The reported crashes were classified into three groups based upon damage type as property damage, personal injury, and fatality. Among the 189 available records of crashes in the DelDOT database, the classification is as follows, 141 (74.60%) property damage, 45 (21.81%) personal injury, and 3 (1.59%) fatal crashes. In terms of the number of people involved, 53 people were injured and 4 people died because of FVCs. By adding the 31 records of crashes from the State Police Annual Reports, the percentage of fatality and personal injury cases will increase to 1.81% and 24%, respectively, which is presented in a pie chart in Figure 1. For comparison purposes, a similar analysis was performed for all crashes in Delaware (including farm and non-farm vehicles) from 2005 to 2019 based upon Delaware’s Annual Traffic Statistical Report [16] which is presented in Figure 2. It can be seen that the percentage of personal injury crashes for FVCs is very similar to the overall state average; however, the percentage of fatal crashes is 3.7 times higher than for all crashes. Several studies reported a fatality percentage close to 2% for FVCs. The percentage of personal injury crashes in Delaware seems to be slightly lower than those reported in Maryland, Iowa, New York, and North Carolina [6] [11] [14] [17].

Figure 3 shows the location of FVCs on a map of the State of Delaware with individual county designations. It can be seen that, Sussex County is highest with 53%, followed by Kent County with 27% and then New Castle County with 20%. The share of each county in FVCs can be partially explained by the number of farm vehicles operating within the county. Figure 4 shows the share of FVCs along with the number of tractors in each county based on the 2017 Census of Agriculture [18]. This comparison shows that the percentage of FVCs for a given county is somewhat proportionate to the number of tractors; however, the share of New Castle county FVCs is slightly higher than its share in the number of tractors which might be attributed to the fact that this county is the most densely populated.

According to the 2010 Census [19], an urbanized area is an area with a population of more than 50,000, and an urban cluster is an area with a population between 2500 and 50,000. Further, a rural area is defined as an area that doesn’t fit the definition of an urbanized area or urban cluster. Figure 5(a) shows the percentage of FVCs in Delaware by urbanization boundary, and Figure 5(b) shows the number of crashes in each boundary zone concerning their severity. The result shows that urban crashes account for about one-third (30%) of all FVCs. This finding is consistent with the findings of [10] which collected the FVCs data from nine Midwestern states. The percentage of fatal and injury crashes for rural areas were 23.5% and 1.5%, and for urban areas 24.5% and 1.7%, respectively. The fact that the percentages for rural and urban crashes were very similar was surprising since rural crashes tend to be more severe than urban crashes. According to NHTSA Traffic Safety Facts [20] in 2017, the fatality rate per 100 million vehicle miles traveled at the national level was 2.1 times higher in rural areas than in urban areas and for Delaware, in particular, the fatality rate was 3.3 times higher in rural areas. One possible explanation for that similarity might be the fact that even FVCs that occurred within urban boundaries take place largely in less populated suburbs that have similar traffic characteristics to rural areas.

The location of the urban and rural FVCs is shown on the state map in Figure 6. Urban boundaries are highlighted on the map by a light pink color. Among urban boundaries, Laurel and Middletown had highest number of crashes; however, as both areas are magnified in Figure 6, it can be seen that the density of the crashes at the vicinity of Laurel is higher.

In terms of time-dependent variables, the analysis includes the year, month, day of the week, and time of the day. The number of the FVCs in each year is presented in Figure 7 based on both the DelDOT database (May 2005 to August 2020) and the State Police Annual Reports (2005 to 2019). It is shown that some

crashes are missing from the DelDOT database, and when considering the higher reported numbers, an average of 14 FVCs occurred each year. Although the number of crashes through this period fluctuates substantially absent any notable pattern, a significant upward trend can be seen from 2016 to 2019; where, the number of crashes almost doubled. Moreover, two of the fatality cases happened within the past two years.

The frequency of FVCs by month is presented in Figure 8. This chart shows that the number of crashes was lower during winter and spring, and higher during the summer and fall which would be considered as the planting, growing, and harvest seasons for farmers. March and June had the lowest and highest number of FVCs, respectively. Although some studies from other states showed that the number of FVCs was higher during the harvest season, the Delaware peak was experienced during the growing and planting seasons (April to August). The same trend was reported by Hughes and Rodgman [13] in North Carolina. Surprisingly, despite the significant change in the total number of FVCs, the number of injury crashes follows a much flatter trend line and stays close to the average of 4 crashes per month.

Figure 9 shows the distribution of FVCs over days of the week. As shown, most crashes happen during the week with fewer on the weekend, especially Sunday. This can be expected since many farmers take Sunday off and the general traffic volume on the roads is less during the weekends. Crashes are distributed evenly during the weekdays and no meaningful differences were observed.

Figure 10 shows the number of FVCs during different times of the day. The number of crashes increases as the farming day starts around 5:00 AM and reaches its peak at noon, and declines from there and reaching its minimum around midnight. In terms of the severity of the crashes, it can be seen that the peak of personal injury crashes doesn’t exactly follow the peak of all crashes and is shifted more towards the darker hours of the day. The median hour for all crashes was noon, while the median for personal injury crashes occurred at 3:00 PM.

Figure 11 represents the number of crashes by the lighting conditions. It appears that the majority of the FVCs (73.5%) occurred during daylight. This result complies with the studies from other states and in particular the Maryland study [17]. While the number of crashes during dark (not-lighted) conditions was only 17% of all crashes, the injury percentage for this category was nearly 47% which is much higher than the 19% injury percentage for the daylight category. Moreover, two of the three fatal cases happened under dark (not-lighted) conditions and the lighting for the remaining fatal case was unknown.

Figure 12 and Figure 13 show FVCs by weather and road surface condition. Nearly 80% of FVCs occurred under clear weather and 14% under cloudy weather. Snow weather had the highest percentage of injury crashes with 47%, following by cloudy weather (33%) and clear weather (19%). About 90% of crashes took place on a dry surface. This finding is consistent with the finding of other studies and correlates with the fact that farming tasks are most likely to be accomplished during clear weather.

Figure 14(a) displays the map of Delaware roads based on their functional class defined by DelDOT [21] and Figure 14(b) shows the percentage of each category in the state’s total miles of roads. Unfortunately, the functional class of nearly 30% of the crashes was unknown. Among the known cases, major collectors and local roads account for the majority of the FVCs with 29% and 21%, respectively, as presented in Figure 15. Considering the low percentage of major collectors of total state’s miles of road (7%), the risk of FVCs in this classification is nearly 14 times that of the local roads. Freeways have the lowest percentage of crashes since they not only comprise 1% of the total state roads but also slow-moving farm vehicles are not allowed on most high-speed high-volume roads.

Local roads, major and minor collectors were the only categories with fatalities. Among them, minor collectors had an injury percentage of 27%, following by local roads and major collectors with 23% and 15%, respectively. Figure 16 also shows that the majority of crashes occurred on non-junction segments (76%) and nearly 16% either happened at intersections or were intersection-related. Gkritza et al. [11] also reported that the majority of FVCs occurred on non-intersection segments of the road in Iowa. However, it should be noted that multiple factors can impact the frequency and severity of crashes in each functional class, such as the speed limit and traffic exposure.

Figure 17 shows the number of FVCs in Delaware by their manner of impact along with the number of fatal and injury crashes, and Figure 18 shows the percentage of each category in a pie chart. By combining the same-direction and opposite-direction sideswipe collisions, four major categories can be seen, each accounting for nearly a quarter of the pie. The most frequent type was front to rear (rear-end) collision which comprised 24% of all FVCs, followed by angle, single-vehicle, and sideswipe same direction crashes which account for 22%, 19%, and 17% of all crashes, respectively. Sideswipe collisions combined account for 27% of all crashes. The single-vehicle crashes category includes collision with any other objects (e.g., fixed objects and animals) and accidents such as rollover with only the farm vehicle being involved in the impact; however, other vehicles could contribute indirectly to the crash. Among single-vehicle crashes, rollover only accounted for 4 crashes (2.12%); however, half of them resulted in personal injuries. Head-on collisions were also few in numbers but had the highest percentage of injury crashes (50%), followed by rear-end collisions (40%), single-vehicle crashes (25%), and angle crashes (20%). The combination of sideswipe collisions had an injury percentage of 12%; however, surprisingly the injury percentage for the same-direction category was three times higher than the opposite-direction.

In order to assess the possible relationship between lighting conditions and different types of crashes, the manner of collision is analyzed based upon the hour of the day and lighting conditions. Figure 19 depicts the distribution of different manners of collisions throughout the day. To avoid crowding the image, only six of the most important categories are shown here. Furthermore, since similar analysis was conducted by Glascock et al. [15], the hour spans are set according for comparison purposes. The result shows that different crash types can follow different trends throughout the day. The peak of angle crashes occurred between 9:00 AM to 12:00 PM, and after that reduced through the rest of the day. Rear-end collision seems to be the only category that shows a similar trend to the one observed by Glascock et al., which is increased in numbers after 12:00 PM and experiences its peak during the darker hours of the day.

Figure 20 shows the percentage of each crash type for different lighting conditions. Further, this figure shows that all of the head-on collisions occurred in conditions other than daylight. Crashes during daylight conditions belong to the most diverse variety of crash types, while the rear-end crashes have dominated

the category of the dark (not-lighted) condition. Rear-end crashes comprise about 65% of crashes under dark conditions while they only account for 15.6% of crashes under daylight. The overrepresented share of rear-end crashes under dark conditions implies that poor lighting and marking of farm vehicles may be a contributing factor, and improvements in that area can potentially reduce the risk of rear-end crashes.