The applied transport model is based on a multimodal and multi-hour assignment, developed with the software VISUM [25].

The transport “offer” (network model) has been built using a graph that includes road network (based on Navteq commercial graph), with information on car, motorbike and pedestrian trips, together with bicycle lanes and LPT transit lanes. To this, the railway and the subway have been added as incremental networks.

In order to build the transport demand, there could be several pathways: for this study, the results of a survey on a sample of around 15% of the urban population (200,000 people), performed by the Municipality of Milan transport office (AMAT - []), has been used. This has allowed to build the Origin/Destination (O/D) matrix for an average working day, disaggregated over 24 hours: it indicates the number of estimated movements of people for each hour and each O/D pair.

Five main transport modes have been considered, according to what emerged from the results of the surveys: car, motorbike, bike, pedestrian, local public transport (bus, trolleybus, tram, subway and railway – LPT). Moreover, a sixth mode, called mixed, that involves the combination of car and LPT in the same trip, has been added.

Once demand and offer have been defined, the modal split (rate of choice for each transport mode) was calculated and assigned to the multimodal transport network in VISUM. The modal split falls within a multinomial Logit model that estimates the use of each transport mode based on “comfort” parameters, mostly related to time (travel time, parking searching time, pedestrian time) and trip costs (cost per passenger-km of private car [26], ticket fares [27], parking fare, traffic pollution charge).

The results are then used as input to assess the energy and environmental performance of the baseline case (and, in a further step, the sustainable mobility scenarios), with the software COPERT (the EU standard vehicle emissions calculator for emissions and energy consumption for a specific country or region [28]). The circulating vehicles are as per the national inventory by ISPRA [29] adapted to Milan Metropolitan Area (MMA).

In a previous study on sustainable mobility in MMA [30], the same transport model was used to simulate several sustainable mobility measures, as scenarios evolving from a baseline case. For this study, by using the same baseline case, further scenarios were analysed, where Ride Sharing was considered in parallel with the other mobility solutions in order to evaluate its applicability in MMA.

As mentioned in the previous paragraph, energy and environmental performances are calculated through the use of the software COPERT.

In order to evaluate the economic performances, the main source are the specific costs reported in the 2016 Sustainable Urban Mobility Plan (SUMP) of the municipality of Milan [31]. The main financial difference, compared to the baseline case, is related to both investment costs and variations in revenues and operating costs for public transport, parking areas and restricted traffic zones, all affecting the Municipality of Milan. With a standard formula, an annual cost (AC) of the Ride Sharing has been defined:

$\mathrm{A}\mathrm{C}=\mathrm{I}\frac{\mathrm{i}}{1-\frac{1}{{\left(1+\mathrm{i}\right)}^{\mathrm{n}}}}+∆\mathrm{O}\mathrm{C}\mathrm{ }$ (1),

where:

• $\mathrm{I}$ is the investment required to implement the mobility measure;

• $\mathrm{i}$ is the discount rate (assumed at 3%);

• $\mathrm{n}$ is the duration of the loan (assumed at 30 years);

• $∆\mathrm{O}\mathrm{C}$ is the variation of the operating cost (negative values mean a net income for the municipality of Milan).

Since sustainable mobility measures (RS in this case) generate also non-financial positive impacts, that benefit society in terms of environmental and energy efficiency impacts and reduction of traffic congestion, a Net Annualized Cost was defined, that rewards the maximization of these types of benefits. The Net Annualized Cost is then normalized on the number of passengers daily traveling in the MMA, in order to obtain the per capita cost for RS:

Net Annual Cost per Passenger=AC-∆Energy-∆CO2-∆PM2,5-∆CongestionNumber of passengers (2)

$∆\mathrm{E}\mathrm{n}\mathrm{e}\mathrm{r}\mathrm{g}\mathrm{y}$ , $∆\mathrm{C}{\mathrm{O}}_2$ , $∆\mathrm{P}{\mathrm{M}}_{\mathrm{2,5}}$ , $∆\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{g}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}$ are calculated as the monetisation of energy savings (with reference to Italian electricity and fuel prices - that can be found in [32], [33] and [34]), CO₂ and PM_2,5 emissions avoided and congestion reduction (with reference to [35] and [36]).

The main results of the baseline case simulation are shown in Figure 2 and Table 1. Private car is the main transport mode (59% of the trips), followed by LPT (24%) and pedestrian (11%). The other transport modes account for less than 5%. In terms of travelled distance, most of it (65.2%) is attributed to individual motorized transport (car and motorbikes), while LPT accounts for 30.1%. Less than 5% is performed through non-motorized transport. In terms of energy consumption, most of the 719 toe/y are due to fossil fuels (647 toe/y). Through the software COPERT, the main pollutants have been quantified in 1,781 kt/y of CO₂ and 738 t/y of PM_2.5.

Figure 2

Modal split in the Milan Metropolitan Area baseline case

The mixed mode doesn’t appear in the table, since all the travelled distances have been divided in the modes composing it.

The RS model for the MMA has been built as one of the evolutions of the baseline case model. Two main parameters have been inserted:

• Number of RS subscribers: the number of subscribers has been assumed equal to 20% of the population travelling in the city everyday (760,000 subscribers out of 3.8 million travellers), of which 50% are riders. 20% has been assumed considering also the high penetration that current car sharing services have in the city (around 30% of the residents are subscribed to at least one car sharing service, according to [21]).

• Review of the transport model: RS has been added to the previously considered transport modes (car, motorbike, bike, pedestrian, LPT, mixed). Two cases have been analysed, one with RS as parallel to the other transport modes, and one as RS as a sub-mode of car, in parallel with private car. The two options are shown in Figure 3 and Figure 4.

Figure 3

Logit model with RS in parallel to other transport modes

Figure 4

Logit model with RS as sub-mode of car

The final adopted model has been the one reported in Figure 4, where the parameters for “old” transportation modes haven’t been changed with respect to the baseline case, while for RS the occupancy rate of the car has been put equal to 2, the unavailability of the car equal to the complementary to the number of RS subscribers, and, for riders, pedestrian time has been reduced and parking time has been set to 0.

In the next paragraphs, results are presented and the impact of RS is assessed in terms of mobility demand, energy and environmental performances and economics, in comparison with the above described baseline case.

Mobility demand, with the final Logit model, gives the result that private car (Priv) is still the most used mode, however 9% of rides is covered by ride sharing, as shown in Figure 5. The 9% of share in RS is mostly related to a reduction in private cars use, while LPT and motorbike lose 1% of share each with respect to the baseline case.

Figure 5

Modal split in the ride sharing case

Considering the split during the day, in Figure 6 it can be noticed that private car is largely used at night, while LPT increases at peak time hours. Ride sharing is used at the same rate at almost all times of the day, with a small reduction at peak time, when LPT service is available at a higher frequency. In general, it follows the trend of the private car.

Figure 6

Modal split during the day

Considering the travelled distances, the main results are reported in Table 2. As already mentioned for the baseline case, mixed mode is here split into the different modes included in it. Moreover, RS and private car are summed-up as “car”.

The changes in passengers transport and traffic flow with regards to the base case are reported in Figure 7 and Figure 8. Ride sharing and private car have been summed up in the “car” mode, since they use the same vehicle. It can be noticed that there is a general increase in the number of passengers using cars, with the same reduction in pedestrians, bike and LPT. The reduction is registered by motorbikes (-18.1%), that appear as more expensive than ride sharing.

Figure 7

Modal split comparison in passenger-km

The increase in number of passengers travelling by car, that can seem contradictory with respect to the goal of RS adoption (that should be a reduction in greenhouse gas emissions) shall be however seen in the overall picture: in terms of number of vehicles, actually, there is a reduction of circulating cars of around 1.1 million vehicle-km, thanks to the increase of the average occupancy rate, for each car, from 1.2 to 1.3. Other positive impacts can be seen in Table 3and Table 4, regarding energy efficiency and environmental pollution.

Figure 8

Modal split comparison in vehicle-km

In terms of energy consumption the results are shown in Table 3 and Table 4.

Analysing the data, there is a slight increase in the consumption of LPT (subway and tram), while fossil fuelled cars overall reduce their consumption of 6% and motorbike of 18%, in line with the results of vehicle-km comparison. Considering fuels, there is a total reduction of 5%, with 6% reduction of fossil fuels use (36 ktoe/y) and a slight increase of electricity use (related to LPT), however lower than 1%.

From an environmental point of view, different pollutants were taken into account, with a focus on the change in emissions of PM₁₀ (coarse particulate matter), PM_2.5 (fine particulate matter), PM_Exhaust (particulate matter at exhaust discharge), EC (elemental carbon), OM (organic aerosol), NO (nitrogen oxide), NO₂ (nitrogen dioxide), NMVOC (Non Methane Organic Volatile Compound), NH₃ (ammonia), CO₂ (carbon dioxide) and in FC (fuel consumption). The comparison between the results for baseline case and ride sharing is reported in Table 5.

The fossil fuels consumption reduction of about 6% is reflected by the lower environmental impact of RS case, in which pollutant emissions are reduced in a range from 4% to 8%.

Costs of implementation of RS have been divided in two categories:

• Investment costs, in order to build the online tools needed to support RS adoption;

• Differential operation and maintenance costs (ΔOC), with regards to the baseline cases.

• Lower income from LPT;

• Positive income from RS management;

• Positive income from parking payment;

• Positive income from the fares of restricted traffic areas (RTA).

In particular, operation and maintenance costs for Milan public administration include:

Costs data have been derived from different sources: the already mentioned investigation [2], Milan SUMP [31], public documents, such as newspapers and magazine articles [37] and [38], and Milan LPT company financial statement [39]. The results are reported in Table 6.

In order to evaluate the long-term financial impact of the adoption of RS, it has been assumed a loan to be refunded in 30 years, with an interest rate of 3%, as illustrated in the Methodology.

The financial results are:

• Yearly instalment: 0.35 M€/y;

• O and M costs: 37.8 M€/y;

• Total costs: 38.15 M€/y.

From the above results, compared also with previous works on RS, it is possible to draw the following recommendations, for a policy maker that would like to implement RS in large cities.

• Safety and privacy:

1. both riders and drivers shall be registered to the app through a verified account;

2. at the start and the end of the ride, the provider shall record the identity of each person on-board the car, through the mobile phone data;

3. at the end of the ride, riders and driver shall give a score to the trip companions;

4. before the ride, rider and driver shall receive each other’s picture and score, in order to recognize each other and evaluate whether to accept or not the match;

5. there shall be the possibility to set some filters (e.g. women can choose to exclude men riders or drivers);

6. the app shall include an emergency button that is directly connected to the police.

• Economics:

1. the cost of the service shall be assumed equal to a standard recognized fee/km, e.g. 0.25€/km ([26]), divided by the number of passengers, in order not to generate a source of income for riders;

2. costs for riders might include access to RTAs or parking payments, depending on the ride;

3. there shall not be cash exchange on-board the vehicle, everything shall be managed through the app that, at the end of each month, calculates the due amount for the riders and the revenue for the drivers, that is the total amount of the performed rides, cut of a % related to the commission for the provider (e.g. 10%).

• Ride sharing process:

1. riders and drivers insert in the app their origin, destination, departure time, number of passengers and time tolerance. The app shall match them automatically and, once the ride is approved on both sides, rider and driver shall meet in the agreed place at the agreed time;

2. the time to agree or disagree on a match shall be limited to maximum few minutes;

3. it shall possible to refuse, for the same trip request, two matches; if a third is refused, the request is automatically deleted;

4. it shall not be possible to modify the origin and destination once the ride has been approved, unless by contacting the provider.

5. for safety reasons, meeting point shall be in places that guarantee the rider to get-in and get-out easily, such as parking spots or public transport stops.