In order to address the issue of planation surfaces mapping in the southwest of Hubei Province, our investigations are based on the visual identification of topographical expressions of planation surfaces from a DEM. The overall methodology involves the analysis of geomorphological characteristics of planation surfaces, the selection of optimal visualization technique of DEMs, field campaigns to set up the training areas for planation surfaces and the generation of a planation surface distribution map. Finally, comparisons between our results and published works are carried out to evaluate the accuracy and effectiveness of this study.
The DEM data used is ASTER GDEM which is generated based on photogrammetry with stereo pairs of near infrared images of ASTER. These data can be freely accessed via the NASA Land Processes Distributed Active Archive Center (LP DAAC). Data are available in GeoTIFF format with geographic coordinates and elevation postings at 1 arc second, or approximately 30 m. This resolution generalizes terrain features and smaller topographic details are lost, but previous works have suggested that DEMs at this scale could provide enough information for the detection of palaeosurfaces (Rowberry et al., 2007). More detailed resolutions will reduce the speed of the process of GIS analysis. The spacing of the grid DEM is nominally 30 m, however, its metric length varies from place to place which reduces the efficiency of spatial analysis. In this paper, we project the original DEM with the geographical coordinate into the 30 m DEM with the UTM (Universal Transverse Mercator) projection by bilinear interpolation method.
The GDEM usually contains a large number of depressions or sinks. These are single or multiple pixels which are entirely surrounded by higher elevation pixels. Some of these sinks are naturally occurring on the landscape, representing inland basins without outlets. In most cases, the sinks are considered spurious, often caused by the interpolation errors and inadequate accuracies in elevation (Martz and Garbrecht, 1998). The spurious sinks will affect the successful application of different visualization techniques to a DEM as they generate anomalies across the DEM surface. Therefore, sinks are usually removed from the DEM before such geo-spatial analysis. Standard GIS procedures have been developed to remove spurious sinks, and a common approach is to raise the elevation values within the sinks until an outflow point is encountered (Jenson and Domingue, 1988). Thus, a sink free DEM is created for landscape visualization and landform mapping.
The term planation surface has been variously used in different senses. Some reserve it for a flattish plain resulting from erosion; others use the term as a synonym for any erosion surface, whether it be a flat or inclined etchplain, pediplain, or peneplain (King, 1949; Wayland, 1934; Davis, 1901). Because genetic implications are so often associated with various names, in this article, we refer to these features as simply planation surfaces.
A planation surface is a remnant of an old landform according to the Davisian cycle. During this cycle, the landscape would go from a brief period of youth, characterized by deep V-shaped valleys and an abundance of waterfalls and rapids; to a longer period of maturity, characterized by a great variety of landforms, including meandering rivers and floodplains; followed by a long period of old age, in which the landscape might be worn down to an erosional surface of low relief that he called a peneplain (Fig. 3). Then, rejuvenation occurs and there is another uplift of mountains and the cycle continues (Keller and Pinter, 1996). Few planation surfaces survive because they have been dissected, but some geomorphologists claim that they can be extrapolated from accordant summits (Zhang, 2008; Blackwelder, 1912). In the study area, the planation surface was found appear as plains, landforms of similar summits and rolling landscapes with gentle slopes, having a low relative relief (typically < 500 m) and characterized by convexconcave hillslopes (Xie et al., 2006; Li et al., 2001). This provides the most basic morphological signatures for planation surfaces identification. With the advancement of geo-spatial information system, these features could be modeled and rendered by a computer with a proper visualization technique of DEM. Therefore, make it possible for planation surfaces mapping on DEMs.
Figure 3. The cycle of erosion. (a) Youth, V-shaped valleys, few or no floodplains, extensive interfuves, many falls and rapids plus some lakes and swamps; incising watercourses; (b) maturity, well-drained terrain, all in slopes except floodplains; trunk and some tributary streams meander; maximum relief; (c) old age, broad, open valleys with widely meandering streams, indistinct divides, erosion remnants of resistant lithologies, surface near erosional base level (from Encyclopædia Britannica).
The 3D visualization of topography belongs to the fields of computer image processing, which aims to enhance the visual appearance of the real world landscape. Smith and Clark (2005) compared a variety of different methods for visualizing DEMs, finding significant differences in mapping between 12 imageprocessing options. They also noted that relief-shading by fixed azimuth is fast and efficient, making it suitable for many environmental applications although it is less effective than other methods as for lineament mapping because of the azimuthal bias. This problem was also stated by some other authors when dealt with linear feature extraction (Pike, 1992; Lidmar et al., 1991). However, most of those authors didn't discuss how relief-shading could be implemented into mapping of large-scale features other than lineaments. In this study, different light source azimuth and altitude angles of relief-shading are tested to determine the best combinations for landform visualization, and then enhanced with color tones for the planation surface mapping.
Relief-shading is also known as shaded relief model (SRM), which is generally intended to mimic the Sun's effects such as illumination, shading and shadows on land surfaces. The geomorphorlogical features can therefore be enhanced or suppressed by manipulating the illumination direction. In practice, azimuth-biasing is most pronounced with respect to linear features mapping and therefore multiple-illumination azimuths were proposed (Loisios et al., 2007). As for this study, a single view appears to be adequate since geomorphological features of planation surface are not linear. Different orientations of the model under varied light directions are quite useful for identifications of features of varied prevailing directions. In this article, several relief shading models with different azimuths and altitudes are examined to find the optimal one for the planation surface mapping. This is achieved with the Hillshade tool available in ArcGIS 9.3.
The SRM is a good representation of landforms, but is still limited to show the subtle relief variations because of its grey tone. On the contrary, colored relief shading, modulated by elevation and by exposure to illumination, presents topography in a particularly vivid and descriptive manner and helps the interpreters distinguish more easily the landform features. Therefore, we added the ability to "colorize" the SRM to create a painted relief model (PRM), which can provide quick visual understanding of the terrain, as different colors allow for better identification of slope angles and changes in elevation than monochromatic images. The painted relief image was created on the basis of SRM by splitting the elevation data into 25 equal levels and assigning a distinct color to each level. The colors are taken from the USGS Color Lookup Table (LUT). This procedure was automated in Erdas Imagine 9.1 software package.
The painted relief model was a key component throughout PS identification because it was the most optimal visualization model for the recognition and interpretation of complex geomorphologic features. In this study, two field trips were launched in August 2010 and August 2011 and the typical geomorphological signatures were carefully observed and analyzed in the field. The coordinates of the visited locations were recorded with a GPS, and then integrated into the GIS and overlaid on the PRM to set up the "training areas". Due to the large extent of study area, sampling sites of planation surfaces from previous studies were also collected. They were digitized on-screen and overlain on the model in order to support the interpretation. Due to the distinctive landforms of the planation surfaces, they are easily identified through visual interpretation of the PRM with the aid of the auxiliary information. The planation surfaces were thus mapped, recorded through an iterative process of on-screen digitizing. In order to demonstrate the distribution and number of planation surfaces, the mapped planation surfaces were further divided into three surfaces or one surface and four sub-surfaces according to the previous studies as shown in Table 1 (Tian et al., 1996; Liu, 1983).
This section carries out quantitative comparison of the mapping results from the PRM with the planation surfaces map made by Tian et al. (1996). Since the present study area is partially overlapped with that of Tian et al. (1996), two common locations are selected for comparisons. Both of the maps are overlain on the PRM to help the comparisons.
Datasets and Processing
Visualization Techniques and PS Mapping Visualization techniques
Comparison with Previous Studies
Landscape visualization is crucial for the geomorphological feature identification. Four data visualization methods were applied to a subset of the study area, which is composed of landforms of similar summits, rolling landscapes and river valleys to determine the most optimal one for mapping as illustrated in Fig. 4. It is found that although the Landsat ETM+ false color composite contains the adequate information of land use/cover, the landform features still can't be easily distinguished (Fig. 4a). The grey tone DEM can present the large-scale landforms such as the river valleys, but fails to depict the subtle features such as the accord summits and rolling features (Fig. 4b). SRM provides a new way for landform visualization, of which the azimuth and altitude of the light is critical for successful geomorphorlogical mapping. It suggested that the best suitable orientation for identification of the PS was the model with 345° azimuth angle of the illumination source and 45° altitude angle of the light source above the horizon after examining several models. This model was oriented in a way to avoid inverse topography or false topography perception (Fig. 4c). Hill shading enhances the visualization of objects perpendicular to the azimuth of the illumination; the azimuth and altitude angles of the illumination source were selected taking into account that important geomorphological features in the study area, such as the rivers and valleys have a prevailing E-W direction. Subsequently, the painted relief image was created on the basis of SRM where the geomorphological features were clearly displayed by different textures and colors (Fig. 4d). It can be drawn from Fig. 4 that the PRM is the best visualization option for planation surfaces mapping.
Figure 4. Comparison of the data visualization methods used in this study for a subset of the main study area. The subset shows landforms of similar summits (upper left part), rolling landscapes with gentle slopes (upper middle and right part), and river valleys (middle part). (a) ETM+ false color composite with band combination of 5 : 4 : 3; (b) grey-scale ASTER GDEM; (c) SRM; (d) PRM.
The planation surfaces were mapped by visual interpretation of the PRM as displayed in Fig. 5. In the spatial domain, the distribution pattern of planation surfaces is closely related to that of the rivers as a result of the constant dissection by streams and erosion by water in the long process of landform evolution. According to the Davisian cycle, the landforms in the study area are in the mature period, attaining maximum relief with well-drained basins. The planation surfaces were further divided into five classes according to Liu (1983) and Tian et al. (1996) and described as follows.
PS1 (1 700–2 000 m) is mainly distributed on the north and south edge of the Qing River basin, and extends along the anticline axis in the north and is dissected into several parts in the south. It also serves as the main drainage divide between the Yangtze and Qing rivers (Fig. 5). The surface is partially reworked by karstification into different landforms, such as karst platforms, rolling valleys and round summits with similar elevation (Li et al., 2001). The surface is composed of the Cambrian, Silurian, Permian and Triassic strata and cut across the structures of different times. Xie et al. (2006) deduced the higher surface was formed in the late Paleogene by analysis of the sediments and the restrictions of tectonic movements.
PS2 (1 300–1 500 m) is the main part of the uplifted mountain area, and is distributed around preceding PS I and has a much wider coverage. This level of PS is prevalently exposed in the anticline axis carbonate zone, and extends along the folds axes. Xie et al. (2006) reported residual PS weathering crust and iron coatings on this stage geomorphologic surface, the analysis shows the Three Gorges area was developed in a lowland environment under which the weathering crust developed. Recently, Wang et al. (2010) have found gravel layers in high altitude areas of Lichuan which represents the evidence of fluvial process.
PS3 (1 000–1200 m) and PS4 (800–900 m) are usually regarded as two sub-levels in one erosion cycle. PS3 (1 000–1 200 m) mainly extends along the subsidiary fold or core part of syncline, and is commonly distributed in the flat watershed area of Qing River branches. This geomorphologic surface tilts from east to southwest, and the height difference of residual hills is generally small, which reveals the relatively high degree of peneplanation. The ESR age of 1 200 m erosion surface is reported as 2.37 Ma by Li et al. (2001), which in accord with previous speculation.
PS4 (800–900 m) is preserved as wide valley among denudation monadnocks, or stripped ridges, island angular mountain summits and other shapes of geomorphologic surfaces. From southwest to east, the height difference between PS3 and PS4 is gradually decreased.
The lowest surface (PS5) below all other planation surfaces, with altitudes ranging from 500 to 600 m, was named by Liu (1983) as the Yun Meng surface. This surface extends along the subsidiary folds axes, it shows more typical planation surface properties in some broad valley areas, and in the narrow terrain it's more like a high altitude river terrace. Tian et al. (1996) proposed that the PS5 was formed at the end of the Early Pleistocene.
Comparisons of the results of this study with that of Tian et al. (1996) for the two locations are presented in Figs. 6 and 7. Shades and textures with colors in PRM in Figs. 6c, 6d, 7c and 7d directly reflect the landform features and are therefore easy for visual comparison. It can be drawn from Fig. 6 that both of the studies successfully mapped PS1 and PS2. But the areas of mapped planation surfaces were quite different. From Table 2, the mapped areas of this study for PS1 and PS2 were 406.74 and 173.95 km2 while Tian et al. (1996) mapped 110.41 and 85.63 km2 for PS1 and PS2. The total mapped area of this study was 580.69 km2, three times of Tian et al. (1996)'s, which is 196.04 km2. Figure 6 also indicates that all the PS of Tian et al. (1996) were included and extended by this study. It's well illustrated in Fig. 6d that some of the unmapped areas show the same landform characteristics as the mapped locations. It is revealed that the previous study just identified part of the planation surfaces of this area, also known as "omission error".
Figure 6. Comparison of mapped planation surfaces by this study (a), (c) with that of Tian et al. (1996) (b), (d). (a) and (b) are the classified planation surfaces; (c) and (d) are the mapped planation surfaces draped over the painted relief model.
Figure 7. Comparison of mapped planation surfaces by this study (a), (c) with that of Tian et al. (1996) (b), (d). (a) and (b) are the classified planation surfaces; (c) and (d) are the mapped planation surfaces draped over the painted relief model.
Table 2. comparison of the areas of planation surfaces mapped by this study with that of Tian et al. (1996) for the two selected locations
As for the second comparison location, both of the studies identified planation surfaces of four classes. Visual inspection of Figs. 7a and 7b shows good agreement between this study and Tian et al. (1996), but with some localized differences. Both studies mapped similar areas as for PS1 and PS4. However, this study identified 175.96 km2 for PS2, almost three times of Tian et al. (1996)'s, and 13.30 km2 for PS3, nearly half of Tian et al. (1996)'s. It's easy to find that the coverage of previous planation surfaces are also included and extended by this study. It's well presented in Fig. 7d that some of PS2 is neglected and failed to be mapped by the previous study. After further examination with the aid of a DEM, it reveals that the previous study confused PS1 with PS2 in some areas. As for this case, it seems the previous study is subject to both omission and commission errors.
In the past few years, it has seen an explosive growth in the availability of free accessed DEMs with almost world wide coverage (e.g., SRTM DEM, ASTER GDEM). DEM-based landform mapping is more and more popular, which provides several advantages over conventional planation surfaces mapping using topographical maps. They include larger coverage that can be held as seamless data, easy integration into GIS due to digital properties, better accessibility and visualization techniques to enhance landscape features particularly with regard to features of variable scales. Due to its 3D nature and rendition with colors, PRM has more power to distinguish detailed geomorphological features than the other three examined methods given the same spatial resolution. The visualization technique presented here is directly transferable to DEMs of other places, and other mapping purposes. Although there may be variability in their suitability to landform mapping applications, this technique provides another choice for consideration. This study provides an effective and accurate approach that is very suitable for extensive planation surface mapping of large areas, however, expert knowledge is required as for the setup of training areas in the field as well as their association with the representation of landforms on DEMs.
As asserted by Hill and Smith (2008) that a truly "objective" method would use quantitative, and therefore exactly repeated, measures to parameterize and directly detect geomorphological features on DEMs. However, this remains a difficult task for landforms such as rounded hills where there is a wide range in amplitude and length, coupled with highly variable shapes resulted from the modification of landforms during the long evolution. Under such a circumstance, visual interpretation remains the best technique. However, new intelligent techniques for automated planation surface mapping should be studied in future work.