CONTEMPORANEOUS AUTOREGRESSIVE-MOVING AVERAGE (CARMA) MODELING IN WATER RESOURCES1

ABSTRACT: The Contemporaneous Autoregressive-Moving Average (CARMA) model is a simple and efficient model that can be used to fit many multivariate hydrological time series. For certain types of multistation river flow systems, the CARMA model is naturally obtained when the physical restrictions of the system or the characteristics of the data are taken in consideration during the formulation of the model. It is shown how the CARMA model can optimally be used to handle multiple time series where the number of observations in each series may be different. Adequate model building techniques, as well as computational and statistical efficient algorithms to estimate the parameters of the model, are given. The methodologies and applications of the CARMA model are illustrated with three examples. It is also shown how the full multivariate ARMA model may lead to losses in efficient of the estimators when the CARMA model is adequate.     

Improving Benefit Transfer Demand Functions: A GIS Approach

Methodologies for transfer of non-market, natural resource recreation benefits are an active research topic. This arises because of the appeal of modelling the impacts of possible changes in site quality or benefits at unsurveyed sites. However, successful benefit transfer must necessarily rely on development of reliable visitor demand functions that incorporate travel time, demographic and substitute factors. Previous efforts to include all of these elements in a single arrivals model are rare. By integrating data from numerous sources within a geographical information system (GIS) we have developed a model to predict the number of visitors to a recreational woodland in eastern England. Variables were classified into discrete groups that were combined into comparatively homogeneous zones from which to calculate visit rates. Poisson regression techniques were then applied in a stepwise procedure to assess the influence of each determinant. Our analysis highlighted both substantial promise and some caveats in using GIS for future benefit transfer work.     

Integrated Assessment: an emerging methodology for complex issues

Over recent years Integrated Assessment (IA) has been presented as a new and innovative approach to modern decision making. In this paper we explore some of the key features that we consider characterise IA; we present IA as a framework that allows a flexible approach to uncertainty in decision making and to which a broad palette of expertise can contribute. The example of the development of an international agreement to reduce acidic deposition is used to illustrate the current practical status of an IA. The scientific process leading to this agreement has previously been described as IA but we argue that, although it marks a change from traditional approaches, further progress is still required to satisfy the requirements for a successful and durable IA.     

Testing the MAGIC acid rain model in highly organic,low-conductivity waters using multiple calibrations

Accurate predictions of acid precipitation effects on water resources are important in order to allow a better understanding of various pollution control strategy outcomes. Dynamic geochemical models have been developed to address this need, but have to be tested under a variety of environmental conditions to provide confidence in their predictions. The most commonly used aquatic acidification model in North America and Europe is the model of acidification of groundwater in catchments (MAGIC). Though extensively used, MAGIC has never been tested in catchments with extremely low ionic strength water and high in natural organic acids (NOAs) from wetlands, two conditions which are common in large parts of Canada. We calibrated the model for two catchments located in Nova Scotia, Canada, which had some of the most dilute freshwaters reported in the literature and very high NOA. We also evaluated the variability inherent in calibration data sets by using five separate data sets collected over a 15-year period at the same sites. We show good model simulations for the main cations and anions in catchment waters. However, modeling pH is more difficult in the highly organic waters and requires modification to the acid dissociation constants. Calculated acid neutralization capacity can also be more difficult to model due to the low ion content making small errors more important. In theory, multiple calibrations of a model at a same site should produce identical hindcasts and predictions. In reality, the multiple calibrations produced a series of similar, but not identical outcomes which give a probable range of past values and future outcomes. We feel that this practical approach to validation is a useful addition to the arsenal of model testing tools.     

Plant growth rates and seed size: a re-evaluation

Small-seeded plant species are often reported to have high relative growth rate or RGR. However, because RGR declines as plants grow larger, small-seeded species could achieve higher RGR simply by virtue of their small size. In contrast, size-standardized growth rate or SGR factors out these size effects. Differences in SGR can thus only be due to differences in morphology, allocation, or physiology. We used nonlinear regression to calculate SGR for comparison with RGR for 10 groups of species spanning a wide range of life forms. We found that RGR was negatively correlated with seed mass in nearly all groups, but the relationship between SGR and seed mass was highly variable. We conclude that small-seeded species only sometimes possess additional adaptations for rapid growth over and above their general size advantage.     

Habitat-mediated variation in predation risk by the American marten

The probability of prey encounter, attack, capture, and kill are often hypothesized to depend on habitat structure, but field evidence in terrestrial systems is rare. We tested whether predation efficiency by the American marten (Martes americana) and fear of predation by their primary prey, the red-backed vole (Clethrionomys gapperi), differed between 20- to 50-year-old regenerating forest stands and older uncut stands. Our results showed that the frequency of prey encounter, prey attack, and prey kill were higher in old uncut forests, despite the fact that small-mammal density was similar to that in younger logged forests. These differences in predation efficiency were linked to higher abundance of coarse woody debris, which seems to offer sensory cues to martens, thereby increasing the odds of hunting success. Red-backed voles in regenerating forest stands exhibited increased wariness compared to voles living in old uncut forest, suggestive of a behavioral response to habitat-mediated variation in predation risk.