A fine construction method of urban road DEM considering road morphological characteristics

Study area

The current study focused on exploring a fine construction method for urban road DEMs. Therefore, the diversity and morphological characteristics of urban roads should not be the only factors to consider when selecting the experimental area, but also the availability of data for the area. To meet these requirements, a part of the central Jianye District of Nanjing City was selected as the study area (Fig. 1). The topography of the study area is significantly affected by human activities, the urban road network is complex, and the spatial elements are complete. Thus, the study area is ideal for investigating urban road DEM modeling.

Data acquisition and preprocessing

Nanjing 1:500 DLG data was used as the experimental data source in this study. DLG data have a massive amount of urban road geographic information, including geometric, semantic, and elemental relationship characteristics. Thus, these data can provide accurate spatial location and attribute information for high-precision urban roads DEM construction, such as road centerlines, road edge lines, road surfaces, and elevation points. However, topographic line data in DLG data only include plane information without the corresponding elevation information, making the continuous spatial characteristics of urban roads difficult to describe. Consequently, the original elevation points cannot be applied directly to high-precision urban road DEM construction. In addition, different types of urban roads in DLG data have varied data collection processes in DEM construction, and thus, urban roads should be hierarchized in accordance with morphological characteristics. Therefore, performing a series of data pre-processing steps on DLG data, including urban road hierarchies, urban road information extraction and supplementation, and elimination of urban road elevation outliers, is necessary.

Urban road hierarchies

Previous urban road classification methods have focused on the expression of road level and utilization function without considering the morphological characteristics of roads. Thus, meeting the current increasingly high-precision and complex urban road DEM construction process is difficult. To express the lateral and longitudinal morphological characteristics of roads completely and restore the actuality of road DEM, the current study classifies urban roads into six categories in accordance with morphological characteristics: main lane, auxiliary lane, single lane, bicycle lane, sidewalk, and separation zone from the perspective of urban road DEM construction (Table 1).

The spatial combination relationship of urban road elements under an ideal condition is illustrated in Fig. 2. From the road cross-section, the main lane is the road for motor vehicles, which is typically the core part of the road, located in the middle of the road, and the key element of road DEM construction. The auxiliary lane is generally located on both sides of the main lane, and it can be used by motor vehicles alone or shared by motor vehicles, electric vehicles, and bicycles, depending on planning needs. It is usually provided for vehicles that will make lane changes or turns. The single lane is located next to the auxiliary lane, and it can be used by bicycles, electric bicycles, and other nonmotorized vehicles. It usually includes the electric bicycle lane and part of the bicycle lane. In addition, separation zones can be appropriately added between the aforementioned categories of roads in accordance with the geometric width and functional needs of roads.

However, many different spatial combinations of urban road elements exist in practice due to the constraints of site and planning factors. The common combination forms are as follows: main lane + auxiliary lane + single lane + bicycle lane + sidewalk + separation zone, main lane + auxiliary lane + single lane + sidewalk + separation zone, main lane + auxiliary lane + sidewalk + separation zone, and main lane + single lane + sidewalk + separation zone. From the DEM modeling perspective, oriented to the multiple spatial combination changes of urban road elements, the main lane, auxiliary lane, single lane, bicycle lane, sidewalk, and separation zone are ranked from high to low in accordance with the level of road elements to divide them into one, two, three, four, five, and six levels, respectively (Table 1). That is, when a lower-level road is gradually integrated into a higher-level road in the extension direction, road morphology follows the principle that a higher-level road is superior to a lower-level road. The higher-level road is used as the DEM construction benchmark. For example, when the auxiliary lane joins the main lane, road DEM is constructed with the morphological characteristics of the main lane as the benchmark.

Extraction and supplementation of urban road information

In accordance with the urban road hierarchical method from the DEM perspective, the information of the main lane, auxiliary lane, single lane, bicycle lane, sidewalk, and separation zone in the study area is extracted separately from the DLG database. Subsequently, road surface, centerline, and corresponding elevation point data were extracted for each level of road elements on the basis of elevation, spatial, and semantic information.

The road surface of the main lane, auxiliary lane, single lane, bicycle lane, and sidewalk with a width greater than 1 m is replenished by using the DLG database of road edge line, road accessory facility surface, and residential surface data. In addition, although the DLG data include sufficient road surface and centerline data, problems, such as overlaps and voids at the junction of various road elements, occur. Therefore, checking the topology of each level of road is essential to ensure coherence and integrity between the individual road surfaces, particularly the core integrity of the main lane. Notably, road information is replotted on the basis of the longer main lane continuity when intersections are encountered. For the road surface after replotting and error correction, the new road centerline is re-extracted to produce complete road centerline data (Fig. 3).

Elimination of road elevation outliers

As a result of various factors, such as measuring instruments, measuring process, and external environment, outliers inevitably occur in urban road elevation data. Urban elevation outliers are the abnormal abrupt change points within a road surface. They include global and local outliers, which are too high or too low than the whole trend surface of the road surface, destroying the morphology of urban roads and leading to the distortion of urban road DEMs. Therefore, eliminating the elevation point outliers of urban roads is particularly significant for urban road DEM construction. The elimination steps are as follows. In the first step, the exploratory spatial data analysis tool in ArcGIS software is used to initially search and eliminate road elevation outliers. In the second step, the remaining elevation points are constructed into a triangulated irregular network (TIN) and then transferred to a raster to produce a preliminary road DEM. Then, this road DEM is analyzed for mountain shadow rendering to eliminate elevation outlier data. In the third step, the improved Douglas–Peucker (DP) algorithm with an elevation change rate³⁰ is applied to eliminate elevation outlier data that are inconsistent with road morphology. This step is repeated until no evident elevation outliers remain. In the final step, the elevation point data that meet the requirements are combined as the final urban road DEM construction elevation point data. The comparison of the original elevation points (OEP) of the urban roads and the constructed model elevation points (CMEP) with the elimination of outliers is depicted in Fig. 4.

Urban road DEM construction method

The morphological characteristics (including longitudinal and transverse morphological characteristics) of different levels of urban roads vary significantly; thus, hierarchically constructing DEMs of different levels of urban roads is necessary.

Construction method for the main lane

The main lane is the core spatial element of the whole urban road network, and an accurate representation of the main lane model provides an important foundation and support to urban road DEM construction. The following steps are used to model the main lane (Fig. 5).

1. (1)

   Elevation points assignment mapping The extracted main lane elevation points are not strictly located on the road sideline, as shown in Fig. 5a. Thus, mapping the extracted road elevation points vertically to the main lane sideline (generated by the main lane road surface) and main lane centerline, as depicted in Fig. 5b, is necessary.

2. (2)

   Elevation point encryption The density of topographic elevation points in DLG data is relatively sparse and insufficient to support high-precision urban road DEM construction. According to the industry standard of “Digital products of fundamental geographic information 1:500 1:1000 1:2000 digital elevation models” (i.e., 1:500 high-precision urban DEM grid size is 0.5 m)³⁹, we use the equal spacing encryption method to encrypt the mapped elevation points at equal spacing of 0.5 m from the longitudinal section (Fig. 5c) and cross section (Fig. 5d) directions.

3. (3)

   Main lane DEM construction The main lane DEM is produced via inverse distance weighting (IDW) on the basis of uniformly distributed elevation points, road centerlines, and urban road surface data after encrypted interpolation.

Auxiliary lane and single lane construction methods

The auxiliary lane and single lane have the same morphology as the main lane in urban roads under most situations, and their construction approach is consistent with that of the main lane. However, some morphological characteristics at the junction of the auxiliary lane and single lane with the main lane do not match the significant height difference of the main lane, which is inconsistent with the actual gradual integration of roads of all levels.

Therefore, to overcome the problem that the end of the road cannot be smoothly transitioned in the DEM construction of the auxiliary lane and single lane, the elevation values at the intersection of the end centerline and the road surface are extracted from the main lane DEM and involved in the classification modeling of the auxiliary lane and single lane to control road end morphology, and finally, construct auxiliary lane DEM and single lane DEM.

Bicycle lane and sidewalk construction methods

The lateral morphological characteristics of the bicycle lane and sidewalk are single flat and straight surfaces, and the construction of the bicycle lane DEM model can refer to the construction method of the main lane DEM. The construction of the sidewalk road DEM model can be divided into four situations in accordance with the number of elevation points.

1. (1)

   Sufficient road elevation points The sidewalk DEM construction method is the same as the aforementioned main lane DEM construction method.

2. (2)

   Insufficient road elevation points and road surface has only one elevation point. A sidewalk generally exists in blocks, and its longitudinal length is shorter than those of other road levels. Longitudinal elevation exhibits nearly no change. A plane can be determined by one elevation point to express the surface of the sidewalk.

3. (3)

   Road elevation points are insufficient, and road surface has only two elevation points. If the two elevation points are located at the beginning and end of the road, then these points can be used to determine a short flat straight plane to express the surface of the sidewalk. If the two points are not strictly located at the beginning and end of the road, then they can be handled in accordance with the second situation, i.e., find the average to determine the sidewalk road surface.

4. (4)

   No elevation points on the sidewalk road surface. No model is constructed for this situation.

Separation zone construction method

Separation zone morphology is evidently different from the other levels of road morphology; it is essentially a road boundary at all levels. Whether the separation zone can be constructed in urban road DEM depends on the effect of its shape or width. In general, only the non-cutaway type of separation zones with a width greater than 1 m can be constructed in urban road DEM. Then, the actual local morphology of the separation zone is constructed in the model by using the corresponding construction method, such as the traditional TIN construction method to build a natural undulating separation zone and the flat surface method to construct an artificially modified separation zone.

Similarly, the distribution of elevation points is uneven in the separation zone surface, and some of the separation zones have a sufficient number of elevation points for normally constructing the model, while others have fewer elevation points and unable to build the model. To address the aforementioned problems, elevation points within the non-urban road surface, i.e., elevation points of non-urban buildings, flat surfaces, slopes, and other special feature elements in urban parcels, and the road boundary line, steep hill line, and slope line elements are used as constraints to automatically encrypt the remaining elevation points for constructing DEM²⁸, and finally, the separation zone model is clipped based on the boundary line of the separation zone.

Merging of models

The aforementioned hierarchically constructed urban road DEMs are merged to generate the complete urban road DEM. In accordance with the core construction principle of the main lane, the main lane DEM is extended to the single lane or bicycle lane, and then the auxiliary lane DEM is binarized and embedded into the main lane extended surface model by using the conditional function tool to obtain the “main lane + auxiliary lane” DEM. The preceding mosaic method is repeated to embed the single lane, bicycle lane, and sidewalk DEMs into the previous mosaic result to obtain the “main lane + auxiliary lane + single lane”, “main lane + auxiliary lane + single lane + bike lane”, and “main lane + auxiliary lane + single lane + bike lane + sidewalk” DEMs. Finally, the separation zone DEM is embedded into the “main lane + auxiliary lane + single lane + bike lane + sidewalk” DEM in accordance with the same method to complete all the merged models. Road surface buffer analysis is performed to eliminate the problem of jagged gaps that occur at different levels of urban roads during the merging of models. In the end, topological checks are used to ensure the correctness of the spatial relationships of all levels of urban road DEM. In addition, slope data are extracted from the constructed urban road DEM for further analysis.