A geographic information

P E E R R EVI EWE D. The ABCs of GIS An introduction to geographic information systems By Russell G. Congalton and Kass Green A geographic information system (GIS) can be defined as a system for entering, storing, manipulating, analyzing, and displaying geographic or spatial data. These data are represented by points, lines, and polygons (fig. 1) along with their associated attributes (i.e., characteristics of the features that the points, lines, and polygons represent). For example, the points may represent hazardous waste site locations; the attributes associated with each site may be the specific chemical dumped at the site, the owner, and the date the site was last used. Lines may represent roads, streams, or other linear features, while polygons may represent areal feature such as vegetation types or land use. Geographic data can be represented in either of two formats. The first, older format is called raster or grid structure. Raster data are stored in a grid or pixel format that is referenced to some coordinate system (e.g., latitude and longitude). The size of the grid can vary; therefore the spatial resolution of the data is determined by the size of the grid. Raster data are computationally easier to manipulate but typically require larger amounts of storage space. Digital, remotely sensed data (satellite imagery) are a good example of raster or grid data. Geographic data may also be repre- sented by vector or polygon data, which use a series of points (x,y coordinates) to define the boundary of the object of interest. This type of data may require less storage and is preferred for display purposes because it maintains a truer rendition of an object's shape. However, some computations are especially difficult and time-consuming to perform on vector data. Figure 2 shows what an object would look like in both raster and vector formats. Note tha! the raster image has lost some of its true shape due to the gridding process. In this example, a large grid cell size was used to emphasize the difference between the two formats. Recent technological developments and refinements in GIS computer hardware, software, and data acquisition techniques have revolutionized land management and land planning. Land managers, planners, resource managers, engineers, Forest stand boundaries are digitized into a geographic information system; the digitized stand boundaries are then linked to associated database information. Digitizing is a common method of spatial data entry. NOVEMBER 1992 13

POINTS LINES POLYGONS Figure 1. The three elements of geographic data. Figure 2. Depiction of the same shape in raster (left) and vector (right) format. and others can use geographic data more efficiently than ever before to analyze management and policy issues. The GIS link between locations and attributes makes it possible for decisionmakers to simulate the effects of management and policy alternatives. GIS is a potentially powerful tool because knowledgeable users can quickly search, display, analyze, and model spatial information. In addition, maps and other data can be updated more quickly and accurately with GIS than by conventional methods. This article highlights basic concepts and fundamentals, and introduces the reader to many aspects of GIS. Data Map projections and coordinate systems. Spatial data are the "lifeblood" of any GIS: 80--90 percent of the effort and money required to run a GIS is used to acquire, input, update, and manipulate data. Therefore, it is critical that the GIS user have a good understanding of the many aspects of data acquisition and manipulation even before the system is turned on. One important aspect of spatial data is ground registration. Failure to register the spatial database to the ground can cause serious problems in the analysis and assessment stages. Surveyors understand this registration process well. In fact, they call a registered spatial database a multipurpose cadastre. In other disciplines, such as geography and the resource sciences, less attention has generally been placed on precise ground registration. However, failure to consideregistration before beginning any GIS project can lead to serious problems during the later analysis stages. Registering spatial data to the ground requires transforming the original data, recorded in digitizer inches, to some ground-based coordinate system. A d g - tizer is an instrument that permits the electronic transfer of x,y coordinates into a computer by manually tracing a map and then sending an electronic pulse of the coordinate location to the computer A coordinate system is simply the two-d - mensional (x,y) values that designate the position of a given point on the ground Common coordinate systems include latitude-longitude, Universal Transverse Mercator (UTM), and state plane coordi- nates. The procedure for registering spatial data to the ground is quite simple--and would be even simpler if the earth were flat. However, because the earth is round, a projection system must be used to make a map flat. A map projection is an orderly, mathematical system of parallels and meridians that allows us to flatten the earth only at the cost of one or more spatial attributes. Depending on the projection used, the attributes affected could be distance, shape, and/or direction. A good example of this projection problem is the large size of Greenland (much bigger than it really is) on world maps that use a Mercator projection. Registering the data to ground coordinates such as UTM or a state plane is a relatively simple process. However, just because all the data have been registered to the same coordinate system does not mean that it will overlay. Careful attention must be paid to the original map projection as well. Knowledge of map projections and coordinate systems is critical because overlay analysis is such an mportantool in GIS. Failure to considerather factor will greatly complicate the analysis using spatial information (Muehrcke 1986, Snyder 1987, Thompson 1988). In addition, the ability to locate the necessary points on the ground (called ground control points) is vital, or it will be impossible to assess the accuracy of the spatial data layer--which makes the database very unreliable for any analysis or decisionmaking. Sources of spatial data. In a GIS, spatial data are expressed as points, lines, or polygons. The spatial relationship of the points, lines, and polygons to one another is called topology. All landscape features 14 JOURNAL OF FORESTRY

can be reduced to one of these three data types using x,y coordinate pairs. The data can then be entered and stored in a computer as topology for future analysis. Collecting, importing, verifying, and updating spatial data is the most expensive component of any GIS. Knowledge of how each data layer is created is critical to the ultimate economic success of the GIS. Before any new data are collected, an exhaustive search should verify that no substitute data exist. One of the most well-known sources of existing spatml data in the United States is the US Geological Survey (USGS). It provides spatial information in the form of hardcopy maps and also in digital format on computer tapes, which can be readily input into a computerized GIS. The USGS produces and distributes two kinds of spatial information: digital hne graph (DLG) data and digital elevation models (DEM). DLG data consist of vector-format information about a certain characteristic found on the maps. Examples include land use, land cover, transportation, ownership boundaries, and hydrography. These data retain the spatial integrity of the map; in other words, everything is in its proper place in relation- sh p to everything else (i.e., topologically structured). The DLG is also referenced to a geographi coordinate system and can be readily tied to other new or exist- mg spatial information. DLG data is available for most but not all of the United States. Digital elevation models, also called d gital terrain models, are digital files that contain a grid pattern of point elevations that can be used to simulate the topography of an area. These data are useful for generating three-dimensional information about a land area such as slope, aspect, volume, and surface profile. The USGS provides DEM data at two resolutions. A complete database for the entire United States was derived from Defense Mapping Agency (DMA) data for 1 ø x 2 ø (l:250,000-scale) maps. The map is split in half, and a digital file created for each half. The grid interval is 3 arc-seconds or approximately every 100 meters. In addition to the DMA data, USGS also has incomplete coverage of the United States using a grid interval of 30 meters. These higher-resolution data, frequently called DEM data, are available in 7.5-minute blocks. Another excellent source of informa- tion is the US Census Bureau Topologically Integrated Geographic Encoding and Referencing (TIGER) system. TIGER contains all the digital data for the 1990 census features including roads, railroads, rivers, census tracts and blocks, political areas, latitude and longitude, feature names, and classification codes. Other government agencies, state agencies, and even local governments may have good information. If no satisfactory data exist, new data must be collected. In this age of satellite imagery, video cameras, and global positioning satellites, it is easy to overlook the advantages and uses of aerial photography as a source of spatial information. However, this would be a huge mistake. Aerial photography will continue to be an important source of spatial information for a long time to come. From a historical perspective, aerial photography is irreplaceable. Any project that reviews changes over time must rely heavily on aerial photographs. Satellite imagery and other remotely sense data are too new to be of much historical significance. In addition, only aerial photographs provide the accuracy, precision, and detail required for many mapping projects. Off-nadir viewing satellites may be able to generate maps for areas of the world where no topographic data are available, but their accuracy and detail does not approach that of aerial photographs. Most important is the ease of use and simplicity of aerial photography with respecto other remotely sensedata. It can be as simple as using the photograph to record some historical event ("a picture is worth a thousand words") to as complicated as digitizing the photo and entering it into a computerized image processing system. Photographs do have some disadvantages with positional accuracy due to distortion and displacement. However, techniques to correct these problems are available and should be used. Perhaps the most exciting development in the area of sources of spatial data is the ability to geocode and/or terrainprocess digital satellite data (Fischel and Labovitz 1987). In geocoding, remotely sensed data is accurately registered to a ground coordinate system. This process NOVEMBER 1992 15

is possible because we can know the precise location of the satellite in space. Terrain processing refers to removing topographic displacement in the satellite data using DEM data. As a result, digital satellite data (such as Landsat and SPOT) are excellent sources of new spatial information. SPOT digital data can also be used in a very special way. As mentioned above, the geocoding process can compensate for terrain parallax effects. However, because the SPOT satellite has an off-nadir (nonvertical) ability, it can produce stereo imagery. DEMs can be derived by using a SPOT stereo pair and working backward through the terrain parallax effects. So, not only can digital image data be easily added to a GIS, but it can also be a source for one of the most important GIS data layers, elevation. Both these facts have far-reaching implications for our ability to update and revise GIS databases quickly and efficiently. Analysis Techniques Once all the necessary data have been collected, they must be registered to a common base map. Such collection and registration can be expensive, time-consuming, and frustrating; however, GIS techniques such as overlay analysis, modeling, buffering, and network analysis cannot be initiated until this process is completed. Overlay analysis. The ability to analyze spatial data separates true geo- Map Map E V Figure 3. Some examples of Boolean operators used in overlay analysis. graphic information systems from mere spatial databases. In early computeraided design computer-aided mapping (CAD-CAM) packages, data were frequently input such that each layer contained only a single attribute or label. For example, instead of all stream types being in one data layer with various labels, each stream type would be in a separate layer. There would be one layer for major streams, another for intermittent streams, and so on. The appropriate layers could then be chosen to derive the desired map. It is easy to see the problems inherent in such an approach. Each unique data layer had to be derived manually and entered in the database. Deriving some layers could be quite labor-intensive. In addition, the number of layers needed could quickly become unwieldy. Clearly, another approach was needed: enter GIS, which could create a new layer of information/ data from two or more existing layers. Extracting specific information from a data layer and combining it with other information from that or another data layer depends on the use of Boolean algebra, in which the operators AND, OR, XOR, and NOT manipulate spatial data by testing to see if a given condition or statement is true or false (fig. 3). Then data layers can be combined to form a new layer. For example, to find all locations where vegetation type B exists with soil type C, one would simply use the statement B AND C. Map E (fig. 3) would indicate all locations where this statement was true. Overlay analysis can be divided into two general categories: point operations, and neighborhood or region operations Point operations can be anything from the Boolean operators to simple weighting function such as multiplication by some factor. In addition, point operations can involve more complex functions such as clustering, discriminant analysis, principal components analysis, and other statistical techniques. Neighborhood or region operations relate a point to its neighbors to a specified region. This process is much more complicated than point operations and involves the spatial component of the data Measures of correlation and diversity, as well as slope and aspect, are common examples of neighborhood operations. The true utility of a GIS shines in these neighborhood operations. Modeling. This term can mean many different things to many different people, but three aspects of modeling spatial nformation will be discussed. The first has been called cartographlc modeling (Burrough 1986). When the user of spatial information is presented with a problem, a common response is to rush to the computer and start to work The proper response, however, is to prepare a careful plan of what needs to be done. Cartographic modeling suggests detailed flow charts and careful plans to decide which data are important and how they should be used. Another aspect is the simulation approach. In this case, the user tries to simulate some complex phenomenon using a combination of spatial and nonspatial information. This approach typically requires a knowledgeable expert--and rarely do any two experts agree on exactly how the model should be built. A good example is evaluating wildlife habitat suitability. An expert might use the spatialayers that contain information on vegetation, elevation, aspect, slope, ownership, roads, and streams. A model would combine this information with weights that prioritize important layers In addition, calculations of distances (e.g., distance from roads, distance to streams) and measures of diversity may be included. This model is then used to evaluate areas of good habitat and determine where the habitat can be improved In predictive modeling, statistical 16 JOURNAL OF FORESTRY

techniques (usually regression analysis) are used to build a model that will use spatial information to make predictions. The first step is to collect information about the phenomenon investigated. A subset of this information is then used to build the model statistically by looking at each layer of spatial information and each component of nonspatial information to see which are correlated to the phenomenon to be predicted. Once the model is built, it is tested using the remaining information. As an example, suppose one wants to predict the amount of snowmelt runoff from a forested watershed. These predictions currently are being made by point samples taken throughouthe watershed. The predictions can be compared to the actual runoff statistics collected by stream gauges. Using spatial data that completely cover the area should lead to better predictions than point samples. Therefore, a GIS would put together the necessary layers to predict runoff: vegetation, slope, aspect, snow extent, elevation, and soil type. Point sample data on snow depth and the amount of water in the snow might also be included. Some of this information remains constant over time while some changes daily. Therefore, a variety of conditions (i.e., years) should be represented in the information collected. In this example, runoff data would be collected for dry years, wet years, and average years. A subset of this runoff data would then be used to develop the model and selecthe necessary spatial and point sample data. The remainder of the data would then be used to test the model. Once the model has proven effective, it can be used to predict snowmelt runoff in future years. Buffering. Buffering is a technique by which a boundary of known width is drawn around a point or linear feature (fig 4). A point buffer may be a zone around a hazardous waste site or the area around a nesting tree for a particular endangered bird. A linear buffer could delineate an area around a stream where logging is prohibited, or around a utility hne where digging is restricted. Network analysis. In network or corridor analysis (fig. 5), a linear path is identified that represents the flow of some objecthrough the area. Network analysis ts especially useful in hydrology, trans- LINEAR BUFFER POINT BUFFER Figure 4. Buffers are boundaries of known width around a point or linear feature. portation, and other disciplines that study the flow of an object. However, it is also applicable for vehicles, utility and communication lines, and animals. Error Analysis In order to make effective use of any GIS, it is important that both users and suppliers understand the errors associated with spatial information. These can be divided into three groups: user errors, measurement/data errors, and processing errors (Burrough 1986). User errors, probably the most obvious, are errors more directly under the control of the user. Measurement/data errors deal with variability in spatial information and the corresponding accuracy with which it was acquired. Processing errors are those inherent to the techniques used to input, access, and manipulate the spatial information. User errors include data age, scale, coverage, and relevance. Errors result when the information used is out of date. A good example of this problem is using old aerial photography because new photography is not available. Error also results when data of the wrong scale are applied. This situation is especially dangerous when small-scale data are used to meet the objectives of some large-scale project--such as using a statewide soils map to obtain soils information about a particular county, or using digital elevation data from a 1:250,000 USGS map sheet for mapping on a 7.5-minute quadrangle. In both cases, the source data are of insufficient detail to meet the required objectives. In addition, errors result from using data sources that do not completely cover the area of interest. Many times only partial areas have been covered by the latest data and the user must choose Figure 5. Representation of a network analysis---a linear path that represents the flow of some object through the area. between full coverage using out-of-date information or using partial-coverage new data. Finally, errors can be caused by using indirect data layers in the GIS, i.e., layers derived from some primary information. Probably the best example is a vegetation classification generated from satellite data or aerial photography. Included within the measurement/data category are errors associated with variation within the data. Possible causes are instrument error, field error, and natural variation. All instruments used to collect information have limitations in ability and quality. Field error measures the human limitations and carefulness of indi- viduals collecting data in the field. And variation is a fact of life in nature. As much as one may hope, nature will not be pigeon-holed into nice, neat compartments. Although such errors are inevitable, efforts must be made to account for them in using spatial information. Measurement/data error is usually quantified in terms of positional accuracy or accuracy of content. Processing error includes precision, NOVEMBER 1992 17

nterpolation, generalization, data conversion, digitization, and other methodologies performed on the data. A fact often overlooked is that a computer is designed for only a certain level of precision, and going beyond that point results in roundoff error. Almost every introductory course in computers discusses precision and the use of significant digits, yet these factors are often not considered when processing information. Interpolation, extrapolation, and generalization techniques are certainly subjecto error when from vendor to vendor, and software selection depends greatly on the needs of the user. Several important factors should be considered when choosing GIS software. Data input and editing functions. A particular software package may offer rapid data query and retrieval, and powerful analysis functions, and yet have difficult data input and editing functions. Novice users who become discouraged because these basic functions are so diffi- cult will never discover the more ad- Although hardware, software, and data are essential, people are the most important component of GIS nformation is derived about an area from a series of sample points (e.g., generation of a digital terrain model from a series of elevation points). Errors are introduced as data is en- tered into the GIS and when data is scanned or digitized. Also, the way in which the data are stored and used in a GIS may include error. This becomes especially obvious when converting between vector and raster data, such as incorporating digital remotely sensed data into a GIS. Digital satellite data is recorded in pixel (raster or grid) format. Other information in the GIS may be recorded in vector or polygon format. It is clear how error could result from making smooth polygons into grid cells or making grid cells into smooth polygons. Additional problems arise when data layers are combined to perform some type of analysis. It is easy to imagine problems associated with overlaying data layers, with boundaries, and with registration. All these factors and more result in error in the processing stage. Hardware and Software GIS software links the attribute data to the geographic features (represented by points, lines, and polygons and their topology) using a database management system. It also provides the input, editing, and analysis functions. The capabilities and costs of GIS software vary greatly vanced aspects. It is important to find software that handles both the simple and the difficult functions in a highly interactive and easy-to-use manner. Indicators of well-developed software are pulldown menus, help screens, system prompts, and single-character command strokes. Analysis functions. Cartographic analysis tools such as polygon overlays, buffer generation, line and area measurement, and map production are essential to any software package. Modeling capabilities such as network analysis and project simulation may also be important. Some packages perform certain functions better than others. The user should identify the functions most needed for the application and choose the system that best performs them. Flexibility. Flexibility refers to a software package's ability to interface with different computer systems and highlevel programming languages. Obviously, data from a variety of sources should be able to be entered into the system quickly and easily. Data conversion is time-consuming, inefficient, and frustrating and should be avoided if possible. Risk. The risk associated with any vendor depends on several factors: length and type of experience with GIS, number of users, customer satisfaction, training offered, and research and development. Cost. The cost to purchase, operate, and maintain a GIS is most likely the major consideration for potential GIS customers. The initial start-up costs for a GIS have droppe dramatically in the last few years, but they still involve a considerable investment. It is important to note that the cost of good people is as vital as the cost of software and hardware. Database management systems. The two database management systems most commonly used in GIS are hierarchical and relational. In a hierarchical system, data are related by level in a manner similar to a family tree. Data access is rigid and restricted to a series of hierarchical pathways between layers. Most GIS software packages use a hierarchical database management system. A relational database allows the user to relate or compare attributes in different data layers regardless of the structure of the data. This approach is more complex and expensive and is presently found in only a few GIS systems. There are as many possible hardware configurations for GIS as there are users As with software, the hardware should be matched to the user's needs. A GIS is now possible in almost every work setting. Computer sizes range from microcomputers and advanced workstations to miniframe and mainframe computers Size is mainly a function of speed, disk space, random access memory, number of users, types of input/output devices (tape drives, streaming tape, scanners, digitizers, plotters, printers), and cost There always seems to be a tradeoff between buying the right computer and then being able to afford the necessary peripherals, like the dilemma facing us when buying a car. Do we buy a basic model with a big engine, or do we settle for a smaller engine and get the stereo, air conditioner, and sun roof?. Careful consideration of applications is critical to buying the right software and hardware (Guptfil 1988, Parker 1989). People An integral and yet largely forgotten and unnoticed component of a GIS is people. Geographic information systems need people in order to operate. Without well-trained individuals and an adequate staff, it is likely that an investment of thousands of dollars on state-of-the-art equipment and data will be wasted. GIS 18 JOURNAL OF FORESTRY

definitely requires a financial and philosophical commitment of human resources. Continued funding is critical to provide training and database mainte- nance. Although this discussion of staff is brief compared to the discussions data and hardware/software, people are the most important resource. Failure to make strong commitments to the people operating the GIS dooms it to failure. Applications No article on GIS would be complete w thout mentioning some of the applications of spatial data. One important appli cation is wildlife habitat assessment. Spat al data can be used to calculate the home range or territory of a particular species. They can also be used to eliminate areas where a certain animal would rarely or never be found (e.g., slope data could be used to eliminate extremely steep slopes, or elevation data could be used to eliminate elevations above a certain level). In addition, by simply moving a 3 x 3 pixel window over the vegetation map, calculations of juxtaposition and interspersion can be made. Juxtaposition can measure the number of different vegetation types in a certain area, while interspersion measures the value of the edges. Both are very importanto wildlife, since food and cover must be in close proximity in order for the habitat to be useful. Obviously, other proximity analyses are possible, such as distances to water and roads. A wildlife habitat assessment model, typically based on knowledge compiled by one or more wildlife specialists, combines the spatial data, its derivatives, and other information necessary to predict habitat quality. The more important variables are frequently given higher weights. This analysis can determine areas of good, marginal, and poor habitat, providing the first step toward intelligently improving the area. Another application, predicting snowmelt runoff from a watershed, involves a statistical model using regression analysis As in the wildlife example, the same spatial data are useful. However, in this case the regression analysis chooses the important variables instead of the wildlife b ologist. In predicting snowmelt runoff the important parameters are the snowcovered area and the snow-water equiva- DslyOl, MAPPING SYSTEM Fast Accurate Easy-to-Use 50 q9 650 651 Overlay Zoom/Refresh Digitize Edit Save Quit Zoom 2X Pixel= 100 Line= 200 X= 650500 Y= 3878500 DMS TM converts your PC to a DIGITAL STEREO IMAGE WORKSTATION DMS now provides a Rigorous Photogrammetric Solution for 2-D and 3-D Mapping, DEMs, Digital Orthophotos, Terrain Visualization, Digital Mosaics and Image Interpretation from Satellite Images and Scanned Aerial Photographs --for less than $5,000. t R-WEL, Inc. Mapping Software for Small Computers P.O. BOX 6206, Athens, GA 30604 USA ' (706) 353-4166 gfax: (706) 549-4674 Desktop Mapping System and DMS are trademarks of R-WEL, Inc. 1992, R-WEL, Inc. NOVEMBER 1992 19

AUTHORIZED DISTRIBUTOR FOR CADCore/ racer TM Maclean,.o. DATA ANALYSIS EXPERTS. su,mnot, S, ' ß Data Acquisition or Purchase ß Digital or Manual Image Processing ß Hardcopy Map Production or Conversion to GIS Database Layer ß GIS Database Needs Assessment, Establishment and Maintenance ß Map to Digital Conversion of Existing Maps ß Large Format Document Scanning ERDAS GIS and Image Processing Sales With thousands of systems installed worldwide, ERDAS provides applicationsolutions for a multitude of disciplines, including natural resource management. The modular design of ERDAS software allows users to tailor the software for each applications. Programs prompts are in plain English with intelligent defaults ensuring meaningful results the first time. ERDAS IMAGINE advanced GIS offers multiple windows, geographically linked, provide different views of the same area, allowing you to deal with geographic data on your own terms! lent. Other spatial data could include elevation, slope, aspect, vegetation, thermal emittance, near-infrared reflectance, and soils. In addition, data must be collected at specific sample locations on the ground. All these data are tested by the regression analysis to see which are s gnificant in predicting runoff; the best model has the highest correlation coefficient. This model is then used to predict snowmelt runoff over the entire water- shed. These two examples demonstrate different modeling approaches using spatial data. It should not be inferred from these examples that GIS only applies to natural resources. GIS has many applications, from aiding the census to evaluating spotted owl habitat. Those starting to use spatial data as a tool for decisionmaking should review the literature in their particular field and learn as much as possible from past successes and failures. Finally, a warning. A GIS is a very powerful decisionmaking tool. However, it is just a tool. There is a great temptation to lean too heavily on the computer and to stop thinking. There is also a tendency to collect too much spatial data. One more layer of data is not always the answer. Be thoughtful and wary, and GIS will serve you well. ß IMAGINE WITHOUT MEASURES ß An ideal tool for surveying, n ß No special targe ß Integrated compass, YOUR LIFE TAPE AND WHEE ß S te-of-the-a Head Up Display.t-' 'gplay ß Unparallelled a uracy in range a g ß G G Link & GIS Sof a omp i ß RS-232 Interface for auto h S A A A taloggifi. ement Literature Cited BVRROV, P.A. 1986. Principles of geographical nformation systems for land resources assessment Oxford Univ. Press, Oxford, England. 193 p FISCHEL, D., and M. LABOV1TZ. 1987. Questions and answers about geocoded products. EOSAT Corp., Lanham, MD. 7 p. Gvvrm, S.C., ed. 1988. A process for evaluating geographic information systems. US Geol Surv., Reston, VA. Open File Rep. 88-105. MUEHRCKE, I C. 1986. Map use: reading, analysis, interpretation. JP Publ., Madison, WI. PARKER, H.D., ed. 1989. The GIS sourcebook. GIS World, Inc, Fort Collins, CO. SNYDER, J.P. 1987. Map projections--a working manual. US Gov. Print. Off., Washington, DC USGS Prof. Pap. 1395. 3 IOMPSON, M.M. 1988. Maps for America. US Gov Print. Off., Washington, DC. Russell G. Congalton is assistant professor, Department of Natural Resources, Universtty of New Hampshire, Durham; Kass Green s president, Pacific Meridian Resources, Emeryville, California. 20 JOURNAL OF FORESTRY