COOPERATIVE STUDY PROPOSAL
Approved by the Inyo Los Angeles Standing Committee March 23, 2000.
Project Title:
APPLICATION OF CANONICAL COMMUNITY ORDINATION (CANOCO) TO ASSESS OWENS VALLEY VEGETATION CHANGE
Principal Investigators:
Inyo County Water Dept. - Sally Manning
Los Angeles Dept. of Water and Power - Dave Martin
Purpose:
The long term effects of fluctuating water tables on Owens Valley plant communities are not known. Although much is known about the short term response of Owens Valley phreatophytes to groundwater pumping, the cumulative effects of pumping over the long term could result in changes at the vegetation community level that would be contrary to the water agreement=s goal of avoiding changes in vegetation. Over the past decade, the Technical Group has collected a vegetation data set that contains information on species abundances. In addition, several environmental data sets have become available in recent years, including the NRCS soils data base, precipitation data, and depth to water estimations. Not only have data become available, but appropriate tools for reducing the complexity of large data sets have recently become widely available. Multivariate data analysis techniques, for example, Canonical Correspondence Analysis (CCA, introduced by ter Braak 1986), provide the means to analyze the vegetation data in conjunction with the environmental influences. A recently released Windows version of CANOCO (Canonical Community Ordination) contains CCA and other multivariate algorithms that can be used to analyze the Owens Valley data set. These techniques are designed to quantify the influence of natural and management-induced environmental characteristics on the measured vegetation conditions. By applying these analyses, the Technical Group will be better able to understand the relationship between environmental variables and vegetation change, the rates of change, and the predisposing conditions that are likely to result in significant long term adverse conditions. In addition, these methods may discern the early-warning signs of change, thus allowing managers to respond and correct problems before they require to costly mitigation.
Background:
Typical plot-based data, like the data collected in vegetation inventories performed in the Owens Valley, consist of lists of species and their respective abundances. To date, these data have been analyzed by tallying the total cover or composition of species groupings of interest, for example, total percent perennial cover or total weed cover. Obviously, much more information exist in the data than these simple measurements. Logical means of extracting additional information from these data include sorting the species into groupings that have ecological meaning (for example, based on species geographical distributions) or sorting plots into groupings that allow for better application of management practices (for example, accounting for the various soil types). Sorting these plot data according to similarity can be performed subjectively, but for improved repeatability and a more complete understanding of criteria used, sorting could be performed using accepted statistical techniques. Sorting typically requires organizational and mathematical procedures that are best performed on a computer.
It has long been recognized that species respond to environmental gradients in a non-linear fashion. For example, the cover of saltgrass (Distichlis spicata) typically increases as soil moisture increases from dry to wet. However, as soil moisture approaches saturation, saltgrass cover may decline. In this example, saltgrass is lost from plots when soil conditions become too dry or too wet (and it is presumably replaced by other species better adapted to those conditions). An optimal level of soil moisture could theoretically be measured which maintains the highest cover of saltgrass. The shape of the saltgrass abundance curve along a soil moisture gradient, thus, would be unimodal. The typical model applied to this type of species response curve is the Gaussian, or normal, curve.
Realistically, however, species are not affected by one easy-to-quantify environmental variable. Using the above example, saltgrass may also respond to concentration of a particular salt in the soil in a similar unimodal fashion. If soil salinity and soil moisture are not linearly related, then the saltgrass cover at a site would be the result of the interaction of the two environmental factors. Add to this, the prevailing temperature regime, the timing of exposure to moisture, herbivory influences and other disturbance patterns, competition effects from other species, and any other single or combination of potential environmental influences, and one begins to appreciate the difficulty of understanding why a particular plant exists at a site.
For this reason, ecologists often rely on ordination techniques to simplify the multi-species plot data to arrange species or plots along one to a few ordination axes or into clusters. Often, when environmental variables are known or assumed, the axes or clusters are interpreted in light of what is known about the environmental characteristics that might be causing the groups to sort the way they do. With older ordination techniques, the meaning of the axes was inferred ex post facto by trial and error with correlations with suspected explanatory variables, a process termed Aindirect gradient analysis.@
The development of Adirect gradient analysis,@ which constrains ordination output to environmental variables that are input simultaneously with the plot data were a welcome solution to the limitations of classical ordination. ter Braak (1986) developed the canonical correspondence analysis (CCA) method where the ordination axes chosen to array plot data were restricted to be linear combinations of environmental variables. By using this technique, ordination output shows patterns that are directly related to the environmental conditions being examined.
The computer program CANOCO (Canonical Community Ordination) is a collection of ordination methods that was developed to allow ecologists to analyze species and environmental data simultaneously (ter Braak 1994). It includes principal components analysis (PCA), redundancy analysis, and canonical correspondence analysis (CCA). The advantages of using the CANOCO algorithms are that it: performs ordination of plot and species data; arranges the ordination (output) according to the environmental variables that accompany the plot data; accounts for both continuous or discrete environmental variables; and calculates regression coefficients relating the plots to the supplied (input) environmental conditions.
CANOCO can be used to investigate a variety of ecological questions (ter Braak 1987). It is frequently used in community classification attempts in which it relates plant and/or animal assemblages to environmental variables (Lyon and Sagers 1998; Witkowski and O=Connor 1996). As a corollary to the determination of natural assemblages, CANOCO has also been used in the reverse sense to reconstruct past environmental conditions based on species abundance data (Janssen and Birks 1994; Wiemann et al 1998) or, when environmental variables and gradients are known over a large region, to produce predictive maps of community or species locations (Weiss et al 1996). Furthermore, CANOCO can be used to provide insight into the impact of natural or anthropogenic disturbances on natural communities. For example Sada and Nachlinger (1996) reduced the species in their plots to a smaller number of biological indicators, to which they assigned weights based on known ecological factors of the species in their plots. In this way, the arrangement of the plots in relation to known disturbances showed changes in ecological significance with increased disturbance. Other researchers have applied CANOCO or related techniques to analyze vegetation change following a disturbance (Coppedge et al 1998; Laine et al 1995; Stromberg et al 1996; ter Braak and Wiertz 1994; Weixelman et al 1997)
Some researchers, however, find that the output obtained using canonical correspondence analysis can be misleading if attention is not paid to certain assumptions (McCune 1997) and data input arrangements (Oksanen and Minchin 1997; Tausch et al. 1995). Therefore, it is important to understand the intricacies of the CANOCO program before output validation.
Procedures:
General Approach
Canonical Correspondence Analysis (CCA) is a powerful algorithm contained in the CANOCO program. CCA performs ordination of plot data (in our case, parcel summaries, for example). The input data consist of an environmental data set, in which variables may be continuous or discrete, and a Aresponse@ data set, such as the plot data summarized by species and percent cover of each species. The ordination axes along which the plots are arranged in the output is therefore constrained by the supplied environmental variables. Regression may be applied to the output to quantitatively estimate the contribution of a particular environmental variable to the resulting plot condition.
In this study, I propose to use CANOCO to investigate at least three lines of inquiry. Different arrangements of the data are required to address each question. The answers are expected to ultimately assist in the development of a more comprehensive model of vegetation status and change given various environmental circumstances. The areas of investigation are as follows:
1) To what degree are measurable environmental gradients responsible for the distribution of different plant species and communities in the Owens Valley?
Potential environmental data sets to employ: baseline depth to water, soil salinity, soil water holding capacity, microrelief, landform, long term average precipitation
Possible response data sets: average parcel cover or relative cover by species, with species grouped by life form, and/ or with species grouped in other ecologically meaningful ways
2) In what ways, if any, do plant communities change when subjected to withdrawal then recovery of the water table?
Potential environmental data sets to employ: duration of water table absence, short term precipitation conditions, initial plant community type, soil type variables
Possible response data sets: changes in species composition, changes in cover, changes in life forms, changes in species indicative of certain environments (for example, change in wetland indicator status or forage desirability)
3) Using multivariate techniques and GIS, can a map be generated showing predicted vegetation conditions? Could this be done based on past, present and future environmental conditions?
Potential environmental data sets to employ: land use history data, soils, hydrologic conditions and/or management, actual or hypothetical weather conditions, results of aerial photography study
Possible response data sets: plant community Acentroid@ values, data on rates of change from one community type to another, or rates of change in particular species
Implementation
Application of CANOCO or other multivariate analyses to the Owens Valley vegetation data is likely to become a routine procedure. Therefore, it is desired to develop in-house expertise in their theory, application, and limitations. Arrangements have tentatively been made with a scientific consultant/advisor who is experienced in the use and application of CANOCO.
The availability of this consultant is somewhat limited. He will not be accessible on a daily basis; rather, the PI=s will schedule formal in-person appointments with him to review the work and discuss future procedures. For this study, Inyo and LA=s PIs will collaborate on development of environmental and response data sets. A relatively small data set will be selected for a preliminary analysis. Upon completion of the CANOCO procedure and a written report, an appointment will be scheduled with the consultant. It is expected the consultant would review the analysis and interpretations and suggest strategies for investigating the questions listed above, given the constraints of the Owens Valley data sets. After incorporating all concerns, a larger data set would then be processed and analyzed to the point where sufficient information would exist for a potential scientific publication (for example, sufficient information to address one of the three basic questions). The consultant would again review the resulting manuscript. All written materials will be presented to the Technical Group for review. Oral progress reports will be made at Technical Group meetings.
Schedule:
| Activity |
2000 spring |
2000 summer |
2000 fall |
2001 winter |
2001 spring |
2001 summer |
develop environmental and response data sets |
X |
|
|
|
|
|
| run preliminary data trials and prepare report |
X |
|
|
|
|
|
| Meet with advisor, revise report and submit to Tech. Group |
X |
X |
|
|
||
| design and proceed with broad investigation and prepare report |
|
X |
X |
|
||
| Advisor/ Tech. Group review of report |
|
X |
X |
Products:
The primary product of this project is to develop in-house expertise in the use of powerful analytical tools for analyzing complex vegetation and environmental data. It is anticipated that these tools will be applied to future analyses of Owens Valley data, and they may serve to detect adverse vegetation trends and changes that are not detected by simple statistical methods. All reports of such analyses will be made available to the Technical Group.
References Cited:
Coppedge, B. R., D. M. Engle, C. S. Toepfer, and J. H. Shaw. 1998. Effects of seasonal fire, bison grazing and climate variation on tallgrass prairie vegetation. Plant Ecology. 139(2): 235-246.
Janssen, C. R. and H. J. B. Birks. 1994. Recurrent groups of pollen types in time. Review of Palaeobotany and Palynology 82: 165-173.
Laine J., H. Vasander, and R. Laiho. 1995. Long-term effects of water level drawdown on the vegetation of drained pine mires in southern Finland. Journal of Applied Ecology 32(4): 785-802.
Lyon, J. and C. L. Sagers. 1998. Structure of herbaceous plant assemblages in a forested riparian landscape. Plant Ecology 138(1): 1-16.
McCune, Bruce. 1997. Influence of noisy environmental data on canonical correspondence analysis. Ecology. 78(8): 2617-2623.
Oksanen, Jari and Peter R. Minchin. 1997. Instability of ordination results under changes in input data order: Explanations and remedies. Journal of Vegetation Science 8(3): 447-454.
Sada, Donald W. and Janet L. Nachlinger. 1996. Spring Mountains ecosystem: Vulnerability of spring-fed aquatic and riparian systems to biodiversity loss. Report to U. S. Fish and Wildlife Service Nevada State Office, 4600 Kietzke Ln., Suite C-125, Reno, Nevada 89502.
Stromberg, J. C., R. Tiller, and B. Richter. 1996. Effects of groundwater decline on riparian vegetation of semiarid regions: The San Pedro, Arizona. Ecological Applications. 6(1):113-131.
Tausch, R. J., D. A. Charlet, D. A. Weixelman, and D. C. Zamudio. 1995. Patterns of ordination and classification instability resulting from changes in input data order. Journal of Vegetation Science. 6: 897-902.
ter Braak, Cajo J. F. 1987. The analysis of vegetation-environment relationships by canonical correspondence analysis. Vegetatio. 69:69-77.
ter Braak, Cajo J. F. 1986. Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology. 67(5):1167-1179.
ter Braak, Cajo. J. F. 1994. Canonical community ordination. Part I: Basic theory and linear methods. Ecoscience. 1(2): 127-140.
ter Braak, Cajo J. F. and Jaap Wiertz. 1994. On the statistical analysis of vegetation change: a wetland affected by water extraction and soil acidification. Journal of Vegetation Science. 5:361-372.
Weiss, Andrew D., Stuart B. Weiss, Alisya T. Galo, Jan Nachlinger, and Daniel Pritchett. 1996. Integrating stratified sampling, canonical correspondence analysis, and GIS for predictive vegetation modeling in the Spring Mts. of southern Nevada. Abstract from the Third International Conference/Workshop on Integrating GIS and Environmental Modeling. January 21-25, 1996. Santa Fe, New Mexico.
Weixelman, Dave A., Desiderio C. Zamudio, Karen A. Zamudio and Robin J. Tausch. 1997. Classifying ecological types and evaluating site degradation. Journal of Range Management 50(3): 315-321.
Wiemann, M. C., S. R. Manchester, D. L. Dilcher, L. F. Hinojosa, and E. A. Wheeler. 1998. Estimation of temperature and precipitation from morphological characters of dicotyledonous leaves. American Journal of Botany. 85(12): 1796-1802.
Witkowski, E. T. F. and T. G. O=Connor. 1996. Topo-edaphic, floristic and physiognomic gradients of woody plants in a semi-arid African savanna woodland. Vegetatio 124: 9-23.