Data for Evaluation of the Latent Health Factor Index as an Indicator of Ecosystem Health

Two related Access databases containing taxonomic information related to sampling of benthic infauna at a dredge disposal site offshore of Yaquina Bay.

Impact/Purpose

Protecting the environment requires evaluation of an ecosystem’s condition, hereafter referred to as ecosystem health. Further, it is important that this evaluation be done with respect to whether ecosystem health is the outcome of external effects on the observed health, particularly if those effects are imposed by human society. The complexity of ecosystems usually necessitates that many measurements, or metrics, of the ecosystem be obtained and collectively evaluated. Given that multiple metrics have been obtained, the investigator evaluating the ecosystem’s health must then synthesize this information. The process of compiling and analyzing this information leads to the idea of an ecological indicator – a summarization of the combined metrics that presumably gauges ecosystem health. Such indicators are appealing because they are readily communicated to stakeholders, provide a means of comparing health across ecosystems, and can provide insight into the factors that affect the observed ecosystem’s health. A common approach to developing an ecological indicator is to develop what are termed “multimetric” indicators. As the name implies, these measures combine the observed metrics into a numerical value, and the scientific literature contains an extensive number of such indicators. Unfortunately, these indicators can contain methodological flaws that limit their utility. First, they may combine disparate metrics that may not have been standardized to a common base, and hence combining them can be nonsensical. The reader familiar with principal component analysis (PCA) knows that if the collection of variables used in PCA are not comparable in terms of their units or magnitude, then the variables must be standardized to have a zero means and standard deviations of one, i.e., the analysis must use the correlation matrix of the variables, and not their covariances. Second, the investigator may introduce arbitrary modifications to the indicator in an attempt to improve its utility. This is the case with the popular AMBI indicator (Borja et al. 2000) which is used to evaluate marine benthic communities. The calculation of the AMBI indicator includes coefficients to each of the components whose values apparently have not been documented or explained by the developer or in papers using the indicator. Third, and most importantly, many ecological indicators lack a firm statistical basis. An informal definition of a statistic is that it is a numerical quantity that is derived from observational or experimental data. Under that definition, all ecological indicators are also statistics. Yet statistical principles are often not applied in the formulation of the indicator, e.g., the indicators are not provided with accompanying measures of their variance or uncertainty. As a result, comparisons within and between ecosystems cannot justifiably be made. For example, the relative importance of an annual change in an indicator cannot be assessed. Was the change in the magnitude of the indicator value large, and a source of concern, or just a minor variation? Grace Chiu, and her coauthors (Chiu and Guttorp 2006; Chiu et al. 2010; Chiu et al. 2013), have presented an approach that overcomes the above concerns. They have named the indicator they developed the Latent Factor Health Index, or LHFI. They consider health, denoted as H, as a latent variable that cannot be observed directly. They present a generalized linear model and use a Bayesian approach to estimate the parameters in the model. Their model can be expressed as: .   There are multiple advantages to this approach. First, a numerical measure of health for each site,  is obtained, and the variance in these measures are available so that objective comparisons among sites can be made. Second, the influence of the individual metrics (the ) on health can be examined. Third, the role of covariates (as ) is also available for examination.

Citation

Power, Jim. Data for Evaluation of the Latent Health Factor Index as an Indicator of Ecosystem Health. U.S. Environmental Protection Agency, Washington, DC, 2025. [DOI: 10.23719/1532333]

DOI: Data for Evaluation of the Latent Health Factor Index as an Indicator of Ecosystem Health