Effects, Monitoring and Management of Forest Roads Using Remote Sensing and GIS in Angolan Miombo Woodlands

Abstract

Angola’s forests are abundant and highly productive with enormous potential to support local needs and exportation. The forests are well distributed across the country, but the existing road network is generally poor and, in some cases, inappropriate. Based on our previous work examining deforestation patterns and the modeling of primary tree attributes of vegetation types, we proposed forest management zones (MZ) for future planning in Huambo province in Angola. Herein, that same framework is applied for the detection of the existing road network in Huambo and the proposal of alternative routes inside the MZ. We used analytic hierarchy process (AHP) and geographic information systems (GIS) to optimize connectivity among the existing forest plantations and their distance to the closest major cities within the province. We developed road suitability maps based on AHP and GIS to ensure safer driving conditions and contribute to the forest planner’s access to the current plantations. According to the suitability map created, 59.51% of the total area is suitable for road development and is counted in classes 4 and 5 in automatic classification. Parameters such as geology, slope, distance from roads to the railway, soil types, elevation, flow accumulation, and aspect were used. We provide a completed assessment of the state of existing roads and evaluate the safety of the observed road sections based on the AHP method. The calculated weights of the factors were all consistent with the model used (consistency ratio was 0.09 < 0.1). Finally, we proposed the best alternative routes to the existing cities, MZ in miombo woodlands, and forest plantations inside the province. Our findings indicated that flow accumulation, soil type, and geology were the most significant factors impacting road construction. Overall, our framework is an important starting point for further research activities towards developing a spatial decision support system (SDSS) for planning road networks in Angola.

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.

2. Methods

2.1. Site Characteristics and Selection

Angola is located on the southwestern coast of Africa (1.24 million km²), and its climate ranges from tropical wet/humid in the north to highly arid in the southwest. According to the World Wildlife Fund, Angola has 15 ecoregions where miombo forests are the most widespread biogeographic unit [34]. Our study focused on Huambo province within the Central Plateau (12°29′59.99″ S; 15°39′59.99″ E) of Angola (Figure 1).

The total size of the Angolan road network is around 76,000 km. The classified network is approximately 43,655 km in length, or 58% of the entire road network. This classified road network includes about 26,000 km of fundamental roads, which connect the capital to the 18 provinces and the main cities, and around 17,500 km of complementary roads. Most of the classified roads are 6 m wide, but a few sections are 4.5 m wide [35]. The distribution of infrastructure networks in Angola follows patterns of population density and natural resources availability. Primary roads are mostly ones that connect cities (Figure 2), while forest roads, or at least tertiary roads, are in many areas non-existent. Nowadays, land managers face difficulties optimizing the existing road network to maintain all roads, trails, and paths for different purposes and reduce negative impacts on the environment [36]. The inadequate conditions of the roads caused by years of destruction and lack of maintenance contribute to the low traffic levels.

The dominant soils in Huambo are ferrosols, usually found at higher elevations, with fluvisols more common at lower elevations. With an area of approximately 34,270 km² and about 1.9 million inhabitants, Huambo province is divided into 11 municipalities. The region has a mean annual temperature of 20 °C and rainfall that ranges from 1200 to 1600 mm per annum [37,38]. The climate is humid mesothermal with dry winters and warm summers. Most precipitation comes during the warmer summer months (regularly between October and April), with heavy rainfall in December and March. Six of the major Angolan rivers originate in Huambo, many of which drain into the Atlantic Ocean. The vegetation in Huambo province consists mainly of four types [39,40]:

(a)

highland forests (Afromontane forests);

(b)

miombo woodlands;

(c)

swamps;

(d)

dry grasslands.

The landscape is mainly comprised of miombo (which is subdivided into closed and open miombo) and savanna woodlands, intercepted by grasslands in lower drained areas. These vegetation types are scattered and patchy and are difficult to effectively map to show their spread and distribution across the whole province. Frequent intense fires, the cutting of trees for charcoal, and clearing for new agriculture fields affect vegetation distributions and structure. Generally, the plantations species were introduced in Angola in 1930 to produce pulpwood and wood fuel production for locomotives. Plantations were distributed in the highlands and the Central Plateau in the provinces of Benguela (Ganda, Babaiera, and Alto Catumbela), Huambo (Kuima, Sanguengue, Ukuma, and Tchinjenje), Bié, and Huila (Bunjei), all along the railway [41]. Huambo province has the greatest number of eucalyptus plantations, and they are in a critical state of degradation. The accessibility to these plantations is of significant concern. In many cases, forest roads or trails are not created for silvicultural treatments or management practices.

2.2. Acquisition of Environmental Layers

Due to a lack of data, Google Earth Pro was used to create the digital elevation model [40,41,42]. First, the keyhole markup language (KML) layer of study area was imported in google earth pro to visualize and identify the area of interest. Using the tool “add path”, elevation data were captured across the entire area in approximately 17,639 points (approximately 100,000 km profile path) (Figure 3a). Then, the KML layer derived from the previous step was converted to GPS exchange format (GPX) using the free GPS visualizer website (https://www.gpsvisualizer.com/elevation, accessed on 20 November 2021). Next, the GPX file was converted to point shapefile in the ArcGIS Pro V2.7.2 (ESRI Inc., Redlands, CA, USA) using a conversion tool called “GPX to feature”.

For converting the point layer containing the elevation to DEM, the geostatistical analyst toolbar in ArcGIS Pro V2.7.2 was used. First, using a histogram of values, the normality of the data was investigated (Figure 3b); then, using a trend analysis tool, we evaluated the existence of any trend in our data in a different axis (Figure 3c).

Finally, the DEM was conducted using the deterministic method for interpolation available in ArcGIS: inverse distance weighting (IDW). Using the following properties, we were able to map the elevation in miombo territory with the lowest root mean square: 50.98, optimized power: 3.75, neighborhood type: standard, maximum neighbors: 15, minimum neighbors: 5, sector type: 4 sectors with 45° offset, major semiaxis: 0.98, and minor semiaxis: 0.98 (Figure 3d).

Using the conducted map aspect, slope (%) and flow accumulation maps were created, with the pixel size of 50 m × 50 m slope and aspect computed using Toolbox “Raster Surface”. For computing the flow accumulation, small imperfections of the DEM were first removed by the tool “Fill” located in the “Hydrology” toolbox. The output of Fill was then used to create a flow direction map using the D8 method, which assigns flow direction to the steepest downslope neighbor. Finally, a raster of accumulated flow was created using the “Flow Accumulation” tool based on the computed flow direction layer. To simplify the results, the log 10 of flow accumulation was first computed; as a result, the values of the entire map were rescaled from 0 to 5, then by writing a conditional equation in the Raster calculator tool, values ≤ 2 were extracted as the final flow accumulation map.

We created a potential average annual solar radiation layer using the “Area Solar Radiation” tool to determine the aspects with higher sunlight. Aspects with higher solar radiation are favorable for road construction because there is less or little soil moisture in areas where the sun shines more. The entire preparation of aspect, slope (%), and flow accumulation was conducted in ArcGIS Pro V2.7.2.

For geological features, we used the map developed by Upton et al. [43] for all of Angola and clipped only our study area, which contained three geological layers. We ranked these three layers, with Quaternary Unconsolidated Sedimentary considered the best layer due to its wide distribution across the country and its physical characteristics [44]. Edaphic features were ranked based on the Angolan soil map developed from previous studies in this area [45,46,47], with the highest rank awarded to ferralsols types [43,44].

2.3. Analytic Hierarchy Process (AHP)

The AHP method was used to find the relative weight and priority of individual factors and subfactors to suggest alternative routes to access the forest plantations in Huambo province. AHP is a decision-making approach using multiple objectives and criteria which allows the user to arrive at a scale of solutions rather than one pulled from a set of alternative solutions [47,48]. It helps decision-makers discover the most suitable routes for road construction and to understand the challenges in the design process.

The first step was to create a pair-wise comparison matrix with the help of the scale of relative importance for each criterion (Equation (1)). The pair-wise comparison was based on the expert knowledge and the guidelines for road construction and maintenance [49,50]. The decision-maker is an expert in Forestry Inventory with practical and technical knowledge to hold responsibilities regarding forest road networks. The components considered for pair-wise comparison (n = 6) were:

• Row 1—Column 1: elevation
• Row 2—Column 2: the slope of the roads
• Row 3—Column 3: soil type
• Row 4—Column 4: flow accumulation
• Row 5—Column 5: geology
• Row 6—Column 6: aspect

  $$A_{ij}=\left[\begin{array}{cccccc}{a}_{11}& a_{12}& a_{13}& a_{14}& a_{15}& a_{16}\\ a_{21}& a_{22}& a_{23}& a_{24}& a_{25}& a_{26}\\ a_{31}& a_{32}& a_{33}& a_{34}& a_{35}& a_{36}\\ a_{41}& a_{42}& a_{43}& a_{44}& a_{45}& a_{46}\\ a_{51}& a_{52}& a_{53}& a_{54}& a_{55}& a_{56}\\ a_{61}& a_{62}& a_{63}& a_{64}& a_{65}& a_{66}\end{array}\right].$$

  (1)

In the next step, we created the normalized pair-wise comparison matrix (NPWM), where each criterion ($a_{ij}$) was divided by the column sum of the pair-wise comparison matrix to produce the normalized scores as follows (Equation (2)):

$$NPWM=\frac{a_{ij}.}{\sum a_{ij}}$$

(2)

We then measured the criteria weights by considering the average values in each row. We calculated the consistency of the given values in each criterion by multiplying the original pair-wise comparison matrix with the criterion in each row. The weighted sum value for each criterion was calculated to extract the ratio. The normalization of the matrix was performed by dividing the given individual scores by the total sum of the scores. The results of the normalized matrix must be less than 1. Finally, we estimated the Lambda max (λmax) as the average of calculated ratios. We then calculated the consistency index as follows (Equation (3), [50]):

$$CI=\frac{\lambda max-n}{n-1}$$

(3)

where Lambda max (λmax) represents the total size of the matrix λ = n and n is the number of criteria used.

The consistency of the matrix (Equation (1)) was evaluated by calculating the consistency ratio (CR) (Equation (4), [45]):

$$CR=\frac{CI}{RI}$$

(4)

where RI is the consistency of the random generated pair-wise matrix, equal to 1.25 for n = 6.

Once the consistency ratio was found to be reasonable (<0.1), we used the values of the pair-wise comparison matrix for further analysis.

To map the final outputs, we used the AHP for the spatial decision-making plug-in in ArcGIS Pro V2.7.2, importing the final calculations of AHP.

2.4. Ranking and Classification

Based on the locality expert knowledge, literature on guidelines for road construction [11,51,52,53], and our judgment, we ranked each class (

Table 1. Rating summary table for all the used criteria.

	Rank
Layers	1	2	3	4	5
Elevation	>2000	2000	1800	1600	1400
Slope (%)	>15	15	10	5	3
Soil Type	Ferralsols Háplics	_	Acrisols Háplics	Arenosols Ferralsols	Cambisols Ferralsols
Flow Accumulation	x > 3	3 > x > 2	-	2 > x > 1	1 > x
Geology	-	-	Volcanic Mesozoic	Precambrian Basement	Quaternary Unconsolidated Sedimentary
Aspect	South-East, South	-	South-West	North-East, East, West	North, North-West

). The values of the matrix indicators are determined as shown in

Table 2. Pair-wise comparison matrix and principal eigenvectors for road analysis applied in the AHP model.

	Elevation	Road Slope	Soil Type	Flow Accumulation	Geology	Aspect
Elevation	1.00	0.14	0.14	0.14	0.20	0.33
Road Slope	7.00	1.00	0.33	0.33	0.33	3.00
Soil Type	7.00	3.00	1.00	3.00	1.00	3.00
Flow Accumulation	7.00	3.00	0.33	1.00	3.00	5.00
Geology	5.00	3.00	1.00	0.33	1.00	3.00
Aspect	3.00	0.33	0.33	0.20	0.33	1.00

. The higher the value, the better the condition for road planning and construction. The category scores were based on the importance of each component towards the construction of forest roads. For instance, the major part (94.89%) of the study area has no more than 10% of the slope.

2.5. Extraction of the Final Map Based on AHP

The ranked layers were multiplied by the criteria weight calculated during the AHP process. Finally, all the weighted layers were combined using the “Union” toolset in ArcGIS Pro V2.7.2. We summarized the features of the existing road network, in connection to the management zones as well as the weighted ranks for each polygon created (

Table 3. Characteristics of the existing road network, as connected to the current forest plantations.

Plantation Code	MZ *	Road Code	Length of Roads (km)	CSDR * (km)	City Point	Minimum Road Slope	Maximum Road Slope	Average Road Slope
1	E	2	87.75	25.50	α	0.023	2.090	0.824
2	A	3	78.47	8.21	β	0.041	6.971	2.323
3	A	3 & 8	70.05	28.94	β	0.041	10.604	3.349
4	A	1 & 6	140.03	99.07	α	0.035	4.973	0.907
5	A	5	52.45	25.60	β	0.238	6.005	2.021
6	C	6 & 9	42.13	14.44	α	0.068	2.652	1.430
7	A	3 & 8	70.05	19.19	β	0.041	10.604	3.349
8	E	4	60.44	57.30	β	0.003	1.885	0.611

* MZ: Management zone. CSDR: Closest straight distance to the railway.

and

Table 4. Weight of criteria for road location.

Layers	Criteria Weights	Weighted Sum Value	Ratio	Λmax
Elevation	0.0295	0.19	6.53	6.59
Slope (%)	0.1256	0.79	6.30
Soil Type	0.3005	2.12	7.04
Flow Accumulation	0.2771	1.90	6.84
Geology	0.2006	1.32	6.57
Aspect	0.0666	0.42	6.29

). Based on the calculated score, we classified the final map into 5 and 20 classes, respectively (

Table 5. Areas of suitability map classes based on AHP detailed in 5 classes.

Rank	Area	Area (%)	Score
1	40.76	0.12	27.71–300
2	1041.33	3.10	300–400
3	12,520.34	37.27	400–500
4	19,730.19	58.73	500–600
5	263.50	0.78	600–632.70
Sum	33,596.12	100

and

Table 6. Areas of suitability map classes based on AHP detailed in 20 classes.

Rank	Area	Area (%)	Score
1	0.73	0.00	175.62
2	0.28	0.00	261.09
3	76.22	0.23	312.90
4	216.40	0.64	344.45
5	324.90	0.97	369.88
6	319.26	0.95	390.62
7	280.23	0.83	407.98
8	330.96	0.99	422.27
9	502.58	1.50	435.44
10	888.39	2.64	447.71
11	1166.58	3.47	459.82
12	2285.73	6.80	471.89
13	1894.89	5.64	484.10
14	4023.09	11.97	496.46
15	4098.35	12.20	509.83
16	5343.20	15.90	524.72
17	4895.48	14.57	540.67
18	3671.97	10.93	559.83
19	2574.28	7.66	584.93
20	702.58	2.09	632.70
Sum	33,596.1	100

). The final map evaluated the current road network and served as a base layer for designing the alternative routes.

Global forest watch (GFW) data were used to enhance the interpretation of the results. Tree cover loss during the years 2001 to 2020, forest gain from 2001 to 2012, and tree coverage (2000) maps were extracted from the GFW website with a resolution of 30 × 30 m [53]. The extracted layers were then overlayed on existing network roads, alternative roads, and AHP final maps.

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.

5. Conclusions

Our research highlights the dangerous and problematic roads in the Huambo province of Angola and suggests alternative routes leading to the existing forest plantations and forest roads. An analysis of the parameters shows that the most significant factors impacting road construction were flow accumulation, soil type, and geology. However, it is worth emphasizing that the presented ranking depends not only on assessing the relative weights of the used factors, but also on the assumed importance criteria. This study proposes a comprehensive and detailed methodological framework whereby dynamic maps of the existing road network were created using GIS in combination with the AHP method. The integrated use of GIS and AHP allowed for the proposal of alternative routes in an optimized manner. The proposed alternative route was (i) to provide access to the forest zones and forest plantation in order to implement management plans and (ii) to promote forest operations that are environmentally sound, socially acceptable, and economically viable. We introduced the concept of AHP for the possibility of constructing alternative routes and to detect all dangerous roads based on the selected parameters used in AHP. The produced suitability map can be a good source of information for decision-makers, land-use managers, engineers, and architects. Overall, our framework is an important starting point for further research activities towards developing a spatial decision support system (SDSS) for planning road networks in Angola.