2.1 Study Area Description
Addis Ababa is fully urban. According to the Central Statistics Agency’s (CSA) 2007 census results, 662,728 peoples were reported as living in 628,984 housing units, with an average of 4.2 persons per household. The city contains 22.9% of all urban residents in Ethiopia and contains 3.7% of the total population of the country. Addis Ababa’s road network is highly affected by its topographic changes, resulting in positive uphill and negative downhills. With an average road gradient of almost 4%, it can be expected that the road gradient has a substantial effect on fuel consumption and CO2 Emissions. Detailed location and map descriptions are shown below in table 1 and figure1.
Table 1:Location Description
Addis Ababa City
|
Coordinates
|
38°44′24″E 9°1′48″N
|
Climate
|
Subtropical highland Climate
|
Area(total):
|
519.482 km2
|
Elevation
|
2,355 m
|
Time zone
|
Universal Time Coordinated (UTC+3)
|
2.2 Road transportation in the city
In the city of Addis Ababa, the dominant public transportation modes are automobiles and dry cargo. Standard city buses and mini-bus taxis are the most available and affordable means of public transportation for the majority of the residents. There are also private limited companies and governmental road freight transportations which give delivery to the customers within a city. As of 2019, a total of 596,084 types of vehicles were available (Addis Ababa city Transport Authority 2019). The proper movement of vehicles poses the greatest challenge for the city to establish a reliable and sustainable transportation system.
A project to improve the transport management systems by making heavy-duty distribution vehicles and public buses to traverse in selected lanes is in progress. The current research may prove instrumental in selecting an environmentally friendly road network for distribution vehicles, that takes into account the gradient and other environmental factors into consideration.
2.3 Data and Data Acquisition
The Data used in this study was obtained from different sources. The Road network data mainly digitized from google earth image of 2019 on the study area at the spatial resolution of 10m, and ortho photos were used to georectify the digitized road networks. Advanced Land observing satellite (ALOS), Phased array L band synthetic aperture radar (PALSAR) of 12.5 spatial resolution was used as a digital elevation model for the study area obtained from Alaska satellite facility. GPS Coordinates of the origin - destination locations were collected using actual measurements in the field. Emission related to the data were obtained from Methodologies to Estimate Emissions from Transport (MEET) emission estimation model and other literatures.
2.4 Methods
This paper primarily espouses a methodology that finds optimal eco route networks that reduce the consumption of fuel and its related CO2. Additionally, other pollutant emissions for transporting cargo from origin to destination and vice versa are considered.
The model application was done in four steps: (1) Establishment of 3D-RN ;(2) Fuel consumption and pollutant emission estimation by using Emission estimation Model; (3) Road network data set development and (4) Optimization for different scenarios and other attribute calculations by utilizing visual basic script options.
Below, Figure 2 shows the general Work flow and procedures that were adopted in this study and form the basis for deriving the Eco-Route of the study area and subsequently the overall findings.
2.4.1. Connecting 2D road Network with DEM: 3D Road Network Generation
At this stage, the 2D road segment consisting of 40,886 polyline records with functional class [motorway, principal arterial, sub arterial, collector and local streets] were pre-processed by adding speed attributes to fit study area speed limits based on the Addis Ababa City Road authority (AACRA )Standard: 80 for Motorway,70 for Principal arterial,60 for Sub-arterial and Collector, 30 for Local Street. The Street Hierarchy was also tabulated based on AACRA’s Street Hierarchy functional classes standard:1 for Motorway,2 for Principal arterial, 3 for Sub arterial, 4 for collector, 5 for Local Streets. It also worked on 2D Road Network. Finally, ALOS PALSAR DEM [12.5m resolution] was used to convert 2D road Networks to 3D Road Network. ArcMap was utilized to obtain gradients of a street network on the first step feature vertices to points on centerline feature class twice; one for beginning and one for end was generated.
Secondly, values are extracted from points on the start and end features classes. Finally, new field gradient were added by subtracting start elevation from end elevation values. Then values were divided by segment length and multiplied by hundred, which ultimately led to inaccuracies in gradient values. Therefore, to obtain more reliable slope it is necessary to determine with a higher resolution.
To obtain granular result, DEM was interpolated to the two-dimensional road networks to create three-dimensional road networks. Then, the created 3D road network was split at each turn, junction, and roundabout. Moreover, the straight road segments were split at every 10meter interval to create vertices at each 10-meter point. Therefore, more precise gradient values and three-dimension length of each edge were added as additional attributes in the road network data set. Finally, to determine the gradient direction from each value data was extracted using add surface information. Hence, the gradients’ direction (+ve and -ve) was obtained based on edges start and end elevation. The results are tabulated as an additional attribute value in the network data set.
The 3D digital model and the road network for the city of Addis Ababa are shown in Fig. 3. The study area is faced with many undulating terrains, particularly on local streets due to DEM resolution on some parts of local street gradient values being unrealistic. To overcome these gradient values for the generated street segments within study boundaries higher than the permissible maximum restriction for road slopes, are reduced according to Bartlett (2015).
2.4.2. Calculation of Fuel Consumption and Pollutants Emission Estimation
For this study, the method proposed in Methodologies to Estimate Emissions from Transport (MEET). MEET is based on road measurements and all its parameters are extracted from real-life experiments. It includes numerous vehicle technologies for different vehicle classes, such as weight classes (< 3.5, 3.5-7.5 ,7.5 -16 ,16-32, and >32 tons) (Demir, Bektaş, and Laporte 2014).
For calculations of emissions and fuel consumptions of a broad range of vehicles and engine types, MEET depends on a database of parameters. The emission factors and the parameters are gained by sampling a range of vehicles (i.e. heavy-duty vehicle range between 3.5- 32 ton) (Kousoulidou et al., 2010). MEET formula methodology for estimating emissions is based on total fuel consumption data. Fuel consumptions and CO2 Rates estimations are speed dependent Regression functions introduced below for all vehicle classes. All calculations and estimations are done using the data available on MEET vehicle weight ranges (i.e. 3.5 up to 32 tones) for three scenarios and each scenario tested with three cases.
The basic fuel consumption and other pollutants expressed only in speed (g/ km), and was obtained from MEET Model has been given below:
Scenario (1):
For V 5 – 60 1425.2V-0.7593 (1)
For V 60 – 100 0.082V2 – 0.0430V + 60.12 (2)
Other Pollutants
CO For V 5 – 100 37.280V-0.6945 (3)
NOx For V 5 – 50 50.305V-0.7708 (4)
For V 50 – 100 0.0014V2 - 0.1737V + 7.5506 (5)
VOC For V 5 – 100 40.120V-0.8774 (6)
PM For V 5 – 100 4.5563V-0.707 (7)
Scenario (2):
For V 5 – 60 1068.4V-0.4905 (8)
For V 60 – 100 0.0126V2 – 0.6589V + 141.2 (9)
Scenario (3):
For V 5 – 60 1595.1V-0.4744 (10)
For V 60 – 100 0.0382V2 – 5.163V + 399.3 (11)
Based on the vehicle’s category, corrections can also be applied to consider the gradient of street and load of vehicles' effects on the emissions.
Gradient factor
A road’s uprising or down rising has the effect of ascending or descending the vehicle's fuel consumption and emission (EEA 1999). For street gradient class, the gradient correction factor calculated using the MEET model. The model provides a gradient correction factor for a range of ( -6 % to 6 %) gradient. To include a broad range of gradient classes in a city road network, the given equations were adjusted to fit an exponential function. The equations (12-14) were also used as the road gradient correction factor for other pollutants They are expressed by the following equation, where RG represents the road gradient in percentage.
For scenario 1: GCF = 0.213e0.165RG (12)
For scenario 2: GCF = 0.419e0.180RG (13)
For scenario 3: GCF = 1.0458e0.187RG (14)
Vehicle Load factor
The higher the vehicle weight the higher its fuel consumption and emission and also the same for vice versa (Tavares et al 2009).
50% of the load for emission factor is corrected to tolerate different load conditions with the use of the following equation (15).
where,
Lcf = emission factor corrected of the fuel consumption in [g/km]
lp = 0 for empty load vehicle and lp = 100 for fully load
cf = load correction factor of the FC which is 0.18
For Loaded vehicles Lcf of 1.18 and for Unloaded Vehicles Lcf of 0.82 was applied
The equation above (15) was also used to consider vehicle load for other pollutant emission estimation calculations. The actual load factor for pollutants is given in MEET and applied for this study. The load factors applied for pollutants is given in table (2) while table (3) shows the load correction.
Table 2: Load factors applied to heavy duty vehicles
Pollutant
|
Load Factor(lp)
|
CO
|
0.21
|
NOx
|
0.18
|
VOC
|
0
|
PM
|
0.08
|
Table 3: Load correction factors applied to heavy duty vehicles
Load Correction Factor (Lcf)
|
Pollutant
|
Empty Load
|
Complete Load
|
CO
|
0.79
|
1.21
|
NOx
|
0.82
|
1.18
|
VOC
|
1
|
1
|
PM
|
0.92
|
1.08
|
MEET Suggests the Total Fuel Consumption in gram as in equation (16) and for all pollutants, (i.e. CO, NOx, VOC, PM) the total Emission per Road length is calculated in gram as in equation (17).
FC = Fcv * GCF* Lcf*D (16)
Total Emission of Pollutant = Pv * GCF* Lcf*D (17)
Where: Fcv = Speed factor for Fc , PV = Pollutants speed factor , GCF = Gradient Correction Factor , Lcf = Load Correction Factor , D = Distance in km
For real application, fuels in grams have been converted to Liter [L].
Diesel for automotive use is around 832gm/liter.
The calculated fuel consumption values are used as an impedance to find eco routes in a network. For comparison, the road networks are completed with time attribute, length and allowed road functional class velocity expressed in equation (19).
L[km] = 3D Distance used for Short Routes
Moreover, CO2 emissions rates are estimated based on fuel consumption only for each scenario. The following equation (20) is applied from the MEET methodology.
Weight of CO2 = 44.011 (wight of fuel/(12.011 + 1.008 *rH/C)) (20)
Where: rH/C = hydrogen to carbon ratio in the fuel ( ~2.0 for diesel)
2.4.3. Road Network Data Set Development
Transport networks are mostly done in a GIS environment by the Network dataset. Network datasets are made of network elements. Network elements are created from point or line features to create the network dataset. Network elements contain attributes that change navigation over the developed network. Network attributes help to regulate traversing ability over the network. In this study, a one-way restriction was applied while solving the analysis. The hierarchy was given according to the road functional class: how drivers generally select the level of a street to travel. It is predictable that drivers choose high order roads rather than low order roads. For this study, vertex connectivity was selected. And finally, Elevation fields were considered to help improve the connectivity at line ends. They contain elevation information that consider vehicles passing overpasses, underpasses and normal streets in the network effectively.
2.4.4. Optimization
The route optimization was done with ArcMap Network Analyst. In this study, the route optimization primarily focuses on minimizing fuel usage to find eco-routes in a network that connects origin-destination points. Additionally, time and distance attributes are added to be used as cost attributes of the network. This study mainly aims to reduce vehicle environmental footprints. Therefore, reduction in fuel is not questionable.
For each step, terrain and fuel consumption module was developed. The modules share information stored in a spatial database created in the GIS environment. Below, Figure 4 shows the structure of the developed model with its spatial data base and modules.
In this study, three scenarios which incorporate the different vehicle weight classes available in Addis Ababa were introduced to test different routing conditions.
Most fuel consumption models concentrate on vehicle, traffic, and environmental influences but do not capture driver related issues which are relatively difficult to measure (Demir et al., 2014). However, this study focuses on environment related influences (i.e. Roadway gradient effects on fuel consumption).
For this study design, the sample vehicle categories with a significant influence on road transportation emissions were selected based on vehicle weight classes. Samples of vehicles were selected from distribution vehicle private limited companies. The companies’ vehicles, including distribution vehicles, were selected, based on their weight (i.e 3.5 -32 tons), engine type (i.e diesel), and production year (i.e 2010 models). Their total number of vehicles was also selected: Company 1- 15 vehicles, Company 2- 13 vehicles and Company 3- 8 vehicles. Then, the vehicle weight classes were grouped into three scenarios. The first Scenario for diesel HDV is from 3.5 to 7.5 tons. The second scenario for diesel HDV is from 7.5 to 16 tons and the third scenario for diesel HDV from 16 -32 tons. All scenarios are implemented in the GIS network analyst environment. Each scenario contains three cases with different loading and routing conditions. The overall implementations are tabulated in table 4.
Table 4:Scenario Implementation for the methodology
Scenarios
|
Cases
|
Weight Class [Ton]
|
Loading Condition
|
Routing
|
Network Analyst Solver
|
1
|
1
|
3.5-7.5
|
Loaded
|
Simple
|
New Route
|
2
|
Full Distribution
|
Closest Facility Solver
|
3
|
Unloaded
|
Back Haul
|
2
|
1
|
7.5 -16
|
Loaded
|
Simple
|
New Route
|
2
|
Full Distribution
|
Closest Facility Solver
|
3
|
Unloaded
|
Back Haul
|
3
|
1
|
16 -32
|
Loaded
|
Simple
|
New Route
|
2
|
Full Distribution
|
Closest Facility Solver
|
3
|
Unloaded
|
Back Haul
|