Assessing and testing prediction uncertainty for single tree-based models: A case study applied to northern hardwood stands in southern Québec, Canada

Estimating prediction uncertainty for a single tree-based model is hindered by the complex structure of these models. In this paper, we addressed this issue with a case study applied to northern hardwood stands in Québec, Canada. SaMARE is a stochastic single tree-based model that was designed for these types of stands. Using a Monte Carlo approach, the model can provide a mean predicted value and its confidence limits for some plot-level attributes.The mean predicted values were compared to observed values in terms of bias and accuracy. In addition to these common statistics, we compared nominal coverage of Monte Carlo-simulated confidence intervals with real (observed) coverage to verify the adequacy of the simulated uncertainty. A comparison was made using several plot-level attributes, which exhibited an increasing discriminative complexity. This complexity ranges from coarse attributes, such as all-species basal area, up to more complex ones, such as basal area for stems of a particular species and with sawlog potential.The results showed that in terms of absolute value, biases were small, but could be relatively high with respect to the average observed value when the discriminative complexity of the attribute increased. The comparison between nominal and real coverage of confidence intervals gave satisfactory results for all-species plot-level attributes. However, for some species-specific attributes, the Monte Carlo-simulated confidence intervals overestimated the real coverage.     

Comparing discriminant analysis, neural networks and logistic regression for predicting species distributions: a case study with a Himalayan river bird

We assessed the occurrence of a common river bird, the Plumbeous Redstart Rhyacornis fuliginosus, along 180 independent streams in the Indian and Nepali Himalaya. We then compared the performance of multiple discrimant analysis (MDA), logistic regression (LR) and artificial neural networks (ANN) in predicting this species’ presence or absence from 32 variables describing stream altitude, slope, habitat structure, chemistry and invertebrate abundance. Using the entire data (=training set) and a threshold for accepting presence in ANN and LR set to P≥0.5, ANN correctly classified marginally more cases (88%) than either LR (83%) or MDA (84%). Model performance was assessed from two methods of data partitioning. In a ‘leave-one-out’ approach, LR correctly predicted more cases (82%) than MDA (73%) or ANN (69%). However, in a holdout procedure, all the methods performed similarly (73–75%). All methods predicted true absence (i.e. specificity in holdout: 81–85%) better than true presence (i.e. sensitivity: 57–60%). These effects reflect species’ prevalence (=frequency of occurrence), but are seldom considered in distribution modelling. Despite occurring at only 36% of the sites, Plumbeous Redstarts are one of the most common Himalayan river birds, and problems will be greater with less common species. Both LR and ANN require an arbitrary threshold probability (often P=0.5) at which to accept species presence from model prediction. Simulations involving varied prevalence revealed that LR was particularly sensitive to threshold effects. ROC plots (received operating characteristic) were therefore used to compare model performance on test data at a range of thresholds; LR always outperformed ANN. This case study supports the need to test species’ distribution models with independent data, and to use a range of criteria in assessing model performance. ANN do not yet have major advantages over conventional multivariate methods for assessing bird distributions. LR and MDA were both more efficient in the use of computer time than ANN, and also more straightforward in providing testable hypotheses about environmental effects on occurrence. However, LR was apparently subject to chance significant effects from explanatory variables, emphasising the well-known risks of models based purely on correlative data.     

The role of echinoderms as bioindicators of seawater pollution: A case study from patraicos and corinthiacos Gulf,N. Peloponnesus,Greece

To determine the role of echinoderms as bioindicators of seawater pollution we collected sea‐water and urchin samples from two sites in N. Peloponnesus, Greece. One site was in Patraicos gulf (Patras) and was influenced by point source pollution (urban sewage outfalls) while the second was in Korinthiacos gulf (Akoli) with no evident habitat pollution.

Three heavy metals (copper, nickel and zinc) have been measured in the gonads of two edible species of echinoderms, Arbacia lixula and Paracentrotus lividus.

Statistical treatment of the results showed statistically significant differences in bioaccu‐mulation related to the dry‐weight of gonads, to species, to sex of echinoderms, to season of sampling and to sampling station.

The concentrations of zinc were higher than those of Ni and Cu at in both sampling sites.

Microbiological analysis showed that contamination of the urchins was correlated to that of the surrounding seawater.

Finally both sampling sites seems to be influenced, directly or indirectly, by heavy metal and bacterial habitat pollution.     

Physical damage to coastal dunes and ecological impacts caused by vehicle tracks associated with beach camping on sandy shores: a case study from Fraser Island,Australia

As coastal populations expand, demands for recreational opportunities on beaches and coastal dunes grow correspondingly. Although dunes are known to be sensitive to direct human disturbance and provide irreplaceable ecosystem services (e.g. erosion control, critical habitat and nesting sites), dunes serve as campsites for large numbers of people (∼90,000 p.a.) on the ocean-exposed shores of Fraser Island, Australia. Campsites are located in the established dunes and can only be accessed with 4WD vehicles along tracks cut directly from the beach through the foredunes. Here we quantified the extent of physical damage to foredunes caused by this practice, and tested whether human-induced physical changes to foredunes translate into biological effects. Of the 124 km of ocean-exposed beaches, 122 km (98%) are open to vehicles driven on the beaches, and camping zones cover 28.7 km or 23% of the dunes. A total of 235 vehicle tracks are cut across the foredunes at an average density of eight tracks per km of beach. These tracks have effectively destroyed one-fifth (20.2%) of the dune front in camping zones, deeply incising the dune-beach interface. There is evidence of accelerated erosion and shoreline retreat centred around vehicle tracks, resulting in a “scalloping” of the shoreline. No dune vegetation remains in the tracks and the abundance of ghost crabs (Ocypode spp.) is significantly reduced compared with the abutting dunes. Because current levels of environmental change caused by dune camping may not be compatible with the sustainable use of coastal resources and conservation obligations for the island (listed as a World Heritage Area and gazetted as a National Park), restoration and mitigation interventions are critical. These will require spatial prioritisation of effort, and we present a multi-criteria ranking method, based on quantitative measures of environmental damage and ecological attributes, to objectively target rehabilitation and conservation measures. Ultimately, coastal management needs to develop and implement strategies that reconcile demands for human recreation, including beach camping, with conservation of coastal dune ecosystems.