Modeling the CNW in Milwaukee in 1957

Computer Fonts that Model Typefaces for Railroad Waybills

Andy Laurent was looking at some of the model waybills I've produced (see, for example https://cnwmodeling1957.typepad.com/cnwmodeling/2018/10/model-waybills-for-frozen-fruits-vegetables-and-foods.html) and he wrote:

"The variety of typefaces for the input data is especially engaging! Which fonts are you using?" and that's what this post is about.  

 

First of all, I composed these waybills in Photoshop on a MacBook Pro running macOS Mojave, so I may be of limited help to those of you running Windows-based systems.  And I'm not an expert on fonts or installing them; I don't know much about True-type or any of the other font descriptors.  But I have been collecting fonts that look like they may be useful for modeling typewriter typefaces.  

As I was working with Andy on a series of posts about the Ahnapee & Western waybill collection I was surprised at the variety of typefaces that were used.  I started looking for and gathering fonts that had the appearance of typewriter typefaces.  Many of these fonts can still be found here:

https://www.dafont.com/theme.php?cat=113

and some of them are even free!

 

I've prepared example Word and pdf files that illustrate the appearance of the fonts I've used, and provided their names:

Download Typewriter-like Fonts (Word - note that this might not display right if all of the fonts are not on your computer)

Download Typewriter-like Fonts (pdf)

 

I hope those of you that are modeling waybills find some of these computer fonts useful as you try to reproduce certain typewriter typefaces.

 

Charles Hostetler

Washington, Ill

 

"deliver something other than livestock"

I've been following a thread on the steam era freight car list that is discussing some of the uses that stock cars were put to other than delivering livestock. I was enjoying the accounts, and since I had all of the relevant data transcribed, I thought I'd put together a quick table to see what the ICC data say about the subject.  

The data in this post are from the ICC's 1% Carload Waybill Survey.  They represent carloads in the sample, and the sample was one out of every 100 waybills terminating on Class 1 railroads.  In this case, the year is 1957.  I made two files, one for those with Excel that like to search, filter, and sort, and one that is a pdf for those that don't use Excel.  They can be downloaded here:

Download Stock Cars and Reefers by Commodity Class 1957 OP  (the pdf version)

Download Stock Cars and Reefers by Commodity Class 1957 OP  (the Excel version)

The first column is labeled CCN with stands for Commodity Class Number, and this is a standard code that was assigned by the ICC.  The second column is a short description of the Commodity Class, and in most cases is self-explanatory.  The abbreviation NOS is used a lot, and it means not otherwise specified.  So for example, 303 is Anthracite coal to breakers and washeries, while 301 is Anthracite coal NOS (i.e., NOT to breakers and washeries, but rather to coal dealers and other consignees.  The third column is the number of carloads of that commodity class that were carried in stock cars in 1957.  And just for grins and giggles, I extended the concept a bit and the fourth column is the number of carload of that commodity that were carried in reefers in 1957.  So you can scroll through this, or if you are using Excel you can sort or filter.  I've also included a few descriptive statistics at the bottom of the file.  

Here are some of my observations:

1. There were almost 4 times as many carloads of reefers unloaded in 1957 than carloads of stock cars.
2. About 1 in every 100 carloads unloaded from stock cars were not livestock.  This was spread out over 19 commodity classes (out of 261 total).  In 1957, it appears that non-livestock loads in stock cars were not common but that they were not exceedingly rare.  
3. I used a pretty generous interpretation of the term perishable as I was considering the refrigerator cars.  If a commodity class could conceivably use any kind of protective service I marked it as a perishable load.  All in all, at least 1 out of every 25 reefers unloaded in 1957 contained commodities that definitely did not need protective service.  Some interesting examples include wheat flour, coffee, peanuts, hides (!!!), asphalt, salt, refined petroleum products(????), aluminum bars, ingots, and slabs, and a lot of paper and printed material.  
4. Look toward the end at commodity class 787, containers returned empty.  These may be crates of beer bottles back-hauled from distributors following a beer delivery.  

I did glance at the data from the earlier 1950s and can say that the use of stock cars for non-livestock shipments and the use of reefers for non-perishables appears to have been even more widespread earlier in the decade.  

 

Charles Hostetler

Washington Ill.  

 

Layout Progress - Different Ballasting Styles

The ballasting I did that is described in the previous post was for an older yard, ballasted with cinders.  Following this effort I decided to work on the Bayview area a bit.  The trackage here is different, being a mix of main and branch CNW lines, as well as a MILW interchange track.  This is the view just north of St. Francis, looking toward Bayview (click to enlarge):

From the backdrop to the aisle, the track are:

1. the CNW westbound main leading toward downtown Milwaukee,
2. the CNW eastbound main leading south toward Chicago (together these two comprise the "Old Line" Subdivision, and remember that the CNW runs on the left-hand side),
3. the CNW branch track (labeled track 4 on track charts) that is used as a transfer track and switching lead for the Jones Island, Updike Yard, Water Street, Fifth Ward, and Third Ward switching areas), and
4. the MILW Jones Island transfer lead (note how it cuts across the three CNW tracks - this was a big controversy in the early 1950s that I may find time to write about later).  The cut of cars just beyond the ballasting tools are on the MILW interchange track leading onto the Jones Island peninsula.

The CNW main lines are raised and well drained and were re-ballasted in 1948, about 9 years before 1957.  I modeled them with code 83 rail on Central Valley mainline ties laid on Woodland Scenics ballast profile pads.  The CNW branch line is an older track that was laid in 1938.  Although there is a short stretch of code 83 track near the MILW crossing in the background, and the track is raised to the same height in this area, most of it is code 70 rail on Central Valley branch line tie strips laid directly on the sub roadbed (the pink styrofoam sheeting).  Finally, the MILW track was installed brand new in 1952 or 1953 and the pictures I have show a clean white ballast. I modeled it with code 83 track, CV mainline tie strips, and laid the tack on top of N-scale cork roadbed that I picked up a while back at a bargain price.  I wanted it to be elevated a bit to represent good drainage but ultimately not as "formal" as the CNW mains.  

This photo shows the end of the modeled track south of Bayview.  South is to the left, and in this "dry" test installation you can see that I was looking to choose ballast textures and colors to differentiate between the tracks:

All are Woodland Scenics "fine" ballasts.  From back to front are the two CNW mains, the CNW branch track, and the MILW interchange track.  After selecting the textures and the colors I applied several thin layers of ballast using the techniques described in the previous post, then did some touch up work and applied an acrylic wash:

One thing I did (to save a bit of money) was to get some ordinary builder's sand at the local hardware store.  I filled in most of the deep valley in between the two CNW mains with this cheaper material rather than the expensive ballast.  You can see this coarse builder's sand in between the two tracks at the back of the photo to the right os the ballasted area.  The Woodland Scenics ballast profile strips look good but the center valley is way to steep and deep from all the photos I've seen of the CNW main line tracks.  

Charles Hostetler

Washington, Ill.

Layout Progress - Ballasting the Car Ferry Yard

Over the holiday season I was able to take all of the Fridays in December off!  I hoarded a lot of my vacation time until late in the year for the first time ever with the intent of undertaking a major layout project.  I started with some ballasting.  Now I had ballasted some track before but never anything as extensive as the car ferry yard on Jones Island.  So it was a learning experience and took some patience and several iterations to get a satisfactory result, but in the end I was pleased with the result and had a good time.  

The track in this area was about 25 years old in 1957.  From the few pictures available in the early 1950s it appeared that the ballast was cinders.  And the track was definitely sunk deep into the sub roadbed with the top of the ties flush with the ground surface at the edges.  I wanted a sunken profile with the tie tops even with the ground and ballast surface, apparent poor drainage, and a mild amount of invasive vegetation.  

I used fairly standard techniques.  The ballast was Woodland Scenics Fine Cinders (B1376) and the bonding medium was white glue in wet water to the consistency of milk, applied with a turkey baster with a diffusing brush.  My first attempt looked pretty good, but I had not taken enough care to get all the ballast off the top of the ties and out from the rail web (click to enlarge):

So I ran a pick along all of the rail webs and dislodged the ballast, then broke the ballast off the top of the rails with a sanding sponge, and applied a second layer.  I eventually figured out that several thin coats were much better and easier to control than one big thick coat.  The next two photos were taken after the third thin coat, the first looking south down the long axis of Jones Island toward the Jones Island classification yard and Bayview:

and the second looking north toward the City of Midland and the Municipal Ferry Terminal:

After the fourth thin layer I had everything pretty much evened out and went to address point flaws, such as thin areas:

"craters" where I had dropped too much water instead of letting it flow out through the diffuser brush:

and holes where feeder wires dropped through the subroadbed:

At the end I thought all of the scraping and rubbing had produced an uneven appearance (reflectance?) on the top of the ties under certain lighting angles:

So I ran an acrylic wash with Liquitex flat pigments in an isopropyl alchohol/water mix.  I also used some mixed Liquitex pigment thinned slightly with water as paint to touch up the ends of the ties in some places, shiny spots on rail joiners, and the ends of feeder wires.  After some detail work I was really pleased with both the close up and far off appearance of the final result:

On to the main line next!

Charles Hostetler

Washington, Ill.

 

 

Some Other Freight Commodity Statistics Related wit Warren's Tank Car Fleet

In an earlier post I provided some railway freight statistics for commodity classes 501 (gasoline, which includes natural or casinghead gasoline) and 507 (petroleum products refined, NOS; which includes LPG).  Those statistics were from the 1% carload waybill survey and included some nationwide data about the frequency, length of line haul, average tons per carload, and revenue generated through the 1950s as well as some state-to-state distributions for the early 1950s.  That post can be found here: 1% Waybill Statistics for Warren.  

In this post I'm going to provide some additional statistics for these two commodity classes.  This first set is also from the Interstate Commerce Commission and is called the Freight Commodity Statistics.  These are ones I just have happened to have transcribed so far (I've been jumping around as I get interested in different commodity classes):

Download 501 Gasoline 1948 FCS

Download 501 Gasoline 1950 FCS

Download 501 Gasoline 1957 FCS

Download 501 Gasoline 1958 FCS

Download 501 Gasoline 1959 FCS

Download 501 Gasoline 1962 FCS

Download 507 Petroleum products refined NOS 1950 FCS

Download 507 Petroleum products refined NOS 1957 FCS

Download 507 Petroleum products refined NOS 1962 FCS

These are all Excel files.  Each file contains a transcription of the freight commodity statistics for one commodity class for a single year.  Note that like the 1% Carload Waybill Survey these data were compiled from all US Class 1 Carriers.  However, unlike the 1% Carload Waybill Survey, each carrier is reporting ALL of the freight it handled for the year, not just a 1% sample.  

The bulk of the transcription is a table contained in columns A through G.  Each row is for a single Class 1 Carrier.  As provided, the carriers are in alphabetical order by reporting mark but of course can be sorted in a variety of ways.  The column headings and meanings are:

• Column A:  the carrier's reporting mark.  
• Column B:  the number of tons of freight that originated on that carrier's lines and terminated on its lines.
• Column C:  the number of tons of freight that originated on that carrier's lines and were delivered to all other connecting carriers.
• Column D:  the number of tons of freight that were received from all connecting carriers and that terminated on that carrier's lines.
• Column E:  The number of tons of freight that were received from all connecting carriers and were delivered to all other connecting carriers (through freight).  
• Column F:  The total number of tons of freight, or the sum of columns B through E.  
• Column G:  The total number of carload handled.  
• Column H:  {not provided in the original hard copy, but calculated in Excel} the average number of tons per carload, obtained by dividing the entry in column F by the entry in column G.  

The last carrier is the Western Railway of Alabama (in row 126 of the 1950 file for commodity class 501, for example).  Below this are some totals calculated in Excel for each of the columns as a QA check.  And below that are the calculated total number of tons originated and total number of tons terminated for the commodity class for that year, again provided as a QA check.  

An interesting note here is that it seems in an ideal world that there ought to be mass balance, i.e., that the total number of tons originated should equal the total number of tons terminated.  In fact, there is a systematic bias and over all the commodity classes across the country for a year there are a few percent fewer terminations than originations.  The reasons for this could be:

• some freight not terminated because of accidents, losses, or damages.
• human error, either by the reporting carriers or by the ICC tabulators.  
• freight originated or terminated on Class 2 or 3 railroads is not counted.  
• freight originated or terminated on Mexican and Canadian railroads is not counted.  

I've been looking into this problem for over a year, and have found some new resources to attack the apparent mass imbalance issue and will be reporting on that in a future post.  

The second set of tabulations is from Canada's Domain Bureau of Statistics and is called Railway Freight Traffic (which I have abbreviated as RFT).  How I found these statistics and what they mean for modelers of US freight commodity flows is another story to be related in a future post, but there are two filed I have transcribed that are relevant to this topic.  They are both from 1957, are organized by commodity class, and are also Excel files:

Download 501 Gasoline 1957 RFT

Download 507 Petroleum products refined NOS 1957 RFT

Like the US freight commodity statistics, the Canadian Railway Freight Traffic statistics are in tons per year, but instead of being organized by carrier they are organized by province.  The column headers and their meanings are:

• Column A:  the province.
• Column B:  the total number of tons of freight loaded in the province.  
• Column C:  the total number of tons of freight that were received from US carriers in the province and that were bound for Canadian destinations (I call these imports).  
• Column D:  the total number of tons of freight that were received from US carriers in the province and that were bound for US destinations (for example, traffic across Ontario from Buffalo to Detroit).
• Column E:  the total of columns B through D.
• Column F:  The total number of tons of freight that were unloaded in the province.  
• Column G:  The total number of tons of freight that were delivered to US carriers.  
• Column H:  The sum of columns F and G

You can see that for gasoline, for 1957, imports to Canada and exports from Canada are a minor component of the commodity flow and the US and Canadian systems operated relatively independently.  This is not the case for commodity class 507, which includes LPG, and suggests that a significant portion of Canada's LPG was imported from the US.  In fact, there were about 154,000 tons imported, mostly in Ontario, compared to a total of about 635,000 tons unloaded in all the provinces.  There were no significant exports of LPG from Canada to the U.S.  If you look on the 507 FCS spreadsheet for 1957 you will see that about 9,483,514 tons were originated.  So about 1 in every sixty or so loads of LPG should be destined for Ontario, and if one is modeling in Canada, about one in every four loads or so could be from a US shipper.  

In my final post on this topic, I'll show how I developed some model waybills for Warren Cars that are relevant to my layout setting using the State-To-State, Freight Commodity Statistics, and Railway Freight Traffic data for 1957 posted here and in the previous post.  

 

Charles Hostetler

Washington Ill

 

 

Some 1% Carload Waybill Statistics Related to Warren Petroleum Corp.'s Tank Car Fleet

Steve Hile is doing a clinic on Warren Petroleum tank cars for Prototype Rails 2019 that is coming up in a couple of weeks.  He asked me if I had any information on rail shipments of Warren Petroleum's products.  I didn't have anything on hand, but I was off from work today with nothing but a company luncheon to attend, so I whomped this up this afternoon.

Steve told me that:

• he was interested in the later 1940's and early 50's,
• he was interested in commodity coding and shipping, and
• Warren's fleet was exclusively ICC 105 tank cars for liquid petroleum gas (LPG) and 104W's for "natural" (casinghead) gasoline.

The first thing to do was figure out the commodity classes involved, as all of the ICC statistics are indexed by commodity class.  It turns out that the commodity class "501 Gasoline" consists of:

• Gasoline, blended, consisting of motor fuels containing 50% or more of gasoline {typically produced at a refinery}
• Gasoline, casinghead {typically produced at the well head and transported to a bulk distribution center}
• Gasoline, natural {a synonym for the preceding commodity}
• Gasoline, not otherwise indexed by name

and commodity class "507 Petroleum products refined NOS" {Not Otherwise Specified} consists of:

• Absorption oil, petroleum
• Belt oil, petroleum
• Benzine (benzene) (benzol)
• Butane gas
• Compression oil, petroleum
• Cordage oil, petroleum
• Floor oil, petroleum
• Harness oil, petroleum
• Kerosene
• Leather oil, petroleum
• Liquified petroleum gas
• Miners' oil, petroleum
• Naptha
• Naptha distillate
• Naptha solvent, petroleum
• Napthenic acid
• Neatsfoot oil, petroleum
• Oil, petroleum noibn
• Paraffine wax
• Pentane
• Petrolatum, noibn
• Petroleum fatty acid
• Petroleum jelly, noibn
• Petroleum wax
• Propane gas
• Putty oil
• Refined oil, petroleum, illuminating or burning
• Refined oil, petroleum, noibn
• Soap oil, petroleum
• Tanners' oil, petroleum
• Tobacco oil, petroleum
• Transformer oil, petroleum
• Wool oil, petroleum

So it appears to me that Warren's 104W tank cars were carrying products in commodity class 501, and Warren's 105 tank cars were carrying products in commodity class 507.  

The most basic statistics in the 1% carload waybill study are contained in the MS-1 tabulations, which are nationwide statistics that were published each year starting in 1947.  I transcribed the statistics for both commodity class 501 and 507, and they are provided as Excel files here:

Download MS1 for 501 1947 through 1960

Download MS1 for 507 1947 through 1960

The first column is the year (all of these statistics are nationwide), the second column is the number of carloads in the 1% sample, the third column is the total number of tons associated with those carloads, the fourth column is the average tons per car, the fifth column is the average line haul (in miles) per car, and the sixth column is the average revenue (in $) per car.  The data are tabulated from 1947 through 1960.  

A couple of things to note:

1. There are roughly 350,000 carloads in the 1% sample, so the expected frequency of occurrence of a freight car carrying gasoline is about 1 in every 100 cars, and this is similar to the expected frequency of occurrence of a freight car carrying LPG.  
2. The number of carloads of gasoline per year is strongly declining over the 14-year study period; while the number of carloads of LPG is rising, reaching an approximate plateau in 1955 or so.  
3. The length of line haul of a typical gasoline load is very short and is generally declining over the 14-year period, while the line haul for LPG shipments is of moderate length and is generally increasing.  
4. In the later years the line haul carriers split almost 3 times as much revenue for a LPG shipment than for a gasoline shipment, even though they weight about the same.  
5. From looking at a lot of these statistics it seems to me like the gasoline business is slowly dying off but that the LPG business was going strong throughout the late steam era.  

A second thing I looked at was the type of car that was recorded as carrying these shipments.  These are in the TC-1 statistics, and I transcribed those for 1950 and 1952.  Those Excel files are available here:

Download TC1 for 501 1950 and 1952

Download TC1 for 507 1950 and 1952

In the first column are mileage blocks, the first is line hauls from 1 to 49 miles, the next is line hauls from 50 to 99 miles, and so on.  You can see that in both 1950 and 1952, the vast majority of gasoline was carried in tank cars (they did not distinguish between the subtypes) but that there were a handful of shipments in some sort of containers hauled in box cars.  The majority of line hauls were tabulated in the 100-199 miles and the 200-399 miles categories of line haul.  A handful of shipments were over 1,000 miles but no shipments were transcontinental (i.e., more than 2,000 miles).  

In both 1950 and 1952, the vast majority of LPG gas was carried in tank cars, with a handful in box, refrigerator, and (even 1) stock car.  A plurality of the line hauls fall into the 600-999 mile category.  The smaller bump in the 200-399 miles category is probably caused by redistribution (for example, shipment of LPG by lake boat to a bulk terminal, followed by further distribution by rail.  

The third thing I looked at from the 1% carload waybill statistics was the state-to-state distributions.  This helps narrow down the national statistics and identify shipments that are most likely to be represented in the location of your particular layout setting.  I didn't have time to tabulate all of these statistics - for major commodity classes like these that takes a week or so.  So I focused on the originating states where Warren Petroleum Corp. had distribution facilities.  Steve may have better insight into this, but according to the Oklahoma Historical Society's web page on Warren Petroleum Company (Warren Petroleum Company), "By 1953 one of the largest producers and marketers of natural gasoline and LPG in the world, the company had more than two thousand employees, six shipping terminals in Texas, New Jersey, Florida, and Alabama, LPG plants in Oklahoma, Texas, New Mexico, Illinois, and Indiana, and the world's largest privately owned fleet of railroad tank cars."  So I focused on tabulating gasoline shipments originating in Alabama, Florida, New Jersey, and Texas for commodity class 501, Gasoline; and on LPG shipments originating in Illinois, Indiana, New Mexico, Oklahoma, and Texas for commodity class 507, Refined petroleum products NOS.  Those Excel files are available for download here:

Download STS for 501 1950 through 1954

Download STS for 507 1950 through 1954

The first column in each of the files is where the shipment originated, and the second is the state in which the shipment terminated.  The third through seventh file is simply a tabulation of how many carloads were loaded in the specified category in 1950, 1951, ... 1954.  Column I is labeled <CL/yr>, which stands for average carloads per year.  In interpreting 1% samples, I have found it useful to average over several years (in this case 5 years) to smooth out sampling artifacts.  Column I is a calculation of the average number of carloads that were loaded over the five-year period.  

If you look at the results for gasoline and for LPG, you will see some stark differences that were foreshadowed by the national statistics presented above.  Looking at the commodity class 501 state-to-state distributions:

• A majority of the shipments are intrastate (i.e., the originating and terminating states are the same).  
• Most of the remainder of the interstate shipments are to a neighboring state.  
• This commodity class doesn't seem to have a lot of reach.

Looking at the commodity class 507 state-to-state distributions:

• A majority of the shipments cross several states.
• A majority of the shipments from a single state end up in a large number of other states.
• There are a lot of "minority" terminating states; Vermont, New Hampshire, Nevada, etc. 
• This commodity class seems to have a lot of reach.  

Steve, I hope you find this background information useful, and best wishes for a successful clinic.  I think the Warren tank cars, particularly the ICC 105 cars carrying LPG are applicable to a lot of layout settings.  And some pretty good models of these cars have hit the streets recently!

Charles Hostetler

Washington Ill.

 

Universal Box Car (for 1957)

In 2013, in an earlier reincarnation of this blog, I wrote what I consider the most original and important piece of analysis that I had done in the world of model railroading.  As I'm getting ready to write some additional posts on this blog about freight car distribution and modeling waybills, I thought it would be good to migrate it over here from the old platform.  I was written just after I had followed a Rutland box car on a series of journeys jumping around the county according to the probability distributions in the 1% Waybill Survey State to State statistics and some basic rules about empty car handling.  It went as follows...

 

In the last post I showed some examples of cumulative statistics that I have been compiling for the travels of Rutland 197, an ordinary 40' steel box car.  The idea is that in each step of the travels, the next shipper and consignee are chosen at random from a probability distribution that is calculated from the 1957 1% Carload Waybill sample and an empty car competition/selection process.  Each step is a small independent sample drawn from a broad spectrum of outcomes that depend on where the car is located and what other empty cars are available to compete for loads.  After sufficiently many steps/samples have been taken, the cumulative statistics should say something meaningful about the average properties of a Rutland box car.  Examples are how often the car is loaded versus empty, how much time it spends in each state, and how frequently it is on a common carrier's rails.

I spent some time last week thinking about comparison, and what it means in the context of this Monte Carlo sampling.  After a statistically meaningful sample has been obtained from the Rutland box car travels, how can it be analyzed and what should it be compared to?

The first comparison that can be made is a comparison of separate subsamples of the cumulative statistics.  For example, imagine that you had a million or so steps compiled.  You might want to look at the subsample of the first 100,000 steps compared to the last 100,000 steps.  If they turn out to be the same (within the sampling error) that is a good sign that your algorithm is stable and that you have adequately sampled the probability space.  This subsample comparison is a technical tool that tells you something about the method.

A second comparison that can be made is a comparison of the cumulative statistics for Rutland 197 versus some other box car (functionally identical, but with different reporting marks, say for example AA 305, ACL 25176, or CNW 4; all of which are 40' PS-1 box cars).  Presumably, if the probability space was adequately sampled, the differences between the cumulative statistics of the different cars (if any!) would reflect something important about freight car distribution.  For example, we might find that CNW box cars spend more time in Illinois, Wisconsin, and Iowa than Rutland box cars (although I rather doubt that this is the case).

This got me wondering whether there was a meaningful absolute comparison (as opposed to the two types of relative comparisons described above).  Sort of like how an ideal gas is a useful concept because it has properties that are relatively easy to calculate and most real gases are close approximations to the ideal gas abstraction.  I call this abstraction the universal box car, and it has some surprising and interesting (at least to me) ramifications.

The universal box car is at home on every carrier's rails; it is never on a foreign road.  It wins every car competition and is selected for a load at each step in its travels.  It never has to return empty via its service route.

One reason that the universal box car is a useful abstraction is that one can quickly and easily calculate its final distribution.  No Monte Carlo sampling is required to figure out where the universal box car spends its time.  In fact, I calculated the frequency that the universal box car spent in each state in 1957 in about 5 minutes using Excel (albeit after 6 months of entering data into Excel!).  Much easier and quicker than calculating a million-step Monte Carlo sample!  Here's a sketch of the method.  Take the number of carloads from the 1% carload waybill survey (for commodities carried primarily in box cars) and turn them into frequencies (or probabilities). Arrange them in a matrix such that each row of the matrix (corresponding to a particular state) is the frequency distribution of box car loads originating in that state.  For example, the 1st row represents Alabama, so the entries in this row represent the cases where the shipper is located in Alabama.  The first column is also Alabama, and the entry in the cell in the first row and first column represents the likelihood of having the consignee be in Alabama GIVEN that the shipper was located in Alabama.  The upper left part of this resulting 49 by 49 matrix looks like this for 1957:

	Alabama	Arizona	Arkansas	California	Colorado	Connecticut	DC
Alabama	0.13727	0.00174	0.00869	0.01390	0	0.00261	0
Arizona	0.01402	0.08411	0	0.06542	0.00467	0	0
Arkansas	0.02185	0	0.17577	0.02284	0.00497	0.00298	0
California	0.00540	0.04005	0.00436	0.43495	0.00892	0.00208	0.00125
Colorado	0.00472	0.01651	0.01179	0.06132	0.25235	0	0
Connecticut	0	0	0.01905	0.10952	0.00476	0.10952	0
DC	0	0	0	0	0	0	0
Delaware	0	0	0	0.02419	0	0	0.01613

 

It turns out that the stationary vector for this matrix, or the eigenvector corresponding to an eigenvalue of one, gives the long-term average behavior of the universal box car.  I calculated the eigenvectors for each state using an interactive technique.  Its most easily understood by looking at a bar chart (click to enlarge):

or a color coded map (click to enlarge):

The universal box car spends almost 12 percent of its time in Texas.  One of the reasons for this, as I mentioned in an earlier post, is that in 1957 Texas had an unusually high proportion of Texas to Texas shipments.  The universal box car also spends a lot of its time in New York, California, Illinois, Ohio, and Pennsylvania.  There are about 50 states (actually 48 plus D.C.) so if a box car just jumped around from state to state at random you would expect to find it in each state about 2% of the time.  Instead, the universal box car spends less time in states where there were relatively fewer shippers and consignees, and more time in states where there are more shippers and consignees.

My working hypothesis for these posts is that Rutland 197 is like every other 40' general purpose box car in every way except for the fact that it says Rutland on its side.  The only way the rail system could have caused Rutland 197 to behave differently than any other box car (if indeed it did) is through the car competition and selection process.  So the universal box car, which is unaffected by car competition and selection (i.e., it wins every competition and is always selected) is a useful comparison standard that allows us to focus on the difference that car competition and selection made to shape freight car utilization and distribution.

The second reason that the universal box car is an important and interesting concept is something I find curious and even a bit paradoxical.  Not only is the universal box car a useful abstraction (in the sense that an ideal gas is a useful abstraction in understanding real gas behavior), it is also a descriptive statistic of the 1957 actual real-world box car distribution (as measured by the 1% carload waybill survey).  The eigenvector of the transition probability matrix was calculated directly from the 1957 state-to-state sample and does not include any variable parameters or model-dependent assumptions.  It is a sophisticated sort of average, and a different way to look at the actual box car distribution in 1957.  It tells us, for example, that on any given average day in 1957, about 12% of the nation-wide fleet of box cars were in Texas, about 6% were in California, about 7% were in Illinois and New York (each), and so on.  Now probably no particular day in 1957 looked exactly like this snapshot, but the average distribution is a useful picture of how the national fleet was distributed during that year.

If one had the time and patience, one could type in the state-to-state distributions for 1950, 1951, ... 1959, form the transition matrices for each year, calculate the eigenvectors, and obtain a really interesting series of snapshots of the year by year evolution of box car distribution during the 1950s.  Or one could focus on  commodities carried primarily by tank cars, or by stock cars, and so on.  The state-to-state distributions are a huge series of dense tables that I find relatively uninformative in and of themselves, but these snapshots of the real-world average distributions would probably be a very visual and accessible depiction of the evolution of freight car traffic in the late steam era.

Charles Hostetler

{then in} Goshen, Ind., now in Washington Ill.

 

 

Painting Turnouts

On Jones Island, my track work is made from Micro Engineering code 70 weathered rail laid on Central Valley 8' branchline ties.  I'm using Micro Engineering #6 code 70 turnouts.  I've been assembling the Central Valley tie strips into three-foot lengths (mostly for ease of handling and installation) and painting the tie strips a flat, dark brown before they are glued into place and the rails laid over them.  When I finished laying most of the track for the Jones Island classification yard and for the ferry yard on the center peninsula, I decided to take a couple of photos to check the light levels and track alignment, and generally to document progress.  This is one of the first photos I took of this part of the layout, late in November (you can click to enlarge):

I didn't realize how much of a contrast there was between the ties under the turnouts and the ties under the other track.  Its distinctly noticeable in all of the ladders, at different sight angles:

So I decided I wanted to do something about this and not just trust that ballasting the track would "tie" everything together and make it all OK.  Also I noted that the sides of the rails of the turnouts were not weathered and I thought this contrast would really be objectionable.  I remembered an article I had read about weathering and improving Micro Engineering turnouts where the author had painted the turnouts, and I decided to plow ahead and see if I could fix this situation before proceeding with ground texture and ballast.  I found it was best to work a single turnout at a time because the paint I use dries pretty fast, so first the areas around each turnout were masked:

I used a flat primer made by Rust-Oleum.  The colors in their camouflage series aren't labeled, but for this track I used the dark brown:

After the paint is applied (from all angles and overhead) the tops of the rails were quickly wiped clear with Q-tips before the paint dried, the masking removed, and each turnout worked back and forth a couple of times.  This is the final result:

And now when you look at the track from different sight angles and different view points you can't tell where the regular track leaves off and the turnouts begin:

I worked pretty slowly on this project at first to develop the technique and make sure I didn't ruin anything, but by the time I got confidence in my approach and into production mode I was working a bit faster.  The main thin was waiting for the paint to dry completely before working on any adjacent turnouts and applying masking over painted areas.  As you can see, I was working on yard ladders a lot, so overall it took me about two weeks to complete this little detail project but I'm glad I did before moving onto ground cover and ballasting.  

Charles Hostetler

Washington Ill

My Track Plan

I've been asked a couple of times about my track plan.  Well I don't really have one, and I'm not particularly embarrassed about it either!  I decided really on I wanted to focus my layout on Jones Island, and only have enough other track to support the operations on Jones Island.  So I selected a very short portion of the Milwaukee to Chicago mainline (about 3 scale miles) and wrapped it around my basement exterior walls.  Here's a schematic from the C&NW's 1959 Track Chart (click to enlarge):

North is to the left in this diagram.  The main line (color coded yellow) runs from Bayshore (at the right) to the Third Ward Yards in downtown Milwaukee.  This is a double tracked main line.  The Jones Island portion that includes the Municipal Ferry Terminal and the C&O's City of Midland Car Ferry is highlighted in green.  

I used the same color coding on this aerial photograph taken in 1955 (again, you can click to enlarge):

Again, north is to the left.  If you enlarge this image and look carefully beneath the left end of the green highlight you will see the City of Midland being loaded by the C&NW.  The GTW's City of Milwaukee is steaming into the harbor nearby.  

So my goal for 2019 is to get the green portion of the layout, roughly from Bayview to the north end of Jones Island into fairly final form and operational.  The track is mostly in place just up to the ferry ramp, and I've been running some test trains to make sure everything is working well.  

This is the view from Bayview looking to the north:

The train in the foreground is Road Special Delivery 3 running from Butler counterclockwise around Milwaukee with cars for Updike Yard that will be picked up by the Jones Island Transfer run.  The Milwaukee Road switcher barely visible in the background has a transfer cut that is coming out of staging and will proceed to Jones Island via the new track in the foreground.  

This is the view from the south end of Jones Island looking to the north:

The Jones Island main classification yard is on the left, and the Milwaukee Road interchange tracks are to the right.  The bulk oil distribution terminal on Jones Island is to the extreme right in the background.  

Another view from about the center of Jones Island looking to the north:

The car ferry loading/unloading yard is in the foreground.  The train headed up by the C&NW switcher is on the Heavy Lift Dock.  It was being used as a test train yesterday and I parked it there temporarily last night to make sure it wasn't in the area where I was spray painting turnouts today.  There's a little bit of ground cover where I was testing out colors and textures, and the City of Midland is barely visible in the background.  

There three photos show the areas I want to get up and running in 2019.  There are going to be a lot of harbor projects, scenery, and track detailing but I think its going to be cool to get some of my car ferry waybills into operation!

Charles Hostetler

Washington Ill

 

Calculating Average Route Complexity Using Freight Commodity Statistics

I've been studying the routes on waybills, and how they might be inferred for commodity classes from waybill statistics, for about almost three years now.  I made a derivation (that is the subject of this post) almost two years ago on a pad of paper that has been sitting around in my layout room at various places.  This winter I am in a general drive to clean up and organize my desk and working area, and to get everything off the layout so that I have an overview of what I'm doing (trying to build a layout here - not an elaborate horizontal storage system!).  So my purpose of this post is to document the derivation in a better place, where it won't get lost, and it won't be cluttering up my workspace, and to free up the pad of paper for more doodling!

Route Complexity - So the basic notion I had was that the first statistic about a route on a waybill should be really simple, and that there's nothing simpler than the number of carriers listed on the route.  This is the easist to grasp notion of complexity, so I called it route complexity (RC).  If the route is "RI" then the RC is 1; if the route is "GN-CNW" then the RC is 2; if the route is "SP-UP-CBQ" then the RC is 3, and so on.  Note that the number of interchanges along the route is one less than the RC.  Figuring out the RC for any particular route specified on any single waybill is just a matter of counting.  

Average Route Complexity for a Group of Shipments - Well what if you have some specific group of shipments, say perhaps all of the shipments of wine during 1953?  This might represent 18,000 or so waybills - is there anything one can infer about the average route complexity of all of these shipments?  It turns out that there is if the statistics are tabulated the way the ICC tabulated the Freight Commodity Statistics (FCS).  

If you have Excel, you can download this file:  Download 747 Wine 1953 FCS.  I have tabulated hundreds of these files from the time period 1948 through 1962 and there are hundreds more to tabulate.  I've been using them to generate synthetic waybills that mimic commodity flows that pass through my layout, but that's another story for another post.  If you don't have Excel, you can click on this to enlarge and follow along anyway:

First the file naming system consists of the ICC commodity class number (in this case 747), the commodity class description (in this case wine), the year, and the type of statistics encoded in the file (FCS is the ICC's Freight Commodity Statistics; STS is the ICC's 1% Waybill State to State; and RFT is Canada's Domain Bureau of Statistics Railway Freight Traffic {which I'll write more about in some subsequent posts}).  Hopefully if these files get scattered around the internet they can keep their pedigree with them ;)

Next the content of the FCS files themselves.  They are all indexed by Class 1 Carrier (column A).  In 1953 there were 122 Class 1 Carriers. This file happens to have been left in alphabetical order by carrier (although you can sort the Excel file any way you like).  The carriers range from AA (Ann Arbor) at the top (row 3) to WRA (Western Railway of Alabama) at the bottom (row 124).  

The next four columns are the key raw data that have been tabulated from the 1953 annual report.  Each Class 1 Carrier had to keep track of, and report, the total number of tons handled on its lines for the year by:

• Tons originated on its own lines and terminated on its own lines (column B)
• Tons originated on its own lines and delivered to connecting carriers (column C)
• Tons received from other carriers and terminated on its own lines (column D)
• Tons received from other carriers and delivered to other connecting carriers (column E)

So in 1953, the SP (row 109) for example, shipped 63,395 tons of wine to its own online customers.  It loaded 233,431 tons of wine on its own line and delivered them to connecting carriers.  It received 28,967 tons of wine from other carriers and delivered them to customers along their lines.  And finally, 48,125 tons of wine were received by the SP from its interchange partners and subsequently passed along to other interchange partners (through traffic).  

The other columns (F, G, and H) are used for QA and to figure out average number of tons per car.  

The total statistics for all of the Class 1 Carriers in the commodity class 747 Wine are provided in row 126.  Note that these totals are a product of a calculation in Excel, they are not provided in the FCS themselves.  After typing in the first 50 or so of these tabulations I got to wondering whether the totals in row 126 could be used to calculate the average route complexity for the commodity class for that year.  With a pad of paper and a little noodling, it turns out that the answer is yes.  I'm reproducing the derivation here for posterity and also so maybe someone else can build or expand on this work.  You can click on the below to enlarge if you want to follow along:

Consider a collection that represents a large number of waybills.  Sort and separate them into piles according to route complexity.  The first pile will be RC = 1, and there will be some number of tons in that pile, say "a" tons.  The second pile will be for RC = 2, and there will be "b" tons represented in that pile.  Continue this up to the largest RC for that year (say for RC = n; whatever n is).  there will be g tons associated with this pile.  Now report to the ICC.  You will report "a" tons as OT (originated and terminated on line; this is what RC = 1 means).  For the waybills in the second pile, you will report "b" tons as originated online and delivered to other carriers (OC) and the receiving railroad will report "b" tons as received from other carriers and terminated on its line.  In row 14 of the derivation above I have shown the total tons that will be reported to the ICC for this commodity class.  

Row 16 simply says that the average route complexity, <RC>, is simply the average of the number of tons weighted by the route complexity.   Rows 17 and 18 are simple algebraic substitutions, and row 19 is the formula that is used to calculate <RC> in my Excel spreadsheets.  So if you run those total tons through the formula, you can see that the average route complexity of all shipments of wine in 1953 was 3.55.  This is a pretty high number indeed and it means that each shipment (typically originated on a handful of railroads in California) there were an average of 2.55 additional appearances of this shipment on other Class 1 Carriers.  I've termed this reach, and I think its one of the things that make a commodity class interesting to model.  

 

Charles Hostetler

Washington Ill