**4. Results and Discussion**

### *4.1. Accident Count over the Monitored Period*

A total of 14 accidents were reported to the monitoring system within the studied period (Table 2). Among these occurrences, 11 involved the transport of hazardous products and three the transport of products potentially harmful to the environment such as vegetable oil, limestone and cement. The largest number of episodes occurred in the morning (from 06:00 to 12:00, 35.71% of occurrences), being followed by the afternoon (from 12:00 to 18:00, 28.57%), night time (from 18:00 to 00:00, 21.43%), and early morning (from 00:00 to 06:00, 14.29%). As regards seasonality, there were nine occurrences in the rainy season (64.29%) and five in the period of low or no rainfall. The results also show that one quarter (28.57%) of all accidents occurred in the section between km 77 and 83 of BR 050. This section is characterized by steep slopes, but paradoxically is also distinguished by fast-speed traffic. Figure 5, generated in Google Earth, shows the elevation profile of km 77–km 83 section, with an elevation of 932 m at km 77 and 810 m at km 83 (2% slope, on average). Ferreira [39] studied the causes of road accidents with hazardous products in the State of São Paulo, Brazil, and observed a greater predominance of accidents in the afternoon (between 12:00 and 18:00) and that the routes between petrochemical poles also influence the number of accidents, due to the greater flow of vehicles transporting dangerous products along these routes. Overall, the results obtained in this study as well as by other authors [2] demonstrate the heterogeneity of factors influencing the incidence of accidents with hazardous products, which hampers the selection of priority sections for hazard management. The alternative path to follow relies on combining the assessment of hazard distribution with environmental

vulnerability assessments, and hence moving from a conventional hazard managemen<sup>t</sup> to the more integrated approach that is risk management. The assemblage of hazard and vulnerability data into a common framework of risk data, especially if using a geographic information system to accommodate and process the maps and associated attribute tables, has the additional virtue to help finding priority sections for management, because risky areas are fewer and smaller than the sum of hazard and vulnerable areas.

**Figure 5.** Elevation profile of the BR 050 highway segmen<sup>t</sup> involved in a large number of road accidents. This segmen<sup>t</sup> is located between km 77 and km 83 of the highway, as illustrated in Figure 1. The elevation profile was generated using the Google Earth software. The accidents are mostly caused by fast-speed traffic in a relatively steep-slope road.

### *4.2. Vulnerability, Hazard and Risk*

The 200 m-buffers around the 14 accidents sum a hazard area of approximately 168.27 hectares. The buffers were defined because large areas around the accident sites can be affected by the road spills, even extending to the entire watershed. Table 3 summarizes the results obtained for vulnerability, hazard and risk in the four predefined scenarios of factor maximization, respectively expressed as vulnerable areas within the BR 050 segmen<sup>t</sup> watersheds (*V*), hazard areas within the 200 m buffers surrounding the accident sites (*H*) and the percent ratio between *H* and *V* (*R* = *H*(%)/*V*(%)). The *V* areas were evaluated within the maps of Figure 4a–d, while the *H* areas were measured within the buffer areas (labeled circles) represented in the same figures. For example, for vulnerability level "strongly vulnerable" (red color in Figure 4) the *V* area is 4337.79 hectares (3.4% of road segmen<sup>t</sup> area; Table 3) in the ground slope scenario (Figure 4a), while raises to 40,073.87 hectares (31.4%) in the geology scenario (Figure 4c). The corresponding *H* areas are 11.50 hectares (6.83%) and 48.75 hectares (28.97%). According to Equation (2) this gives a risk *R* = 2.01 for the ground slope scenario and *R* = 0.92 in the geology scenario. In case ground slope is adopted as reference scenario for decision making on soil and water protection, then road accident sites located where the environment is strongly vulnerable are considered risky because *R* > 1. In general, as regards vulnerability the areas were mostly classified as weakly vulnerable or vulnerable, for the ground slope and drainage density factors, and as vulnerable or strongly vulnerable for the soil class and geology factors. The coverage by extremely vulnerable areas or invulnerable areas was insignificant. The results obtained for the areas where the accidents have occurred (hazards) follow the general results obtained for vulnerability, because the percentage of area ascribed to the vulnerability classes are similar in both cases. The exceptions occur for the scenarios where ground slope or drainage density factors were maximized, because in some cases the areas where the accidents have occurred are more vulnerable than the general vulnerability areas in those scenarios. As mentioned above, for the scenario that maximized ground slope the areas classified as strongly vulnerable along the highway watersheds represent *V* = 3.4% of the total watershed area while the homologous areas around the accident sites represent *H* = 6.83%. The same holds for the strongly vulnerable areas in the scenario that maximized the drainage density factor, which rise from *V* = 4.06% to *H* = 6.83%. Put another way, the strongly vulnerable areas in these two scenarios can be classified as risky, because *R* = *H*/*V* > 1 in both cases, namely 2 and 1.7 (Equation (1)). In that context, these vulnerability levels and corresponding areas of influence would deserve special attention in risk managemen<sup>t</sup> plans.

**Table 3.** Vulnerability assessments within the watersheds that surround the studied segmen<sup>t</sup> of BR 050 highway (*V*), considering the four scenarios. Vulnerability assessments within the 200 m buffers that surround the 14 road accidents (also termed hazard assessments; *H*), considering the same scenarios. Risk assessments (*R = H/V*, in percent ratio).


Figure 4a–d display the vulnerability maps obtained at the 14 sites where the accidents occurred during the studied period. The maps also represent the surrounding watersheds, because they can also be environmentally affected. In all cases, these figures provide visual information for rapid environmental risk assessment of accident sites, enabling immediate prevention and alerts for the sites classified as vulnerable or strongly vulnerable. The representation of accident sites in a color scale related to environmental vulnerability validates the method of [15] as support for an expeditious risk managemen<sup>t</sup> tool, and hence represents the achievement of a proposed objective. In that context, it is important to note the large number of accident sites located in strongly vulnerable areas, especially in the southern sector watersheds and when the focus of vulnerability is put on the catchments' soil and geologic characteristics. An environmental alert is due in these cases to ensure the safety of soil and water quality within the involved watersheds.
