Forest Change in Northern Virginia
1937-1998

Introduction
Description of the Study Area
Data and Methods
     -see animation of land cover change between 1937 and 1998
     -see animation of forest change between 1937 and 1998
Land Cover Change
Forest Loss
Riparian Buffers
Conclusions
References

Introduction

The capacity of humans to alter the natural environment is realized in the modern urban environment, where roads, buildings, and natural processes occur in the same space.  Understanding the interactions between humans and natural processes in built environments is important given the ubiquity of these environments and the alteration of natural cycles that accompanies urban development.  Forman (1999) notes that nature in suburban landscapes is very different from that which occurs in landscapes with a lesser human presence, and has characteristics such as high species richness due to the presence of nonnative exotic species; a predominance of generalist species and a scarcity of specialists; less interior forest area; high inputs of minerals, nutrients, and toxins in the soil; and high levels of soil erosion and nutrient and pollutant runoff.  Forests are also an important element of the cultural landscape in suburban areas, especially on the east coast of the United States (Williams 1989).  This study will begin by assessing land use change in one part of Fairfax County, VA and will specifically focus on patterns of forest loss and preservation.  Based on this assessment, links to land use history, socio-economic history, population growth and conservation policies can then be made.

Study Area

For most of its history, Fairfax County has been dominated by forest.  At the end of the nineteenth century, Fairfax County supported a forestry industry and was a primary source of timber for the small urban center of Washington, D.C. (Hall et al. 1907; Fairfax County Chamber of Commerce 1928).  Although small suburban settlements were common by the early 1900's, the economy of Fairfax County was based on agricultural and forestry products until the middle of the century.  The population boom in the post-World War II years, the growth of the federal government, and the advent of the automobile as the primary mode of personal transportation spurred the suburbanization process, and Fairfax County, along with other surrounding counties, evolved into residential communities for Washington, D.C. (Netherton et al. 1978).  Today, Fairfax County is densely populated and is fully integrated into the regional economic fabric, hosting a variety of economic activities (Knox 1993).  The study area (see Figure 1 below) is located in the Pohick Creek watershed, a few miles south of Fairfax City, just north of the Burke Lake Reservoir and covers an area roughly 4 square miles in size.  It is currently a mixed land use area, although the residential and commercial developments have only occurred in the past three decades, replacing forest and agricultural lands.  Many of the forest preservation policies applied to this area have come about as a result of water quality issues.  Poor water quality and flooding had become serious countywide issues by the 1970's due to the rapid increase in impervious surfaces as the county became more developed (Anderson 1970).  By the early 1980's, the county had implemented best management practices (BMPs) in some of the most threatened watersheds, such as the Occoquan watershed, in order to improve stream water quality.  These BMPs consisted primarily of low-density residential zoning and the creation and/or maintenance of vegetation buffers around streams.  In 1993, these BMPs were established countywide, establishing stream corridors as Resource Protection Areas (RPAs) (Stormwater Management Branch 2001).

Data and Methods

Land use/land cover (LULC) data were derived from a time series of historic air photos for the study area.  Six air photos at a scale of 1:20,000 and a resolution of roughly one meter were acquired that centered on the study area: 1937, 1953, 1962, 1978, 1988, and 1998 .  Each photo was digitally scanned, geo-referenced to a USGS digital orthographic quarter quadrangle (DOQQ), and printed at a large-format scale of 1:5500 to facilitate visual classification.  The classification was performed manually using a drafting film overlay, digitized, and then rasterized to a three-meter pixel size.  An Anderson level I classification scheme was used (Anderson et al. 1976).  The results of the classification are shown in an animation of land cover change between 1937 and 1998.  To model LULC change over time, a Markov transition matrix was derived for each time step (Urban and Wallin 2000).  To specifically address patterns of forest cover, the LULC data were reclassified into two categories: forest and non-forest (see animation of forest change between 1937 and 1998).  These maps were analyzed using RULE (Gardner 1999) and observed  changes in forest patterns were compared to population growth.  Riparian buffers were also analyzed.

Land Cover Change

The results of the Markov transition model (see Figure 2) illustrate rates of land use and land cover change in yearly time steps from 1937-1997.  Although the shapes of these lines are highly dependent on the sequence of air photos used in this case, general trends are easily identifiable.  For example, it is clear that there were low transition rates among the land cover types used in this study between 1937 and 1953.  A decline in agricultural land use begins in 1953, remains relatively constant until 1978, and then rapidly falls to zero by 1998.  Decline in agriculture between 1953 and 1962 is due to an increase in forest cover caused by farm abandonment and an increase in developed land, primarily low density residential.  This trend reflects the economic shift away from agricultural practices and the beginnings of residential development.  Gains in forest made during this time period are lost between 1962 and 1978 as developed land uses increase.  Between 1978 and 1988, the dramatic decrease in forest is mirrored in the increase of developed land.  This "boom" in the 1980's was a regional phenomenon, and has been related to regional economic growth (Masek et al. 2000).  Rates of development and forest loss slow between 1988 and 1998.

The focus on developed land uses in this area is notable.  By 1998, almost 80% of the total area has been developed.  Agriculture has disappeared completely, and almost 20% remains forested.  The remaining 2% of the land area consists of water bodies or "open urban land," which includes recreational facilities such as golf courses and sports fields, as well as power line rights of way.  Although conclusions drawn from these findings are limited by the extent of the chosen study area, they do point to the lack of open space preservation for which Fairfax County is frequently noted in the conservation literature (Environmental Defense Fund 1996; Frankel and Fehr 1997; Noble 1999).  Parkland acquisition is not comprehensive (Division of Historic Preservation 1982), and planning efforts for open space preservation have not been successful (Fairfax County Planning Division 1962).  A large recreational area, Burke Lake Reservoir, is located just south of this area, but very few neighborhood parks or recreational areas exist in this four square mile area.  This has implications for both the biophysical environment as well as the social environment.

Forest Loss

There is little doubt that increasing population is a driver of forest loss (Olorunfemi 1984; Arizpe et al. 1994).  When forest change is observed in a historical context with population, however, this relationship is not as straightforward as it seems at first glance (see Figure 3).  Between 1937 and 1962, for example, modest increases in population are accompanied by an increase in the percentage of forest in the landscape.  Likewise, between 1988 and 1998, it seems that forest loss occurs at a faster rate than population increase.  Although it is clear that the remarkable population increase between 1962 and 1988 is associated with a dramatic decrease in forest cover, it also seems that population increase is not the sole driver of forest loss.  Minimum lot sizes, non-cluster development patterns, and homebuyer preferences may contribute to patterns of forest loss.

There are two more interesting metrics of which to take note.  First, the pattern of change in the number of forest clusters in the landscape is striking (see Figure 5).  The number of clusters decreases between 1937 and 1962 with the period of forest regeneration.  The spike in cluster number that occurs in 1978 coincides with the increase in developed land uses.  This period is actually characterized by a great deal of transitional land and captures the time of maximum land use change.  The number of clusters drops again in 1988 to a level similar to that in 1962, although the proportion of forest is much lower at this time.  The second point is that this landscape never percolates.  A landscape is said to percolate if a habitat patch stretches from map-edge to map-edge.  The percolation frequency is dependent on the proportion of habitat, and critical thresholds have been observed where the percolation frequency suddenly drops.  For the next-nearest neighbor rule, the critical threshold is 0.4073 (Gardner 1999).  The proportion of forest in this landscape does not drop below 0.40 until 1988, yet it is not observed to percolate prior to that date.  This result can be an indicator of the highly fragmented forested landscapes that are produced through human intense interaction with the landscape.

Riparian Forests

In order to investigate the role of conservation policies in determining the spatial arrangement of forests, forested riparian buffers were analyzed.  Like the landscape metrics discussed above, patterns in forest buffers are highly dependent on the proportion of forest in the landscape.  Figure 6 shows the fraction of forest buffer area that is forested for both 50 and 100-meter buffers.  Between 1937 and 1962, increases in the amount of forest contained within the buffers coincides with the increase in the proportion of forest that occurs in the landscape.  Likewise, the decrease in the amount of forest within the buffers that occurs after 1962 is related to the decrease in the total amount of forest.  The forest within 50-meter buffer shows an interesting increase between 1988 and 1998 that would coincide with the conservation policies instituted in Fairfax County (Stormwater Management Branch 2001).  The fact that this trend is not evident in the 100-meter buffer may indicate the scale of conservation.  Streamside forest preservation is illustrated more clearly in Figure 7, which shows the fraction of the total remaining forest that is contained within the buffers.  The 50 and 100-meter buffers show the same trend: preservation of forests along stream is clearly evident after 1978.  Whether or not this trend is the result of active protection of riparian buffers or an avoidance of these areas by developers needs further investigation.

Conclusions

This study investigated patterns of land use and land cover change in a small area in northern Virginia, where population increase and residential and commercial development have drastically altered the landscape.  Forests were specifically investigated, as were forested riparian zones.  Landscape ecological methods were found to be useful in understanding the influence of humans on the environment, and warrant more attention by social scientists studying the processes of land cover change.  The observations of forest change provided in the air photo time series reveals the complex human-environmental interactions that have been at work in this area.  First, it is clear that even moderate human population levels still produced a highly fragmented landscape.  Even when the proportion of forest was high, the level of fragmentation is indicated by the lack of percolation.  Second, the relationship between forest loss and population increase is more complicated than it may seem at first glance.  Metrics such as the largest cluster size, area-weighted mean cluster size, number of clusters, and largest cluster fractal dimension reveal some of the effects of different social processes on the landscape.  Finally, the effects of riparian forest preservation at some scales may be evident.

References