Network Analysis: Sand Transportation Impact on the Roads.

Introduction

This exercise takes part of a bigger project analyzing the impact of sand mining in Wisconsin, in this case, dealing specifically with the road deterioration. At the moment of the county roads construction, it was not predicted a growing mining market, so they don’t necessarily have the appropriate infra-structure to support the sand transportation. Trucks loaded with sand are extremely heavy and can compromise the quality of the roads. The counties need to be aware of the cost related so they can apply the appropriate taxes to the sand mines, which will be used to road management. Therefore, the goal is to provide the cost per year for each county, related to the sand transportation.

The use of Geographic Information Systems is essential to solve this problem. Thanks to the previous exercises, the sand mines are already geocoded, consisting in a point feature class. Railroad terminals are available and are the destination for the trucks. Network Analysis provides the most likely routes taken between these features and then it’s possible to estimate the cost by county.

Methodology

Firstly, the geocoded mines need to be merged to the existing mines on the Trempealeau geodatabase. With the data prepared, it’s necessary to run the network analysis tool. There are some different options of network analysis, such as: New Route; New Service Area; New Closest Facility; New OD Cost Matrix; New Vehicle Routing Problem; and New Location-Allocation. Since first a railroad terminal needs to be assigned to each mine, and then the routes need to be applied; the New Closet Facility is the appropriate tool to be used (Figure 1).

The distance from each route can be applied to a formula to estimate the impact cost, however, the results of this tool are in time units. The feature can then be exported as a feature class, which will contain a field with the shape length. It’s extremely important to use the object-oriented model in this case, to guarantee the reliability on this field. A projected coordinate system needs to be applied to this feature class; otherwise, the shape length will be in decimal degrees – meaningless for this project.

The routes need to be divided by county so the appropriate government can deal with the road management. A summarized inside join is not appropriate because it would maintain the shape of the original route, even if it’s over the county boundary. The routes need to be clipped, but the use of the clip tool repeatedly is not a convenient way to do this. The identity tool is appropriate for this case, since it has the same idea as the clip, but works for more than one county. Multiple routes will be resulted for each county, reason why it’s then necessary to run the summarize tool. A similar result could be obtained by using the intersection tool. However, some of the routes are located outside Wisconsin and would be disregarded by this tool.

Figure 1 – Data Model for Network Analysis

The determination of the cost in dollars is going to be made using a hypothetical calculation, since the math involved in this estimative is complex and it’s not the focus of this project by now. An estimative of 50 truck trips per year and a cost of U$0.022 per mile is applied, both for loaded truck – in direction to the railroad terminal, as well as unloaded truck, in the way back. Therefore, the formula used is the distance of all the routes, per county, in miles, multiplied by 2.2 (50 truck trips * 2 ways * U$0.022).

However, since the coordinate system has meter units, first it’s necessary to create a need field and use the calculate geometry tool to provide its distance in miles. Only then the table can be exported as a .dbf file to be manipulated on Microsoft Excel in the production of graphs. Joining the cost stand-alone table to the spatial feature class for the counties is interesting to show the spatial distribution of the cost. It supports the analysis of why some counties have higher cost than others and its relation to the amount of mines and railroad terminals.

In this analysis, the use of statistical techniques such as correlation with Pearson’s I is used to test the hypothesis of a relation between the variables from the project. A null hypothesis will always state the lack of relation between the given variables, while the alternative affirms the presence of a linear relationship.

Results

Intuitively, it was expected to find higher costs in counties containing a higher number of mines. However, a primary analysis already showed that was not the case. Trempealeau is the county containing more mines in the geodatabase created – although this is due the better quality of the source, and not necessarily because the county does contain the most mines in the real-world. It was expected to find the higher cost in this county, but by looking at the resulting table (Figure 2), La Crosse appears to be the highest cost.

Figure 2 – Cost, mines and railroad terminals by county.

A graph can potentiate the examination (Figure 3), where it’s noticed that although it’s not in the first place, Trempealeau does have a high cost as well. The counties with the highest costs are La Crosse, Trempealeau, Eau Claire and Chippewa. All the other counties’ cost is lower than U$1,000 per year.

Figure 3 – Sand Transportation Impact per County

Comparing these four main counties to the amount of mines, only Trempealeau has a significant number of mines (64). La Crosse – where the highest cost is – contains only one mine; while Eau Claire and Chippewa contains 3 and 15 respectively. A map showing the cost distribution overlaid by the routes, mines and railroads terminals can elucidate the reasons for this pattern (Figure 4).

Figure 4 – Sand Mine Flow in Wisconsin and Related Cost

Although the low amount of mines in La Crosse, it contain the only railroad terminal in the within all its neighbor counties. Since Trempealeau county is at north of La Crosse, all the numerous mines located at the south of the Trempealeau will be directed to the railroad terminal in La Crosse. This only railroad terminal attends three counties (Buffalo, Trempealeau and Monroe).

The same pattern explains the high cost for Eau Claire and Chippewa: Eau Claire railroad terminal is located close to the boundary with Chippewa. Therefore, this terminal attends 10 counties besides Eau Claire and Chippewa (Burnett, Washburn, Barron, St. Croix, Dunn, Pierce, Pepin, Jackson, Trempealeau and Buffalo). Then, the counties’ roads in the way of a route from a distant mine to get to this terminal will also be compromised.

Then, it’s necessary to have a perspective of this issue not only focused on the amount of mines, but also in the presence of railroad terminals. However, this doesn’t mean that the pure existence of railroad terminal will consist in a high cost: that will only happen when there are mines around that are not served by other railroad terminals. There’s no linear relationship between railroad terminals and cost without including other variables, which can be noticed by the Pearson’s I results (Figure 5). A significance level of 0.71 fails to reject the null hypothesis, which would affirm the existence of an association between the variables.

Figure 5 – Correlation results

In the other hand, the number of mines is, indeed, statistically related to the final cost. A scatter-plot graph can illustrate better this relation (Figure 6).  As imagined, the higher the number of mines, the higher the cost. However, although the significance level of 0.00 allows the rejection of the null hypothesis with a confidence interval of 99%, the pearson coefficient of 0.56 determines only a moderate relation. Therefore, the spatial element is essential to find the inter-relations between all these variables together, instead of dealing with them isolated.

Figure 6 – Positive trend line between cost and number of mines.

Conclusion

The main goal of this project is to be able to determine how the counties should apply their taxes, which will be directed to the road management. However, if some counties that don’t even contain a mine are having their roads affected, like Washburn, the application of taxes to the county mines gets compromised. Even though the railroad terminals centralize the routes to their county, it wouldn’t make sense to charge them, since the interest of using the roads is actually in the mines.

The main problem in this scenario is the lack of railroad terminal in western Wisconsin, which makes really distant mines have to count on a few locations in La Crosse or Eau Claire. The counties should encourage, thus, encourage the creation of railroads in their own county, instead of depending on a far location. However, this is expansive and might not be economically worth, not being exactly interesting for them.

Therefore, it’s suggested that this issue don’t be restricted by the county boundaries. A state level management can be helpful to deal with the relations between the counties containing railroad terminal and the ones who doesn’t, but do have lots of mines. It would be interesting if the collection of taxes could be done accordingly to all the degradation each mine is responsible for. Then, the cost could be distributed by the counties where the routes fall on to.

However, this issue involves a complex inter-relation of agreements and laws that cannot necessarily fit the ideal scenario. The registration of mines being recently done by the government can later provide a more accurate and complete set of data, similar to the one found for Trempealeau. This would allow a more realistic answer to this issue.

Also, alternative transportation methods might be used, beyond railroads. In the same perspective, the commercial network needs to be considered: if the sand mines are providing sand to a facility in their own counties, or closer than the railroad terminal, there wouldn’t be the need to use those routes. Therefore, this project has important findings in the elements that need to be taken in consideration and encourages more study on this subject.