser altimetry (de Vries et al., 2003) or synthetic aperture radar (SAR) interferometry (Massonnet and Feigl, 1998) are the most important approaches used at the catchment scale and above, measurements at the plot or micro scale are predominantly performed with laser scanning technologies. Earlier works already approved the feasibility of laser scanner devices to detect elevation differences within a mm range (Huang et al., 1988, Römkens et al., 1988, Huang and Bradford, 1990, Bertuzzi et al., 1990). Jester and Klik (2005) performed a thorough comparison of different data collection and processing methods quantifying soil surface roughness in the laboratory. One of their major results was that laser scanners providing millimeter resolutions in horizontal and vertical direction are particularly well suited for representing fine structured areas compared to other mechanical (pin meter, roller chain) and optical (photogrammetry) approaches. However, being designed for laboratory work, they concluded that such devices cannot be used efficiently in extensive field studies.

Laser scanners have in some cases been applied to quantify surface topography and roughness in field environments. Huang and Bradford (1992) analyzed the impact of different precipitation patterns and agricultural treatments on surface roughness measured with laser scanner devices. They found a strong scale-dependence of roughness and argued that its quantification depends on a chosen scale, which in turn is related to the process of interest. Flanagan et al. (1995) developed a scanning system for the microscale to quantify erosion processes. Beyond laboratory experiments, their device was successfully applied within a set of field experiments on a 3 × 10 m erosion plot measuring soil microtopography.

In most of these studies, laser scanner measurements were performed by using devices that collect point data in a specified grid. While the technology allows a flexible determination of the spatial resolution ranging from millimeter to centimeter scale, the setup and calibration of these devices is more complex compared to other approaches. Coherent three-dimensional datasets gained from a different type of terrestrial laser scanner allowed Schmid et al. (2004) to compare roughness and volume balances in great detail and with relatively low operating expense in a forest test site before and after severe logging activities. Their study, however, was restricted to the comparison of microtopography at a single location between two points in time.

Surface roughness is reflected by the spatial heterogeneity of elevation values at a pre-defined scale and can thus be derived from microtopography data. Its quantification depends on the dimensionality and resolution of the data as well as the desired expressiveness of the index. The most common parameter applied in recent studies is the standard deviation in vertical direction from a single mean value (root-mean squared height RMSH) calculated for a regular raster dataset of n × m pixel values. van Wesemael et al. (1996) successfully derived RMSH values from a previously detrended surface in order to separate multi-scale effects from each other, with the remaining so-called random roughness representing spatial variations in the sub-millimeter range.

The root-mean squared height represents a single global value representing surface roughness for a two dimensional dataset of arbitrary size. However, variations in height at different scales interact in this index (Huang and Bradford, 1992), making its interpretation difficult. While detrending can be applied to remove the effect of larger scale roughness patterns (i.e. slope or curvature), it is obvious that multiple interfering scale-dependent phenomena cannot be represented by single values. Consequently, a set of indices quantifying roughness on different scales needs to be collected to represent surface roughness comprehensively. Bertuzzi et al. (1990) combined RMSH calculations with minimum, maximum, skewness and kurtosis criteria from their sample datasets for this purpose. Huang and Bradford (1992) followed a more complex approach based on semivariograms and statistical models representing multi-scale roughness by two parameters. However, information on the spatial distribution of roughness was not provided, although necessary for studying surface processes in greater detail.

While several studies proved the general applicability of laser scanning methods to generate surface microtopography models, to derive roughness parameters and to quantify surface properties, their contribution for the understanding of surface processes has not yet been analyzed systematically. The aim of this study is to evaluate a laser scanning technology and a subsequent quantification method for the purpose of analyzing surface changes for long time periods in the field, where erosion can be linked to processes of the hydrological cycle. For the first time, long-term multi-temporal microtopography measurements were performed in the field based on a mobile laser scanning device. Novel multi-scale roughness parameters were developed and derived from the generated microtopography models in form of spatial distributions. Based on the study design and methods applied, the analysis enters new territory by focusing on the detection of spatial surface roughness heterogeneities and the formation of distinct roughness patterns over time. Meteorological input data that have been measured concurrently in the field are related to the results. Finally, the feasibility of monitoring long-term microtopography changes with such a technology in the field to gain a deeper understanding of underlying surface processes is discussed.