An Evaluation of Environmental Impats of Different Truk Sizes in Last Mile Distribution in the ity of São Paulo, Brazil Nathalia C. Zambuzi 1, Claudio B. Cunha 1, Edgar Blano 2, Hugo Yoshizaki 1, Carla D. Carvalho 1 1 Universidade de São Paulo, São Paulo, Brazil 2 Massahusetts Institute of Tehnology, Cambridge, USA {nathalia.zambuzi@usp.br, bunha@usp.br, eblano@mit.edu, hugo@usp.br, darval@gmail.om} Abstrat. Motivated by regulations that have prohibited the traffi of large freight vehiles in an inner area of 100 km2 in the ity of São Paulo, Brazil, in this paper we evaluate the impat of this poliy on CO2 emissions by delivery vehiles. These emissions are estimated for two different sizes of freight vehiles under different operational senarios: the so-alled VUCs (4 ton Urban Freight Vehiles) and larger rigid truks (9 ton). Using different ustomer densities, drop sizes, and distribution enter distanes obtained from a sample survey, as well as Daganzo s ontinuous approximations for trip lengths and the NTM method, trade-off urves have been onstruted. From these urves, it an be seen that bigger vehiles are, on most ases, a better hoie than VUCs regarding emission, showing that this poliy annot be generalized for every supply hain. Keywords: Urban Freight, Environmental Impats, Freight Poliies, CO2 emissions. 1 Introdution Aording to the United Nations, São Paulo is the largest ity in the Southern Hemisphere [1]. With more than 4.5 million ars, truks and buses flowing every day [2], heavy ongestion, noise, vibration, traffi aidents and air pollution affet everybody. Aording to the Environment Ageny of the State of São Paulo [3], motor vehiles were responsible for the emission of 260,000 t of air pollutants in 2014, almost 90% of total emissions in the metropolitan region of São Paulo. Truks generated 11% of that total (28,000 t), but aounted for less than 3% of the total fleet. Additionally, high levels of air pollution in São Paulo redue life expetany by one and a half years [4]. Even though freight vehiles have a signifiant ontribution to road ongestion and CO2 emissions, they also supply goods to more than 200,000 ommerial establishments aross the muniipality of São Paulo [5]. These goods satisfy, every day, the onsumption and servie needs of its 11 million itizens. Planning and implementing ations and poliies that seek to balane the need for goods and environmental onerns, have inreasingly been identified as a fundamental role in a modern ity [6]. To redue the ongestion and mitigate its environmental impats, São Paulo authorities have implemented initiatives suh as liense plate rotation for passenger ars and ommerial vehiles, and restrited hours in whih is allowed the irulation of truks. Roughly, those restritions apply within the Maximum Traffi Restrition Zone (Zona Máxima de Restrição de Cirulação - ZMRC). The ZMRC omprises an inner and denser area of 100 km2, with a higher onentration of traffi, stores and servie in São Paulo City. Traditional truks are prohibited to irulate within or aross the ZMRC during speifi periods: Mondays to Fridays between 5 am and 9 pm and on Saturdays between 10 am and 2 pm. Sine 2008, smaller truks, known as Veíulos Urbanos de Carga or VUC (aronym for urban freight vehile) are allowed to drive throughout the ity with no time restritions, exept the liense plate rotation. They are muh more agile than larger truks, thus merge better with ar traffi, but are also less produtive. As most ustomers insisted in daytime deliveries, this has resulted in a signifiant growth of the VUC fleet: 67% over the last ten years [7], against an inrease of 6% of other freight vehiles. In this ontext, this paper aims to analyze and estimate the environmental impat of this poliy, from a CO2 emissions point of view, whih in pratie substituted VUCs for regular truks. The remainder of the paper is organized as follows. Setion 2 presents essential literature of both arbon emissions and length of delivery routes estimations. In Setion 3, we detail the proposed model for estimating CO2 emissions. Setion 4 desribes the values used for the model parameters. Results are presented and disussed in Setion 5. Finally, setion 6 makes final remarks and outlines some diretions for future researh. - 1 -
ILS Conferene 2016, June 1 4, Bordeaux, Frane 2 Carbon emission methods in transportation ativities and delivery route length In this paper, we use CO2 emissions estimation methods that are ativity-based, i.e., the emission fators are estimated based on a referene model and database that relates a partiular transportation ativity to the emissions aused by that ativity. In the GHG Protool methodology, the required parameters in order to estimate emissions (if fuel onsumption information is not available) are freight distane traveled (ton x kilometer) and type of vehile (individual type emission fators). The most ommon database is the data provided by the United States Environmental Protetion Ageny - EPA (GHG Protool Calulation Tools) [8]. Unfortunately, EPA uses the same emission fator regardless of the type of vehile or type of road and travel speed, by assuming an average unity fuel onsumption for the diesel truks [9][10]. This methodology does not inlude a speifi load fator, and therefore is based on an average utilization per truk. An alternative ativity-based methodology is the one developed by the Network for Transport Measures NTM in Europe [11]. The NTM methodology, at the highest level of aggregation, is also based on an average utilization per truk, but type of road and type of vehile are required to estimate fuel onsumption. At the next level of detail, NTM methodology requires detailed parameters: fuel onsumption, distane traveled and weight per shipment. Fuel onsumption is a funtion of the type of truk/trailer, the load fator and the type of road. Velásquez et. al. [12] have shown that higher resolution modeling of CO2 emissions is relevant for deision making. For the purpose of this paper, we will use NTM method at this higher aggregation level. With respet to delivery tour lengths, Daganzo [13] proposed an analytial strategy to determine good traveling salesman tours and estimate their lengths in zones of irregular shapes. Daganzo and Newell [14] analyzed the osts involved in distributing items from a warehouse or depot to randomly sattered ustomers on a daily basis, and derived expressions to determine tour lengths based on density of stops and their average distane to the depot when the loation of the delivery points annot be determined. Motivated by the fat that in most real problems, the road network is a rough ombination of Eulidean, retangular, or ring-radial metri, Galvão et al. [15] applied a multipliatively weighted Voronoi diagram formulation to finding a near optimal partition of the served region into delivery zones or distrits. The Voronoi diagram approah aims to smooth distrit ontours. The method was applied to solve the parel delivery problem that had been previously analyzed by Novaes and Graiolli [16] [17]. The resulting distrit ontours are smoother and loser to onfiguration ontours enountered in pratial situations. The resulting region partition led to a more balaned time/apaity utilization (load fators) aross the distrits. These approximation methods form the basis for our trip length estimations. 3 Estimating CO2 emissions in the ZMRC 3.1 Assumptions on Traveled Distanes and Customers Demands A delivery vehile must visit a ertain number of ustomers in its delivery route, aording to the demand and to its load apaity. For the same ompany, it is assumed that all ustomers' orders have the same basi harateristis, i.e., items delivered belong to the same ategory and drop sizes are omparable during the entire route, i.e., they do not differ signifiantly among deliveries and thus an be used to determine how many vehiles are required to servie an area, as detailed in the following subsetion. All the deliveries are assumed to be made by driving the vehile to the ustomers to be servied. The vehile starts its route from the ompany s depot/distribution enter and travels to the delivery area. After serviing all the ustomers, the vehile must return to the depot/distribution enter. These two legs of the round trip are assumed to be of same length, named here as line haul distane (d l ), see Figure 1. The average distane between two onseutive ustomers d is a funtion of a density parameter, desribed in the model as ustomers per blok. We estimate d using the formulation proposed by Daganzo [13] [14] developed an approximate formula for expeted tour length in zones of irregular shape that is a funtion of the area density levels. Assuming N points uniformly and independently sattered on a region of area A Daganzo (1984) shows that the expeted tour length L is given by: 2 L l NA (1) - 2 -
ILS Conferene 2016, June 1 4, Bordeaux, Frane Figure 1: Traveled distanes for a delivery vehile route Where: : tour length fator, given as a funtion of the shape/density onstant ; : density of points; l : smaller side of a retangular-shape zone where points are loated N : number of points loated in a onneted region of a plane, onsidered here as the total number of visited ustomers; A: area of the plane where the N points are loated, i.e., the delivery area. The author shows that for zones with l 2 12, the tour length fator should be equal to 0.9: L 0. 9 NA (2) Using a density parameter of ustomers/blok and onsidering a blok a surfae of 100 m x 100 m, the average distane between two onseutive ustomers (d ), is given by Expression 3: where: d 0. 9 * 100 d : average distane between two onseutive ustomers [m]; : density of ustomers [ustomers/blok]. Thus, the total traveled distane by a delivery vehile, an be then obtained by Expression 4: d 2 d n (4) l d where: D : total traveled distane by the delivery vehile; d : line haul distane; d l : the average distane between two onseutive ustomers; n: number of ustomers visited by the vehile on the delivery route. The CO 2 emissions are then estimated using the NTM methodology. For a vehile type v, travelling on a road of type r, CO 2 emissions ECO (, ) 2 v r depends on three main parameters, as in: E EF FC D (5) CO2 ( v, r ) ( f ) ( v, r ) ( v, r ) Where: EF ( f ) : emission fator of CO 2 [g/l], for fleet's fuel type f; FC ( v, r ) : fuel onsumption rate [l/km] of fleet's vehile type v, driving on a road type r; D the distane traveled [km] by vehile type v on a road type r. ( v, r ) In summary, using Expression 4 to estimate the total traveled distane of the delivery vehile and onsidering a homogenous vehile fleet and a single type of road, the total emissions E ( ) CO 2 V for an entire daily delivery operations of a ompany is given by: E CO2 ( V ) EF FC 2 V d ( f ) ( v, r ) l N d (3) (6) - 3 -
ILS Conferene 2016, June 1 4, Bordeaux, Frane where: V: total number of delivery vehiles required; d : line haul distane (road type r); l d : the average distane between two onseutive ustomers (road type r). The daily number of ustomers varies signifiantly by different produts and ompanies. Thus, the total number of servied ustomers (N) is a funtion of the density parameter and the delivery area (A): N 100 A (7) where: A: delivery area [km 2 ]. 3.4 Delivery Vehiles Assumptions Two type of vehiles are ompared for the analysis: VUC and a standard truk named TOCO. As mentioned before, a VUC is a small vehile speially designed to meet São Paulo's regulations, inluding a smaller payload. It is the only type of delivery vehile allowed to irulate within the ZMRC of São Paulo during daytime. TOCO is a larger vehile ommonly used in urban freight in Brazil. When a poliy restrits the maximum size or weight of delivery vehiles, it restrits the load apaity of these vehiles and, therefore, the number of ustomers they an visit on a single route. Thus, if N is the total numbers of ustomers to be visited in a daily delivery operation (required by Expression 7), the number of required vehiles (V) is: N s 1 V, V V W / ' m i n n Z n V (8) Where: s : drop size of ustomer (assumed to be the same for all ustomers); W : maximum load apaity of the vehile, in ton or m 3, whihever is more binding. 4 Model parameters 4.1 Vehile Parameters Table 1 presents the assumed values for the parameters of the two vehile types used for the quantitative analysis. As dimensions and load apaities vary, a standard VUC and a standard TOCO are adopted. The standard VUC and TOCO are represented by models ACCELO 815/31 and ATEGO 1419/48 by Meredes Benz. Fuel onsumptions and emission fators are based on the GHG Protool Brazilian Program: for VUC and TOCO they are based, respetively, on the average fuel onsumption for a light truk (TGW between 3.5 and 10 tons) and a medium truk (TGW between 10 and 15 tons). The CO 2 emission fator, in turn, depends only on the type of fuel used by the vehile. Table 1: Vehile Parameters Parameter VUC TOCO Average external dimensions 6.134 x 2.176 x 2.48 8.175 x 2.486 x 2.713 (lenght x width x height) [m] Total gross weight (TGW) [t] 11 14.3 Average load apaity [t] 4.93 9.59 Fuel Diesel Diesel Fuel onsumption [l/km] 0.131 0.180 CO 2 emission fator [kg/l] 2.67 2.67 Soure: Meredes Benz and GHG Protool Brazilian Program (2013) - 4 -
ILS Conferene 2016, June 1 4, Bordeaux, Frane 4.2 Delivery Parameters Table 2 lists the parameters required to estimate the total number of vehiles for a daily delivery operation. They are based on several observations, detailed field surveys and route shadowings performed in the ity of São Paulo in 2012 and 2013. Delivery time parameters (a and b) are based on average values from data olleted. We onsider the same type of road for the entire delivery route as São Paulo is mainly omposed of urban streets and the traveled distanes on not-urban roads observed are irrelevant. The adopted value for vehile's speed is of 50 km/h, onsidering the historial average speeds for the morning period (Map data 2013 Google, MapLink) and the speed limit of 60 km/h for most of São Paulo s urban roads. The delivery area is 968 km 2, whih is the urban area of the ity of São Paulo (Embrapa, 2013). This does not negate the insights of the analysis, one it only influenes the proportions of the delivery operation. In other words, for a given density of ustomers per blok, the number of ustomers is proportional to the extent of delivery area, as it is the number of vehiles required and their total travelled distane. The last three parameters of Table 2, ompany's line haul distane, ustomer density and ustomers' drop size are the hanging parameters for the base senario of analysis, as explained in the following setion. Table 2: Delivery Parameters Parameter Adopted value Delivery time, fixed[min] a 4 Delivery time, variable [min/kg] b 0.1 Vehile speed [km/h] v 50 Delivery area [km2] A 968 Line haul distane [km] d l Variable Customer density [ustomers/blok] Variable Drop size [kg] Variable 5 Senarios and results We aim to examine senarios with the perspetive of different ompanies urban delivery systems. A ompany's urban delivery operation is omposed of three basi omponents: (i) distribution enter (DC) loation, (ii) delivery area and (iii) type of good delivered. In this sense, the base senario for a "Vehile Size/Weight Restrition" poliy is based on three parameters, ompany's line haul distane, ustomer density and ustomers' drop size, whih are strit related to these basi omponents. For eah senario analyzed, CO2 emissions from the two types of vehiles (VUC and TOCO) have been omputed using the parameters shown in Tables 1 and 2. From the base senario others are originated in whih drop size is replaed for others parameters suh as fuel onsumption, and the lowest CO2 emission type of vehile for the whole fleet is hosen as a funtion of line haul distane and ustomer density. Eah urve is alled indifferene urve, and represents the points where CO2 emissions are the same for VUC and TOCO, i.e., below or above the indifferene urve, one of them is the less pollutant hoie. In addition, the analysis also takes into onsideration two different types of delivery-profiles, based on the density of ustomers from two ompanies making deliveries in São Paulo: a beverage ompany and a small groery retailer. These ompanies are represented in the following harts as a point representing their density and line haul distane. The beverage ompany delivers to more than 3,500 ustomers per day, whih orresponds to a density of about 0.037 ustomers/blok, and has its distribution enter loated inside the ity limits, only three kilometers far from ity enter. The groery retailer ompany attends approximately 100 stores per day, resulting in a density of about 0.0011 ustomers/blok, and its distribution enter is loated 21 kilometers far from ity enter. The "Vehile's size/weight restrition" poliy is strit related to vehile's load apaity, whih is the only determinant parameter for the total number of vehiles required for a daily delivery operation. Thus, onsidering the same number of ustomers and drop size, and no other restrition (suh as driver's working hours or ustomer s reeiving hours), more small vehiles are needed than bigger ones. - 5 -
ILS Conferene 2016, June 1 4, Bordeaux, Frane 5.1 Base Senario The base senario examines the effets of different drop sizes on vehile's hoie. As shown in Figure 2, TOCO outperforms VUC when the number of ustomers dereases due to larger drop sizes. On the other hand, VUC is more suitable for areas with smaller density of ustomers and shorter line haul distanes as well. Still, as drop size inreases, TOCO would be the best option, if the poliy that restrits the traffi of larger vehiles was not effetive. For referene, atual values for two types of industries are depited in Figure 2a. Figure 2b shows a bigger sale for density values, but does not add muh to the disussion. Following figures will not show this full density sale. Density [ustomers/blok] 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 Drop Size 50 kg 100 kg 300 kg 500 kg 700 kg 1000 kg 0.1 0.05 Beverage Company Groery Retailer Company 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Line haul distane [km] Figure 2a: Indifferene urves: base senario - "vehile's size/weight restrition" poliy Density [ustomers/blok] 5 4.5 4 3.5 3 2.5 2 1.5 Drop Size 10 50 kg 100 kg 300 kg 500 kg 700 kg 1000 kg 1 0.5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Line haul distane [km] Figure 2b: Indifferene urves: base senario - "vehile's size/weight restrition" poliy - 6 -
ILS Conferene 2016, June 1 4, Bordeaux, Frane 5.2 VUC's Capaity Senario "VUC's apaity" senario analyzes the effets of different load apaities for VUCs on vehile's hoie. Eah indifferene urve denotes a speifi apaity with respeted to the adopted average VUC weight apaity of 4.93 t. In this ase, the average drop size is fixed and equal to 300 kg (Figure 3). Load apaity does influene vehile hoie. Smaller apaity vehiles would be preferred for smaller ustomer density, as well as for shorter distanes. As these parameters inrease, the advantage of higher apaity vehiles also inrease. For even larger VUC apaities (higher than 6.8 ton) VUC is always the best hoie, sine its lower fuel onsumption ounterbalanes the additional number of vehiles, resulting in a lower total emission when ompared to TOCO. Thus, vehiles slightly larger than a VUC ould be a good alternative in terms of reduing air pollution and the total number of delivery vehiles on streets. Density [usomers/blok] 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 VUC's Capaity 4 tons 4.93 tons 6 tons 6.5 tons 6.8 tons 0.1 0.05 Beverage Company Groery Retailer Company 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Line haul distane [km] Figure 3: Indifferene urves: VUC's apaity senario 0.5 Density [ustomers/blok] 0.45 0.4 0.35 0.3 0.25 0.2 0.15 VUC's Fuel Consump on 0.1 l/km 0.11 l/km 0.12 l/km 0.131 l/km 0.14 l/km 0.16 l/km 0.1 0.05 0 Beverage Company Groery Retailer Company 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Line haul distane [km] Figure 4: Indifferene urves: VUC's fuel onsumption senario - 7 -
ILS Conferene 2016, June 1 4, Bordeaux, Frane 5.3 VUC's Fuel Consumption Senario "VUC's fuel onsumption" senario examines the effets of different VUC's fuel onsumption values on vehile's hoie (Figure 4) onsidering a fixed value for drop size (300 kg). The range of values for VUC's fuel onsumption is based on standards developed by EPA and the National Highway Traffi Safety Administration (NHTSA) for future vehiles. One an note that as VUC beomes greener (due to dereased fuel onsumption rate), it beomes a better hoie than TOCO for most ombinations. 6 Final Remarks Larger truks (TOCO) have showed to be the best hoie in terms of CO2 emissions in most ases. Given its larger apaity, a smaller number of vehiles an servie more ustomers in a single route and, one VUC and TOCO use the same type of fuel, lower CO2 emissions are expeted from TOCO. Also, by onsidering different VUC apaities and fuel onsumption rates it is possible to onlude that, in order to allow the traffi of slightly larger vehiles than VUC, or to inentive the use of greener large vehiles ould be good alternatives in terms of reduing the air pollution and the total number of vehiles. Thus, this poliy of "Vehile Size/Weight Restrition" does not look appropriate from an emissions standpoint. For future works, we intend to onsider the effet of urrent truk restrition time windows, as well as atual average truk speeds obtained in the night deliveries pilot projet in São Paulo. The effets of parking availability for different truk sizes (VUC s perform better in this ase) will be also evaluated. Aknowledgments The authors thanks Itau Foundation, MISTI-Brazil and CAPES. Referenes 1. United Nations. Department of Eonomi and Soial Affairs/Population Division. World Urbanization Prospets: the 2011 Revision - Highlights. New York: United Nations publiation (2014) 2. CET-SP, Companhia de Engenharia de Tráfego, available at http://www.etsp.om.br (2013) 3. CETESB. Companhia Ambiental de Estado de São Paulo. Governo do Estado de São Paulo: Qualidade do ar no estado de São Paulo 2014, http://www.etesb.sp.gov.br/ar/qualidade-do-ar (2015) 4. Saldiva, P.: Pobres são os mais atingidos pela poluição urbana, available at http://www.artamaior.om.br/templates/materiamostrar.fm?materia_id=20651 (2012) 5. Department of Finane, Government of the State of São Paulo: Registered establishments, available at http://www.nfp.fazenda.sp.gov.br/lista_aprovaao.shtm (2014) 6. Taniguhi, E., Thompson, R.G.: City Logistis: Mapping the Future. Taylor and Franis, Boa Raton (2015) 7. Detran-SP, Departamento Estadual de Trânsito de São Paulo: Estatístias de trânsito: frota de veíulos em São Paulo por tipo, available at http://www.detran.sp.gov.br/ (2015) 8. EPA - U.S. Environmental protetion ageny: EPA Road: Climate hange Greenhouse gas emissions, on the Road, available at http: //www.epa.gov/limatehange/emissions/ind_road.html (2011) 9. GHG Protool Core Module Guidane. Climate leaders. Greenhouse gas inventory protool ore module guidane, http://www.epa.gov/limateleaders/douments/resoures/mobilesoure_guidane.pdf (2008) 10. EPA Environmental Protetion Ageny: Metris for GHG emissions, http://www.epa.gov/autoemissions (2011) 11. NTM Network for Transport Measures: Environmental Dara for International Cargo Transport. Calulation methods and default data. Road Transport Europe, available at www.ntmal.org (2010) 12. Velázques-Martínez, J.C., Fransoo, J.C., Blano, E., Mora-Vargas, J.: The impat of arbon footprinting aggregation on realizing emission redution targets. Flex. Serv. Manuf. J. 26, issue 1-2, 196-220 (2014) 13. Daganzo, C.: The length of tours in zones of different shapes. Transp. Res. B 18, 135-145 (1984) 14. Daganzo, C., Newell, G.C.: Physial Distribution from a Warehouse: Vehile Coverage and Inventory Levels. Transp. Res. B 19, 397-407 (1985) 15. Galvão, L. C., Novaes, A. G., De Cursi, J. S., & Souza, J. C.: A multipliatively-weighted Voronoi diagram approah to logistis distriting. Comput. Oper. Res. 33(1), 93-114 (2006) 16. Novaes, A.G.N., Graiolli, O. D.: Designing multi-vehile delivery tours in a grid-ell format. Eur. J. Oper. Res. 119, 613-634 (1999) 17. Novaes, A.G.N., Cursi, J.E.S., Graiolli, O. D.: A ontinuous approah to the design of physial distribution systems. Comput. Oper. Res. 27, 877-893 (2000) - 8 -