One of the biggest problems with long term ecosystem monitoring is the analysis of the data for use in management decison making. A land condition trend analysis (LCTA) package has been developed for the Camp W. G. Williams Army National Guard installation in Utah which contains a set of tools written in Avenue. These tools analyze vegetation and land characteristics such as cover, bare ground, canopy structure, fuel load, and species distributions. The tools perform the analyses using tabular data (vegetation transect data, floristic survey, and fuel inventory) in combination with spatial data (environmental, military, utility, and cultural layers) to produce tables, charts and statistical data. This package is being used to help close the gap between monitoring and management. It helps managers to analyze survey data, create reports of trends and conditions, and make informed management decisions from the monitoring data.
The Integrated Training Area Management (ITAM) program was developed by the U.S. Army Construction Engineering Research Laboratory (USACERL) to meet environmental compliance and maintain or improve the health of military lands while meeting the overall training mission of the military. ITAM consists of five components: (1) a standard Land Condition-Trend Analysis (LCTA) monitoring program, (2) an Environmental Awareness (EA) program, (3) a Land Rehabilitation and Maintenance (LRAM) program, (4) a Geographic Information System (GIS), and (5) Training Requirements Integration (TRI). This paper will concentrate on analysis of LCTA data.
The LCTA program is the Army National Guard (ARNG) standard for land inventory and monitoring. The monitoring program should accomplish two fundamental objectives: (1) provide a baseline against which land managers and decision makers can compare future inventories as they continue to monitor land conditions; and (2) evaluate the effectiveness of management activities. To date, the first of these has been the focus of large comprehensive natural resources monitoring programs. However, it is the second objective that will take monitoring beyond simply a requirement of environmental compliance and make it a part of adaptive resource management in support of the military training mission.
Elements of the LCTA program include a floristic survey and vegetation transects, breeding bird surveys, and small mammal trappings that are performed annually. Vegetation monitoring include 100 meter line, aerial and belt transects. Line transects measure ground data such as bare ground, litter, and species types as well as ground disturbance every meter. Aerial transects measure canopy structure by recording species intercepts straight up at every meter along the transect. Belt transects record the location and height of each woody plant in a six meter swath about the center line of the transect. Data collection methods and naming conventions are standardized to facilitate analysis.
However, land managers at military installations have had difficulty analyzing the data collected in the LCTA program. This is due to several factors. First, the data is collected in large databases either in SQL or Dbase. Data analysis quickly becomes overwhelming becuase of the great spatial extent of most military installations and the large amounts of tabular information accumulated in environmental monitoring. Not only has the sheer size of some of the databases been prohibitive for searching and analyses, but also the repetitive and cumbersome calculations must be done by hand. Land managers are generally persons with training in soils, ecology, fisheries and wildlife, botany, or natural resources management. They do not normally have training in computer programming or database management and querying. A package to help them run ecological analyses thus must be user-friendly and geared toward the unsophisticated computer user. Trend analysis should also have the flexibility to take advantage of spatial relationships; military land managers often need to inspect environmental or land use trends in certain training areas.
ArcView provides a user-friendly environment that handles large spatial and tabular data. Customization of the interface allows complicated analyses to be performed on LCTA tables while being invisible to the user. Charts and maps facilitates report making and analysis of trends. The LCTA analysis package created in ArcView contains more than 60 Avenue scripts with over 6000 lines of code that enable the user to look at trends in vegetation, land condition and land use, run searches on small mammal trapping data, and analyze diversity of vegetation and breeding birds from field data.
Camp Williams occupies 25,000 acres on the southern end of Salt Lake Valley in north central Utah. The camp is situated on the western portion of the Traverse Mountains, which range in elevation from approximately 4200-7200 feet above sea level. Vegetation on and around the camp is dominated by oakbrush and tall shrubs at upper elevations, and sagebrush/grass communities at lower elevations. The temperate desert climate at Camp Williams averages 11 inches of precipitation annually with most falling as snow in the winter.
A suite of Avenue programs allow the user to run a set of analyses and searches on vegetation transect, breeding bird transect, and small mammal trapping data.
A plant list table within the data set contains species codes (which are used in the vegetation transect tables), species, genus, family, longevity, growth form, common name, and origin (native or introduced) for all plants found on the installation. In this way, a set of vegetation species can be analyzed by selecting all species codes that are, for example, in the same family. All the vegetation analysis programs use this method.
In the modified GUI, an analysis menu allows the user to choose from a set of analyses: cover index, frequency of occurrence, diversity and evenness, campwide searching, ground disturbance, and certain special analyses.
Message boxes query the user as to the type of a species or set of species that they want to search or analyze. Finally, a list of the possible choices is gathered by the program from the plant list table and listed in a message box. For searches, the search is run on all vegetation transects, breeding bird transects, or small mammal traps. For analyses, the user selects a set of vegetation or breeding bird transects for running the analysis.
The cover index is calculated by counting each hit in the aerial vegetation transect for the species or set of species selected. In this aerial table, vegetation is recorded by a 4 letter species code (usually the first two letters of the genus and the species). For cheat grass (Bromus tectorum), the species code is BRTE. So, if the vegetation information at 1.5 meters on the transect had species codes and heights that were: .1 BRTE, .4 BRTE, and .7 ARTR, and we were searching for BRTE, then the count for this point on the transect would be 2. All hits at each point on the transect are added up to create the cover index for that plot. For a check of annual species change, the program gathers the codes for all species listed as annuals in a plant list table, puts them into a list, and then checks for each species in the aerial transect table, which is linked to the feature attribute table for the LCTA points coverage.
To check the presence of annual species on plots in the impact area, under the Analysis Menu, Cover Index is selected, then Longevity, then Annual. From the chart, we can see that annuals have increased slightly over the last three years.
Changes seen in trends are often hard to interpret. To help determine if changes are meaningful, significance tests have been included in the cover index programs to evaluate the statistical significance of the change in cover index. Knowing the statistical significance of change can help make management decisions and justify management action by adding credibility to the analysis of the user's data.
Significance tests for cover index are done using a paired t-test. The paired t-test allows a comparison of each LCTA transect by year; thus significance is based on the difference for each transect for each year. The null hypothesis is assumed to be no change, or xbar = 0. The alternative hypothesis is that xbar <> 0, since change can be either an increase or a decrease in cover index. Therefore, the test is a two-tailed test.
The user can also change the alpha level from the Analysis Menu under the Set Alpha Level option, and the t-distribution table is included so the user can examine the significance level.
Frequency is calculated by counting the presence of a species or a member of a set of species at each point on the vegetation transect. For example, at 1.5 meters on the transect, the species and heights may be: .1 BRTE, .4 BRTE, .7 ARTR. If we were searching for BRTE, then the count for this point on the transect would be 1. All points with hits on the transect are added up to create the frequency for that plot. The maximum value for frequency for a single transect is 100 (meaning that the species was found at every point on the transect). This differs from cover index which is calculated by adding all the hits at a particular point on the transect.
To get the change for introduced species, from the Analysis menu, Frequency of Occurrence is selected, then Native/Nonnative, then Introduced. From the charts, we can see that introduced species are slowly but steadily increasing in the impact area.
These programs calculate the Shannon index of diversity on bird transects. The Shannon index is based on the proportional abundances of species. It takes both evenness and species richness into account. No assumptions are made about the shape of the underlying species abundance, so it is referred to as a non-parametric index.
The Shannon index is an information theory index which is based on the
rationale that the diversity, or information, in a natural system can be
measured in a similar way to the information contained in a code or message.
It assumes that individuals are randomly sampled from an
indefinitely large
population.
The index also assumes that all species are represented in the sample
(Magurran, 1988).
A summary table is produced that calculates bird diversity and evenness by
habitat type and year.
Campwide searches allow the user to select a year to search for a plant, set of plants, bird, or small mammal. The program checks all tables associated with vegetation transects, breeding bird transects, or small mammal traps for that year, then links the table with the feature attribute table for the coverage so that the sites where the species or set of species were found are highlighted.
In this example, under the Analysis menu, Campwide Distribution is selected, then Species Code Plant, then HEAN, a sunflower species and indicator of disturbance, then 1993. From this type of search, the user can then further analyze these areas to determine if disturbance did exist in 1993, and if so, whether there have been any other possible concerns on these sites.
The ground disturbance programs analyze data collected at vegetation transects on water erosion, wind erosion, military use, non-military use, maintenance, and ground disturbance. These programs gather the data from the tables and summarize and display the data in charts for a selected set of transects. This chart shows observed water erosion for a set of vegetation transects.
To provide this system to users, a number of steps were required. For example, the database tables had to be converted to a different format. Tables with many columns had much slower searching speeds, so they were split into multiple tables based on data type. Also, searching became difficult with large input fields, so Avenue scripts were written to convert the tables from their original form into single character true/false fields whenever possible.
Although multiple programs are used to access the different types of analyses and searches, functions that were needed by more than one program were generalized and consolidated into a single program that is called by many programs. For example, many of the programs provide a list of options to choose from, such as small mammals, plant family names, or bird and plant species codes. These lists are generated from a list of all species (plants, birds, or small mammals) found at the installation. So, a single program was written that accepts the name of a virtual table and a field from the calling program and then creates a unique list that is returned when the program closes, as seen in the code below.
The result is that the programs that are called by the menu and selection message boxes are mostly a series of calls to other generic programs. The following is the program that is used for calculating the cover index for a common name of a plant. Notice that the database tables are not part of the project. This makes the project smaller and removes the need to ensure that tables are saved as part of the project, saving time in both project creation and testing. Database tables can simply be stored on disk and brought into the project as virtual tables when needed.
This suite of programs is expected to run in both the UNIX and DOS environments. This product was developed on a UNIX system. Charts that are created by the various reporting sections of the programs were originally over-written each time the program was run. However, when the project was transferred to the DOS environment, the programs, project, ArcView itself, and occasionally the entire system would crash. The problem was that DOS would not release the memory space being used by the chart as long as the chart existed in the project. The original fix was to check for the chart at the beginning of each program, and delete it and flush memory if it existed. However, this did not allow the user to compare similar reports from different sites. So the final fix solved this problem by creating a different chart with its own name each time the program is run.
This suite of programs requires a number of calls and returns between programs. However, the breakpoint in ArcView does not function when the program being tested is called by another script. So message boxes reporting variables and providing possible exits were very useful. In general, message boxes with exit loops were best for debugging, and were heavily used during development.
While the analyses discussed are common to installations nation-wide, there are certain needs that may be specific on a regional or installation level. For example, other data analyses have been included for Camp Williams. Special analyses include a fuel inventory for fire behavior models, noxious weed search, and precipitation summaries. Data for special analyses can come from many different sources. The fuel inventory was added on to the LCTA transects during the first year of monitoring. The noxious weed list came from the State of Utah, and the precipitation data came from several weather stations around the camp. The GUI can also be customized to include these regional or installation specific analyses.
This analysis application allows for easier analysis of LCTA data. It also makes it possible for military land managers to use their monitoring data to initiate management action. The LCTA analysis application is one part of an overall CD package developed for the National Guard. This overall package includes standard data naming conventions and structure, programming, GUI, views, and documents. It also represents special regional or installation needs with the inclusion of additional specific programs and views. This package will be encorporated for 54 National Guard installations nation-wide and is already complete for several of them.
Funding for this project was provided by the Department of Defense National Guard Bureau (NGB-ARE/ARO) through a contract (Research and Development for a Nation-Wide Decision Support System Using GIS and Remote Sensing) with Allan Falconer at the USU Department of Geography and Earth Resources and through the Camp Williams contract with John Crane.
The majority of the programming for the Analysis Package was done by Tom Van Niel (Camp Williams) and Kimberly Patraw (USU / N.E.D.). Additional programming was also done by Brian Biggs (USU / N.E.D.), Joel Godfrey (USU / Camp Williams), Eric Olsen (USU / N.E.D.), and Burk Royer (USU / N.E.D.).
The functional design of the these tools was aided by consultations with Joel Godfrey (USU / Camp Williams), Doug Johnson (Camp Williams), Jim Long (USU / Camp Williams), Nick Nydegger (Orchard Training Area), Dana Quinney (Orchard Training Area), and Leila Schultz (USU / Camp Williams).
The installation database and primary data layers for Camp Williams were constructed by Tom Van Niel (Camp Williams). Additional layers were obtained, where available, from USGS, DMA, GAP Analysis, the state Department of Natural Resources (or its equivalent), Esri ArcUSA, and other sources. Quality control was done by Eric Olsen.
The authors would like to thank Burk Royer, an undergraduate computer science student who joined the project late, but has already contributed significantly to the effort.
Magurran, Anne E., 1988. Ecological Diversity and Its Measurement. Princeton University Press, New Jersey.
Tom Van Niel, staff remote sensing and GIS specialist
email: ned4@nr.usu.edu.
work: (801) 797-0741
Jim Long, professor
email: fakpb@cc.usu.edu
work: (801) 797-2574
Department of Forest Resources
Utah State University
Logan, UT 84322-5215
fax: (801) 797-4040
Kimberly Patraw, graduate student
email: kimp@nr.usu.edu
Allan Falconer, professor and department head
email: als@nr.usu.edu
Department of Geography and Earth Resources
Utah State University
Logan, UT 84322-5240
fax: (801) 797-4048
work: (801) 797-1790
John Crane, Environmental Director Utah National Guard
Utah National Guard
Environmental Department
Draper, Utah 84020-1776
work: (801) 576-3960