image of forest profile overlain on dollar bills

Home | Mission | Research | Products | Data/Tools | Personnel
Forest Economics and Policy Research Unit

Mill Locations

PURPOSE: These data were assembled as part of a study to define hardwood timber markets in the Southern Appalachians (see the link below). Portions of that study are described in an online presentation below but briefly the initial portion of the analysis required estimating the cost of transporting harvested timber from FIA plots to the nearest mill. The GIS coverage of mill locations was used to quantify direct distance from plot to nearest mill using functions in ArcView (ESRI, Inc.) which were later converted to road distance using a road circuity factor. While our initial study was confined to the Southern Appalachians we found substantial interest in mill information covering a broader area and have expanded our collaborations to encompass the entire eastern US. We intend to offer a more formal metadata description meeting Federal Spatial Data Infrastructure requirements, but hopefully this description will suffice until the more formal documentation is complete.

You can view Eastwide maps generated from these data and an example study using the data at:

You may also be interested in similar datasets describing Southern chip mills: SOURCE and SCOPE: This dataset is based on information collected by Forest Inventory and Analysis (FIA) Units of the Southern Research Station in Asheville, NC (formerly the Southeastern and the Southern Forest Experiment Stations), the Northeastern Research Station in Newtown Square, PA, and the North Central Research Station in St. Paul, MN. Data from the Southern Research Station's FIA Unit did not include information on mills in Texas--this information was obtained directly from the State of Texas. The North Central data were provided in mid-1999, the remainder in late 1998 with some updates since then. While we have attempted to use the best information available, mill data from some states may be several years out of date. Together the data span from Texas and eastern Oklahoma northward through North Dakota and eastward through to Maine and Florida. Note that some of the mills reported by Texas occur in Louisiana.

FIA Units obtain from the various States information on mills to support periodic estimates of wood processing activity. The NE and SO regions and Texas provided us with information on mill type, name, and address. We used a geocoding service provided by ETAK, Inc. (now Tele Atlas North America) to estimate the latitude and longitude for each mill plus identify from this the county and Census tract and block identifier corresponding to that location. Note that many if not most addresses referred to Post Office Box numbers. If the geocoding service could not identify an actual street address, it provided a location based on the zipcode. Therefor points in the coverage can be shared by multiple mills. Put another way, the coverage frequently contains "stacks" of mills with the same location. Data from the North Central Station were obtained more recently and were handled somewhat differently as noted below in the "Lon, Lat" section.

The mill locations are provided in several formats, all unprojected ("geographic projection"): ArcInfo (uncompressed) Export and ArcView Shape files. Each supplier's data are provided as a separate coverage as well as in a single coverage consisting of mills from all four sources.

So what if I don't have ArcView or ArcInfo? Most GIS programs have the ability to import these formats, but GIS programs are expensive. The publisher of ArcView and ArcInfo offers a free downloadable program called ArcExplorer for viewing and analyzing files with these formats.

RECORDS and VARIABLES: The ArcInfo coverage (eastmill.e00) contains minimal information on each mill, represented as points. More detailed information on each mill is available in the dBase IV file "eastmill.dbf." Each record in a coverage corresponds to one mill. Record numbers created for each mill ("mill-id") uniquely identify the mill and match the mill-id variable in the dBase file. Records also contain an identifier for the source which provided it. The included location data represents information on 8,423 mills from the geographic area defined by the North Central, Northeastern, and Southern Research Stations of the U.S. Forest Service, plus all of West Texas. Data reporting quality varies by state, so state-by-state comparisons should be approached with caution. In particular, many mills in Minnesota had no address information and could not be assigned a location, although these mills are still listed in the dBase table. Errors can arise from different sources including errors in mailing addresses and in geocoding. While we've tried to minimize our own procedural errors and corrrect a few of the more obvious errors from other sources, there are doubtless errors still remaining in some of the records, over and above the imprecisions inherent in geocoding in general and on accepting Post Office Box addresses in particular.

ArcView shapefiles begin with several GIS-standard variables describing shape, perimeter, and internal record ID, but these are of limited use in this case. More generally useful are the remaining variables in the shapefiles, and the corresponding fields in the dBase table. Each mill has one row of data, and each row in the dBase file contains fifteen variables (text and numerical). The tabular data can be joined to any of the coverages using the mill-id. Indeed, the shapefiles were generated by joining the dBase file to the ArcInfo point coverage and then subsetting the resulting theme based on region_id (see below). In order, the variables in the table and shapefiles are:

Mill-id Assigned by J. Prestemon, this identifies a unique mill with a number that will never be repeated.

Lon, Lat This is our best estimate of the mill longitude and latitude, respectively. At worst, no coordinates are provided. This was the case for many mills with no address data. Nearly all of the non-addressed mills, however, were located to the county level (hence have FIPS codes--see below). The best estimates of longitude and latitude were GPS coordinates from Missouri, where state officials have reportedly provided the North Central Research Station with GPS coordinates. The next best coordinates, with one exception, were obtained by submitting mill address information to Etak, Inc. (now Tele Atlas N.A.), in January - May, 1999. The quality of the latitude and longitude data when evaluated by Etak, Inc. can be judged by MAT_TYPE, explained below. Finally, the exception to the Etak results was for the North Central Station’s mills. In most cases in that Station, their FIA Unit provided latitude and longitude coordinates to J. Prestemon and J. Pye. These coordinates were those of the town nearest the mill. In cases where Etak match codes were higher than 3, the North Central Station's estimate of the mill latitude and longitude was taken as the best estimate of the mill's location. These are the locations provided in the accompanying data set.

Region_ID Identifiers of who supplied the mill address data. Note that some mills identified as originating from one region were actually located in another region. This occurred where the physical locations of mills were different from the company headquarters of the mill. Regions and their included states were as follows:

Name Name of the forest products business.

Type_New Kind of mill, recategorizations based on identifiers provided by the four regions. Six types were identified:

Town The town corresponding to the mailing address of the physical location of the mill

State The state corresponding to the mailing address of the physical location of the mill

MAT_ZIP The zipcode corresponding to the mailing address of the physical location of the mill

MAT_TYPE Codes include:

0Non-matchAddress could not be matched to any database.
1Block FaceStreet segment exact address match.
2Near matchAddress matched to a single street segment but the exact address number was not found.
3Zip+2 CentroidA point representing the aggregate of all geocoded ZIP+4's for the ZIP+2 of the input address.
45-Digit ZIP CentroidA point representing the aggregate of all geocoded ZIP+4's for the 5-digit ZIP of the input address.
53-Digit ZIP (SCF) CentroidA point representing the aggregate of all geocoded ZIP+4's for the 3-digit ZIP of the input address.
6Ambiguous matchAddress matched to more than one street segment, a centroid of all segments is returned.

Note: The Zip 7, Zip 5, and Zip 3 Centroid is the geographic center of zip codes that share the first three, five, or (in the case of the four-digit zip code extension) seven digits.

Match 0 (nonmatch) do not have latitude and longitudes, except for some cases in the North Central Station states. Match 1 (exact match) is likely to contain precise coordinates, within a city block or so of distance from the exact location. Successively higher numbers indicate less and less precise geographic coordinates. In only 19 cases were matches as poor as code 5. In only 6 cases were Match 6’s encountered, and for even these Etak produced coordinates.

FIPS_Code The format of this code is SSCCC, where SS refers to the state FIPS code and CCC refers to the county FIPS code.

CEN_TRACT This is the census tract code provided by Etak, Inc., when the mill address was geocoded. The code begins with a two-character state abbreviation and is followed by:

  1. a three-digit county FIPS code,
  2. a four-digit census tract code (for which many detailed census data are available)
  3. a two-digit street segment code, which further refines the census tract (combined with the four-digit tract code, the two-digit segment code provides the finest level of disaggregation of U.S. census data)

Street The mailing street or P.O. Box or rural route of the physical location of the mill.

Precise_Type The reporting agency’s classifier of the mill. Contact the Station mill data managers or Texas Forest Service for the code definitions. But the code definitions are fairly self-explanatory and intuitive.

FIA Survey Year The year that the mill address information was recorded. It varies by state.


Entire Eastwide coverage: Regional shapefiles with attributes, in Zip format:

If you only need information on Texas mills, try searching the Directory of Forest Products Industries of Texas.


modified: 4-FEB-2005
created by: John M. Pye
Home | Contents