1. Introduction
Roads are critical components of civilization because they contribute to the development and maintenance of economic activities vital for the quality of modern lifestyles. Roads are the most important public investment of a country, with significant funds allocated for the construction and maintenance of safe and efficient roadways [
1,
2,
3]. The management of forests is highly dependent on roads and their accessibility. The location and construction of forest roads significantly impacts harvest efficiency, operational costs, and environmental impact. The speed of trucks in various classes of functional roads, including the duration in different operational phases, is an essential parameter for analyzing network services, timber prices, and transportation costs. Forest road is the basis for developing forest productivity models, fleet management, and decision support systems [
4,
5,
6]. Increasing the area of forests accessible to management increases national forests’ productivity and sustainability to satisfy the needs of local populations and forestry companies. A forest road network provide access to different parts of a forest for management activities such as protection against fire, pest invasion, logging operations, and rehabilitation of mountainous areas [
7,
8,
9,
10,
11]. Forest roads provide means of communication within a forest for the management and utilization of forest space, including hunting, cattle breeding, recreation, tourism, trekking and mountaineering. Kleinachroth et al. [
12] reported that annual deforestation rates between 2000 and 2017 were highly increased for areas that were 1 km away from roads and were highest for old roads, lowest for abandoned roads, and generally higher outside logging concessions. Therefore, it is important to evaluate the forest road network and wisely allocate the scarce resources available for the construction and maintenance of forest roads.
In Angola, roads have perpetually been a subject of discussion with complaints from local populations, with the extent of roads in the country underestimated for several decades. Road network planning has been impacted by several factors, including the ecological conditions of the regions, forest type, settling economic areas, market for forest products, topographical attributes, and forest exploitation goals.
High-resolution satellite imagery is a valuable source of data for detecting and mapping features such as roads, vegetation, and building footprints [
13]. In remote sensing, ground truth (information collected in locus) allows image data to be related to real features and materials on the ground. For this study, ground truth was not used. The development of roads based on satellite images depends fundamentally on drainage conditions, soil properties, hydrologic factors, stability of the land slope, vegetation, building construction, and landform. Roads are not designed only to connect two localities; they require a lot of data about the area to fulfill other functions. Due to the uniqueness of roads, planning an alternative road network can be a difficult task because one specific limitation of the data availability can compromise the entire planning process. A number of decisions have to be made in the planning process in relation to the type of road, location, and design. Moreover, the planning process varies between different countries, making it difficult to compare results from different studies.
A recent study discussed the linear disturbance in forests such as roads, trails, and asset corridors over large-scale areas [
14]. The abundance of road disturbances usually cut through the forest to allow the placement of geophones, facilitate access and the extraction of wood, and the exploration of minerals. Moreover, the process disturbs regimes in adjacent plant communities by creating gaps, changing plant composition, altering environmental conditions such as light, soil moisture, and bulk density [
8,
15,
16].
The topographical shape, river network, vegetation type, distribution, and the erosion condition of the forest area whose roads are realized can be determined directly using satellite images.
Over the years, several studies related to the road planning process using the multi-criteria decision-making (MCDM) system have been conducted, more of which rely on the use of one-dimensional variables such as distance or time as a cost function. The use of one-dimensional variables can easily lead to unrealistic results with different factors affecting the decision in selecting the most suitable road [
17]. Therefore, the use of the multi-criteria decision-making (MCDM) system, which allows for the use of heterogeneous data, is needed for the planning of road networks [
18]. A common assumption in the MCDM system is that none of the variables selected are expected to be better than all the others in an absolute way. The selection of the variables depends mainly on the objectives set; for instance, variables related to the maintenance of forest road infrastructures will be different from those related to the creation of alternative road networks.
Knowledge of forest road planning is currently based on local foresters’ experience and many field trips through forested areas. Thus, planning forest roads in largely inaccessible areas is inefficient and error-prone. Most of the existing roads that provide access to miombo forests and plantations zones in Angola are selected not because they are the best for accessibility, but because sufficient time was not available to find alternative routes leading to the same locations. Remote sensing techniques and spatial data analysis through geographic information systems (GIS) have been used to map, identify, and access forest roads. Satellite and airborne image analysis have offered valuable thematic information referring to lithology and altered zone mapping from photointerpretation and digital classification. Transport and highway engineering are one of the fields affected by developments in remote sensing and GIS aspects. Using the available spatial data, such as digital elevation models (DEMs) [
19], makes road planning easier. Using a computer to convert terrain data from existing analog maps and photos into digital files, the planner can rapidly develop and evaluate many route alternatives [
20]. The modeling process of road networks leads to graphing theoretical and optimization problems [
21,
22]. Many studies employed models’ GIS to develop programs to determine road locations automatically, such as TRACER, a decision support tool that provides a quick evaluation of alternative route paths [
23]; ROUTES, a tool developed to automate the road pegging process using a large-scale contour map and a digitizer; and PEGGER, an ArcView GIS plug-in that automates the route projection [
24]. GIS is essential in trail route planning; one can apply and evaluate a GIS-based methodology and MCDM systems for determining optimal recreational trail routes using important information items [
25]. In the field of object modeling, the proper representation of objects in the real world within a GIS environment is crucial [
26]. The number of criteria and approaches that have been adopted in these studies vary significantly. While in some studies only a single criterion is used (e.g., landslide), others evaluate several criteria simultaneously. The selection criteria were made in some studies among the alternatives created by assessing the effects of the proposed road network, whereas, in others, the locality where the road will be built was evaluated in terms of multiple criteria decision systems after which the best alternatives were later designed [
27]. Here are some examples of MCDM systems used in infrastructure planning and applied to road planning. Marcelino et al. [
28] used the MACBETH approach for pavement maintenance decision-making at the network level. The study refers to the definition of priorities in pavements of the Portuguese road network, considering different criteria and budgets constraints. Mosadeghi et al. [
18] compared the outcomes of different MCDM techniques in the context of urban expansion along a major transport corridor between two large cites in south-east Queensland, Australia. The results demonstrate that the use of a simplified method such as AHP can be sufficient. Several other studies used MCDM techniques, all for the purpose of the construction of a road network and suitability map based on multiples variables in the model [
7,
9,
27,
29,
30,
31]. Multiple criteria decision-making (MCDM) systems combined with GIS can be used to derive priority road maps and determine the effectiveness of the old and suggested alternative routes. The analytic hierarchy process (AHP) method is a suitable technique for determining the proposed roads’ criteria or significance [
21,
27].
In recent years, biodiversity field research has been carried out in Angola, especially in Huambo province [
32,
33]. There have been difficulties accessing the forest to conduct the national forest inventory (NFI). Roads, in general, do not reach the forest zones where management practices are required. However, no study on roads has been carried out to determine the accessibility of roads using criteria such as elevation, slope, soil type, flow accumulation, geology, and aspect. New roads are needed, especially those that lead to forest zones.
In this context, the objective of this work is to combine modern remote sensing (RS) methods of forest science and geoinformation science and technology to:
- (i)
detect the existing road network and
- (ii)
propose alternative routes inside the management zones that can lead to the current forest plantation locations using multi-criteria decisions, hoping to overcome the shortcomings and limitations of the traditional road planning process. This will be the starting point for further research activities towards developing a spatial decision support system (SDSS) for planning road networks in Angola. To understand the need for this study, it is worth considering the following: basic rural infrastructure was severely damaged during the civil war in Angola, particularly in the most war-affected provinces in the Central and Northern plateaus. Bridges and roads were severely damaged and destroyed, and in many parts of the country, landmines are still an issue. However, some of the mines were removed to allow for roads and bridges repair in most areas. Despite security conditions throughout the country after the war ended in 2002, many rural roads are only passable during the dry season, resulting in inferior road locations.
3. Results
Planning a road network begins with researching the topographic and geologic conditions. For this paper, we considered six essential parameters for the analysis. According to
Figure 4a, the minimum elevation was 1245 m and the maximum 2467 m. The slope ranged between 0–77.78% classified into 5 classes (
Figure 4b). The best slope classes for constructing alternative routes are between 0–3% and 3–5%, which suggests that about 75% of the entire province is suitable for road construction. The comparison between aspect and solar radiation maps indicated that southern and southeastern areas receive the lowest amount of sunlight. In contrast, the northern and northwestern areas receive the highest amount of sunlight (
Figure 5).
The most extended road length was 140.03 km, which was 87.75 km from the city point α and 25.5 km to 99.07 km closer to the railway passing through the A and E MZ to connect the forest plantations (see
Table 3 and
Figure 6a). Two main roads connecting plantations, labeled as numbers 1 and 5, were found to pass through areas with higher elevations (
Figure 4a). According to the soil type map, the highest rank was awarded to ferralsol soil types. Clearly, the Arenosols Ferralsols was the dominant soil type across the province, followed by Cambisols Ferralsols. Acrisols Háplics and Ferralsols Háplics occupied a much smaller percentage in the study area, thus receiving the lowest ranks (see details in
Figure 6b).
The direction of the flow accumulation goes mainly from north to south in classes 3 and 4; minor classes (1 and 2) are scattered across the province (see details in
Figure 7a). Quaternary Unconsolidated Sedimentary was the dominant geology layer, followed by Precambrian Basement. Quaternary Unconsolidated Sedimentary geological types are more permeable to water and dry out more quickly. The least common layer was Volcanic Mesozoic, mainly at the central northern part of the province at the western border with Cuanza Sul and Benguela provinces (
Figure 7b).
The created risk map shows the potentially hazardous road sections in the entire province (
Figure 8). In addition, it identifies the areas near forest plantations and denotes which road sections should be avoided.
Proposal of Alternative Routes Based on AHP
In this study, four main factors influenced road construction suitability in the province: soil type, flow accumulation, geology, and slope. The least important factor was elevation (
Figure 9 and
Table 4).
The consistency ratio was found to be 0.09 < 0.1, meaning that the matrix was consistent. In addition, the
λmax was equal to 6.59 (see details in
Table 4).
According to the suitability map, 59.51% of the total area is suitable for road development and is counted in class 4 and class 5 in the automatic classification (
Table 5 and
Figure 10a). While in the classification with 20 classes, the suitable zones for road construction go from 15–20 classes, comprising about 51.15% of the total area (
Table 6 and
Figure 10b). Nonetheless, it is interesting to observe the results of forest gain location, where gain is seen more in zones with less road access (
Figure 10).
Figure 11 illustrates the entire road network, including the alternative routes based on the AHP method. In addition, the road network was divided into four main categories: main roads, secondary roads, tertiary roads, and rural roads. The alternative routes are flexible, more realistic, and feasible since they allow connectivity with the existing roads (see details in
Figure 11).
4. Discussion
In the past decade, Angola has prioritized the repair, expansion, and modernization of infrastructure as a central element of post-civil war reconstruction and economic development. Roads have been the principal priority of the Angolan government’s reconstruction plans [
54].
The existing road network in Angola and, in particular, Huambo province was initially constructed to cover the local population’s basic needs. The civil war that ended in 2002 played a determinative role in the existing road networks. About one-third of the entire road network in Angola can be described as good; the rest is either considered to be fair or poor [
54,
55]. In the vicinity of Huambo province, there is no appropriate access to the forest areas and connectivity with the main road network to support local economic activities, such as wood extraction and transportation to the processing mills around the major cities.
In general, the determination and planning of a road is a complex process that requires the consideration of several variables that need to be analyzed simultaneously. During the design of potential routes, all available factors (e.g., slope, contour lines, aspect, geology, protected areas, soil quality, natural reserves, etc.) must be primarily determined using weighting coefficients, and they should also be evaluated and analyzed as a whole [
27]. For road construction, soil or geology showed the highest weight. Soil is a primary engineering material for road construction and the main properties required of a road embankment are minimal potential for movement and erosion [
56].
Slope stability is considered the basic requirement of any road built on an inclined plane in road planning and construction. At least 75% of the surface area in the province is covered with little slope (0–5%) (
Figure 4b), thus denoting favorable areas for road construction because they have lower levels of rain impact, run-off velocity, and soil erosion. To avoid or minimize erosional environmental impacts, careful attention must be given to road development and planning, especially in steep terrain. It is much better to have a poor road in a good location because the poor road can be fixed, but an inferior road location cannot [
52,
57]. The location of potentially dangerous slopes on existing roads (
Figure 8) provides a good overview of where to securely open new forestries considering all the variables that influence the construction of roads (
Figure 9 and
Figure 10). This means to potentially access environmental risks and identify construction difficulties at the road planning stage.
It is crucial to locate roads on stable ground, on moderate slopes, in dry areas away from rivers, drainages, and other problematic and difficult areas. In our study, northwestern aspects showed a high percentage of solar radiation (
Figure 5), meaning that those sites can be recommended for road development. Studies confirmed that some meteorological events, such as the direction of the rain, amount of sunshine, the morphologic structure of the area, the direction of the aspect towards a river, and roads make an area suitable for road construction [
51,
58]. The delineation of potential solar radiation (
Figure 5b) and water flow direction (
Figure 7a) indicate the potential areas for road construction, where geological features were also considered.
The drawback of some parameters used is that in Angola, the content related to the planning of forest roads is not yet legally defined. We expect that our study will contribute relevant information towards the issue of access to conservation MZ in the entire province (
Figure 6a), and it will be of general interest to those concerned with the effects of roads construction and environmental policies. This type of research was also suggested by Duarte et al. [
59] when they addressed Angola’s Lobito Corridor from reconstruction to development. At least two roads connecting plantations passed through high-elevation areas, with alternative routes suggested to better facilitate access to plantation areas (
Figure 11). This quality and detail of information should act as a signal to policy-makers and forest managers to revise the existing road network leading to the forest plantations and MZ.
Generally, roads in mountainous areas follow contours to minimize construction and traveling efforts. The AHP used in our study allowed us to track the dangerous and problematic road sections and determine the best alternative areas where roads should pass (
Figure 11). The road suitability map area values ranged between 40.76 and 19,730.19 and were reclassified into 5 classes: (1) terrible, (2) very bad, (3) bad, (4) very good, and (5) excellent. Based on the results (see details in
Table 5 and
Table 6), very good and excellent areas suitable for roads represent 37.27% to 58.7% in the province.
The AHP model found the relative weights of the main factors determining the most suitable location for road construction (
Figure 9 and
Figure 10).
Figure 9 and
Figure 10 are intended primarily for engineers, foresters, technicians, consultants, and government regulators with experience of forest access planning.
As shown in
Table 3, a pair-wise comparison matrix for the factors and their relative weights was processed based on Saaty [
45]. The consistency ratio calculated for all factors was less than 0.10, which means that the attributed weights were suitable and reliable for a road suitability map produced in ArcGIS Pro V2.7.2. A detailed reclassification showed 20 classes, with the last 5, numbering from 15 through to 20, the most suitable for road construction (see details in
Figure 10b and
Table 6).
The AHP model used a rating system based on experts’ opinions. However, experts’ opinions may vary between individuals; therefore, we analyzed the spatial data between road condition factors and the locations of roads. The result demonstrated that GIS accuracy can be used in combination with AHP to derive satisfactory road suitability maps.
According to our road suitability map (
Figure 10), the southwestern part of the region showed more limitations for road construction. Many of the existing roads showed potentially dangerous and inappropriate slopes (
Figure 8). Therefore, alternative routes to access plantation areas and cities are highly recommended. A similar methodology was used by [
60,
61] to create landslide susceptibility maps. Nonetheless, other studies demonstrate the efficiency of RS data and GIS–AHP-based models to design and optimize new routes [
62,
63,
64,
65]. Forest road network planners often need to make decisions on several objectives. Therefore, one of the most useful models is a GIS–AHP. For example, land-use planning in a remote desert zone is usually dependent on an efficient corridor and main road network system.
New alternatives routes were proposed to access forest plantation areas and facilitate access to cities. We believe that the alternative roads suggested based on the AHP methodology can be considered as a framework for forest road locations in the region, whereby factors such as elevation, slope (%), aspect, geology, and flow accumulation are considered using AHP and GIS techniques.
Figure 10 shows an overlay of AHP information with forest gain where it is observed that forest gain is mainly achieved in zones with less road access, while forest loss can be observed all along in the areas with access (
Figure 12). The results are important for monitoring purposes of tree cover (
Figure 13) in the future, where we can follow the trends for forest gain and forest loss. The main concern is whether building a road network in non-suitable areas can improve forest conservation, regeneration, and management (
Figure 12 and
Figure 13).
This study’s main motivation was the lack of a good road network that allows a good assessment and management of local forest resources. For that, a methodology based on several types of geographic information data and using the AHP method was developed. The importance of this study is reflected in its output, which can be extrapolated to other southern African countries (e.g., Zambia, Namibia, Botswana, etc.), which intend to realize NFI in the near future. The prime intent is to refine the current standards of practice that govern activities for road engineering. Such understanding of forest accessibility can guarantee a good efficiency in data collation and the creation of forest maps.