Eliza Davenport Whiteman
UEP 232 – Final Paper
December 13, 2012
BACKGROUND & PROJECT DESCRIPTION
In an attempt to boost regional economies and ensure greater food security, some in the
food sector are working to examine the viability of regional food systems. As part of this
inquiry, some researchers have started to assess the capacity of states or regions to meet
their own food needs, including the slaughter and processing of livestock raised in the
area.
This analysis examines the geographic distribution of cattle production and slaughter
facilities within Maryland as a means of assessing state-level slaughter capacity for
farmers who are selling their beef regionally. The analysis was conducted at the request
of Johns Hopkins Center for a Livable Future (CLF) and is based off a similar assessment
of beef and apple production and processing in New York State conducted by Columbia’s
Urban Design Lab (UDL).
The key question posed by CLF was around the capacity of Maryland slaughter facilities
to service all cattle farms, given that there are no facilities in the southern and Eastern
Shore regions of the state. Additionally, I had an interest in assessing the capacity of the
existing slaughter facilities to handle all of the cattle raised in state if it were to all be
slaughtered and processed locally.
METHODOLOGY
Data Processing Steps
 Put all data layers into 1983 Maryland State Plane Feet coordinate system
 Select by attribute to use only slaughter facilities that slaughter beef
 Create base map using county boundaries, point locations for slaughter facilities
and cattle farms, roads data layer and cropland data layer
Analysis Steps
To assess the capacity of Maryland slaughter facilities to service cattle farms selling
within the state, I used a series of network analysis tools. I started by creating Thiessen
polygons around each of the slaughter facilities, to show the proximity of slaughter
facilities to each other. I then created a Euclidean distance buffer with 15, 30, 45 and 60mile break points around each of the slaughter facilities. Lastly, using the ESRI streets
data network, I conducted a drive time network analysis on using the slaughter facilities
as the origin and the cattle farms point data as the destinations. Thirty and sixty minute
drive time service areas were created around each slaughter facility (time selection based
on the UDL analysis). The comparison of which farms were serviced by Euclidean
buffers and drive time network analysis demonstrated the importance of using street data
with speed limits when evaluating service areas for business destinations.
1. Thiessen polygons map to show proximity of slaughter facilities
 Generate Thiessen polygons
2. Euclidean distance buffer map
 Generate buffer around slaughter facilities with 15, 30, 45 and 60-mile
break points to show how many farms each facility serves/which facilities
are close to lots of farms
3. Service area polygons map using drive time analysis to show proximity of
slaughter facilities to farms
 Use Network Analyst to create a drive time service area polygon to show
drive time from slaughter facilities at 30-minute and 60-minute intervals
 Set up parameters for analysis, using 30 and 60 default breaks, do not
allow U-turns, provide output in time and mileage
 Run to compute the service area
 Use Properties: Symbology to denote the areas that are closer in and
farther away from the facilities
 Overlay farm locations to show how well each slaughter facility serves the
farms
4. Origin-Destination network path analysis to show all farms within one hour
drive time of slaughter facilities.
 Create an O-D cost matrix analysis layer to find the shortest route from
each slaughter facility to each farm
 Add origins and destinations
 Set up parameters for analysis – no U-turns and results in time and
distance (miles)
 Run to calculate O-D matrix
 Create summarize tables to find the average, minimum and maximum
drive time and mileage to each slaughter facility.
For those farms not within the one hour service areas, a closest facility network analysis
was performed to show the shortest driving route (by time) from each farm to the closest
slaughter facility. Median centers were also produced for these farms (one in the southern
counties and one on the Eastern Shore) to determine the point representing shortest
3
Euclidean distance to all farms in the service area. These calculations can be used for
assessing aggregation strategies for farms outside the service areas.
5. Closest Facility network analysis
 Select farms that are outside 60-minute drive time service area and create
new layer
 Using this point layer, set parameters for analysis – no U-turns and results
in time and mileage
 Run to determine shortest driving routes from each farm to its closest
slaughter facility
 Create summarize table to see average, minimum and maximum drive
time and mileage.
6. Median Center
 Use layer for farms outside of 60-minute drive time service area - select
farms in southern counties and create new layer, then select farms in
Eastern Shore counties and create new layer
 Use spatial statistics tool: Measuring geographic distribution – median
center to determine median center for southern counties and for Eastern
Shore counties
Finally, a series of maps were created to show the distribution of beef cattle across the
state by county using three different categories (cow/calf inventory, cattle on feed and
cattle sold for slaughter). This data was gathered from the 2007 National Agricultural
Census (NASS) and is not comprehensive for each county.
7. Distribution mapping for cattle production
 Join and relate NASS data using county FIPS codes for:
o Beef cow/calf inventory (point-in-time)
o Beef cattle on feed (point-in-time)
o Cattle on feed sold for slaughter (annual)
 Add all values in quantities and display using quintiles
4
DATA LAYER/DATA SET
Beef cattle per county:
 Cow/calf
inventory
 Cattle on feed
 Cattle on feed
sold for
slaughter
Current MD slaughter
facilities
Street Data with speed
limit information
SOURCE
NASS - 2007
YEAR
2007
FORMAT
CSV file
NOTES
Gives numbers for cattle raised on feed
– no information for cattle raised purely
on pasture. Does not have information
for all counties.
CLF
2012
Shapefile
Location, services offered and weekly
slaughter numbers.
ESRI Street
Network
2010
Cropland data layer for
pasture/hay/alfalfa land
Location of MD cattle
farms that sell within MD
Cropscape
via CLF
CLF
Street
Network
Dataset
Shapefile
2012
Shapefile
Maryland County Outlines
MD Dept. of
Planning
ESRI
2010
Shapefile
2010
Shapefile
State Outlines
This data was collected based on
information from online farm listing
databases. It is not comprehensive and
does not necessarily align with NASS
data for number of head of cattle on
feed.
CHALLENGES + LIMITATIONS
To execute this project, it had been my hope to map the location and distribution of all
cattle production in the state. Initially I had planned to join the NASS data with the farm
data from CLF and distribute the cattle across the geocoded farms, but it became clear
that joining these two datasets would create very misleading results. The CLF data layer
includes the location of cattle farms in Maryland that sell within the state, but does not
include all beef farms. The NASS data, on the other hand, is for cow/calf counts, cattle on
feed and cattle on feed sold for slaughter by county, but not all counties are available.
These two data sources are not necessarily correlated, since some of the farms in the CLF
layer are likely grass-fed operations and are not a full count of all cattle farms in the state.
I had also considered matching the cattle counts with the cropland data layer, but given
that the cattle counts are for cattle on feed (meaning feedlots) it seemed unlikely that the
cropland data would be an accurate representation of the true location of most cattle
operations. Given these challenges I opted to just represent the distribution of cattle
numbers within the state (by county) without joining these to the service area analysis
using the CLF cattle farms data. This led to a somewhat less thorough analysis, than I had
hoped, but ultimately served as a tool to show the limitations of conducting research
using agricultural data in the U.S.
5
Data Limitations
 Farm-level data
Geospatial analysis of farm production in the United States is challenging because of
limited data availability. Due to privacy concerns, there is no farm-level data through
NASS and counties with small levels of production also typically lack data. This poses a
challenge for accurately representing the density of cattle production.
While there is some farm-level information on cattle farms in Maryland (used in this
analysis), the data is not comprehensive for the state, which makes accurately joining
census data to farm locations problematic.
 Alternative livestock systems
Assessing the production levels and slaughter capacity for alternative livestock systems
poses another limitation for analysis. NASS collects data on cattle raised on feed (mostly
in feedlots), but does not track numbers for cattle raised on pasture. This makes
determining how well slaughter facilities are serving grass-fed producers challenging.
 Tracking cattle numbers
Following the path of cattle through the production chain is difficult. NASS provides
numbers for cow/calf operations, which are much higher in MD than cattle on feed or
cattle sold for slaughter. This suggests many more cows are born in MD than are brought
to full weight and slaughtered in state. Many cows born in MD are likely sent to other
states in the Northeast or Midwest for backgrounding, finishing and slaughter. Given the
gap between the number of cattle slaughtered yearly by the MD facilities and the number
sold for slaughter, it is safe to assume that many of the cows slaughtered in MD facilities
are from other states. Also, the cow/calf and cattle on feed data are point-in-time
numbers, while the sold for slaughter is annual data, making these different measures
difficult to compare.
CONCLUSIONS
The drive time analysis demonstrated that there are 10 southern and Eastern Shore
counties in the state either partially or entirely outside any of the one-hour service areas
for cattle slaughter facilities. While the bulk of cattle production (based on 2007 NASS
data) is in the northern part of the state, where most of the slaughter facilities are, there
are still significant areas of cattle production that lack adequate proximity to facilities.
According to slaughterhouse managers in the state, the problem is not one of capacity to
handle the number of Maryland cattle slaughtered in state (Table 1), but rather an issue of
proximity to services. Instead of putting in additional facilities, it would make more sense
to aggregate transportation of cattle from farms outside of service areas (using median
center and closest facility calculations) so as to reduce drive time.
6
Table 1
The summary for drive time analysis (Table 2) shows that the minimum average drive
time to each slaughter facility ranges from 70-219 minutes. The three facilities with the
shortest average drive time are also the facilities with the largest capacity, suggesting that
they are well located to service many of the farms used in this analysis. Two of the three
slaughterhouses found in the closest facility analysis are also the facilities with by far the
largest capacity in the state, which means they would likely be the best suited for taking
on additional clients.
Table 2
If I were to continue with this research or had more time to conduct it again, I would like
to collect more data on grass-fed production in the state and would also like to have
designed a better analysis to determine the point location of farms so as to more
accurately distribute cattle density within the state. If this was combined with population
and consumption data, we could get a more thorough analysis of Maryland’s capacity to
cover its own cattle production, slaughtering and processing needs
SIMILAR STUDIES
Health -> Infrastructure, Columbia Urban Design Lab
The primary study that I used as the basis for my project is the report put out by the
Columbia University Urban Design Lab. This report is intended to look at the
correlations between the obesity epidemic and our food system in the U.S. The
researchers posit that a regional food systems approach will help to ameliorate the health
crisis we are currently experiencing, by localizing our food supply and creating greater
access to healthy, fresh foods. They argue that in order to draw more significantly on our
7
regional food supply resources, infrastructure (for slaughter, processing and transport)
needs to be improved.
The primary mapping analysis in their project looks at cattle production and slaughtering
facilities in New York State. The final map is a weighted overlay of the optimal sites in
the state of New York for siting additional slaughter facilities. This overlay includes
cattle production density, slaughter facilities with one-hour drive time analysis to farms,
and major urban areas (>50,000 population) with one- and three-hour drive time analysis
to slaughter facilities.
Mapping potential foodsheds in New York State
This study, conducted by researchers from Cornell, examines the potential for New York
State to meet all of its own food needs from within the state. The analysis was done by
creating a spatial-optimization model. First they collected data to assess production
capacity for agricultural land in New York, actual food yields, population density, and
per capita food needs. They then used these datasets to create an optimization model that
demonstrated production capacity to meet the needs of populated areas, while minimizing
driving time.
This study is only peripherally relevant to my project, as it is focused on yield potential
and is looking at a much more comprehensive set of crops, rather than focusing
specifically on the production and processing of one crop or animal. I thought that the
evaluation of distance to population centers might provide useful insight, but they used
Euclidian distances, which are not as accurate as the drive time analysis I conducted.
Also, since they were working with data that expressed actual food needs, they were able
to do a more nuanced optimization for assessing ideal distribution systems.
Synthesized Population Databases: A Geospatial Database of US Poultry Farms
This study looks at the siting of poultry farms in order to study infectious disease. While
the focus of their research is quite different from mine, their methods for determining the
location and size of poultry farms in the country were useful in understanding the options
for analysis. They collected farm data from the Ag Census (number of poultry farm, farm
types and size of operation by county). Then in order to determine the actual siting of the
farms, they determined the primary factors that affect poultry farm location (via literature
review) which they determined to be:
1. Zoning and local regulations
2. Available affordable land
3. Poultry industry infrastructure (hatcheries, slaughter facilities, etc.)
4. Physical constraints for siting poultry facilities (lakes, steep slopes, etc.)
They then used these factors and various data layers to determine the likely location for
poultry farms in the U.S. Following that they looked at actual poultry farm locations in
order to quantify the importance of each of the determining factors as compared to the
likelihood of finding a farm at any given location.
8
Farm Animal Demographics Simulator Aids in Disease Modeling
This article was referenced in the previous study as a GIS methodology used for
approximating livestock farm locations and numbers of animals per farm. The author of
the poultry study points out that this method has several limitations, including having
exaggerated farm density and overly large spatial distribution, but it does at least provide
another specific technique for how to address the issue I faced in my analysis.
In this model, a mask was created to determine where farms could be (i.e. not in
waterways, parks, federal land, urban areas and within 300 meters of the road). The
locations of the farms were then randomly assigned (with the correct number per county)
using the “generate random points” tool. Because the Ag Census provides the number of
farms per county, as well as the number of each size (by category) of farms and the total
number of cattle per county, the researchers were then able to randomly assign a size
category and a total number of cattle to each of the farm points in each county. This
model does not provide a good suggestion for how to handle counties where the data is
missing (which is an issue with the Maryland data).
Additional Maps
Sources: MD Department of Planning, ESRI, CLF
Resources
9
Bruhn, M., B. Munoz, J. Cajka, G. Smith, R. J. Curry, D. K. Wagener, W. D. Wheaton.
(2012). Synthesized Population Databases: A Geospatial Database of US Poultry
Farms. RTI Press, MR-0023-1201. Research Triangle Park, NC: RTI Press.
Conard, M., K. Ackerman, D. Gavrilaki. (2011). Infrastructure -> Health: Modeling
production, processing and distribution infrastructure for a resilient regional food
system. Urban Design Lab at The Earth Institute. New York, NY: Columbia University.
Accessed at http://admin.urbandesignlab.columbia.edu/sitefiles/file/optimizationmodel.pdf.
Geter, K. (2006). Farm Animal Demographics Simulator Aids in Disease Modeling. USDA,
National Animal Health Surveillance System, Outlook, Quarter Two. Retrieved at
http://nsu.aphis.usda.gov/outlook/issue10/outlook_apr06_fads.pdf.
Peters, C., N. Bills, A. Lembo, J. Wilkins, & G. Fick. (2009). Mapping potential foodsheds
in New York State: A spatial model for evaluating the capacity to localize food
production. Renewable Agriculture and Food Systems, 24, 72-84.
10