NCRS - The Changing Midwest Assessment

Data Information

This page contains information about how changes in "Land Cover" and "Forest Characteristics" were calculated. Land cover change was calculated using remotely sensed data according to the method described in Bergen et.al., 2002. Changes in Forest Characteristics were calculated using Forest Inventory and Analysis (FIA) data. See [http://www.ncrs.fs.fed.us/4801/overview/sld001.htm] for a general overview of the FIA Program, and Miles et al., 2001 for a technical description of the FIA Database and sampling and estimation procedures.

We combined these discussions because of the inherent confusion that surrounds "land cover" and "land use" data. There are three factors that contribute to this confusion. First, the difference between "land cover" and "land use" is often not explicitly stated. Further, land cover and land use vocabulary overlap, or are sufficiently similar that people use them interchangeably. Finally, unrealistic expectation often lead to the misinterpretation and misapplication of land cover and land use data.

What is the Difference Between Land Cover and Land Use

Much of the confusion surrounding land cover and land use data is scale related. Land cover data provide a general description of the condition of the landscape, classifying parcels of land as, for example, Agriculture, Forestland, Urban, Wetland, Water, or Barren. Land use data provide a more specific description of the condition of the landscape, classifying the Forestland cover type, for example, as "Deciduous, Evergreen, or Mixed Forest," each of which can be further subdivided by "Forest Type Group, Forest Type, and stand size class." Land use data provide us with "stand, plot, and tree" level information.

Land Cover and Land Use Vocabulary

In addition to the issue of scale, semantics contribute to the confusion between land cover and land use, particularly when it comes to "forestland." For example, Forest Land (two words) is a "land cover" type described by both Anderson et. al. (1976) and the United States Geological Survey (see http://landcover.usgs.gov/classes.html). Forest and Forestland (one word) are used by FIA (see Miles et. Al., 2001) and others to describe a type of "land use". Throughout this web site, "Forestland" is considered a "land cover" type, whereas "forest" is considered a type of "land use."

What is Forestland?

"Forestland" is defined as a cover type that is characterized by: 1) tree cover sufficient to produce timber or other wood products, and 2) an aerial density (crown closure) of at least 25 percent (Anderson et. Al., 1976). However, by definition, "Forestland" also includes certain lands that have a canopy closure of less than 10 percent, including those that have been clear-cut, so long as they have not been "developed" for other uses. In other words, "Forestland" can be used to describe land that is completely forested as well as land that is completely void of trees. Because "Forestland" is measured using a combination of aerial photography and remotely sensed data, it makes good "methodological" sense to base the definition on "aerial density." However, in the spirit of providing a non-technical definition that the general public can relate to, we defined forestland conceptually as land that is suitable for growing trees.

What is a Forest?

"Forest" is defined as a tract of land with a minimum stocking level of 10 percent of live forest trees of any size. Further, the tract must be at least 1 acre, and at least 120 feet wide. To put this in perspective, consider that 15 seedling/saplings per acre constitutes a "minimally stocked" forest.
The term "forest" can also be use to describe land that formerly had such tree cover, so long as it has not been developed for nonforest use. In other words, as was the case with "Forestland," the term "forest" can be used to describe land that is completely forested and land that is completely void of trees.

Interpreting Land cover and Land use Data

Finally, in addition to the scale related and semantic confusion surrounding "Forestland" as a land cover type and "forest" as a type of land use, people often interpret and apply these data inappropriately. Land cover data provide an excellent snapshot of the spatial distribution of each of the major land cover types within the 7-Midwestern states that comprise the North Central Region. Secondarily, they provide a rough estimate of the area of land within each land cover type. Conversely, land use data provide extremely reliable estimates of the area and volume of forests by forest type group, tree size class, and ownership class.

Land Cover

For a technical description of the data we used and how we calculated land cover change, see:

Bergen, K.M., Brown, D.G., Rutherford, J.R., and Gustafson, E.J. 2002. Development of a method for remote sensing of land-cover change 1980-2000 in the USFS North Central Region using heterogeneous USGS LUDA and NOAA AVHRR 1 km Data. Proceedings, International Geoscience and Remote Sensing Symposium, Toronto, CA, 2002. HTML or PDF

In addition to the methodological discussion in Bergen et.al., 2000, it should be noted that land cover classes used for labeling AVHRR pixels were defined as:

These definitions are based on: Anderson, J.R., E.E. Hardy, J.T. Roach, and R.E. Witmer. 1976. Land Use and Land Cover Classification System for Use with Remote Sensor Data, Geological Survey Professional Paper HTML or PDF

Data Reliability

Finally, it should also be noted that the accuracy of our analysis was better than 80 percent, Region-wide, in comparison with air photo interpretation. For those who are not familiar with land cover accuracy standards, 80 percent represents a very reliable estimate. While we are confident in this product, we recognize that the context in which it is presented creates the possibility for comparisons that could lead to confusion. Specifically, we understand and expect that people will compare the land cover and land use data presented on this web site, and in so doing that they may discover what appear to be contradictory findings. For example, the "Landscape Change" slide show indicates that there was a 25.3 percent decrease in "Forestland" in Illinois between 1980 and 2000; the "Area of Forestland" slide show indicates the area of "forest" in Illinois increased by 16.1 percent.

This seemingly contradictory condition is largely the result of comparing "apples" (land cover) with "oranges" (land use). In measuring land cover we used aerial photography and satellite imagery to identify land cover types (i.e Agriculture, Forestland, Urban) that were dominant at a resolution of 1 kilometer. Conversely, in measuring land use (i.e. "forest") we used on-the-ground field-sampling techniques to collect information about the type, size, and condition of stands, plots and individual trees. To put this difference in perspective, consider that a 1km2 parcel of land is equivalent to 247 acres. Further, consider that a canopy closure of at least 25 percent is generally necessary for "Forestland" to be identified as the dominant land cover type. In other words, the satellite will not "see" Forestland as the dominant cover type unless forests are in blocks of at least 60 acres (e.g. 25 percent of the area of a 1 km2 parcel). This is "problematic" in Illinois because the forests are relatively small - many being less than 40 acres - which is only 15 percent of a 1-kilometer parcel. Therefore, in Illinois, much of what a person without the benefit of an aerial view would call "Forestland" was classified as Agriculture or Urban. Hence, while the "Forestland" and "forest" data we reported for Illinois are counterintuitive, they are both reliable. "Forestland" and "forest" data for Illinois should be interpreted as follows: the amount of "Forestland" identified as dominant at a resolution of 1 kilometer decreased - in spite of the fact that the total area of "forest" increased - because the average size of individual forests decreased.

Given that land cover data provide information at a scale of 1 kilometer and land use data provide tree level information, an obvious question is why would we use satellite imagery? There are at least three advantages to using satellite imagery to complement the work that we do "on-the-ground." First, it is more efficient. Further, satellite imagery provides and excellent snapshot of how each of the major land cover types is distributed across the landscape - something that cannot be done with FIA data. Finally, satellite imagery provides us with a rough estimate of the area of land within each cover type - FIA data provide estimates of "forest" that are more reliable, but they do not allow us to estimate the area of land in nonforestland cover types or nonforest land uses. Simply put, they are different tools that, when used appropriately, allow us to tell a more complete story of how the landscape of the North Central Region is changing. For more detailed information about the extent and condition of "forests" in the Region see the "Forest Characteristics" portion of this site.

Forest Characteristics

We used USDA Forest Service Forest Inventory and Analysis periodic inventory data (approximately 1980 and 2000) to calculate changes in Forest Characteristics:

For a complete description of the FIA data base and sampling and estimation procedures, see:

Miles, P.D., Brand, G.J., Alerich, C.L., Bednar, L.F., Woudenberg, S.W., Glover, J.F., And Ezzell, E.N. 2001. The Forest Inventory and Analysis Database: Database Description and Users Manual Version 1.0

In addition to the methodological discussion in Miles et.al, 2001, it should be noted that the Forest Service Handbook (FSH 4809.11) mandates that the sampling error for Area of Forestland cannot exceed 3 percent per 1 million acres of timberland. Hence, we are confident in the reliability of our estimates for large sampling areas. For example, the maximum allowable sampling error at the Regional level is 0.34 percent:

Maximum allowable sampling error for the Region = (.03 X (1,000,000) 0.5)/ (76,000,000) 0.5)).

However, sampling error increases as sampling area decreases. For example, the maximum allowable sampling error in Illinois (a relatively lightly forested state) is 1.4 percent. The following link depicts maximum allowable sampling error on a county-by-county basis for 1980 and 2000, Sampling Error Maps

Get Acrobat Reader to free to read doucments on this page.